guidelines for conducting nutrition and mortality surveys
TRANSCRIPT
Guidelines for Conducting Nutrition
and Mortality surveys in Nepal
April 2013
FINAL
1
TABLE OF CONTENTS
GLOSSARY OF ACRONYMS ................................................................................................. 4
1. INTRODUCTION .............................................................................................................. 5
2. PLANNING THE SURVEY .............................................................................................. 6
2.1 Co-ordination of nutrition surveys in Nepal .............................................................. 6
2.2 Procedures to undertake surveys in Nepal ................................................................. 6
2.3 Decision/justification to conduct a survey ................................................................. 7
2.3.1 Review of secondary information ...................................................................... 7
2.4 Defining the goals and objectives .............................................................................. 8
2.5 Defining geographic areas and population group ...................................................... 8
2.5.1 Geographic area ................................................................................................. 8
2.5.2 Population groups .............................................................................................. 9
2.6 Timing of the survey ................................................................................................ 10
2.7 Meet the community leaders and local authorities .................................................. 11
3.1 Fundamentals of sampling ....................................................................................... 12
3.2 Representativeness and randomness ........................................................................ 12
3.3 Sampling error, probability, and confidence intervals ............................................. 12
3.4 Calculating the sample size ...................................................................................... 13
3.4.1 Expected prevalence (or expected death rate for death rate surveys) .............. 14
3.4.2 Precision level .................................................................................................. 14
3.4.3 Design effect .................................................................................................... 15
3.5 Correction for small population size ........................................................................ 16
3.6 Converting sample size in number of individuals into number of households ........ 17
3.6.1 Percentage of non-response ............................................................................. 18
3.7 Sample size calculation for nutrition surveys .......................................................... 18
3.8 Sample size for the death rate surveys ..................................................................... 18
3.8.1 Recall period .................................................................................................... 19
3.9 Reconciling sample sizes in combined surveys ....................................................... 21
3.10 Sampling methodologies .......................................................................................... 22
3.10.1 Simple random sampling ................................................................................. 22
3.10.2 Systematic random sampling ........................................................................... 25
3.10.3 Cluster sampling .............................................................................................. 27
3.11 Important considerations when selecting subjects ................................................... 35
3.11.1 Polygamous families ........................................................................................ 35
3.11.2 No substitution ................................................................................................. 36
3.11.3 Measure all the children ................................................................................... 36
3.11.4 No children....................................................................................................... 36
3.11.5 Empty houses ................................................................................................... 36
3.11.6 Absent children ................................................................................................ 36
3.11.7 Disabled children ............................................................................................. 37
3.11.8 Child in a centre ............................................................................................... 37
4. MEASUREMENT TECHNIQUES .................................................................................. 38
2
4.1 Nutrition survey data................................................................................................ 38
4.1.1 Inclusion criteria .............................................................................................. 38
4.1.2 Estimating age .................................................................................................. 38
4.1.3 Measuring weight............................................................................................. 39
4.1.4 Measuring length or height .............................................................................. 41
4.1.5 Measuring nutritional oedema ......................................................................... 43
4.1.6 Measuring Mid-upper arm circumference (MUAC) ........................................ 44
4.1.7 Estimating the nutrition status (for referral) .................................................... 46
4.1.8 Recording anthropometric information............................................................ 46
4.2 Death rate survey data .............................................................................................. 46
4.2.1 Crude Death Rate: household census ............................................................... 46
4.2.2 Common problems in recording individual information for mortality ............ 49
4.3 Additional data ......................................................................................................... 50
4.3.1 Deciding what additional information to collect ............................................. 50
4.3.2 IYCF data ......................................................................................................... 51
4.3.3 Food security data ............................................................................................ 51
4.3.4 Health data ....................................................................................................... 51
4.3.5 WASH data ...................................................................................................... 53
4.3.6 Additional qualitative data ............................................................................... 53
5. SURVEY IMPLEMENTATION ..................................................................................... 55
5.1 Preparing for data collection .................................................................................... 55
5.1.1 Obtaining and preparing equipment, supplies, and survey materials .............. 55
5.1.2 Surveyor's manual ............................................................................................ 56
5.2 Selecting and training the survey team .................................................................... 57
5.2.1 Selecting the survey teams ............................................................................... 57
5.2.2 Training survey team members ........................................................................ 58
5.2.3 Standardization of weight, height, and MUAC measurements ........................ 59
5.2.4 Pre-testing ........................................................................................................ 62
5.3 Managing the survey ................................................................................................ 63
5.4 Enhancing the accuracy of the data collected .......................................................... 63
5.5 Supervising data collection team ............................................................................. 64
5.6 Minimising Bias ....................................................................................................... 64
5.7 Ethical considerations .............................................................................................. 65
6. DATA ENTRY AND DATA QUALITY CHECK .......................................................... 67
6.1 Data entry ................................................................................................................. 67
6.1.1 Data Entry: Nutrition Survey Data .................................................................. 67
6.1.2 Data Entry: Death Rate Survey Data ............................................................... 73
6.1.3 Using ENA for double-entry ............................................................................ 76
6.2 Determining nutritional status of individuals and populations ................................ 77
6.2.1 Nutrition indices............................................................................................... 77
6.2.2 Mid-upper arm circumference (MUAC) .......................................................... 78
6.2.3 The reference population curves ...................................................................... 78
6.2.4 Expression of nutrition indices ........................................................................ 78
6.3 Assessing data quality .............................................................................................. 79
3
6.3.1 Outliers (flags) ................................................................................................. 79
6.3.2 Distribution ...................................................................................................... 80
6.3.3 Sex ratio ........................................................................................................... 81
6.3.4 Age distribution ............................................................................................... 81
6.3.5 Digit preference for height and weight ............................................................ 82
6.3.6 Skewness .......................................................................................................... 82
6.3.7 Kurtosis ............................................................................................................ 83
6.3.8 Analysis by team .............................................................................................. 84
6.3.9 Overall data quality .......................................................................................... 84
7. DATA ANALYSIS .......................................................................................................... 85
7.1 Data Analysis: Nutrition Survey Data ..................................................................... 85
7.1.1 Classification of malnutrition .......................................................................... 85
7.1.2 Nutrition survey results .................................................................................... 87
7.2 Data Analysis: Death Rate Survey Data .................................................................. 88
7.2.1 Calculating death rates ..................................................................................... 88
7.2.2 Using ENA to analyse death rate survey data .................................................. 89
7.3 Data analysis: Other data ......................................................................................... 91
7.4 Data Analysis: Qualitative data ............................................................................... 92
8. INTERPRETATION OF RESULTS ................................................................................ 94
8.1 Interpreting the results ............................................................................................. 94
8.1.1 Comparing the results with establish thresholds .............................................. 94
8.1.2 Comparing results with previous survey results .............................................. 96
8.1.3 Analysing the context ...................................................................................... 96
8.1.4 Using UNICEF conceptual framework ............................................................ 97
8.2 Presenting the results, writing the report ................................................................. 97
8.3 Making recommendations ........................................................................................ 98
8.4 Planning the response .............................................................................................. 99
Annex 1: Survey proposal format .......................................................................................... 100
Annex 2: Preliminary survey report format ........................................................................... 101
Annex 3: Random number table ............................................................................................ 102
Annex 4: Decision tree for selecting households at the last stage of cluster sampling ......... 103
Annex 5: Local events calendar ............................................................................................. 104
Annex 6: Weight-for-height z-score table, WHO 2006 Child Growth Standards ................. 105
Annex 7: Nutrition and death rate survey sample questionnaire ........................................... 107
Annex 11: Example of standardization test data collection forms ........................................ 123
Annex 12: Cluster control form ............................................................................................. 124
Annex 13: Final survey report format .................................................................................... 125
REFERENCES ...................................................................................................................... 129
4
GLOSSARY OF ACRONYMS
95% CI 95% Confidence Interval
BCG Tuberculosis vaccine
CBS Central Bureau of Statistics
CDC Centre for Disease Prevention and Control, Atlanta, USA
CDR Crude Death Rate
DHS Demographic and Health Survey
ENA Emergency Nutrition Assessment software
EPI Expanded Program of Immunization
GAM Global Acute Malnutrition
GPS Global Positioning System
HFA Height-For-Age
HH Household
HMIS Health Management Information System
ID Identification number
IDP Internally Displaced Person/People
IYCF Infant and Young Child Feeding
LZ Livelihood Zone
MAM Moderate Acute Malnutrition
MHP Ministry of Health and Population
MSNP Multi-Sectorial Nutrition Plan
MUAC Mid-Upper Arm Circumference
NCHS National Centre for Health Statistics (USA)
NGO Non-Government Organization
PPS Probability Proportional to Size
OTP Outpatient Therapeutic Programme
PSU Primary Sampling Unit
SAM Severe Acute Malnutrition
SD Standard Deviation
SFP Supplementary Feeding Programme
SMART Standardized Monitoring and Assessment of Relief and Transition
TB Tuberculosis
UN United Nations
UNHCR United Nations High Commission for Refugees
UNICEF United Nations Children Fund
WFA Weight-For-Age
WFH Weight-For-Height
WFP World Food Programme
WHO World Health Organization
5
1. INTRODUCTION
The guidelines for conducting nutrition surveys in Nepal provide step-by-step instructions on
how to conduct a nutrition survey in Nepal. The aim of the guidelines is to provide guidance
on how to conduct nutrition surveys with internally agreed standards in Nepal, standardise the
survey protocols, and improve the quality of the nutrition surveys conducted in Nepal.
These guidelines are based on the latest Standardized Monitoring and Assessment of Relief
and Transition (SMART) methodology recommendations and reflect the current international
standards for conducting surveys adapted to the Nepalese context.
The nutrition survey guidelines are intended for small scale surveys where estimates are
required for smaller geographic units such as districts for which estimates are not usually
calculated from the national level surveys. Some of the contexts in which these guidelines
could be used include emergency situations, baseline assessments, programme monitoring,
and end line assessments. Nevertheless, basic principles and measurement techniques can be
used in a survey of any scale.
The guidelines also include information on how to conduct a mortality survey. However, it is
not necessary to carry out a mortality survey along with every nutrition survey. The decision
to include mortality during a nutrition survey should be made prior to the survey and
justified. In most cases, the mortality survey would be conducted along with a nutrition
survey following an emergency.
Nutrition survey findings are interpreted in light of other contextual factors such as feeding
practises, food security, health, water and sanitation. Wherever applicable, the relevance of
colleting additional information has been discussed in the guidelines. A sample survey
questionnaire is included in the guidelines. This questionnaire must be reviewed before each
survey and revised taking into consideration of the local context and other programmatic
issues before it is used in a survey.
The guidelines recommend the use of the Emergency Nutrition Assessment (ENA) for
SMART software in planning, implementing, analysing, and reporting on nutrition surveys
and guidance is provided throughout the manual on how to use the software. For additional
information on the ENA for SMART software and the SMART methodology, users are
referred to the following web page: www.smartmethodology.org
The guidelines were developed by the Child Health Division, Ministry of Health and
Population and the Central Bureau of Statistics of Nepal with financial and technical
assistance from UNICEF Nepal.
6
2. PLANNING THE SURVEY
2.1 Co-ordination of nutrition surveys in Nepal
Ministry of Health and Population (MoHP) and the Central Bureau of Statistics (CBS) are
responsible for nutrition and death rate surveys conducted in Nepal. These government
entities should be consulted prior to undertaking any surveys in Nepal.
The Nutrition Working Group (NWG) that has been established under the Department of
Child Health will co-ordinate and technically oversee all nutrition surveys carried out in the
country with technical assistance from the CBS.
2.2 Procedures to undertake surveys in Nepal
Any agency that is planning to carry out a nutrition survey at the district level should contact
the Department of Child Health at least one month in advance regarding the planned survey.
The Department of Child Health will inform the NWG at the national level and co-ordinate
the surveys.
As soon as an agreement is reached with the Department of Child Health authorities
regarding the type of survey and the type of information to be collected, the survey
implementing agency should submit an electronic copy of the technical proposal (see annex 1
for the format of the technical proposal) to the Department of Child Health. The proposal will
be reviewed by the NWG members, discussed at a meeting, and feedback, if any, will be
provided to the agency within 2 weeks of the receipt of the proposal.
At least one member of the Department of Child Health and CBS should be involved in every
step of the nutrition survey. Every effort should be made to build the capacity of the
Department of Child Health and CBS staff during the survey.
As soon as the survey results are available, a meeting should be organised at the field level
(where the survey had taken place) with the MoHP and CBS officers to discuss the findings
(see section 8.3 for details). The survey report should take into account the outcome of this
meeting.
A preliminary report summarising the survey findings should be prepared within 2 weeks of
completing the data collection using the survey format in annex 2 and submitted to the NWG
for technical review and validation. The members of the NWG group will review the report
and provide feedback, if any, to the survey implementing agency. The final report along with
the survey datasets should be submitted to MoHP within one month of receiving technical
clearance for the preliminary report from the NWG.
7
Co-ordination of nutrition surveys in MSNP districts:
All nutrition surveys conducted in the MSNP districts should be initiated at the district level
before the Department of Child Health is involved. The district level MoHP and CBS
representatives in the MSNP districts should be consulted and the details of the surveys
should be agreed upon prior to contacting the Department of Child Health representatives at
the national level. The district level MoHP and CBS officials should be involved in every
step of the survey, including data analysis and reporting. The results of the survey should first
be presented and discussed with these district level staff before a preliminary report is
prepared and submitted to the NWG. Every effort must be made to build the capacity of these
district level staff to independently carry out surveys by themselves in the future.
2.3 Decision/justification to conduct a survey
Conducting a nutrition survey is a time and resource consuming exercise. The decision to
carry out a nutrition survey should be carefully considered and justified. The rationale behind
the need to undertake a survey may differ depending of the utilization of survey results:
- The need to obtain baseline nutrition information;
- The need to have nutrition information to monitor an intervention or assess its
impact;
- The need to get hold of nutrition information to confirm an emergency and/or
advocate for a response;
- The need to have disaggregated information to identify high risk groups, to estimate
the number of beneficiaries, or to better target a response.
Regardless of the purpose of the survey, the available information regarding the current
situation should first be reviewed before conducting any nutrition survey. The information
that should be reviewed includes previous surveys in the area, national surveys such as Nepal
Demographic and Health Survey, Nepal Multiple Indicator Cluster Survey etc., health
management information system, food security and livelihood assessments, weather
conditions, security, rapid assessments (if any), etc. Additionally, it is also important to
collect information about the population characteristics and figures, livelihood patterns, etc.
of the survey area.
The following should be kept in mind when deciding to carry out a nutrition survey:
If the needs of a specific population are clear, interventions rather than surveys need
to take precedent. In these situations, a rapid assessment will be sufficient to start the
intervention while a nutrition survey can wait till the immediate needs of the affected
population have been met.
Surveys should be carried out only if there is capacity to respond if the survey results
warrant a response. It is unethical to carry out a survey when there is no capacity or
resources to respond if the survey results show a deteriorating situation.
If there is no access to the survey population due to security or other reasons, no
survey can be carried out.
2.3.1 Review of secondary information
Additional information on other factors such as health, water and sanitation, food security
and livelihood, etc. is required to put the malnutrition prevalence in context and interpret the
8
survey results accordingly. This additional information should ideally be collected through
review of secondary data such as food security and WASH sector assessments, other surveys
in the area, health facility data, programme data and plans etc. wherever possible. In
situations where it is necessary to collect this additional information during the survey, the
first choice should be to collect them through qualitative methods such as focus group
discussions, key informant sessions, etc. It should be kept in mind that each additional piece
of information added to the survey questionnaire compromises the quality of data collected.
Therefore, additional information included in the questionnaire should be reviewed carefully
and justified (see section 4.4.1 for additional details).
The review of secondary information along with the information needs for programmatic
decisions should also inform whether a nutrition survey alone will be adequate or there is a
need to combine the nutrition survey with a death rate survey as well. It should be noted that
although these guidelines describe how to conduct an integrated nutrition and death rate
survey, it is not necessary to always carry out nutrition and death rate surveys together.
Similarly, it is not mandatory to collect information on every single indicator that may
have impact of malnutrition in every survey.
2.4 Defining the goals and objectives
Once the decision to conduct the survey has been justified and all available secondary
information is reviewed, the goals and objectives of the survey should be clearly defined and
stated.
The goal (sometimes also referred to as general objective) is a more general statement about
the survey. For example, a goal of a nutrition survey may be to measure the severity of
malnutrition in an area.
The objectives are more specific statements about what the survey intends to do. A clearly
stated objective should include information about the outcome to be measured, the target
group, and the survey area. For example, a clearly formulated objective would be to
measure the prevalence of Global Acute Malnutrition among the people living in Achham
district.
2.5 Defining geographic areas and population group
2.5.1 Geographic area
The survey area should be clearly defined and a map of the survey area should be provided. If
there are sections within the survey area that are inaccessible due to security and/or other
reasons, it should be clearly mentioned and identified in the map. The survey findings cannot
be generalised to the areas that were not included in the survey although nutrition situation in
the areas excluded from the survey maybe discussed in the survey report.
Surveys may be conducted in areas where an agency already has a programme or in new
areas based on anecdotal evidence that the malnutrition situation maybe deteriorating in those
areas. This anecdotal evidence may include worsening situation reflected in a rapid
assessment or food security assessment, outbreak of diseases, increased admission into
9
hospital based feeding centres, etc. The decision to conduct a survey should always be
provided and justified.
Traditionally, surveys have been confined to administrative areas (e.g. districts). While this is
useful for several purposes, the homogeneity of the district in terms of malnutrition levels
(i.e. GAM) should be taken into consideration when surveys are planned. If there are
different population sub-groups within an administrative area such as urban and rural or high-
land and low land and GAM levels between different sub-groups are expected to vary
significantly, separate surveys should be carried out. This is because, if there are two
livelihood zones in an administrative area with varying levels of GAM levels between them,
one survey conducted at the administrative level will only provide an aggregated estimate for
both areas. This will not reflect the actual GAM situation in the survey area. In these cases,
two separate surveys should be carried out in the district to find out the GAM levels in both
populations. The two surveys can then be combined to get the district prevalence (using
appropriate statistical procedures), as described in the analysis section. Similarly, if the issue
of interest, for example, is malnutrition among urban slum population, separate surveys
should be carried out.
It should be noted that once a survey is carried out it is not possible to disaggregate the
survey estimate and get different estimates for different population sub-groups. This must be
taken into consideration at the planning stage of the survey.
The information about the different GAM levels in different population sub-groups in a
survey can be obtained from different sources. Some of these sources include previous
surveys (e.g. high design effect, bi-model distribution of GAM, etc.), food and livelihood
situation in different livelihood zones, and hospital and feeding centre statistics, or
discussions with key informants.
As much as possible, agencies implementing programmes or planning surveys in different
parts (e.g. districts) of a wider administrative unit (e.g. county) should plan together and
conduct surveys at the same time so that the results can be combined to get a wider area
estimate. For example, if there are 6 districts in a particular region and 3 agencies are
conducting surveys in different district it is recommended that the surveys are planned and
implemented together so that in addition to the district level estimates, a region level estimate
can also be obtained.
2.5.2 Population groups
The target group for an anthropometric survey is usually children between 6-59 months, and
for a crude death rate survey is the entire population. The target group for other indicators
will vary. For example, target groups for IYCF indicators vary depending on the type of
indicator – e.g. 0-5 months for exclusive breastfeeding, 6-23 months for minimum dietary
diversity, etc. Although the aim is to get an overall idea about the nutrition situation in the
whole population of interest, the survey is conducted among children 6-59 because this
population sub-group is considered to be the most sensitive to acute nutrition stress.
Additionally, this population group is relatively easily accessible, agreed standards exist for
10
interpreting survey results from this population, and there is often baseline data for this age
group.
Although acute malnutrition prevalence among children 6-59 months acts as a good proxy for
the nutrition situation in the entire population in which the survey is carried out, in some
cases, other population sub-groups such as adolescents, adults, the elderly, etc. may be of
particular concern. In these cases separate surveys are recommended to assess the nutrition
situation in these population subgroups using the principles outlined in these and other
international guidelines.
However, it must be noted that although surveys generally are conducted only among
children 6-59 months, it cannot be used to justify confining interventions to this age group. If
surveys need to be conducted and prevalence estimates need to be available for each
population sub-group before it received an intervention, surveys would become very
cumbersome. Every malnourished individual should be eligible for interventions. If specific
information is needed for a particular population sub-group for programmatic purposes,
however, a separate survey can be carried out.
2.6 Timing of the survey
The timing of an assessment largely depends on the objectives of the survey (baseline survey,
response to a crisis, on-going annual monitoring, etc.). Although the exact dates of the survey
needs to be discussed with the community leaders and local authorities, some broad time
period should be set based on the survey objectives. Some things to consider when deciding
on the timing of the survey include accessibility during rainy season, planting and harvesting
seasons in agricultural areas, and migration of people in pastoral settings. The timing of
certain contexts in which surveys often need to be carried out is described below.
In slow-onset emergency situations, surveys should be conducted at the beginning of the
onset or at the beginning of the ‘hunger season’. This will give time to plan and mobilise
resources to respond to the situation should there be a need to respond. The seasonal calendar
of the survey area should be used as the basis for the survey in these situations.
The ‘hunger season’ varies from region to region in Nepal. Thus, if a survey is carried out to
assess the nutrition situation in a slow-onset emergency, the seasonality of the survey area
should be reviewed and surveys should be carried out at the beginning of the ‘hunger season’.
In case of a rapid-emergency situation, the survey needs to take place as soon as the need to
conduct such a survey is justified regardless of the seasonal calendar or timing.
If a nutrition surveys is carried out to monitor or evaluate a nutrition programme
intervention, the timing of the survey needs to be set accordingly. In these cases, the timing
of the baseline survey, the seasonal calendars, etc. needs to be reviewed carefully to plan the
survey and draw appropriate conclusions.
11
2.7 Meet the community leaders and local authorities
Once the geographical area of the survey is defined and objectives set, and time period have
been planned, it is crucial to meet with the community leaders and local authorities to discuss
about the survey. In general, the following items should be discussed at this meeting:
- Why the survey is carried out- i.e. the objectives of the survey. If the community does
not understand the objectives of the survey, they may not cooperate.
- Obtain a map of the survey area to plan the survey. Use this map during the
discussions with the local authorities and community leaders.
- Obtain detailed information on population figures at ward level.
- Obtain information on security and access to the prospective survey area.
- Obtain letters of permission from the local authorities, addressed to the ward leaders.
The letters should explain why the survey is conducted and ask for the population’s
cooperation.
- Agree upon the exact dates of the survey data collection with the community and local
authorities to avoid market days, local events, food distribution days, and other times
when people are like to be away from home.
- Agree how the results will be used. In particular, realistically discuss the prospects for
intervention, steps that will be taken, and types of programs that are likely to be
implemented if the situation is found to be as poor as expected. Do not make promises
that may not be fulfilled.
Survey managers must make sure that the community meetings include representation from
women and women groups in the community and that their opinions are sought. This is of
particular importance when the actual dates of the data collection are set. Survey managers
may need to work with the community leaders in advance to ensure women’s participation in
these meetings.
Note that every effort must be made to provide feedback to the ward regarding the survey
results as soon as they are available.
12
3. METHODOLOGY
3.1 Fundamentals of sampling
Surveys are carried out to measure the prevalence of an indicator (e.g. GAM) in a defined
survey population. Note that death rate surveys are different in that they measure rates rather
than prevalence in the survey population.
One way of getting the prevalence information for an indicator is to measure all eligible
subjects in the target group and calculate the prevalence. In the case of nutrition surveys, it
would mean that measuring every single chid between 6-59 months and for death rate
surveys, collecting information on everyone in the survey population. The prevalence of an
indicator (e.g. crude death rate, GAM, etc.) will then be calculated by using specific criteria
(e.g. z-score<-2 and/or oedema) and classifying the proportion of people who meet the
criteria.
The process of measuring every single eligible child in the survey area is called census or
exhaustive survey and the prevalence obtained from an exhaustive survey is true
population prevalence. However, exhaustive surveys are usually long and costly and
therefore only a group of individuals from the survey population is usually selected and
surveyed. The results from these surveys are then generalised to the entire survey population.
The process of selecting a group of individuals from the survey population is called sampling
and the group that is selected at the end of the sampling process is called sample.
3.2 Representativeness and randomness
Since inference is made about the survey population by studying the sample, it is essential
that the sample is representative of the entire survey population – i.e. the characteristics of
the sample must be similar to that of the survey population. If the sample is not representative
of the survey population, the estimate obtained from the survey cannot be generalised to the
survey population and that the results obtained from the survey will be biased. A sample is
representative if each individual or household in the population has an equal chance of being
included in the sample, and if the selection of one individual is independent from another
individual. A representative sample can be obtained by selecting the sample randomly (see
section 3.11 below for details).
3.3 Sampling error, probability, and confidence intervals
Since only a fraction of the survey population is selected in a sample survey, a sample survey
only provides an estimate of the true population prevalence and the estimate obtained from a
sample survey is almost always different from the true population prevalence. The difference
between the prevalence estimate obtained from a sample and the true population prevalence
is called sampling error1. The sampling error can be reduced by increasing the sample size
but it cannot be completely eliminated in a sample survey. There will always be some
uncertainty about a result obtained from a sample survey due to sampling error.
1 Sampling error is not the the only reason for a difference between the survey estimate and the true population
prevalence. Another reason for this difference is bias, which is discussed in section 5.6.
13
The size of the sampling error can be estimated using statistical procedures and is presented
as confidence intervals (i.e. a range of possible values). The confidence interval shows that
if a survey were repeated many times with the same sample size and method, 95%2 of the
prevalence estimates would fall within the 95% confidence interval. In other words, you can
be 95% sure that the (unknown) true population prevalence will be within the confidence
interval calculated for the survey estimate. The confidence interval is thus an expression of
how certain we are that the actual result in the population is similar to the result obtained
from the survey.
Example 3.1: The malnutrition prevalence measured by a sample survey is 10.5% and the
95% confidence interval is 9.2–13.5%. This means that it is 95% certain that the actual
prevalence (true population prevalence) of malnutrition in the survey area is between 9.2%
and 13.5%.
The confidence interval is an estimate of precision of the result – i.e. how similar the results
would be if the survey were repeated over and over. If the confidence interval is wide,
sampling error may be responsible for a substantial difference between the estimate
calculated from the survey and the true population prevalence. Precision is increased, and the
confidence interval narrowed, with larger sample sizes. The larger the sample size, the
narrower the confidence interval and, if there is no bias, the more certain we are that the
survey result is close to the true population prevalence. Statistically, a large sample size is
preferable. However, it takes more time and resources to manage surveys with larger sample
sizes. Therefore, at the survey planning, it is important to decide on the precision needed
based on the survey objectives and the expected prevalence and calculate sample size
accordingly (see below).
3.4 Calculating the sample size
The sample size is the total number of individuals to be included in a survey to represent the
survey population. Although a larger sample size will achieve greater representation of the
survey population, a larger sample will also prolong the survey, require more resources, and
delay the survey report.
The calculation of the sample size depends on the following factors:
1. Expected prevalence (or estimated death rate in the case of death rate surveys)
2. The width of the confidence interval. This determines the minimum precision around
the estimate
3. Design effect (if the survey is to use cluster sampling)
Calculating a sample size is almost always a trade-off between the ideal and the feasible. On
the one hand, a sample size that is too small gives results with limited precision and therefore
questionable usefulness. On the other hand, increasing sample size beyond a certain level
produces only small improvements in precision, but may imply a disproportionate increase in
cost.
2 Although confidence intervals can also presented in other values such as 99%, 90%, etc. 95% is used here as it
is the only level that is used in most survey contexts.
14
ENA for SMART software should be used to calculate the sample size. Information on how
to calculate sample size using ENA for SMART software is given in Box 3.3 below using the
ENA 2011 (version: Oct. 27, 2011).
3.4.1 Expected prevalence (or expected death rate for death rate surveys)
The expected prevalence is the prevalence that is to be estimated by the survey such as the
malnutrition prevalence, prevalence of exclusive breastfeeding, etc. (for death rate surveys, it
would be expected death rate). The expected prevalence can be estimated from prior surveys
conducted in the same area, from a survey conducted in a similar adjacent area, from
routine/surveillance data, or from the results from a rapid assessment. Note that the
prevalence information from previous surveys and other sources needs to be carefully
reviewed based on the current situation in the survey area before it is used. For example,
seasonality of the survey must be taken into account when using prevalence information from
a previous nutrition survey as malnutrition prevalence is affected by seasonality. Rapid
assessment results should be treated with caution as they are usually not representative. If
previous survey results are not available, the current prevalence can be estimated by
consulting key informants in the area. In these cases, a range rather than a single value
should be explored (for example, malnutrition maybe 10-15% rather than 12%) and the upper
value of the range should be used to calculate the sample size.
3.4.2 Precision level
The precision indicates the width of the confidence interval of the survey estimate. A higher
precision (i.e. narrower confidence intervals) requires a larger sample size. In general, the
lower the prevalence the greater the precision needed.
Example 3.2: A precision of ±5% is not meaningful in a survey area where the expected
malnutrition prevalence is 5%. This is because the confidence intervals for the survey
estimate will be around 0-10%, encompassing a situation where there is no malnutrition to
one in every 10 children is malnourished. Similarly, if the expected malnutrition prevalence
is 40%, a precision of 5% would be unnecessarily too high.
There is no ‘standard’ precision for any given situation. The survey objectives should be used
to decide on the level of precision. For example, if a survey is meant to simply quantify the
level of malnutrition in an area, a low precision level (5-10%) may be sufficient. However, if
the survey results are to be compared to a baseline or a follow-up survey, a higher precision
level (2-3%) is necessary in order to ensure that any differences between two or more
situations are detected. Table 3.1 below shows examples of the precision needed at various
levels of malnutrition prevalence. These should, however, be reviewed in light of the survey
objective before they are used to calculate the sample size for a survey.
15
Table 3.1: Example of precision needed at various levels of malnutrition prevalence
For death rate surveys, it is not usually possible to achieve a precision much greater than 0.3
deaths/10,000/day with a survey of a reasonable size and a three-month recall period. If
higher precision is required, the recall period would need to be lengthened. The table 3.2
below provides examples of precision usually needed at the various level of estimated CDR.
Table 3.2: Example of precision needed at various levels of death rate
CDR
(Crude Death
Rate∕10,000∕Day)
Confidence Interval
Desired Precision
(±Precision∕10,000∕Day)
0.5 0.2 – 0.8 0.30
1.0 0.6 – 1.4 0.40
1.5 1.0 – 2.0 0.50
2.0 1.25 – 2.75 0.75
3.0 2.0 – 4.0 1.00
3.4.3 Design effect
When calculating the sample size for surveys using cluster sampling design (see section
3.11.3 below), a correction factor accounting for heterogeneity of the outcome being
measured (i.e. GAM or death rate) among clusters in the population must be used. This
correction factor is called design effect. The design effect is low in homogeneous populations
and high in heterogeneous populations.
For example, if the prevalence of malnutrition among different wards (i.e. clusters) in a
survey area is not vastly different from one another, the survey area is considered as
homogeneous in terms of malnutrition prevalence. In these cases, the design effect will be
low. On the other hand, if some wards (i.e. clusters) have high malnutrition prevalence while
others have very low levels of malnutrition, the survey area is heterogeneous with regards to
malnutrition. In this case, the design effect is likely to be high.
The design effect can be obtained from previous surveys conducted in the area or from
surveys carried out in similar areas. However, it should be adjusted for any possible changes
Expected Prevalence (%) Confidence Interval Desired Precision (± %)
5 3 – 7 2.0
7.5 5 – 10 2.5
10 7 – 13 3.0
13 10 – 16 3.0
15 11 – 19 3.0
20 15 – 25 5.0
30 22.5 – 37.5 7.5 40 30 – 50 10.0
16
that may have increased or decreased heterogeneity in the survey area after the previous
survey. Table 3.3 below provides examples that can be used based on the context.
Table 3.3: Example of design effects
Design Effect Context
1.0 - 1.5 Slight differences seen between clusters.
1.5 – 2.0 Differences seen between clusters
2.0 - 2.5 High variation between clusters, such as population from different
livelihood zones.
2.5 -3.0 Some clusters are not affected and others are severely affected
The design effect depends on the expected levels of prevalence and the size of clusters.
- The design effect is low at low levels of prevalence. For example, if expected
prevalence of malnutrition is around 10%, the design effect may be around 1.5. When
the malnutrition prevalence is around 25-30%, the expected design effect may be
around 1.6-1.7.
- The smaller the cluster (i.e. number of children per cluster), the smaller the design
effect. For example, when there 15 children per cluster, the design effect may be
around 1.5, whereas when the number of children per cluster is 25-30, the expected
design effect may be around 1.7-1.8.
Note that, for a death rate survey, if violence accounts for most of the death in the survey
population the design effect can be very high (up to 10). This is because violence is very
rarely evenly distributed in time or place. Such high design effects require very large sample
sizes if meaningful data are to be produced.
It should be noted that surveys should only be carried out in reasonably homogeneous
population. Since nutrition is the primary outcome of a nutrition survey, the heterogeneity of
the malnutrition prevalence between clusters in the survey area should be used to determine
whether one survey would provide reasonable estimate for the entire survey area or multiple
surveys are needed. As described in section 2.4.1, surveys should not be carried out in highly
heterogeneous populations such as populations living in different geographical zones. If a
survey is conducted in more than one livelihood zone in which nutrition status are probably
different, the design effect will be high. If the design effect is suspected to be higher than 2.0,
it is recommended that separate surveys with lower design effect and subsequently lower
sample size be carried out.
3.5 Correction for small population size
The size of the survey population does not affect the sample size for large populations.
However, sample size calculation needs to take into account the population size when the
total population in the survey area is small. Sample sizes should be adjusted to take into
account the population size whenever the total population size in the survey area is less than
10,000 individuals. The ENA software (version: Oct. 27, 2011) can be used to adjust sample
sizes when the population size is small.
17
3.6 Converting sample size in number of individuals into number of households
Once the sample size is calculated in number of target individuals (i.e. number of children
between 6-59 months for nutrition surveys, number of people for death rate surveys, etc.), it
should be converted into number of households. The reasons for this exercise are as follows.
a) During the second stage sampling in cluster surveys, it is highly recommended that
simple or systematic random sampling technique is used (see section 3.11.3 below).
In order to carry out the simple random sampling, a list of all children or households
in the selected cluster area is required (for systematic random sampling, either a list or
a systematic arrangements of households is required). It is easier to obtain or create a
list of households rather than list of children in a ward. Thus, having a number of
households to visit rather than number of children to survey makes it practical.
b) When combined surveys are carried out, sample size is calculated not only for
nutrition surveys but also for death rate surveys (although it is not required to always
integrate all 2 of them). To compare and reconcile sample sizes for different
components of the survey, it is important to use one unit, household.
c) Often additional information such as death rates, food security, water and sanitation,
etc. is collected during a survey. When enumerators are provided with a number of
target children to reach in a cluster, they may only visit the households with children.
This will introduce bias in the estimates obtained for the other indicators.
There are various ways to convert the number of target individuals into number of households
based on the type of information available. The ENA 20113 software uses the percentage of
children under 5 and average household size in the survey to covert the sample size in
number of target individuals into number of households because this information is usually
widely available. The ENA 2011 software should be used to calculate the number of target
individuals into number of households for nutrition and death rate surveys.
Information on average household size and percentage of children under 5 can be obtained
from various sources such as CBS, national level surveys (e.g. DHS and MICS), or localised
surveys (e.g. SMART).
Note that once the percentage of children under 5 is entered, the ENA 2011 software will
automatically calculate the 6-59 month age group assuming that 90% of the children under 5
are 6-59 month age group. This will then be used to calculate the number of households.
One implication of converting the sample size into number of households is that not all
clusters will have the exact number of children. This should be clearly explained to the
survey enumerators to avoid confusion and unnecessary stress during data collection - the
survey teams should only be given a target number of households to survey in each cluster.
However, the total number of children included in the survey and the sample size originally
calculated in terms of number of children should roughly be the same.
3 Note that the ENA software is updated frequently; it is recommended that the latest version available at the
time of the conducting the survey should be used.
18
3.6.1 Percentage of non-response
It may not be possible to visit or collect data from all the households during a survey due to
absence of children, inaccessibility, or refusal of households to participate in the survey,
inability to measure all children in a household, etc. Sometimes, it may also be necessary to
exclude some data from the analysis. These cases are collectively described as non-response.
The survey should anticipate these situations and increase the sample size to account for these
contingencies. Note that non-response can be obtained from previous survey reports.
Failure to account for this may result in reduced sample size during analysis, which will lead
to decreased precision of the estimates. Once the expected non-response is entered in
percentage, the ENA software will automatically calculate and display the final sample size
taking into account the non-response rate.
3.7 Sample size calculation for nutrition surveys
Box 3.1: Steps to follow in calculating sample size for a nutrition survey
[The steps below provide instructions on calculating the sample sizes and converting them
into number of households for malnutrition using the ENA software (version: Oct. 27, 2011)]
Decide on the expected malnutrition (i.e. GAM) prevalence
o This can be estimated from a previous survey in the area, from a similar area, or
from a regional or national survey (note: this needs to be reviewed and adjusted as
needed)
o If no information is available, get an estimate from key informants.
Decide on the precision needed
o It should be decided based on the objectives of the survey.
Decide on the design effect if it is a survey using cluster sampling design
o For most surveys, the design effect will be around 1.5
o If there have been surveys done in the same area in the recent past, use design
effect from this
Estimate the average household size in the survey area
o This can be obtained from a previous survey in the area, from a similar area, from
a regional or national survey, from census data, or from the CBS.
Estimate the percentage of children under 5 in the survey area
o This can be obtained from a previous survey in the area, from a similar area, from
a regional or national survey, from census data, or from the CBS.
Estimate the expected percentage of non-response households
o For most surveys the non-response will be around 3-5%
3.8 Sample size for the death rate surveys
Calculation of sample size for death rate surveys is similar to the calculation of sample size
calculation for nutrition surveys. However, for death rate estimates, one additional factor,
called recall period, must also be taken into account as death rate surveys estimates rates
rather than prevalence.
19
3.8.1 Recall period
The recall period for the death rate survey is the time interval over which deaths are counted.
The length of the recall period is thus a critical factor in determining the death rate. Other
things being constant, increasing the recall period will reduce the sample size and reducing
the recall period will increase the sample size.
There are important considerations when setting the recall period. The first question should
be: "what is the period most relevant to the purpose of the survey, the risk of death rate being
measured, and the context of the study?" Example 3.3 below shows a scenario where
depending on the recall period selected different mortality events are captured in the survey.
Example 3.3: Selecting a recall period for death rate survey
A death rate survey is planned in a rural population. Events which likely to have had an impact on deaths include
the following:
1. Seasonal food shortages from May to October but particularly from August to October
2. Outbreak of malaria in the season of heavy rains in January and Feb
3. Flare-up in ethnic conflict in March and April, leading to
4. Population displacement in April through June
Event Time
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Food shortage X X X XX XX XX
Malaria outbreak X X
Conflict X XX
displacement X XX X
Assuming that the survey is to be carried out in October, if you set a recall period of 3 months,
you will obtain death rate data that captures the worst of the food shortages. Similarly, if you use
a recall period of 6 months, you will obtain death rate data that is affected by food shortages
(moderate and severe) and by effects of displacement.
In rapid-onset emergencies, shorter recall period are advised to capture the change in death
rates that would have happened from this particular emergency. In slow-onset emergencies,
recall period can be longer. However, this should be deliberated and decided for each survey
based on the survey objectives and context. A recall period of around 90 days represents a
compromise between the number of households to be visited, the precision of the data
generated and the estimation of the death rate that is close enough to the current situation to
allow for planning health and nutrition interventions.
Having a clear starting date is one of the most important aspects in defining a recall period in
order to reduce recall error. The beginning of the recall period should always be a date that
everyone in the survey area can remember, e.g., a local event, a major holiday or festival
(new year, Christmas, beginning of Ramadan, etc.), an episode of catastrophic weather, a
political event (election, political decree, etc.), or similar memorable event. The beginning of
the recall period should be the same for all the survey population, so care should be taken for
events that may have occurred at different times in various parts of the survey area, such as
onset of the rainy season or taking in the harvest. The same event should be used as the
beginning of the recall period throughout a survey.
20
The end of the recall period should be the mid-point of the period of survey fieldwork. For
example if the data collection period is set for 5 days 3rd
day of the data collection would be
the end of the recall period. The exact number of days of the recall period therefore needs
to be counted for each survey, and used in the calculation of death rate survey results.
For each individual interview, the endpoint of the recall period is the time of administering
the questionnaire. However, the endpoint to calculate the number of days used for the recall
period is the mid-point of the data collection.
Example 3.4: Recall period for a survey
A survey was planned in Mugu district of Nepal. The data collection was to take place during the
2nd
week of April, April 11 being the mid-point of the data collection. The beginning of the recall
period was set to be the new year’s day. The recall period was therefore 102 days.
Box 3.2: Steps to follow in calculating sample size for a death rate survey
[The steps below provide instructions on calculating the sample sizes for malnutrition and
converting them into number of households using the ENA software (version: Oct. 27, 2011)]
Decide on the expected death rate
o This can be estimated from a previous survey in the area, from a similar area, or
from a regional or national survey (note: this needs to be reviewed and adjusted
before use)
o If no information is available, get an estimate from key informants.
Decide on the precision needed
o It should be decided based on the objectives of the survey.
Decide on the design effect if it is a survey using cluster sampling design
o The recommended value to use for design effect for sample size calculation for
death rate is 1.5, which is sufficient in most contexts, especially if violence-related
is limited
Estimate the average household size in the survey area
o This can be obtained from a previous survey in the area, from a similar area, from
a regional or national survey, from census data, or from the CBS.
Estimate the expected percentage of non-response households
o For most surveys the non-response will be around 3-5%
21
Box 3.3. Using ENA to calculate the sample size for the nutrition and death rate surveys
Steps to follow in calculating the sample size in ENA
1: Select the type of sampling (random or cluster) and tick the small box for correction for
small population size if it needs to be applied (i.e. population in the survey area is <10,000
individuals).
2: Enter the estimated values for the nutrition survey. The sample size is automatically
calculated both in terms of children and households.
3: Enter the estimated values for the death rate survey. The sample size is automatically
calculated both in terms of population and households.
Note that these default values are set for reference purposes only. They need to be adjusted
based on the survey objectives and context.
3.9 Reconciling sample sizes in combined surveys
If combined surveys are carried out, when sample sizes in number of households are
calculated for nutrition and death rate surveys, they will most likely be different for different
indicators. In these cases, the highest sample size should be used as the final sample size in
the survey. The sample sizes for the other indicator should then be compared with this final
sample size and the total number of households that needs to be visited to collect information
on the other indicators should be decided.
1
2
3
22
3.10 Sampling methodologies
There are three main sampling methods traditionally used in nutrition and death rate surveys.
They are: 1) simple random sampling, 2) systematic random sampling, and 3) cluster
sampling. The sampling method is determined mainly by the size of the population and the
spatial distribution of the households (i.e. physical organization of the households).
Each of the sampling method is described below but definitions of some terms used with
different sampling methods are first given below:
Sampling universe: The entire survey population
Sampling frame: Description of the sampling universe, usually in the form of a list of
sampling units (for example, wards, households or individuals)
Sampling unit: The unit selected during the process of sampling. If districts are selected
during the first stage of cluster sampling, the sampling unit at the first sampling stage will be
the district (in this case they are also called primary sampling unit). If households are selected
from a list of all households in the population, the sampling unit will be the household.
Basic sampling unit or elementary unit: the sampling unit selected at the last stage of
sampling. In a multi-stage cluster survey, if wards are selected first and households are then
selected within the selected wards, the basic sampling unit would be the household.
3.10.1 Simple random sampling
The process of randomly selecting sampling units (i.e. sample) from a sampling frame is
called simple random sampling. For example, simple random sampling is used when a
specific number of households are selected from a list of all the households in a survey area.
An up-to-date sampling frame must be available to carry out simple random sampling.
Simple random sampling is usually used when the survey population is small (e.g. less than
10,000 individuals) since a list of all households can either be obtained or created relatively
easily when the survey population is small.
Example 3.6: The following is the list of indicators and their respective sample sizes
calculated for an integrated nutrition and death rate survey:
Indicator Sample size in number of households
Nutrition 522
Death rate 721
The highest sample size of 689 households calculated for the death rate survey will be the
final sample size for the survey. Assume that 30 clusters are to be selected for the survey.
- 25 households (721/30) would be selected for the survey to collect information on death
rate per cluster.
- The first 18 households (642/30) will be interviewed for the nutrition surveys (the death
rate survey questionnaire will be administered in all 25 households regardless of whether
there is a child in the household or not but the 6-59 month form will be administered in
the first 18 household only if the selected household has a child 6-59 months)
23
Box 3.5: Steps to follow when selecting a sample using simple random sampling
Define the survey area and population
Calculate the sample size using the ENA software
o Note that the option ‘random’ under the ‘sampling’ need to be selected when
calculating sample size for the survey sing simple random sampling
o If the population size is less than 10,000 individuals in the survey area, a
correction for small population size must be applied
Obtain or create a list of every household in the survey area and number them from
1 to N.
o Make sure that every household in the survey area is included in the list,
especially when an existing list of households is used
Select the households to survey using the Random Number Table option in ENA (see
box 3.6 below)
Visit all the selected households and collect information based on the survey
objectives
o For nutrition survey, measure all eligible children (i.e. 6-59 months)
o For death rate survey, administer the death rate questionnaire on every selected
household regardless of whether there’s a child in the household or not
The random number procedure in the planning screen of the ENA software can be used to
randomly select households. This is explained in box 3.6 below using an example.
Note that the below example is also used to illustrate the sample size calculation for small
population size (<10,000 individuals).
The total population size in the survey area is 5,902 individuals
The small box for correction for small population size is selected along with the type
of sampling (i.e. random)
Note that in order for the software to apply the correction, the total population needs
to be entered in the table for cluster sampling
24
Box 3.6: Using ENA to select a sample for a survey using simple random sampling [The use of ENA software to select a sample using simple random sampling method is
described below using the following example: An integrated nutrition and death rate survey
is planned in a newly opened IDP camp in Nepal where a total number of 5,902 individuals
(976 households) is residing. The list of all the 976 households was obtained from the camp
management. The sample size calculated for the integrated survey is 325 households. Note
that the same planning screen that is used to calculate the sample size for the survey is used
to generate random numbers for the sample.]
1: Enter the household range from which the sample is to be selected
Enter 1 and 976 as the lower and upper limits
2: Enter the total number of households to be selected
Enter 325
3: Click on ‘Generate Table’ (A table with 325 numbers will open up in a Microsoft word
document)
Mark the numbers on the list of households; these households must be visited for the
survey
Random Number table
Range: 1 to 976, Number: 325
630 946 686 696 422 501 265 645 110 649 975 295 80
22 754 262 209 253 222 800 963 424 720 966 164
96 386 668 59 276 203 888 904 556 931 612 728
242 237 584 408 521 538 893 659 245 589 912 718
63 685 49 806 285 352 101 261 26 219 834 644
437 214 151 689 158 973 389 949 374 577 196 354
426 791 375 355 37 740 55 298 506 539 474 555
140 125 322 643 714 687 135 40 866 348 868 574
258 85 427 454 201 291 1 551 327 503 586 561
677 111 697 179 42 150 366 475 226 743 495 68
636 531 882 337 711 615 753 516 187 113 351 34
608 502 508 798 494 748 118 745 960 954 916 588
889 240 215 600 499 458 128 221 681 275 592 524
410 737 29 703 457 953 771 854 316 527 36 606
333 812 334 466 857 671 7 362 321 418 190 515
469 833 799 701 598 529 70 405 670 66 230 167
436 699 4 41 412 619 120 482 9 336 305 441
99 347 639 646 416 892 74 664 852 909 155 741
781 662 913 273 60 211 350 473 595 625 132 828
443 185 266 169 695 809 5 576 928 523 134 107
53 390 433 549 455 921 315 269 299 282 751 116
290 200 90 313 614 819 694 924 783 734 760 136
853 400 642 146 676 396 633 500 849 27 871 859
815 485 370 323 891 463 877 330 704 72 149 526
178 6 719 346 941 810 505 744 568 602 813 733
533 776 961 583 326 280 837 170 432 445 236 830
727 543 860 933 601 490 161 293 207 802 838 962
25
In practice, surveys are usually conducted in large populations that are scattered in relatively
large geographical areas. Reliable sampling frames are not usually available for these
populations and developing a sampling frame is not practical in these settings. Therefore, the
simple random sampling is rarely used. However, simple random sampling is widely used to
choose households as part of the 2nd
stage sampling in surveys using cluster sampling method
(see section 3.11.3 below).
3.10.2 Systematic random sampling
Systematic random sampling is a sampling method that is used to select households at a
certain predetermined interval called the sampling interval. The sampling interval is
determined by dividing the total number of sampling units (i.e. total number of households)
in the survey area by the required number of sample size (i.e. sample size in households). In
systematic random sampling, the first household is chosen at random and the subsequent
households are visited systematically using a sampling interval.
Example 3.7: Calculating sampling interval
If a nutrition survey is to be carried out using the systematic sampling method in a population
with 1500 households, and the sample size calculated for the survey is 300 households, the
sampling interval is calculated as follows:
=
= 5
Systematic random sampling is usually used in relatively small geographic areas where there
is an orderly layout of the houses that make it possible to go systematically from one house to
another, in order, without omitting any of the houses (see example 3.8 below). Such a
situation may occur in a camp where tents are pitched row after row, in blocks of flats, where
streets are laid out in a grid, or where the houses are all along the edge of a river, coast, road,
or other major feature. Thus, the main advantage of the systematic random sampling is that it
can be used even if a list of sampling units is not available. Nevertheless, systematic random
sampling is also possible in other settings where there is a complete list of all the sampling
units is available.
26
Example 3.8: a community setting where systematic random sampling is feasible
Box 3.7: Steps to follow when selecting a sample using systematic random sampling
[The steps in selecting a sample using the systematic sampling method is illustrated here
using the following example: An integrated nutrition and death rate survey is to be conducted
in an area where the total population is 3,400 households. The final sample size calculated
for the survey is 250 households.]
Define the survey area and population
Calculate the sample size using the ENA software
o Note that the option ‘random’ under the ‘sampling’ needs to be selected when
calculating sample size for a survey using systematic random sampling
o If the population size is less than 10,000 individuals in the survey area, a
correction for small population size must be applied
On the map of the site, trace a continuous route that passes in front of every
household (example 3.8 above)4.
Determine the sampling interval by dividing the total number of households by the
number that must be visited
o The sampling interval using the above example is 26 (3,400/250=13.6~13
Select the first household to be visited. o The first household is randomly selected within the sampling interval by drawing
a random number between 1 and the sampling interval 13 (using the example
above)5.
If the random number chosen, for example, is 11, the list household to be
visited in number 11.
Select the subsequent houses to be visited
o The next household to be visited is found by adding the sampling interval to the
first household selected (or counting the number of households along the
prescribed route)
4 If the households are in neat rows, such as tents in a refugee camp, it is not necessary to draw a map. This step
is not applicable if you are using systematic random sampling to select a sample from a sampling frame. 5 This can either be done with ENA software (using the function described in Box 3.6, using the Range from 1
to sapling interval and Numbers =1), or by using a random number table in Annex 3.
27
o Using the above example:
1st HH to be visited: 11
2nd
HH to be visited: 11 + 13.5 = 24.6 ~ 25
3rd
HH to be visited: 24.6 + 13.6 = 38.2 ~ 38
Etc.
Visit the selected households and collect information based on the survey objectives
o For nutrition survey, measure all eligible children (i.e. 6-59 months)
o For death rate survey, administer the death rate questionnaire on every selected
household regardless of where there’s a child in the household or not
Systematic random sampling is usually used for small-scale surveys of limited areas (1,000 –
5,000 households). It may also be used to select the households during the second stage
sampling in a survey using cluster sampling method (see section 3.11.3).
3.10.3 Cluster sampling
Cluster sampling is used in large populations where no accurate population register is
available and households cannot be visited systematically. This is the most commonly used
sampling method in nutrition and death rate surveys. Cluster sampling usually reduces the
distances the survey team has to walk. However, the sample size is always greater than in
random sampling so that more households need to be visited.
Most surveys using cluster sampling method are done in two stages. In the first stage, the
whole population is first divided, on paper, into smaller discrete geographical areas, such as
wards, enumeration area, administrative unit, etc. whose population is known or can be
estimated. Clusters are then randomly selected from these wards with the chance of any ward
being selected being proportional to the size of its population. This is called sampling with
"probability proportional to size". In the second stage, the individuals are chosen at random
from within each cluster area or ward. This means that each person in the whole area has an
equal chance of being selected6.
3.10.3.1 Determining the number of clusters
The number of clusters needs to be decided first before clusters are selected from the
geographical areas. There are some important considerations when deciding on the number of
clusters.
The number of households to survey in each cluster should be chosen so that one team can
complete one cluster per day. For example, if a team can only survey 20 households per day,
the number of clusters should be determined accordingly. To determine the number of
households that can be surveyed in a day, the following should be considered: a) the time
needed to travel back and forth the survey area, b) the time to meet the ward leaders and
introductions, c) the time to conduct the sampling (i.e. household listing and selection), d) the
time to travel from one house to the next, and e) time for lunch and other break. Some houses
6 Although larger wards are more likely to be selected to contain a cluster than smaller wards, individual
households within the larger ward are less likely to be sampled than a household from a small ward. These
effects balance each other so that each household in the whole population has an equal chance of being selected.
28
will have to be revisited at the end of the day to measure children that were missing during
the first visit. See example 3.9 for details.
Example 3.9: Determining number of household that can be surveyed per day
If the team leave base at 8 am, takes one hour to reach the cluster site and another hour to
introduce itself and select the first house, then measurements will start at 10 am. The team
will need two refreshment breaks of 15 minutes each, one hour for lunch, and will need to
leave to get back to base before dark, say about 4pm. This means that the team will have 4.5
hours to measure children and interview household heads.
If the survey takes around 7 minutes in each household, plus 2 minutes to reach the next
house and introduce the team to the new household, 30 households can be visited in a day. If
13.5 minutes are necessary in each household (including walking to the next house), then 20
households can be visited. With 18 minutes spent in each household plus walking to the next
house, 15 households can be visited per day.
The average time necessary to administer the survey questionnaire and take anthropometric
measurements should be assessed during training and pre-testing. The time necessary to walk
from one house to the next should also be estimated depending on the terrain (presence of
hills) and the organization of the houses (how spread houses are). These practical points
should be considered when designing the survey. If the distances between houses are not
great and there is no insecurity, more children can be included in a cluster.
Although children’s nutritional care is a joint responsibility of both parents and ideally both
parents should be present during the interview, in most households the mother will need to be
interviewed and she should also be with her children when they are measured if they are not
to be frightened by the team. Thus, the different components of the survey will often need to
take place consecutively even if there are additional team members to provide information.
The interview should always take place before the anthropometric measurements. During the
interview, the children will "settle down", see that the mother interacts with the team
harmoniously and be more amenable to being measured.
If the time scale simply cannot be kept, there are two choices. The team could either use two
days to survey one cluster, which will double the time taken to collect the data. This is
undesirable. Alternatively, the number of households in each cluster could be reduced and the
total number of clusters increased, affording more time to carefully collect data. This is a far
better option. Thus, if data from 30 households, for example, cannot be collected in one day,
the number of clusters should be increased and the number of households in each cluster
correspondingly reduced. To avoid "shortcut" bias, it is better to measure fewer children
accurately than overstress the team so that the measurements are not made accurately.
The design effect is smaller with a larger numbers of clusters, meaning that although there
may be more clusters, fewer total numbers of households are likely to be needed per cluster.
29
Thus, sampling 45 clusters of 20 children is more efficient than 30 clusters of 30 children.
Each survey should have at least 30 clusters. As the number of clusters decreases, the design
effect increases rapidly. Fewer than 25 clusters can yield unreliable results and should not be
intended.
3.10.3.2 Stage one: selecting the clusters
In stage one the entire survey population is first grouped into smaller geographical units
called primary sampling units (PSU) – these are usually wards, enumeration areas, or
divisions. PSU should be the smallest possible unit, provided population data is available and
geographical unit has a name to locate it. The PSU should be the wards from the census data
in Nepal.
It is important to go through the list of wards in the survey area with someone who knows the
area well and make sure that each ward can be identified. Note that sometimes the names
might be misspelt or a different name may be used making it difficult to identify the ward. It
is crucial that each ward in the sampling frame can be identified as it will not be possible to
change a ward once clusters are assigned.
Each ward should have at least the number of households required to form a complete cluster.
If there are insufficient houses in a ward, two adjacent wards should be combined at the
planning stage. If this combined ward is selected as a cluster, the number of households
should be allocated proportionally between the 2 wards. Similarly, if there are wards with
very large population size, it is recommended that they are divided into smaller wards using
boundaries such as sub wards, north/south, etc. to make the population sizes roughly equal in
all wards. This is to avoid getting 2 clusters in one ward. However, if this is not possible
before the clusters are assigned, the ward containing more than 1 cluster should be divided
into sections geographically and the required number of clusters should then be chosen
segmentation (see section 3.11.3.4 for details). The selection of clusters will be done with
probability proportional to size (PPS).
In a stable population, such as a drought-affected region with little in- and out-migration, a
census that is several years old may still be acceptable as a base for PPS sampling. However,
in refugee situations where influx continues, reliable up-to-date counts are important for a
valid sample. Alternatively, if no population data are available, the relative size of the
population living in each section of the map can be estimated using key informants.
30
Box 3.8: Steps to follow when selecting a sample using cluster sampling
Steps to follow in choosing the clusters
Define the survey area and population
Calculate the sample size using the ENA software
Obtain the best available census data for each ward
o Go through the list of wards with someone who is knows the area and make sure
each ward can be identified
o Either data for total population or population under-five can be used, as far as the
count used is consistent for the entire sampling frame.
o If there are wards with less number of households than the number required to
complete a cluster, combine it with a neighbouring ward
o If the population size is too large in a ward, divide the ward into smaller
geographical units before clusters are assigned
Select the clusters using ENA software as detailed in Box 3.3
It will not be possible to change a cluster site once it is selected. If the survey is to be
unbiased, the selected site must be visited. Thus, it is important to define the survey area in
the planning stage very realistically, taking travel, security, and any other factor that could
influence your ability to get to the cluster site into account before listing the sites in the
planning table.
Box 3.9: Using ENA to assign clusters [The ‘Table for Cluster Sampling’ in the planning screen is used to assign clusters]
Enter the names of the ward under the column ‘geographical unit’ and the
population sizes under the column ‘population size’
o This can be manually entered or copied and pasted from an Microsoft Office
document (make sure the headings are excluded when copying) using the paste
icon
Enter the number of clusters to be assigned in the box provided for number of
clusters (refer to section 3.11.3.1 for details on calculating the number of clusters)
Click ONCE on assign cluster
o Click on the Excel icon to get the selected clusters on a Microsoft Excel file
o Click on the Print icon to print the assigned clusters
31
Note that in addition to the specified number of clusters some additional clusters will also be
assigned and noted as RC (Reserve Clusters). The use of these RC is described in section
3.11.3.7.
3.10.3.3 Stage two: selection of households within selected clusters
Once clusters have been selected, households have to be selected. There are several methods
of choosing households from a cluster. The best way is to treat each cluster as if it is a "small
population" and to select the houses using the simple or systematic random sampling methods
described above.
Definition of a household
In order to select households to include in the survey, it is crucial to carefully define a
household. The following household definition should be used when defining a household for
Geographical unit Population sizeCluster
Daicha(Baaye) 763 1
Toricha 560
Adhe 520 2
Mandete Kuro 800 3
Diba Okotu 1062 RC,4
Boji 870 5
Mathare 990 6,RC
Guyo Roba 930 7
Gamura 572 8
Duke 751 9
Lalasalama 422
Chira 533 10
Chile 614 11
Boji 142
Rage 310 12
Kutur 553
Ollum 222 13
Balal 227
32
survey purposes in Nepal: a group of people who live together and share a common cooking
pot.
Survey teams have to be clear on who is, and who is not, part of a household. The definition
has to be clarified at the planning stage, and the same definition applied consistently by all
teams and throughout the survey.
3.10.3.4 Segmentation
If the clusters correspond to a large population, the first step of stage two is to subdivide the
population into different segments before applying any type of sampling method. The
purpose of segmentation is to get smaller survey areas (about 150-200 households) so that
households within a small segment can be listed and selected using simple or systematic
random sampling method.
After segmentation, if each segment has equal number of households, one of them can be
selected for data collection at random. This can be done by numbering each segment and
selecting one of them using a random number table (see annex 3). However, if the segments
have different number of households, selecting a segment at random will violate the equal
probability of selection. PPS technique should be in these cases applied to choose a segment.
This is explained n example 3.10 below.
Example 3.10: Applying PPS technique to select a segment.
During a nutrition survey, a survey team segmented and listed a ward as follows:
Segment Number of households
A 70
B 100
C 30
D 190
Of the 4 segments, the team needs to one segment to survey. Since each segment has different
number of households, PPS technique needs to be applied to select a segment. The steps in
applying the PPS technique are as follows:
1: Calculate the cumulative population and determine the population intervals
Segment Number of
households
Cumulative
population
Population interval
A 70 70 1 – 70
B 100 170 71 – 170
C 30 200 171 – 200
D 190 390 201 – 390
2: Calculate the sampling interval by dividing the total number of households by the
number of segments to select. Since one cluster needs to be selected, 360 must be divided
by 1. The sampling interval will therefore be 390.
33
4: Select a random number between 1 and the sampling interval. That is, in this case, a
number between 1 and 390.
5: Choose the segment that the random number selected falls under. Suppose the random
number selected is 99. Since 99 falls under the range, 71 – 170 (i.e. segment B), segment B
will be chosen for the survey.
In segmentation, a ward is reduced to an area containing up to, say, 150-200 households.
These households can then be listed and the required number of households can be selected
from the list by simple or systematic random sampling.
3.10.3.5 Household listing and selection of households
When a cluster (or segment in the case of large cluster) is around 150-200 households, all
households in the area can be listed with the help of the chief of the ward (note that if a list of
households already exists, it can be used but the list must be reviewed and updated as needed
before it is used.).
During listing, divide the cluster (or the selected segment in the case of large wards) further
into north, south, east, and west and, with the help of the ward chief, list the names of the
heads of the households in each division on a piece of paper. Number the names of the
households from 1 to N. Either simple or systematic random sampling method can be used to
select the required number of households from the list.
Survey teams can be given a sheet with numbers already printed from 1 up to 200; the team
can take out the required numbers (i.e. number of households in the cluster/segment) and
select the required numbers (i.e. number of households to be sampled in the cluster) using the
‘lottery’ method. It is useful to inform the ward chief in advance of the exact day of data
collection and ask him to be present in the ward on that day to facilitate the listing. Note that
if households in a cluster are arranged in some logical order, it is not necessary to list the
households. Required number of households can be selected using systematic random
sampling method as described above.
Simple or systematic random sampling should be the method of choice in selecting
households in a survey using cluster sampling method. Every effort must be made to use
either simple or systematic random sampling method to select households and the use of any
other method (such as modified EPI method) should be justified.
3.10.3.6 Modified EPI method
If it is not possible to select the households using a simple or systematic random sampling
method, the modified EPI method7 can be used as a last resort. Although this method is
7 Modified EPI method has been developed from the EPI method to overcome some of the problems inherent to
the EPI method; samples selected by the modified EPI method are better than the ones from EPI method,
modified EPI method still produces statistically less desirable sample.
34
simple, easy to train, and rapid, it results in a somewhat biased sample. The modified EPI
method is described below.
Box 3.10: Steps to follow when using modified EPI method
1. Go to somewhere near the centre of the selected cluster area.
2. Randomly choose a direction by spinning a pen on the ground and noting the direction
it points when it stops.
3. Walk in the direction indicated, to the edge of the ward (from a to c in direction b).
4. At the edge of the ward, spin the pen again until it points into the body of the ward.
5. Walk along this second line (from c to d in direction e) counting each house on the way
(both left and right side) until you reach the other edge of the ward.
6. Using a random list of random numbers (annex 3) or the lottery method, select the first
house to be visited by drawing a random number between 1 and the number of
households counted when walking. For example, if the number of households counted
was 10, then select a random number between one and 10.
7. If the number 7 was chosen, go back to the seventh household counted along the
walking line. This is the first house that should be visited. Go to the first household
selected and complete the survey questionnaire.
8. The subsequent households are chosen by proximity.
a. In a ward where the houses are closely packed together, choose the next house
to the right8.
b. If the ward is spread out, choose the house with the door closest to the last
house surveyed, whether on the right or left; this saves a lot of time in an area
where the dwellings are spread out.
c. The same method should be used for all the clusters.
9. Continue in this direction until the required number of households has been visited.
As described in section 3.11.3.3, simple or systematic random sampling method should be the
method of choice when selecting households from a cluster. Modified EPI method should
only be used as the last option. The decision tree included in annex 4 should be used in each
selected cluster separately to decide on the sampling method for that cluster. It may mean
that, in one survey, while simple or systematic sampling is used in some clusters, modified
EPI may need to be used in other clusters.
Note on sampling unit: house vs. household
8 Or left, but this should be decided during the planning stage and the same rule should be used by all the teams.
It is more convenient to always go to the right for every survey.
35
The household definition must be upheld throughout the survey. According to the household
definition, there can be more than one household living in the same house/compound.
Selection should be based on households, not on houses. With the EPI method, within a
compound, households should be selected according to the same rule as for other households.
With the household listing, surveyors compile a list of households. If more than one
household live in the compound, all of them would be listed. Only one household may be
selected in a given house/compound.
3.10.3.7 Reserve clusters
The sampling frame must only contain sampling units that are accessible and can be visited if
they are selected as clusters. However, in certain circumstances, certain clusters may become
inaccessible after they are selected as clusters due to various reasons such as insecurity.
These circumstances should be anticipated and preparation should be made at the planning
stage.
As described above, when cluster assignment is done in the ENA software, the software
assigns a few more clusters than the number of clusters specified and includes them in the
final list of clusters selected. These additional clusters are called Reserve Clusters,
abbreviated as RC in the list of clusters selected. The ENA software assigns RCs by taking
10% of the total number of clusters specified and rounding it up to the higher whole value.
For example, if 32 clusters are specified, 4 additional clusters are selected as RCs and if 44
clusters are specified 5 additional clusters are selected.
RCs are surveyed only if 10% or more clusters that were originally planned to be are not
visited. In this case, all the RCs will be surveyed. If less than 10% of the clusters are not
surveyed, there is no need to visit the replacement clusters. For example, if it is not possible
to visit one cluster during the survey, there is no need to survey one of the RC. The sample
size will still be enough to carry out the analysis and achieve the survey objectives. It is not
acceptable to just visit a neighbouring, similar cluster if one cannot be accessed, because it
violates the principle of equal probability of selection.
3.11 Important considerations when selecting subjects
There may be special cases in the field during data collection such as disabled children,
absent children, empty households, etc. These cases must be treated in a standardised manner
throughout the survey. Recommendations as to how to deal with these cases are described
below. It is crucial that survey teams are trained on how to deal with these cases so that the
procedures are standardised and there is no ambiguity in the field during data collection.
3.11.1 Polygamous families
Household definition should be the basis for dealing with polygamous families. Families
should be counted as one household as long as they are living together and sharing a common
cooking pot. If polygamous families form different households based on the household
definition they should be treated as separate households. This will need to be explained to the
community leaders prior to data collection.
36
3.11.2 No substitution
Whenever a household is selected according to the rules, there should not be a substitution for
this household for any reason. House occupants sometimes refuse to be measured, the staff
sometimes fear dogs, local people may try to direct the team to include particular houses and
omit others, or houses may be deserted or physically difficult to reach (up a steep hill for
example). If, for any reason, the selected household is not included, the team must make a
note and go to the next household according to the rules. Another household should never be
substituted for the properly selected household. This is not usually a problem with the EPI
method, because the rules say that the nearest house to the right should be the next selected.
(In this case, however, a house to the left should not be substituted.)
3.11.3 Measure all the children
Before the household is visited, it is not known how many children are present, or whether
there are any children at all. All the children part of the household in the correct age range
should be included in the sample and measured. If two eligible children are found in a
household, both are included, even if they are twins. This is extremely important, as it
ensures that every child has the same chance of being selected, which is a basic principle of
the survey design. Detailed analysis has shown that there is little correlation between the
nutritional status of children living in the same household. Individuals, rather than
households, seem to become malnourished – note, however, that the malnutrition prevalence
should be se disaggregated in the report.
3.11.4 No children
When there are no children under age 5 in a household, the selected household should remain
a part of the sample that contributes zero children to the nutritional part of the survey.
However, it is very important to include this household for the other data being collected (e.g.
death rate). Survey teams should record the household on the nutritional data sheet as having
no eligible children and proceed to the next house according to the rules.
3.11.5 Empty houses
If the house is empty, the neighbours should be asked about the family that lives in that
house. On the data collection form, record why the house is empty (if this can be
determined). If the residents are likely to return before the team leaves that cluster, the team
should return to the house to include the residents in the survey. If the house is permanently
empty or the residents will not return before the team must leave, this house can be skipped
and a note made. Again, a house that is not in the original sample should never be a substitute
for the empty house. If more than 5% of the households in a selected cluster are not found,
the teams should revisit the area at another time to see if they can complete the sample. The
total number of absent households should be included in the survey report.
3.11.6 Absent children
If a child lives in the house but is not present at the time of the survey, this child is recorded
on the datasheet when the house is visited. The weight and height of course cannot be entered
37
at that stage. The team should inform the mother that they will come back to the house later
in the day, after all the other houses have been visited in the cluster. The team should go back
to the house to find the child. The team should continue to look for missing children until
they leave the survey area. There are always some children who cannot be weighed or
measured, and this needs to be recorded and reported. The team should not simply take
another child and forget about the child that is missing.
3.11.7 Disabled children
Disabled children that would otherwise be eligible should be included where possible. If it is
not possible to measure height and weight due to deformity or other abnormality, the child
should be given an ID number and the data recorded as missing (and a note taken). With
missing height, they will not be included in the final sample unless they have oedema.
3.11.8 Child in a centre
If a child has been admitted to a hospital or feeding centre, the team must go to the centre and
measure the child. This is critical as such a child is very likely to be severely or moderately
malnourished. If it is impossible to visit the centre (it may be many miles away), the child
should be included in the datasheet and a note added that the child was in a feeding centre
and probably severely malnourished. In reality, the child may or may not be severely
malnourished. If there are a large number of such children, and the centres cannot be visited
to complete the measurements, then two rates of severe malnutrition can be calculated, one
assuming that these children are all severely malnourished, and the other excluding these
children from the survey.
38
4. MEASUREMENT TECHNIQUES
This section describes practical issues related to data collection and measurement during a
survey and provides information on standard procedures that should be used. For
anthropometric data collected during a nutrition survey, this section provides information on
the necessary field equipment and methods for taking measurements.
4.1 Nutrition survey data
Nutrition surveys entail collecting information on anthropometry among children 6-59
months. Taking accurate anthropometric measurement is a skill requiring specific training.
Step-by-step procedures and standardizing methods are necessary to ensure that the
measurements will be correct, which makes comparisons possible. Equipment used to take
anthropometric measurements in surveys should be standardised.
4.1.1 Inclusion criteria
The target group for a nutrition survey is children between 6-59 months. Before a child's
measurement is taken, the survey team should ensure that the child fits in the age criteria for
being included in the survey, either by converting his date of birth into months with the help
of the local calendar of events or by estimating his age, as described below.
Although height (65-110 cm) has also been used in the past along with age as an inclusion
criterion, this practice is not recommended since it often causes confusion in the field and
children from malnourished population are often stunted, and height criterion for inclusion
may bias the sample towards older children. Age should therefore be the only inclusion
criterion and when accurate age information is not available, survey teams need to estimate
age using a local events calendar. Determination of age should be a major component of
training for surveyors.
4.1.2 Estimating age
The age of children is needed not only to know whether a child meets the inclusion criteria
(i.e. between 6-59 months) for the survey but also to calculate nutrition indices such as
height-for-age and weight-for-age.
In estimating the age, two situations can be found:
1. The child has an official document stating his or her date of birth (birth certificate,
baptism certificate, immunization card9, etc.). In that case, the surveyors should verify
that the child is above 6 months and below 60 months and record the exact date of
birth on the survey questionnaire. This is the ideal situation, but it might be rarely
found in rural areas.
2. If the child does not have an official document mentioning his date of birth, the age of
the child should be estimated in months with the help of a local calendar of events (see
annex 5). If the age is estimated without an official document, the estimated age
should be recorded on the questionnaire. The following methods can be useful:
9 Note that the DoB on the immunization card is not always accurate, and therefore should be verified
39
a. If the mother knows the age or the date of birth but does not have an official
document to verify, estimate the age in months with the calendar of events,
verifying at the same time the plausibility of the information given.
b. If the age of a neighbour’s child is known, ask other women whether their
child was born before or after the "reference" child. The younger the child, the
more accurate estimate of his or her age.
c. In the absence of any of the above, the local event calendar will be used to
estimate the age of the child by asking the mother whether her child was born
before or after a certain event, and repeating the same procedure until reaching
a fairly accurate age estimate.
For any given child, either the date of birth or the estimated age should only be recorded so
that the proportion of children with an exact age can be computed.
4.1.2.1 Local events calendar
A local events calendar shows all the dates on which important events took place during the
past 5 years prior to the survey, giving equivalent in dates. In addition to seasonal patterns
(rainy season, harvest time, etc…) and major festivals/holidays, the local calendar can show
local events that will be known by the population of the area, such as: holidays, hailstorms,
the opening of a nearby school or clinic and political elections, etc. It is used to estimate the
age of the child based on proximity with event of known dates. It also serves as an "age-
converter", giving the age in months when either the date of birth or a reference event around
the birth of the child is known. An example of a local events calendar is given in annex 5.
In places where surveys are periodically conducted, event calendars may already exist. In
these cases, the calendars must be reviewed and, if needed, updated prior to the planned
survey. It may be necessary to develop a new events calendar in places where surveys are
conducted for the first time. To develop the calendar, start with the survey date and go
backwards into 5 years listing the major events in the survey area that the survey population
can remember/relate to. The local events calendar format in annex 5 can be used to list the
local events and develop the survey calendar. The events calendar should be finalised prior to
the survey and should be discussed during the training for the survey enumerators.
4.1.3 Measuring weight
The weight should always be measured to the nearest 100g for all children. Children should
be measured naked. If for any reason children cannot be measured naked, they can be
measured with clothes however, the average weight of the clothes should be calculated so
that it can be adjusted for when calculating nutrition indices using ENA (see section 61.1.3
for details).
4.1.3.1 Equipment for measuring weight
An electronic scale with double-weighing function (such as the UNISCALE or SECA scale)
should be used for weighing. Although fairly expensive and fragile, this type of scale has the
advantage of allowing easy but precise measurement, especially for young children who can
be weighed in the arms of his/her mother or measurer. The measurement is made at the closer
40
100g and is easy to make with both younger and older children. It is recommended to make
some wooden board of the size of the scale to use for stabilization of the scale.
The use of 25kg hanging spring scale (such as Salter scales) graduated at 100g level is not
recommended as weighing a child with a hanging scale is quite traumatic, both for the child
and the mother, and it is often difficult to get a good measure when the child is moving (and
subject to rounding errors).
Calibration of the scales should be checked each morning using a standard weight (standard
5-10kg weight).
4.1.3.2 Using an electronic mother/child weighing scale
The mother/child electronic scale, like UNISCALE or SECA scale, can be used either as a
standard bathroom scale if children can stand still on it, or can be used with the "double-
weighing" function for younger children who cannot stand on their own. The double-
weighing requires an adult (the mother or the assistant-measurer) and the child to be weighed
simultaneously.
Although fairly expensive, mother/child electronic scale should be the only choice for
weighing children in nutrition assessment because of the ease of use and the quality of the
data collected. The mother/child electronic scale can be obtained from health facilities,
organizations with stocks or purchased.
Steps in weighing with an electronic mother/child scale
Remove the clothing on the child.
Ensure the scale is not over-heated in the sun and is on an even surface enabling the
reading to be clear.
The assistant measurer or the mother stands on the scale.
The measurer presses the "double-weighing" button (or briefly covers the captor in a solar
scale).
The assistant or the mother takes the child to be weighed and holds the child tightly, as
shown on the figure below.
The measurer reads and records the reading with one decimal point (e.g. 5.1 Kg)
41
Source: Cogill, 2003
Refer to the manual of the scale since there might be some slight differences depending on
the model of the scale.
4.1.4 Measuring length or height
Children’s length or height should be measured accurately to the nearest 0.1cm. Measurement
errors of 2-3cm can easily occur when measuring length or height and cause significant errors
in classifying nutrition status.
4.1.4.1 Height vs. length
Children <24 months should be measured lying down, and children >=24 months should be
measured standing up. In situations where it is not possible to measure a child < 24 months
lying down, the child can be measured standing up but a note should be made on the
questionnaire indicating the measuring position. Similarly, if a child > 24 months needs to be
measured lying down it can be done so with a note on a questionnaire. A correction factor
will be applied before the appropriate nutrition indices are calculated for these cases at the
analysis stage.
4.1.4.2 Equipment for measuring height/length
A measuring board used for children aged 6-59 months is at least 130cm long, is made of
hardwood and has a hard water resistant finish. The board should have a metal tape-measure
attached to it, which should be marked out in 0.1cm graduations. The head-board must be
movable and the foot-board must be large enough for a child to stand on it. Measuring board
can be obtained from health facilities, organizations with stocks or purchased.
4.1.4.3 Using a measuring board to measure length
Children <24 months should be measured lying down (i.e. length should be measured). The
procedures for measuring the length are described below.
42
Steps to measure the length of a child
Explain the procedure to the child’s mother or carer.
Remove the child’s shoes and hair accessories.
Place the child gently onto the board, with the head against the fixed vertical part, and the
soles of the feet near the cursor (moving part). The child should lie straight in the middle
of the board, looking directly up.
The assistant should hold the child’s head firmly against the base of the board, while the
measurer places one hand on the knees (to keep the legs straight) and places the child’s
feet flat against the cursor with the other hand.
The measurer checks the child's position, reads and announces the length to the nearest
0.1cm.
Source: Bruce, 2003
4.1.4.4 Using a measuring board to measure height
Height (i.e. standing up position) should be measured for all children aged 24 months or
more (i.e. 24-59 months). The procedures for measuring the height are described below.
43
Steps to measure the height of a child
Explain the procedure to the child’s mother or carer.
Place the measuring board upright in a location where there is room for movement
around the board.
Remove the child’s shoes and hair accessories.
Stand the child on the middle of the measuring board.
The assistant hold the child’s ankles and knees against the board.
Ensure that the child’s head, shoulders, buttocks, knees and heals touch the board.
The measurer should hold the chin to position the head of the child.
The measurer should position the head and the cursor at right angles — the mid-ear and
eye socket should be in line and hair should be compressed by the cursor.
The measurer checks the child's position, reads and announces the height to the nearest
0.1cm.
Source: Bruce, 2003
4.1.5 Measuring nutritional oedema
Oedema is the retention of water in the tissues of the body. Bilateral oedema is a sign of
kwashiorkor, a form of severe acute malnutrition. Children presenting oedema must be
referred to the closest health centre or a feeding centre.
To diagnose oedema, normal thumb pressure is applied to the tops of the feet for about three
seconds (if you count “one thousand and one, one thousand and two, one thousand and three”
44
in English, pronouncing the words carefully, this takes about three seconds). If there is
oedema, an impression/indentation remains for some time (at least a few seconds) where the
oedema fluid has been pressed out of the tissue (see steps to measure oedema).
The child should only be recorded as oedematous if both feet present pitting oedema.
Steps to measure oedema
Explain the procedure to the child’s mother or carer.
Ask the mother or caregiver to hold the baby in a sitting position on their lap.
Apply constant pressure on both feet of the child constantly for about 3 seconds
Release the hand and check if there is any impression/indentation that remains
If there is an impression/indentation, record the child as having oedema on the
questionnaire
Refer the child to the nearest health facility or feeding centre
4.1.6 Measuring Mid-upper arm circumference (MUAC)
4.1.6.1 Equipment for measuring MUAC
Mid-upper arm circumference measurements should be made using a flexible, non-stretch
tape. Only tapes specially designed to measure MUAC with appropriate graduation and
colours should be used. The colour should be coded as follows: red: <110 mm, yellow:
between 110 mm and 125 mm, and green: >125 mm.
4.1.6.2 Measuring MUAC
MUAC should be measured on the left arm, using a flexible non-elastic tape, at the mid-point
of the upper arm, with the arm hanging freely by the child’s side. Measurements should be
made to the nearest millimetre. MUAC should be measured for all children aged 6-59
months.
The decision to include MUAC in SMART (as an independent indicator for wasting) is based
on the recognition that agencies frequently use MUAC in rapid assessments, screening and
45
referral of cases in the community. MUAC is also a better predictor for risk of death than
weight for height.
Steps in measuring MUAC
Explain the procedure to the child’s mother or carer.
Keep your work at eye level – i.e. while taking the measurement, your eyes should be
parallel to child (if the child is sitting on the mother’s child, sit down next to the child
while taking the measurement).
Ask the mother to remove clothing that may cover the left arm of the child.
Calculate the midpoint of the child’s left upper arm by first locating the tip of the
child’s shoulder (Arrows 1 and 2) with your finger tips. Bend the child’s elbow to make
a right angle (Arrow 3). Place the tape at zero, which is indicated by two arrows, on the
tip of the shoulder (Arrow 4) and pull the tape straight down past the tip of the elbow
(Arrow 5). Read the number at the tip of the elbow to the nearest centimetre. Divide
this number by two to estimate the midpoint. As an alternative piece of string can also
be used for this purpose.
Mark the mid-upper arm point with a pen (Arrow 6).
Straighten the child’s arm and wrap the tape around the arm at midpoint. Make sure the
numbers are right side up. Make sure the tape is flat around the skin (Arrow 7).
Read the measurement at the window of the tape measure.
Record the measurement to the nearest 0.1cm.
Note: MUAC measurement is fast and simple, but not easy, and variations in measurements
often occur between different measurers. This is mainly related to how the tape is pulled or
“squeezed” around the arm.
Source: Cogill, 2003
46
All children with the MUAC measurements less than 125 mm should be treated as having
acute malnutrition and should be referred to the nearest feeding centre for further assessment
and treatment if there is a feeding centre programme in the area and if they are not enrolled in
such a programme). If there is no feeding centre programme in the survey area all children
with MUAC less than 115 mm should be referred to the nearest health facility.
4.1.7 Estimating the nutrition status (for referral)
When measurements are taken, the weight-for-height z-score should be calculated for each
individual child in order to determine the nutritional status of the child and be referred if
necessary. See annex 6 for a copy of weight-for-height table (WHO, 2006).
As described above, children should also be referred for nutrition care if they have MUAC <
125 mm and/or bilateral pitting oedema.
Steps to calculate the WHZ
Find the child’s length or height in the middle column of the table.
If the length or height is between those listed, round up or down as follows: If the
height/length is 0.5 cm or more than the next lower height/length, round up. Otherwise,
round down.
Then look in the left columns for boys or the right columns for girls to find the child’s
weight.
Look at the top of the column to see what the child’s z-score is.
Note: The child’s weight may be between two weights listed in the table and therefore
between two z-scores. If so, indicate that the weight is between these scores by writing less
than (<). For example, if the score is between −1 z-score and −2 z-score, write < −1 z-score.
4.1.8 Recording anthropometric information
The anthropometric measurements should be recorded in the child 6-59 months form (up to
CH09) of the sample survey questionnaire (see annex 7) or a similar form that is designed to
match the ENA data entry pane, and with the conventional units.
4.2 Death rate survey data
4.2.1 Crude Death Rate: household census
To estimate a death rate from a survey, the total number of people at risk and the length of
time over which they were at risk need to be known. However, the composition of some of
the households will have changed during the recall period (due to death, birth, migration into
and out of the household). Thus, the number of people within each household will not have
been constant during the recall period.
47
Example 4.1: Change of number of household members during the recall period
The diagram below depicts an example of a household in a death rate survey. At the
beginning of the recall period, the household had three members, and at the end of the recall
period the household also had three members but only one person was in the household
during the entire interval. At one time, the household had six members.
In calculating a denominator for this household, people joining or leaving the household
during the recall period should be taken into account. In an emergency, it is likely that people
will both leave and join households at an increased rate. If the in-migration and out-migration
are significantly different from each other, this will have an effect upon the calculated death
rates.
Sometimes in death rate surveys, the respondent is simply asked to state how many people
are in the household. Although this is quicker, it is much less accurate than asking the
respondent to list all household members and therefore this method is discouraged. It is
recommended that the household members be enumerated through a household census.
Crude death rate and 0-5 death rate are the 2 basic death rates that are usually calculated in a
death rate survey. However, the death rates can be disaggregated further and age and sex
specific death rates can also be calculated given that the age and sex information of
household members are collected. Additionally, causes of deaths can also be investigated.
The need for the sex and age disaggregated death rate information and the causes of deaths
should be reviewed and decided at the beginning of the survey and objectives need to be set
accordingly. The survey questionnaire should then be designed accordingly to collect the
needed information.
It should be noted that conducting verbal autopsy requires advanced training, which is not
covered in a standard surveys’ training. Therefore, the causes of deaths collected during a
death rate survey can only provide a rough idea about the causes of death in the survey area.
48
Survey managers should be aware of this limitation when planning to collect information on
the causes of death.
To calculate the crude death rate and 0-5 death rate, the respondent is asked to:
1. List all the household members at the time of the survey and indicate whether each of
these household members were present at the start of the recall period
2. List all members of the household that were present at the start of the recall period but
have left the household during the recall period
3. Indicate whether the individual is above or below age 5 (to derive 0–5 Death Rate) and
whether young children were born during the recall period
4. Indicate all deaths that occurred in the household during the recall period
Two additional questions are asked if age and sex specific deaths rate are to be calculated
5. The age of each member (to confirm if an individual is above or below age 5 and allow
a demographic pyramid of the population to be constructed)
6. The sex of each member (only necessary if sex-specific death rates are required)
Finally, cause of death may also be asked10
if the objective of the survey is also to investigate
into the causes of deaths.
These data are collected on a form, using a separate sheet for each household. An example of
the form is given in annex 7. The death rate form has 2 tables: 1) table for household census
and 2) summary table.
4.2.1.1 Table for household census
The table for household census is used to collect information about the sex, age, and in and
out migration details about individuals who were part of the household during the recall
period. The data on the table can be directly entered into the ENA 2011 software under the
Data Entry Individual Level data entry template and overall as well as age and sex specific
death rates can be calculated. Furthermore, causes of deaths can also be analysed.
4.2.1.2 Summary table
The summary table is used to summarise the data on the table for household census. The
summary data can be entered into the ENA 2011 software under the Data Entry household
Level data entry template and both crude and 0-5 death rates can be calculated. Note that it is
not possible to calculate age or sex specific death rates or analyse the causes of death by
using the summary table data. For supervision and data quality assurance purposes, it is
recommended that the summary table is filled out by the survey team leader at the end of
10
Where there have been an unusual number of deaths due to a single event, such as a natural disaster or a
violent attack, it is inappropriate to calculate a death rate (deaths per unit time) to estimate the effect that
happened at a single point in time. In these circumstances, deaths at the time of the event or shortly thereafter
(the time interval needs to be defined) are recorded and expressed as a proportion of the population that died
associated specifically with the event itself. It is also very important to record whether the death was directly
due to the disaster/war/violence. When examining such an episode, we also want to estimate the CMR and 0–
5MR before and after the event as well as the proportion who died during the event.
49
each household interview - even if the data on the summary table is not used at the data entry
(i.e. the data from the table for household census is used for data entry and analysis).
See section 6.1.2 for details on how to enter and analyse data using the data from the table for
household census and summary table.
4.2.2 Common problems in recording individual information for mortality
4.2.2.1 Mass migration
In a rapid-onset emergency situation, there are likely to be whole families that arrive in the
survey area during the recall period. Part of their experience will have been in the study area,
and part in the area from which they migrated. The death rate in the camp itself is likely to be
very different from the death rate before arrival in the camp. In addition, the various
households will have arrived at different times.
Under these circumstances, if we take a fixed recall period, some of the respondent
households will have been in the camp for the whole period and some will be new arrivals
that have spent most of the recall period elsewhere or on the journey.
It is therefore desirable to derive separate death rates: one for the time that the population was
in the camp, and another one for the time before the displaced households reached the camp.
Death rate since arrival in the camp
To calculate the death rate in the camp, the number of person-day at risk has to be
determined. Since families will have arrived at different times, the recall period (or "period
considered at risk") is different for each household. The date of arrival should be recorded for
each household, and the time period used in the equation should be the average number of
days each household has spent in the camp.
Death rate before arrival in the camp
To derive the separate death rate for the time before arrival, the fixed recall period is used, as
in the standard method, and the average time spent in the camp subtracted from this time.
Deaths are recorded as occurring in the camp or before arrival but after the start of the recall
period. The "before arrival" death rate is much more susceptible to serious sampling error
because the households are self-selected in terms of those that have the means, opportunity,
and composition that enable them to migrate, and the households may have arrived from a
wide variety of different geographical areas. The "rate before arrival" in the camp only
applies to the migrants who have reached the camp and should not be extrapolated to the area
of origin.
It is much more difficult to calculate the sample size needed to separate crude death rate into
two components—before and after arrival. There is an added variable in the calculation: the
average length of time households have spent in the camp. If the average length of time in the
50
camp can be obtained from the camp’s administrators, this is used as one of the "recall
periods" in the calculation.
4.2.2.2 Neo-natal deaths
In keeping with basic protocols for registering vital events, a live birth should be recorded as
a birth and a death that follows during the recall period should be recorded as a death - they
are two separate events and should be recorded as such in the household enumeration tables.
In the summary table, for purposes of entry into ENA, however, it is important that an infant
birth and death should be recorded only as a death and not as a birth and a death. Similarly,
when the Data Entry Individual Level template in ENA 2011 is used to enter death rate
survey data, the neo-natal deaths during the recall period should only be recorded as death
and the birth column should be left blank for that individual record.
If a birth and death were entered for the same person, the two events would cancel each other
out in terms of contributions of “person-time” of exposure.
4.2.2.3 In- and out-migration
In many societies, even under ordinary circumstances, movements in and out of the
household are routine occurrences. While it is important to measure migration into and out of
the household, however, it is also reasonable, under most circumstances, to assume that short-
term movements in and out of the household will not significantly affect the death rate
estimates.
Thus, for purposes of simplification, it is recommended that:
- In-migration only measures those who entered the household during the recall period
and stayed (either up to the current time or until time of death)
- Out-migration only measures those who left the household and stayed away (if they
died while away from the household, that would not be counted as a household death).
As with neo-natal death, for in-migration, a person who enters the household and
subsequently dies during the recall period should have both events recorded but for purposes
of entry into ENA, it is important that is recorded only as a death.
4.3 Additional data
4.3.1 Deciding what additional information to collect
As described in section 1.2, it is critical to understand that each additional piece of data
collected degrades the accuracy of the whole dataset and prolongs and complicates the
survey. Any additional information to be collected should be justified in the objectives and
have a realistic prospect of leading to a meaningful intervention. If such data are definitely
needed, consideration has to be given to whether the information could be collected more
efficiently in other ways (for example from health clinics, sentinel sites or a surveillance
system, or from focus group discussion), or whether it would be better to conduct a separate
survey to collect the supplementary information.
51
If additional information is to be included in the survey it must be quickly and reliably
obtainable during a short visit to the household.
4.3.2 IYCF data
The target group for most of the IYCF indicators are children < 24 months of age. Since the
age group is different from nutrition surveys, separate forms have been developed to collect
information on IYCF indicators. The sample survey questionnaire includes examples of data
collection sheets to collect information on these indicators (see annex 7).
Note that the sample size will be usually inadequate to provide meaningful IYCF estimated if
IYCF data is collected on a sample that is calculated for nutrition survey. If precise estimates
are required for IYCF indicators, this needs to be taken into account during survey planning
and sample sizes need to be calculated accordingly.
4.3.3 Food security data
In order to explain malnutrition levels (i.e. the GAM prevalence obtained from the survey)
and plan for appropriate interventions, food security information available from different
sources should be collected and reviewed. Reports on food security assessments conducted
monthly can be obtained from WFP in Nepal. Additionally, reports from the rapid food
assessment conducted from time to time can also be accessed from the food security sector.
These reports provide detailed analysis of the food security situation by district and by
livelihood zones and classify areas into various categories ranging from extremely food
insecure to food secure.
Food security data can also be collected through focus group discussions and key informant
interviews with people or groups from the same community during the survey. Those data
should be collected at the same time from the same population, but preferably by separate
teams using different methods. Although there are some food security questions included in
the sample questionnaire, food security data should be obtained mainly from key informant
interviews and focus group discussions and review of secondary data.
4.3.4 Health data
4.3.4.1 Morbidity data
Even during famines, people rarely die as a direct result of famine – people die because they
catch infectious diseases (measles, acute respiratory infections, diarrhoea and malaria). These
diseases may spread more rapidly because of conditions found during famine, and also may
be more severe or of longer duration because people are malnourished. Of most immediate
importance are recent or current outbreaks of disease that may be contributing to excess
deaths and/or malnutrition. Information on which diseases are most common will help plan
an intervention.
Unfortunately, good data on morbidity is difficult to obtain. Different people understand
different things by diarrhoea or fever, so standardised case definitions should be used. Also,
52
some symptoms (like diarrhoea and fever) are associated with more than one disease (like
malaria and measles).
Probably the best way to get information on morbidity is from the MHP staff and through
discussions with women or community leaders. They can tell you if there have been any
outbreaks and what the major illnesses are at the time of the survey.
Data on morbidity of children can also be collected during a nutrition survey, but the
interpretation of this data should be done very carefully. It is most useful to collect
information only on very common diseases, or very well-known diseases. Thus, questions
about measles, diarrhoea and fever are commonly included. This type of information should
always be crosschecked with Ministry of health staff and key informants.
4.3.4.2 Measles immunisation
Measles and malnutrition are closely associated: poor nutrition makes children more
susceptible to measles and makes the attack of measles worse. In turn, measles leads to
increases in malnutrition because of diarrhoea and fever. Information on previous measles
immunization campaigns or routine vaccination can be found from Ministry of health staff
and discussion with community leaders.
It is however common to add questions about measles vaccination to nutrition surveys for
children aged 9-59 months. Information should be collected from 1) the record on the
immunization card, and 2) the recall of the carer. If the rates of vaccination are low, then a
measles vaccination campaign is always advisable.
4.3.4.3 BCG vaccination
A BCG vaccination prevents an individual from developing TB. Children should receive a
BCG injection soon after birth. BCG vaccinations are not normally given during vaccination
campaigns (unlike measles), but are routinely administered by the Ministry of health when
the child visits the clinic, or during routine EPI work. Measuring the rate of BCG vaccination
gives an indication of how well the health system is working in a given area. In addition, TB
is associated with chronic (long-term) malnutrition in both adults and children.
BCG vaccinations are relatively easy to detect from a scar present on the upper arm. The scar
is normally on the right arm, but may be on the left, so both arms should be checked.
4.3.4.4 OPV Vaccination
Oral Polio Vaccine (OPV 1, 2 and 3) is provided to prevent polio. OPV is mostly given
through routine vaccinations but is also given through polio campaigns in Nepal. The
information about whether or not the child received OPV can be obtained from the child’s
immunisation card or from the mother. Information about polio campaigns in the survey area
should be obtained from the Ministry of Health staff.
53
4.3.4.5 Vitamin A supplementation
Vitamin A deficiency is associated with increased mortality, especially when children have
low WFH. Low WFH is usually associated with low vitamin A body stores and often with
frank vitamin A deficiency. Furthermore, vitamin A requirements are greatly increased
during nutritional rehabilitation.
Vitamin A deficiency is difficult to detect without special training. However, information on
supplementation can determine whether or not a vitamin A distribution is necessary. When
asking a mother about vitamin A supplementation, it is normally easier to bring capsule
samples (red, blue, and yellow) with you to show to the mothers/care takers. Show the mother
the capsule and ask her if her child has taken one of the capsules in the past year (the capsules
are normally distributed in conjunction with vaccination campaigns).
4.3.4.6 Deworming
Soil-transmitted helminths have a significant impact on the growth and development of
children and affect their cognitive development and long-term economic prospects. Periodic
deworming is one of the routine health programmes in Nepal. The coverage of deworming in
a survey area is estimated by collecting information on the number of children who received
the treatment.
4.3.5 WASH data
Water and sanitation practices are related to incidence of diarrhoea, which in turn causes
malnutrition through the pathways explained above. It is important to know about the access
to water and sanitation as well as hygiene practices to be able to explain the malnutrition
prevalence and also to plan interventions.
4.3.6 Additional qualitative data
Additional qualitative data can be collected through Focus Group Discussions (FGD), key
informant interviews, and direct observations to triangulate the findings of a nutrition survey
and investigate the immediate and underlying causes of malnutrition.
4.3.6.1 Focus group discussion
A focus group is a small group of 8-12 people led through an open discussion by a skilled
moderator and assistant. The moderator facilitates the discussion while the assistant takes
notes. FGD is an informal way of obtaining additional information to explain the malnutrition
situation in the survey area. It helps gathering qualitative information that is not captured
through quantitative data. See annex 8 on how to organise a FGD and record discussion
points.
4.3.6.2 Key informant interview
A key informant interview is an in-depth interview of selected people for their first-hand
knowledge about a topic of interest. The interviewer probes for feelings, opinions and views
of the key informant.
54
The interviews are semi structured, relying on a list of issues to be discussed, allowing for a
free flow of ideas and information. Key informants in nutrition surveys include staff of
Ministry of health, Ministry of Agriculture, and staff from organisations implementing
health/nutrition, food security, and WASH in the survey area as well as officials and leaders
of women groups. There must be efforts to ensure that women are included as respondents in
the various key informant interviews to provide gender perspective to the problems. See
annex 9 for a sample key informant questionnaire.
4.3.6.3 Direct observations
Direct observation involves watching people and events to see how something happens rather
than how it is perceived. It facilitates confirmation of some of the qualitative information
given through focus group discussion or confirmation of some quantitative data like presence
of sanitary facilities, water points, etc. Annex 10 provides a list of things that can be observed
during a nutrition survey to collect additional information.
55
5. SURVEY IMPLEMENTATION
5.1 Preparing for data collection
5.1.1 Obtaining and preparing equipment, supplies, and survey materials
During the preparatory phase of the survey, an inventory of all the material resources
required and available should be completed. Measuring instruments, questionnaires, means of
transport, fuel, safety equipment, and other material necessary for the proper functioning of
the teams should be clearly identified and budgeted for before the start of the survey.
Measuring material, scales, and height boards should be in good condition. During the
survey, scales should be checked each day against a known weight (standard weight).
Similarly, both measuring boards and MUAC should also be checked. If the measure cannot
be made to match the standard measure the equipment should not be used. Spare equipment is
needed to allow for damage or loss. Equipment and supplies needed for the survey include
transport, fuel, paper and pens, per diem, and recording forms. Copies of questionnaires,
absentee forms and forms for referral of malnourished cases should be prepared. A list of
inventory of common materials needed for a nutrition survey is given in Table 5.1.
56
Table 5.1: Example of equipment and materials needed for a nutrition and death rate survey
Category Item Quantity
Per team
Weighing Electronic mother/child weighing scales 2
Standard weight 1
Support board 1
Batteries Tbd
Measuring Height board 1
Height stick 1
MUAC Paediatric and Adult MUAC tapes 10
Plastic pipe of 200 mm in circumference for
calibration
1
Recording Questionnaires Tbd
Enumeration forms (depending on sampling method) Tbd
Folders/file box - plastic document holder 1 per cluster
Laptop (if data entered in the field) 1
Laptop charger (if data entered in the field) 1
Clipboard 1
Black Pen 3
Red Pen 1
Notebook 1
Document bag 1
Reference
documents
Surveyor's manual 1
Official letter of introduction 1 per team
Calendar of events (preferable laminated) 1
Weight-for-height z-score look-up tables 1
Referral forms Tbd
Map of the area 1 per team
Absentee form Tbd
Daily equipment checklist 1 per cluster
Logistics Vehicle 1 per team
Fuel Tbd
Camping and cooking gears
First aid box
Tbd
1 per team
Tbd= to be determined, depending on sample size and number of clusters.
5.1.2 Surveyor's manual
The surveyor's manual is a reference document during training and during data collection.
The surveyors’ manual should be used as the basis for the training and each survey team
should have at least one surveyors’ manual in the field for reference during data collection.
57
The following information should be included in the manual:
Survey process: how to introduce the team to the authority, how to get help from
local officials, how to use identification numbers, etc.
Sampling: how to delineate the area to survey, how to select the households, etc.
This has to be in line with the method chosen.
In the household: how to introduce the team, how to proceed for the mortality
interview, how to select eligible children, proper measurement techniques and
recording, appropriate coding to use, etc.
Data entry: if data are entered by the survey team, give details on how to proceed.
Job Descriptions, instruction on how to fill the questionnaire (if any, based on the
training)
5.2 Selecting and training the survey team
5.2.1 Selecting the survey teams
Proper screening of enumerators who are fluent in English/Nepalese and the local language
of the target area and are also physically fit (there is usually a lot of walking), is essential.
Selecting enumerators from the same community is advised as they are usually better
accepted by the community. Enumerators should be able to grasp the main concepts behind
nutrition surveys during the training and carry out their tasks accurately.
The survey should be made up of an overall survey manager and the survey teams. It is also
useful to have field supervisors during data collection to ensure the households are selected
properly and measurements are done using the correct protocols. The number of survey teams
required to conduct the survey will depend on the sample size, the time available and the
logistical and material resources for implementing the survey. Each survey team should be
made up of at least 3 people: 1 team leader and at least 2 enumerators (i.e. measurers). If the
questionnaire section of the survey is carried out separately, a fourth enumerator may be
needed. Every effort must be made to get a survey teams that are gender balanced.
Table 5.2: Roles of different members of a survey team
Member Role
Survey manager Responsible for training team members, visiting teams in the
field, ensuring that households are selected properly, and
ensuring the equipment is functioning and calibrated and that
measurements are taken and recorded accurately; ensuring
questionnaires are fully and correctly filled, and providing
feedback to the team on data quality (e.g. based on plausibility
check)
Field supervisors (optional) Responsible for monitoring and supervision during the data
collection along with the survey managers; ensuring
questionnaires are fully and correctly filled
Team leader Responsible for the quality and reliability of the data collected,
including appropriate sampling procedure
enumerators responsible for taking and recording anthropometric
measurements
58
Depending on the size of the questionnaire and on the repetition of tasks, either the team
leader or one of the enumerators can be trained in administering the questionnaires.
In addition to these three (or four) members, it is recommended to include a representative of
the ward chief in the survey team as a guide. This person can introduce the survey team to the
households and assist in guiding the team around the location. Wherever possible, the guide
should be identified before the data collection (i.e. during community meeting to decide on
the timing of data collection).
Two to six teams may be needed depending on the number of households to be visited, the
size and the accessibility of the area covered. To reduce the time of data collection, the
number of teams can be increased. However, more teams will be more difficult to train,
organise (logistically) and manage, and may result in increasing the variation in the precision
of the results.
It is always useful to recruit and train 2-3 more enumerators than the number of enumerators
required as stand-by enumerators. In case if there is any problem with any of the enumerators
selected, the stand-by enumerators can be used.
5.2.2 Training survey team members
The training of enumerators is essential in ensuring that accurate data are collected. Such
training should be conducted before each survey. Every team member should undergo exactly
the same training, whatever their former experience, to ensure standardization of methods. In
large-scale surveys with a great number of teams, it is recommended to split enumerators in
groups of 10-15 people to increase the effectiveness of training.
The training should be tailored down to the level of tasks expected of the field staff. Note that
topics such as causes and consequences of malnutrition, protocols and treatment options for
malnourished children, etc. are unnecessary and should be avoided during the training. The
duration of each aspect of the training depends on the experience of the staff and the design
of the survey. The following should at least be included in designing a training program for
survey enumerators:
1. Theoretical sessions (1 day)
A clear explanation of the objectives of the assessment.
A clear explanation of roles and responsibilities of each team member.
An explanation of the sampling method that stresses the reasoning behind and
importance of each child and household member having an equal chance of being
selected (including households without children for the death rate survey, if death
rate is being measured).
Measurement techniques.
Training on using a calendar of event.
Conducting focus group discussions, key informant interviews, and observations,
including gender issues
59
2. Practical sessions (1 – 1.5 days)
Using the field questionnaires, data entry forms and other survey tools.
Training on filling out the mortality form.
If possible, visit a nutritional rehabilitation unit to see and feel children with severe
malnutrition especially oedema.
Taking anthropometric measurements.
Use of weight-for-height z-score tables for identification of acute malnutrition and
referral for nutrition care.
3. Standardization of measurements11
(0.5 – 1 day)
To ensure that enumerators take measurements of good quality
To be conducted with all the enumerators with 10 children, as described below.
4. Pre-test (1 day)
To ensure proper organisation of the team and of the material
To ensure good understanding of the sampling method
To estimate the time spent in each household
To ensure that teams are properly organized
5.2.3 Standardization of weight, height, and MUAC measurements
The objective of a standardization test is to assess whether or not the enumerators are taking
the measurements in a standard and accurate way, and to test their precision in taking
measurements. This test must be conducted with 10 children before each survey.
A standardisation test involves repeating a measurement twice on 10 different children, with
a time interval between measurements on the same child. For each enumerator, the difference
between the two measurements is calculated to assess the precision, and a mean of the
measurement is calculated to assess the accuracy – this can be compared with the
supervisor’s values or the mean value of all enumerators.
The equipment used in the exercise should be the same equipment used to measure children
in the survey itself. The team members will rotate but the equipment should not, so that each
child is always measured with the same equipment (the team is being tested not the
equipment).ENA software should be used to calculate precision and accuracy of height,
weight, and MUAC measurements.
Steps for conducting standardization test:
Select 10 children whose ages fall within the range for the study (6–59 months), and
given them an ID number.
The supervisor carefully takes weight, height, and MUAC measurements on each child
without allowing the trainees to see the values.
Each child, with his/her mother, remains at a fixed location with the ID number clearly
marked. The distance between each child should be far enough to prevent the trainee
11
The purpose and procedure of the test should be explained at the beginning of the test
60
from seeing or hearing each other’s results.
Each pair of trainees starts with a different child. The trainees should carefully conduct
the measurements and clearly record the height, weight, and MUAC measurements on
their form.
When each member of the pair has done the measurement, they should move on to the
next child.
After a break, the process should be started again. Without seeing the measurements
they previously made, each enumerator measures each child a second time.
Example of data collection forms for the standardization test is given in annex 11.
5.2.3.1 Outputs of the standardization test
The data obtained from each child by both the supervisor and the enumerators should be
entered into the ENA software and analysed. There are 2 types of reports/results that can be
generated once all data are entered. They are: a) report from previous ENA versions and b)
new report.
The test of standardization allows survey coordinator to identify enumerators that take good
measurements. If some enumerators performed poorly, a number of actions can be taken:
When extra enumerators were trained, the standardization test can be used for the
final selection of enumerators, only the best performer being included in final
survey team.
Enumerators who performed poorly can be given tasks within the survey team that
are not related to the measurement, such as data recording or measurer-assistant.
Additional training can also be provided to enumerators who performed poorly. A
subsequent test should then be administered to them to make sure that their
performance reached an acceptable level.
Box 1. Using ENA 2011 to assess the outcome of the standardization test
In the Training screen of ENA, enter the data obtained for each child, measure 1 and 2,
starting with the supervisor and then with each enumerator. Note that MUAC data must
always be entered in millimetre to get the correct results (ENA will produce reports for data
entered in other units such as centimetre but these results will be wrong).
61
Click on "Report from previous ENA versions" to get evaluation of each enumerator. ENA
will generate an evaluation report in a separate Microsoft Word document.
Click on "New report" to get evaluation of each enumerator. ENA will generate an evaluation
report in a separate Microsoft Excel document.
Report for Evaluation of Enumerators
Weight:
Precision: Accuracy: No. +/- No. +/-
Sum of Square Sum of Square Precision Accuracy
[W2-W1] [Superv.(W1+W2)-
Enum.(W1+W2]
Supervisor 0.17 4/5
Enumerator 1 0.21 OK 0.30 OK 4/5 4/4
Height:
Precision: Accuracy: No. +/- No. +/-
Sum of Square Sum of Square Precision Accuracy
[H2-H1] [Superv.(H1+H2)-
Enum.(H1+H2]
Supervisor 2.94 4/4
Enumerator 1 3.24 OK 5.24 OK 6/3 7/3
Enumerator 2 4.31 OK 11.95 POOR 6/4 8/1
MUAC:
Precision: Accuracy: No. +/- No. +/-
Sum of Square Sum of Square Precision Accuracy
[MUAC2-MUAC1] [Superv.(MUAC1+MUAC2)-
Enum.(MUAC1+MUAC2]
Supervisor 9.00 4/2
Enumerator 1 126.00 POOR 1995.00 POOR 4/2 10/0
Enumerator 2 34.00 POOR 149.00 POOR 2/5 7/2
Enumerator 3 16.00 OK 2123.00 POOR 2/5 10/0
Enumerator 4 33.00 POOR 162.00 POOR 1/7 7/3
Enumerator 5 29.00 POOR 126.00 POOR 1/8 5/3
For evaluating the enumerators the precision and the accuracy of their measurements is calculated.
For precision the sum of the square of the differences for the double measurements is calculated.
This value should be less than two times the precision value of the supervisor.
For the accuracy the sum of the square of the differences between the enumerator values
(weight1+weight2) and the supervisor values (weight1+weight2) is calculated. This value should be
less than three times the precision value of the supervisor.
To check for systematic errors of the enumerators the number of positive and negative deviations
can be used.
62
Note that although reference values are available at the end of the Microsoft Excel output
for the new report, detailed guidance on how to interpret these results is still awaited.
5.2.3.2 Practical tips to conduct the standardisation test
Standardisation test is a - very difficult to conduct in practice, but is it extremely important.
In order to facilitate its implementation, it is recommended to plan the standardization in a
location where there is enough space. Open-air area might be more appropriate than a closed
room. Since many children need to be involved, conducting the test in a community rather
than bringing children to a training centre might ease the availability of children and reduce
the noise/stress resulting from a confined space with many people/children.
In order to reduce the burden on children (who each have to be measured twice by the
supervisor and then twice by each enumerator), children can be taken into "batches". A batch
of 10 children can be taken for 3 or 4 enumerators. The number of measurements made by the
supervisor will be higher. However, although more children will be involved overall, taking
batches will considerably reduce the pressure on each individual child for whom this exercise
is very unpleasant.
5.2.4 Pre-testing
Field training is practical and not confined to the classroom. It takes place after the teams are
able to make accurate and precise measurements, have "passed" the standardization test, and
have formed teams that have practiced working together. For field testing the teams go to a
convenient, local ward that has not been chosen to contain a cluster. The teams should go
through all the steps in conducting the survey. They practice selecting the houses that will
Standardisation test results Precision Accuracy OUTCOME
Weight subjects mean SD max Technical errorTEM/meanCoef of reliabilityBias from supervBias from medianresult
# kg kg kg TEM (kg) TEM (%) R (%) Bias (kg) Bias (kg)
Supervisor 10 12.8 4 0.3 0.1 0.7 99.9 - 0.2 TEM acceptableR value goodBias reject
Enumerator 1 10 12.9 4 0.2 0.1 0.8 99.9 0 0.3 TEM poor R value goodBias reject
enum inter 1st1x10 12.9 4.1 - 0 0 100 - - TEM good R value good
enum inter 2nd1x10 12.9 4.2 - 0 0 100 - - TEM good R value good
inter enum + sup2x10 12.8 4 - 0.1 0.7 100 - - TEM good R value good
TOTAL intra+inter1x10 - - - 0.1 0.8 99.9 0 0.2 TEM acceptableR value goodBias reject
TOTAL+ sup2x10 - - - 0.1 1 99.9 - - TEM acceptableR value good
Height subjects mean SD max Technical errorTEM/meanCoef of reliabilityBias from supervBias from medianresult
# cm cm cm TEM (cm) TEM (%) R (%) Bias (cm) Bias (cm)
Supervisor 10 84.2 2.3 1.2 0.4 0.5 97.2 - -0.8 TEM good R value acceptableBias good
Enumerator 1 10 84.3 2.2 1.2 0.4 0.5 96.7 0.1 -0.7 TEM acceptableR value acceptableBias good
Enumerator 2 10 84.6 2 1.2 0.5 0.5 94.8 0.4 -0.4 TEM acceptableR value poorBias good
enum inter 1st2x10 84.5 2.1 - 0.5 0.6 93.5 - - TEM acceptableR value poor
enum inter 2nd2x10 84.4 2.1 - 0.4 0.4 96.8 - - TEM good R value acceptable
inter enum + sup3x10 84.4 2.1 - 0.4 0.5 95.6 - - TEM good R value acceptable
TOTAL intra+inter2x10 - - - 0.6 0.8 90.8 0.2 -0.6 TEM acceptableR value poorBias good
TOTAL+ sup3x10 - - - 0.6 0.7 91.8 - - TEM acceptableR value poor
MUAC subjects mean SD max Technical errorTEM/meanCoef of reliabilityBias from supervBias from medianresult
# mm mm mm TEM (mm)TEM (%) R (%) Bias (mm) Bias (mm)
Supervisor 10 116.7 7 2 0.7 0.6 99.1 - -1.3 TEM good R value good
Enumerator 1 10 123.2 8.5 7 2.5 2 91.2 6.5 5.2 TEM rejectR value poorBias reject
Enumerator 2 10 117.5 6.3 3 1.3 1.1 95.8 0.8 -0.5 TEM poor R value acceptableBias good
Enumerator 3 10 123.5 6.6 2 0.9 0.7 98.1 6.8 5.5 TEM good R value acceptableBias reject
Enumerator 4 10 117.3 6.3 3 1.3 1.1 95.8 0.6 -0.8 TEM acceptableR value acceptableBias good
Enumerator 5 10 116.8 6.6 3 1.2 1 96.7 0.1 -1.3 TEM acceptableR value acceptableBias good
enum inter 1st5x10 119.3 7.7 - 4.1 3.4 71.4 - - TEM rejectR value reject
enum inter 2nd5x10 119.9 7.2 - 3.8 3.2 72.5 - - TEM rejectR value reject
inter enum + sup6x10 119.1 7.4 - 3.8 3.2 73.9 - - TEM rejectR value reject
TOTAL intra+inter5x10 - - - 4.2 3.5 67.4 3 1.1 TEM rejectR value rejectBias poor
TOTAL+ sup6x10 - - - 4.1 3.4 70 - - TEM rejectR value reject
Suggested cut-off points for acceptability of measurements
Parameter MUAC mmWeight KgHeight cm
individual good <1.0 <0.04 <0.4
TEM acceptable<1.3 <0.10 <0.6
(intra) poor <2.1 <0.21 <1.2
reject >2.1 >0.21 >1.2
Team TEMgood <1.3 <0.10 <0.5
(intra+inter)acceptable<2.1 <0.21 <1.0
and Total poor <3.0 <0.24 <1.5
reject >3.0 >0.24 >1.5
R value good >99 >99 >99
acceptable>95 >95 >95
poor >90 >90 >90
reject <90 <90 <90
Bias good <1 <0.04 <0.4
From sup if goodacceptable<2 <0.10 <0.6
outcome, otherwisepoor <3 <0.21 <1.4
from medianreject >3 >0.21 >1.4
63
form the cluster, approaching mothers and explaining the purpose of the survey, making the
measurements, and conducting the interviews. This step is essential for the teams to feel
confident when they begin conducting the actual survey.
The field training data from each of the teams should be entered into ENA and analysed. The
teams should each have selected different households from the ward (otherwise it is likely
that the selection was not random). Each team’s results will be slightly different; this is used
as a practical demonstration of the effect of sampling error and the importance of taking a
random sample.
There should be a time gap between the pre-test and the beginning of the data collection so
that if there is anything that needs to be fixed before the actual data collection, it can be done.
5.3 Managing the survey
The coordinating survey manager has the overall responsibility for training team members,
visiting teams in the field, ensuring that households are selected properly, and ensuring the
equipment is checked and calibrated each morning during the survey and that measurements
are taken and recorded accurately.
Unexpected problems nearly always arise during a survey, and the survey manager is
responsible for deciding how to overcome them. Each problem encountered and decision
made must be promptly recorded and included in the final report. The survey manager is also
responsible for overseeing data entry and for the analysis and report writing.
Where possible, the survey manager should organize an evening wrap-up session with all the
teams together to discuss any problems that have arisen during the day12
.
5.4 Enhancing the accuracy of the data collected
There are several ways to improve the quality of the data collected during a nutrition survey:
Ensure errors in the field are minimised by using good quality equipment that is
regularly calibrated.
Check the forms for blank entries before leaving a household to make sure no data
is left out. The team leader should review all questionnaires before leaving an area
in order to make sure no pieces of data have been left out. If there are any problems
the team should return to the household as soon as possible and fix them.
Check for data collected. Each evening, or during the next day while the teams are
in the field, the survey manager should arrange for data to be entered into the
computer. Recording errors, unlikely results, and other problems with the data may
become clear at this stage. ENA software will automatically flag abnormal values as
data are entered. Each morning, before the teams set out for the day, there should be
a short feedback session. If any team is getting a large number of “flagged” results,
the survey manager should accompany that team the next day. If the results are very
12
This may not be possible if the survey area is large and teams are widely separated or remain in the field for
several days. Communication with teams in the field is often very difficult. In such circumstances, team leaders
must be sufficiently trained to make decisions independently.
64
different from those obtained by the other teams, it may be necessary to repeat the
cluster from the day before.
Apart from the evening and morning meetings, survey team members should be encouraged
to regularly discuss their experiences and findings together. This often brings out important
points, and sometimes shows where survey methods need to be modified.
It is highly recommended to use a cluster control form (see annex 12) during the data
collection. This can be used to summarize what has been done in each cluster, keep records of
household that refused to participate, keep track of absent households and households that
had absent children (to be visited again), etc. The cluster control form will also be good tool
for the survey manager and field supervisors when conducting field monitoring.
5.5 Supervising data collection team
Field supervision is important in ensuring valid data collection and minimising bias. The
survey manager should:
Make frequent unannounced spot checks on the teams in the field.
Ensure that the methodology is closely followed and document any deviations.
These need to be corrected as soon as possible.
Check all forms to ensure that all sections are accurately completed.
Ensure that all instruments to be used by the survey teams are calibrated every day.
It is particularly important to check cases of oedema, as there are often no cases seen during
the training and some team members may therefore be prone to mistaking a fat child for one
with oedema (particularly with younger children). The survey manager should note teams
that report a lot of oedema, and visit some of these children to verify their status. Any reports
of measles (death/illness) should also be verified.
Although spot checks are necessary throughout the data collection, it is of particular
importance during the first couple of days of data collection as survey teams are still getting
used to the survey procedures. Survey teams should be closely monitored and any deviation
from the standard survey procedure should be immediately corrected.
5.6 Minimising Bias
Bias is anything other than sampling error that causes the results of the survey to be different
from the true population prevalence. Bias cannot be calculated nor its effect upon the result
assessed. It is the main reason why surveys may not give an accurate result.
As bias cannot usually be calculated or corrected by the computer after data collection is
finished, it is critical to avoid bias during sampling and data collection. Bias is minimized by
adequate training and use of good technique.
However, the quantitative data can be examined using ENA to see if there is likely to be
some form of systematic bias. The teams should be aware that such techniques will be
applied during the analysis to discourage their succumbing to the temptation to take shortcuts.
65
Examples of bias
1. Because the foot piece of a length-board was loose, one team systematically measured
the height of each child 1 cm taller than he or she really was. Even though weight was
accurately measured, each child’s WFH z-score was lower than it should be and the
prevalence of wasting was exaggerated. Any inaccuracy in the equipment or
measurement technique will lead to systematic bias.
2. Inaccurately taken weight and height – even when the inaccuracy is random and evenly
distributed between over- and under-measurement – results in systematic
overestimation of the prevalence of wasting. This overestimate is greater for severe
malnutrition than for moderate malnutrition, and relatively greater when the true
population prevalence is low than when it is high.
Shortcuts are likely to be taken if the survey teams are required to work too hard, if there is
inadequate time allocated to rest periods and refreshments, or if the time that can be spent in a
particular household to administer the mortality form and measure the children properly is
insufficient. Therefore, the data may be much more accurate if there are fewer, rather than
more, households in each cluster. This tends to be more common in rough terrain or when
there are long distances to walk.
The following are some of the sources of bias that occur during the interview.
Recall error: Respondents often fail to recall all deaths during a given recall period.
Infant deaths, in particular those within a short time after birth, are particularly
under-reported. Respondents may also misreport ages, dates, and salient events.
"Calendar" error: Respondents may report events as happening within the recall
period when they did not (or vice versa) due to lack of clarity about dates.
“Age heaping”/digit preference: Respondents may round ages to the nearest year i.e.
12, 24, 36 and 48 months.
Sensitivity/taboos about death: In general, the death of a household member is not a
subject discussed readily with strangers.
Deliberate misleading: In some populations, with experience of relief operations,
some respondents may deliberately give incorrect answers in the expectation of
continuing or increased aid.
Interviewer error: Enumerators may ask questions or write down answers
incorrectly, skip questions, assume answers, or rush respondents in an effort to
complete the interview quickly.
5.7 Ethical considerations
Although nutrition assessments would not qualify for research, data should still be collected
in an ethical manner. Some ethical issues are highlighted here:
1. Provide sufficient information to local authorities about the survey. Such information
includes the purpose and objectives of the survey, the nature of the data collection
procedures, the targeted subgroups in the community. Where possible, survey
procedures and copies of survey questionnaires should be available to the community
leaders for their comments prior to the survey.
2. Verbal consent must be obtained from all adult participants and parents guardians for
children in the survey. Every individual has the right to refuse to participate in the
survey. Such a decision should be respected.
66
3. The confidentiality of survey data should be protected by ensuring that information
leading to identification of individuals is not shared, especially in the communities.
4. Referrals for survey participants who show signs or symptoms that require immediate
clinical attention should be made. Although a nutrition survey differs from a nutritional
screening, children who show signs or symptoms that require immediate clinical
attention should be referred to the closest health centre. Team leaders should refer
children if:
a. They have bilateral oedema;
b. Their weight-for-height is below -2.0 z-score.
c. Their MUAC is less than 125 mm.
67
6. DATA ENTRY AND DATA QUALITY CHECK
6.1 Data entry
Nutrition and death rate survey data can be entered directly into ENA software. Other data
can be entered into ENA software (in the Data Entry Anthropometry screen), into ENA for
EPI INFO software (the hybrid version of ENA and EPI INFO software), or into any other
appropriate software.
6.1.1 Data Entry: Nutrition Survey Data
The nutrition survey data can be directly entered into ENA using the Data Entry
Anthropometry screen. The sections below describe how to enter nutrition survey data into
ENA.
6.1.1.1 Preparing for data entry in ENA
Before starting the data entry in ENA, the Options screen and Variable View of the Data
Entry Anthropometry screen must be visited and appropriate parameters must be set and
defined. If there are more than one data entry clerk, parameters must be set identically for all
of them prior to starting data entry. How to set the Options screen and Variable View of the
Data Entry Anthropometry screen is described below.
Box 6.1 Setting ENA Options screen for data entry
For ease of reference, the Option screen has been divided into 9 smaller sections and
described below.
1. Automatic fill out section can be used to automatically fill data – the survey date, cluster
no., and team no. will be repeated and ID and HH will be incremented. Since it may cause
problems when entering data from twins in one household, it is not recommended to
select household no. for automatic refill, especially when the data entry staff are entering
data in ENA for the first time
2. This should be decided based on the survey context – e.g. if majority of the surveyed
children have age information in months, months should be selected.
3. Data can be either entered directly or selected from a calendar (using pull down editors).
This should be decided based on what the data entry staff are familiar with.
4. If any of the surveyed children are measured with clothes, height measurement is taken
against the standard (i.e. a 37 month old child measured by lying down), or any weighing
1
2
3
4
5
6
9
8
7
68
needs to be applied, this checkbox must be selected so that additional columns will be
displayed on the Data Entry Anthropometry screen. If children are weighed with
clothes, average weigh of the clothes should be entered in grams.
5. Depending on the programme that is used in the computer either MS Office or Open
office should be selected as the programme to generate reports
6. The age groups included in the survey report and plausibility check report will be based
on the age group set in the Options screen. For standardisation purposes, all surveys
conducted in Nepal should follow the default age group settings in ENA (i.e. as shown
above).
7. Although limits for anthropometric data analysis can be set based on either age or height,
in line with the inclusion criteria, age (6-59.99 months) should be selected. If survey is
conducted on other age groups such as 0-5.99 months, this should be set accordingly.
8. MUAC cut-offs should be set as shown above in line with the national standards.
9. Based on the option selected in this section, the z-scores included in the analysis will
vary. SMART flags should be maintained
Note: any change in the Options screen must be saved using the Save button at the bottom
of the screen to effect the changes.
6.1.1.2 Defining variables in the variable view of the data entry anthropometry
screen
In order to limit data entry errors, it is advised to set adequate limits and "checks" before
starting data entry, especially if additional data are being entered in ENA. The variables are
defined in the variable view of the data entry anthropometry screen (see box 6.2 for details).
During data entry, values that do not match the variable type or range will be highlighted.
For example, range values for weight are intended to highlight values that are abnormal for
children aged between 6 and 59 months. Data are therefore highlighted to check for data
collection or data entry error, and allow for easy identification of data that should be verified.
Similarly, the purpose of setting range values for height is to identify extreme values that are
highly impossible for a population of children aged 6 to 59 months.
Default values are already set for the following variables: cluster, team, age, weight, height,
and nutrition indices (WHZ, HAZ, and WAZ). These values should be reviewed and
modified, if needed, for each survey. Default values for the ranges of nutrition indices allows
for identification and verification of measurements that are extreme and probably result from
measurement errors:
WHZ: <-5 SD or >+5 SD
HAZ: <-6 SD or >+6 SD
WAZ: <-6 SD or >+5 SD
Those values correspond to the values of flags in EPI INFO software and should not be
changed.
69
Box 6.2: Defining variables in ENA for data entry
Variables are defined in the Variable View of the Data Entry Anthropometry screen.
Steps to define variables in the “Variable View” in ENA
1. Review each variable that is already included in the screen and modify ranges as necessary
- To change a value, double click on the current value, delete it, and enter the new value
- Do not modify range values for nutrition indices (WHZ, HAZ, and WAZ)
- Note that if the type of variable is shown as character it has already been defined
2. Enter additional variables as required and define them
- Enter the name of the data under the column Name
- Select the variable type (n: numerical, c: character, d: date)
- Specify the Range Low and Range High
1: Variable view where variables are defined
2: Parameters that are used to define variables
3: Variables that have already been defined but need adjustment for each survey
4: Example of new variables added
1
2
3
4
70
6.1.1.3 Entering data – Data Entry Anthropometry
The Data View of the Data Entry Anthropometry screen is used to enter the data into the
software. The first 19 columns of the Data Entry Anthropometry screen is pre-defined
however, in default setting, only the first 15 columns are visible (to see the rest of the
columns, the checkbox for ‘showing columns for measure, clothes, and weighing
variables’ in the Options screen needs to be selected). The nutrition indices are
automatically calculated and filled in the grey cells as data are entered.
The first 5 columns (SURVDATE, CLUSTER, TEAM, ID, and HH) are already filled out
with data when the software is opened. These values need to be reviewed and changed as
needed before data entry is started. Note that the ward that has been chosen for the cluster is
already in the software (i.e. planning screen) – the same cluster number should be used in the
column for cluster.
As the data are entered, data in SURVDATE, CLUSTER, and TEAM columns will default to
the last entered information and the ID and HH number will increment automatically by one
for each new record (if automatic fill out option for these variables are selected in the
Options screen). If automatic fill out is set for HH, household number should be reviewed for
each record before the rest of the information is entered as there may be more than one child
in the same house – in this case, each child in the same household should be given the same
household number.
Either the birth date or the age in months should be entered for the age of the child. If the
birth date is entered, the age will be automatically calculated. If age is entered the birth date
field will be left blank and age in month is entered, the birthdate field will be empty. It is not
necessary to enter an age to proceed. If age is not entered, WFA and HFA indices will not be
calculated but WFH index will be computed.
Weight should be entered in kg, height in cm, and MUAC in mm. ENA will analyse the data
entered in other units (e.g. MUAC in cm) but the results will be wrong. Sex should be filled
in with "m" for male and "f" for female, and oedema must be filled in "y" for presence of
oedema and "n" for absence of oedema. If the oedema field is not entered, it defaults to
oedema being absent during analysis. ENA does not calculate WHZ and WAZ for children
with oedema.
If length measurement were taken on children who are older than 24 months, this must be
specified by entering “l” in the column, MEASURE, for those children so that ENA will
adjust the length measurement before calculating nutrition indices for these children.
Similarly, if height measurement was taken on children less than 24 months of age, this must
be indicated as “h’ in the column MEASURE when entering records for those children. There
is no need to enter “l” under the column MEASURE for children under 24 months old who
were measured lying down or children older than 24 months whose height measurement was
taken. ENA will calculate the indices using the standard protocols (i.e. without adjusting
measurements) if the column MEASURE is left blank.
71
If children were measured with clothes, this should be specified in the column, CLOTHES, as
“y” (note that the average weight of the cloths must be specified in the Options screen –
under ‘weight for subtraction of clothes’ – for ENA to deduct it from the total weight
measurement before calculating the nutrition indices). If the CLOTHS column is left blank it
is assumed that weight measurement was taken on the child with no clothes on.
If weighing needs to applied, weights need to be calculated and entered for each record in the
column WTFACTOR so that it can be taken into account when the overall nutrition
prevalence estimates are calculated.
Example 6.1: A situation where weighing needs to be applied
Two nutrition surveys were conducted using the same methodology in one region in Nepal.
Of the total population of about 190,000 people, one survey covered about 20% of the
population while other survey covered the rest of the population. Separate estimates were
calculated for each survey. For programmatic purposes, the MMP would like to calculate one
estimate for the entire region using data from the 2 surveys. Weighing needs to be applied as
the probability of selection is not the same in both populations.
If there appears to be an error in the data entered, the field will turn pink. The cut-off points
to alert the person entering the data can be set in the options screen/variable view as
discussed previously in this chapter. When a field turns pink, the first thing to do should be to
check whether it is not due to data entry. If data is entered as recorded on the questionnaire,
and if the team is still in the area, the team can return to the household to retake the
measurements.
Non-anthropometric data can also be entered in the Data Entry Anthropometry screen as
described above.
72
Box 6.3: Using ENA for data entry – Data Entry Anthropometry
Steps to enter data in Data Entry Anthropometry screen:
1. Go to Options screen and set parameters to fit the survey data
2. Go to Variable View of the Data Entry Anthropometry screen and define variables
3. Go to Data View of the Data Entry Anthropometry screen and enter data
1: Data view that is used to enter data
2: Variables that have already been defined
3: Example of variables that have been newly added
4: Buttons that are used to add, delete, sort, or filter variables (i.e. columns) and rows; the
Report Plausibility check button can be used to generate plausibility report
5: Percentiles of an individual record
6.1.1.4 Missing data
If certain critical pieces of information are missing from a child’s survey record, it will not be
possible to include the child in some of the anthropometric data analyses:
Age: If information on age is missing, wasting and oedema can be assessed for the
individual child. The child however will not be included in the analysis and overall
estimation of malnutrition in the area as the inclusion criteria for analysis is 6-59.99
months (set in Options screen).
Sex: If information on sex is missing you should still include the child in the
assessment of oedema.
Height: If information on height is missing you cannot include the child in the
assessment of wasting. However, the child can still be included in an analysis of
oedema, because any child with oedema is severely malnourished.
Weight: If information on weight is missing you cannot include the child in the
assessment of wasting. However, the child can still be included in an analysis of
oedema.
2 3
5 4
1
73
MUAC: if MUAC is missing, the child can be included in the assessment of wasting
and oedema.
As long as MUAC and age data are available, the GAM by MUAC can be calculated even if
all other information is missing.
6.1.1.5 Data outside the required range
Most nutrition surveys measure children aged 6–59 months. Children outside these ranges
should not be included in the results. These values depend upon the defaults that have been
set in the Variable View panel as well as the Options screen of the software. Any child
outside the age range will be marked by the program.
If height data are missing, the anthropometric indices of interest cannot be calculated.
However, if the age is within range, the child can be included if there is oedema. The
accepted height range can be altered in the variable view sheet, for example, to change the
range to 60cm–100cm if the population is very stunted. These choices must be included in the
report.
Thus, by default, if a child is 55 months old and 112cm the child will be included. However,
if the child is 65 months old, it should not be included and the computer will automatically
exclude the child in the results.
6.1.2 Data Entry: Death Rate Survey Data
Death rate survey data can be directly entered into ENA using the Death Rates screen. The
data can be entered either by individual or by household – the summary of information is
used when data is entered by household. The sections below describe how to enter data at
individual as well as household level.
6.1.2.1 Data entry individual level
In addition to the identification variables such as survey date, cluster, team, and household,
data on 6 other variables need to be entered when data are entered at individual level. If death
rate survey is conducted alongside a nutrition survey and nutrition survey data have already
been entered, the identification variables can be imported into the death rate data entry screen
from the Data Entry Anthropometry using the shortcut button on top of the death rate Data
Entry Individual Level panel. The rest of the information on the death rate survey form can
then be entered by selecting the respective household. As the data is entered for each
household, the household level data will be automatically computed and displayed.
In addition to entering data required to calculate the crude and 0-5 death rates, the Data
Entry Individual Level can also be used to the enter additional data such as cause of death
and location of death. The individual level data entry also allows you to compute age and sex
specific death rates.
74
Box 6.4: Using ENA for entering death rate survey data – data entry individual level
1: Data Entry Individual Level panel that is used to enter individual level death rate data
2: Shortcuts, including an a option to import HH and Cluster information from Data Entry
Anthropometry Screen
3: Recall days that is required to calculate different death rates. This is linked to the Recall
period in days in the Planning screen
4: Data entry screen
5: References for cause and location of death. The options can be modified and new
categories can be added
6: As data are entered in the data entry screen (4), household information will be computed
automatically and displayed here. Note: if identification information (Date, Cluster, Team,
and HH) is imported from the Data Entry Anthropometry, the HH should be selected from
this screen before rest of the data are entered in the data entry screen above (4)
7: Age of children less than 12 months should be recorded as 0 (years)
6.1.2.2 Data entry household level
The summary of full household census is entered when death rate data is entered at the
household level. In addition to identification data (household and cluster numbers), 9
variables need to be entered per household. Crude and 0-5 death rates are the only 2 estimates
that can be obtained when household level summary data is entered. Age (other than 0-5)/sex
specific deaths or cause/location of death cannot be entered or analysed when data entry
household level pane is used.
1
2 3
4 5
6
75
Box 6.5: Using ENA for entering death rate survey data – data entry household level
Death rate survey data summarized by household should be entered in the Data Entry
Household Level screen of ENA. Data from households is entered in rows. The cluster and
household numbers should be the same as the ones entered for the anthropometry data. These
can be imported from the Data Entry Anthropometry screen if nutrition survey is also
conducted alongside the death rate survey and nutrition survey data has already been entered
into the software.
The following 11 variables should be entered for each household:
Cluster No.
Household No.
1: Data Entry Household Level panel that is used to enter household level death rate data
2: Shortcuts, including an a option to import HH and Cluster information from Data Entry
Anthropometry Screen
3: Recall days that is required to calculate different death rates. This is linked to the Recall
period in days in the Planning screen
4: Death rate survey data entry screen
It should be noted that the Recall days from planning screen (3) is given here for reference
purposes only and cannot be changed. ENA takes the number of recall days specified in the
Planning screen to calculate the different death rates. The same ENA file that has been used
to calculate the sample size for a death rate survey should be used to enter death rate survey
data from the survey. If this is not possible, the Planning screen should be visited and the
Details Total U5
Current HH Members
Arrivals during the Recall period
Number who have left during Recall period
Births during recall
Deaths during recall period
1
2
3
4
76
recall period must be set accordingly before starting data entry (for both individual as well as
household level data entry).
6.1.3 Using ENA for double-entry
It is common practice, where the results of the survey are critical and there is sufficient time
available to enter all the data twice and compare the two resulting datasets. The data should
be saved in two separately named files. The data can then be compared automatically using
the "check for double entry" in the Extras menu of ENA, as shown in box 6.6. When there
are discrepancies in data, the questionnaire should be verified to decide which data to
maintain in the final database.
Box 2. Using ENA for double-entry
Steps to check for double entry
1: Select the Check of double entry option from the Extras menu
2: Select the 2 datasets to compare, check whether anthropometry or death rate survey data
should be compared, and click on OK.
3: The double entry report will allow you to see discrepancies in data entry
Check of double entry
Anthropometric Data
Difference in Line: 1
Dataset A .as 25/4/2007 23 4 6 6 f 31 14.1
1
2
3
77
- If there are discrepancies check the original questionnaires to identify the actual value)
- Make the corrections in only one dataset that will be used for data analysis.
6.2 Determining nutritional status of individuals and populations
Nutritional status of individuals and populations are determined either by calculating
nutrition indices or comparing Mid-upper arm circumference (MUAC) values with
established thresholds. Survey results should always be reported by nutrition indices as well
as MUAC thresholds.
6.2.1 Nutrition indices
To calculate the nutrition indices, data on the weight, height, age, and presence of oedema are
needed. The relationship of these measurements to each other is compared to international
reference standards. The three nutrition indices that are typically appraised through nutrition
survey are: height-for-age (HFA), weight-for-height (WFH), and weight-for-age (WFA).
WFH, HFA, and WFA are calculated for individuals and groups using ENA software.
6.2.1.1 Height-for-age (HFA)
Growing children get taller, and the height of a child in relation to a "standard" child of the
same age gives an indication of whether the growth has been normal or not. This index of
growth is called height-for-age. Children who have a low HFA are referred to as stunted.
Growth is a relatively slow process, and if a child of normal height stops growing it takes a
long time for that child to fall below the cut-off point for stunting13
. For this reason, HFA is
often used to indicate long-standing or chronic malnutrition. If the insult that led to stunting is
in the past, it is possible that the current growth rate is actually normal (although this is
unusual without a change in the family circumstances). Stunting may also be due to
intrauterine growth retardation followed by normal postnatal growth.
6.2.1.2 Weight-for-height (WFH)
A child getting taller will also gain weight if body proportions remain normal. A thin child
will weigh less than a normal child of the same height. Weight-for-height is a measure of
how thin (or fat) the child is. Because weight gain or loss is much more responsive to the
present situation, WFH is usually taken to reflect recent nutritional conditions. Being
excessively thin is called wasting. It is also often termed "acute malnutrition", although
individual children may have been thin for a long time.
6.2.1.3 Weight-for-age (WFA)
Neither stunted nor wasted children weigh as much as normal children of the same age.
Weight-for-age is thus a composite index, which reflects both wasting and stunting, or any
combination of both. In practice about 80% of the variation in WFA is related to stunting and
about 20% to wasting. It is not a good indication of recent nutritional stress. It is used
13
A child who is 100% of normal growth who falters to 70% of normal will take up to half his life to fall below
the usual cut-off point and be labelled as moderately stunted. Thus, a 1-year-old child who is gaining height at
70% of normal will not be designated as stunted for six months.
78
because it is an easy measurement to take in practice, and can be used to follow individual
children longitudinally in the community.
6.2.2 Mid-upper arm circumference (MUAC)
MUAC directly assesses the amount of soft tissue in the arm and is another measure of
thinness (or fatness), like WFH. Although it is easier to measure MUAC than WFH, it is
more difficult to make a precise measurement. MUAC is the best indicator for mortality. It is
commonly used in the community (for screening) to identify individual children in need of
referral and as an admission criterion for feeding programmes. Because MUAC is used in
programmes, it is useful to know the relationship between WFH and MUAC in a particular
community to establish a full nutrition program including screening. MUAC data should
always be collected and reported as part of a nutrition survey.
6.2.3 The reference population curves
To assess malnutrition as determined by WFH, WFA, and HFA, individual measurements are
compared to an international reference standard. There are 2 references that are commonly
used to compare individual measurement with. They are: a) "NCHS reference" derived from
surveys undertaken in the United States (NCHS/WHO/CDC reference table, 1977) and b) the
"WHO standards" (WHO Child Growth Standards, 2006) are developed from the Multicentre
Growth Reference Study conducted on populations from 6 different ethnic backgrounds and
cultural settings (Brazil, Ghana, India, Norway, Oman and the USA).
The reference values are used as a standard to compare nutritional status in different regions,
and in populations over time. All survey results must be reported using the WHO standards in
Nepal. The results based on the NCHS reference may be included as an annex to the survey
report.
6.2.4 Expression of nutrition indices
When anthropometric data are analysed with the WHO standards, nutrition indices are
expressed in z-scores however, with NCHS reference, nutrition indices are expressed in two
ways: as z-scores derived from the reference and as the percentage of the median value of the
reference. The z-sore is described below.
6.2.4.1 The z-score
A z-score is a measure of how far a child is from the median WFH of the reference (often
written as WHZ). In the reference population, all children of the same height are distributed
about the median weight, some heavier and some lighter. For each height group, there is a
standard deviation among the children of the reference population. This standard deviation is
expressed as a certain number of kilograms at each height. The z-score of a child being
measured is the number of standard deviations (of the reference population) the child is away
from the median weight of the reference population at that age group.
79
The WHZ is based upon the child’s weight, the median weight of children of the same height
and sex in the reference population, and the standard deviation of the distribution of weights
in the reference population for children of the same height and sex.
( )
For example, consider a male child who measures 84 cm and weighs 9.9kg. As shown on the
table in annex 6, the reference median weight for boys of height 84 cm is 11.3kg. The
standard deviation from the reference distribution for boys of height 84 cm is 0.908 kg. This
child's WHZ = (9.9 – 11.3) / 0.908 = -1.54 z-score.
These calculations should all be done by computer using ENA software, but it is useful to
understand the basis for the calculation.
6.3 Assessing data quality
The data collected during the survey should be of good quality if meaningful conclusions are
to be reached and appropriate programmatic decisions are to be made. The data quality
should be of high priority throughout survey planning and implementation.
There are several data quality checks that are automated in ENA and reported in the data
quality check report. The plausibility check ensures that the quality of the data is sufficient to
be used for planning interventions.
Box 6.7: Using ENA to generate plausibility report
To obtain the plausibility report, click on the "Report Plausibility check" button in the Data
Entry Anthropometry screen. Alternatively, you can also select ‘Plausibility check’ under
the Extras menu to generate this report.
6.3.1 Outliers (flags)
In ENA, there are two methods of identifying results from children that are unlikely to be
correct measurements. If flagged values or indices cannot be corrected, they should be
excluded from the analysis, but never removed from the survey dataset.
80
6.3.1.1 Exclusion of z-scores in Variable View (WHO flags)
In the data entry screen, WHO flags (coloured pink) are by default the following resulting z-
score in relation to the reference mean:
WHZ <-5 SD or >+5 SD
HAZ <-6 SD or >+6 SD
WAZ <-6 SD or >+5 SD
Those values are defined in the Variable View of the Data Entry Anthropometry screen
and excluded automatically in all the calculations by selecting the "Zero (reference mean)
WHO flags" in the Options screen. This is the same "flagging system" that the one used in
Epi-Info software. These flags identify the values that are absolute abnormalities for
individuals as the data are being entered. The absolute values have been chosen because they
are so extremely abnormal that they are very unlikely to be correct – indeed some are hardly
compatible with life.
The purpose of these flags is to enable correction in data entry or re-measurement while still
in the field. If there has been a recording error, it should be corrected. If it is a measurement
error, the team should go back to the household to correct the measurement or the estimation
of age.
Any uncorrected values should definitely NOT be used for the analysis – and the number of
exclusions recorded in the report (they should not be simply eliminated from the data-base).
6.3.1.2 Exclusion of z-scores in Data Quality check (SMART flags)
In the plausibility report, the program will list and query any value that is (by default) ± 3
standard deviations of the observed mean. The default values can be changed in the Options
screen and excluded automatically in all the calculations by selecting the "Observed mean
(SMART flags)".
The purpose of these flags is to allow additional data cleaning before running analysis and to
exclude from analysis cases that are not plausible (i.e., cases that probably resulted from data
collection errors). The computer examines the data to see if there are values outside the
expected range to exclude from the analysis data that are "more likely to be errors than real
values".
For the final analysis it is recommended that the exclusions based upon the Plausibility Check
should be used and the numbers excluded reported. The values should not be removed from
the survey dataset. This "cleaning" is done automatically during the analysis.
6.3.2 Distribution
Most children with wrongly measured data give values that are within the plausible range.
Inclusion of such errors can be suspected from examination of the distribution of the data.
6.3.2.1 Mean
The mean value is robust to random measurement errors. It is affected by non-random errors.
Note that even a small systematic error can have a surprising effect.
81
If children are routinely weighted in underclothes that weigh about 30 grams, even though
this is less than the precision of the scale, it will make a difference. This is because it
changes the rounding of the numbers being read on the scale – more of the children will be
rounded up to a higher figure than the lower figure.
There are other systematic errors that cannot be discerned from examination of the data
unless there are marked differences between the teams (e.g., one team has a scale that has not
been properly zeroed – see below).
6.3.2.2 Standard Deviation
The standard deviation of the z-scores for WFH and HFA should be examined. As explained
in the section on outliers, this tells if there is substantial random error in the measurements. If
the standard deviation is high (over 1.2), it is likely that there are a lot of extreme values and
values more than ±3 z-scores of the mean.
The SD of WFH should not exceed 1.2 in a good survey (there is no lower limit, although it
is generally above 0.8). When the SD is below 1.2, then the data can be analysed the
conventional way and the counted prevalence should be reported. Otherwise, the calculated
prevalence should be reported along with the counted prevalence.
6.3.3 Sex ratio
The sex ratio should be approximately 1.0, i.e., 50% male and 50% female. If it deviates
markedly from 0.9 – 1.1, either overall or within age groups, then there has either been a
sampling problem or there is a social problem.
For example, when house-to-house visits are not made, but the mothers asked to bring their
children to a centre for measurement (this is a short-cut unsupervised teams sometimes
employ), then the mothers can bring their boys rather than their girls (or vice versa) and this
can show as an abnormal sex ratio. If there is an excess of girls this is often due to failure of
the teams to warn the ward of their arrival or failure to get the children outside the household
to come for measurement.
As another example, if the sex ratio is near equality in the younger age groups but deviates in
the older age groups then this is sometimes due to the 4 and 5 years old children being
occupied in the fields or outside the house.
6.3.4 Age distribution
The distribution of age should be examined for any age heaping. Age heaping typically
occurs at the ages of 12, 24, 36, and 48 months as mothers and/or surveyors usually tend to
round the age to the nearest year. Age heaping may be of concern especially if age is
primarily estimated using events calendar.
82
Surveyors should be adequately trained in the use of events calendar prior to the data
collection. As part of the data quality check, the age distribution should be checked on a daily
basis using the plausibility check option in ENA and enumerators need to be retrained on
collecting age information if there is any age heaping.
6.3.5 Digit preference for height and weight
A common mistake when recording measurements is to round them to the nearest round
number – usually a zero or five as the last figure with weight and height and 12, 24, 36, 48
for age in months.
A large digit preference can have a large effect upon the result – but minor degrees of
"rounding" will not affect the result even though it might be statistically significant.
6.3.6 Skewness
Skewness is a measure of the direction and degree of any asymmetry that there is in the data.
A value of zero means that the distribution is symmetrical distribution. A positive value
indicates skewness (long tail) to the right. A negative value indicates skewness to the left, as
shown below.
Example of skewness in distribution of survey data
The values of the statistic increase as the distribution becomes more and more skewed. There
is no general agreement at what level one would declare the results to be sufficiently skewed
to cause a problem in analysis. It is however recommended that values below ± 1.0 always be
accepted as normal, that up to ±3.0 the data are probably not sufficiently skewed to cause
concern. As the vales increase above ±3.0 the data can be said to be skewed.
Skewed data are not necessarily due to poor quality of data collection. Skewness can be
generated if there are subgroups within the population that are sufficiently different from the
rest of the population to form a distinct subgroup. When the populations are almost equal in
size and markedly different this can result in a bimodal distribution (a curve that has two
peaks and a wide SD). If the subgroup is smaller in size than the main population and
sufficiently similar so that a bimodal distribution is not generated, then the distribution is
likely to be skewed (depending upon whether the subgroup is better or worse than the general
population.)
83
If the data are greatly skewed then great care needs to be taken with interpretation. It is likely
that there are distinct subgroups within the population that should have been identified and
surveyed separately during the planning phase of the survey.
6.3.7 Kurtosis
Kurtosis is a measure of the "peakedness" of a distribution. When there is a Kurtosis problem,
although the distribution is symmetrical and can appear like a bell shaped curve – it is still
not normal. This statistic measures whether there are too many values in the tails and not
enough in the shoulders of the distribution or conversely whether the shoulders of the curve
have too many children and the tails are missing, as shown below.
A Normal Distribution with a bell-shaped curve has kurtosis of zero. A positive kurtosis
indicates a sharper peak with longer/fatter tails – like a Mexican hat - and relatively more
variability due to extreme deviations. A negative kurtosis
coefficient indicates broader
rounded shoulders with shorter/thinner tails – like a pudding.
A positive kurtosis is often generated by large numbers of outlying values – this can occur
from errors in reading the scales or measuring board, recording the measurement, transferring
the recordings or entering them into the computer. If there are large numbers of flags it is
likely that there will be a high positive Kurtosis. Frequently, the kurtosis falls as one cleans
the data from the raw values to the Epi-Info style flags to the plausibility check flags. If, after
applying the plausibility check cut-offs, the kurtosis remains high then the data are flawed.
A negative kurtosis is less common. It usually indicates that the data have been "over-
cleaned" or that the teams have not taken values that they themselves think might be extreme
– so that there are far too many values clustered around the mean value.
Examples of kurtosis in distribution of data
Again there is no general agreement about the values where one would reject survey data.
The same authorities give the same cut-off values for Kurtosis as they do for skewness. Thus
<±1.0 is always acceptable. From ±1.0 to ±3.0 the data are probably normal but should be
taken with some caution. Above 3.0 the data are not normally distributed.
84
6.3.8 Analysis by team
The problems with measurement usually do not involve all the teams. Often it is due to one
poorly trained team or team member that can affect the overall results of the survey. Even if
the overall analysis is acceptable, there may be an aberrant values arising from a particular
team – this may go unobserved if there are a large number of teams so that the contribution of
an individual team to the overall result is diluted.
All the tests that are applied to the overall results are also applied to each individual team.
If any particular team has obtained data that is very different from the other teams, it is likely
that this team’s technique has created a systematic bias. If there is time, the aberrant team’s
clusters should be re-sampled using a different team and the new data substituted for the
aberrant data. If re-sampling is not feasible within a reasonable time, the data should be
analysed with and without the aberrant clusters, and both results reported with a
recommendation from the survey manager indicating which result is likely to be more
reliable. There has to be a full report of such occurrences and how they are resolved (e.g.,
perhaps the team’s equipment is faulty or their training has been inadequate).
6.3.9 Overall data quality
It is very difficult for a non-mathematician without electronic equipment to fabricate data that
forms a normal distribution without Skewness, Kurtosis, an acceptable SD, and without digit
preference. If these values are all within acceptable limits it can be assumed that the data
have been well taken and entered into the computer, and the analysis acceptable.
Based on the statistics that are generated from the data, ENA generates an overall score for
the survey data, which is summarised in the findings of the plausibility check report. The
overall quality is assessed for 9 categories such as: 1) missing/flagged data, 2) sex ratio, 3)
age distribution, 4) digit preference for height, 5) digit preference for weight, 6) standard
deviation (WHZ), 7) skewness, 8) kurtosis (WHZ), and 9) Poisson distribution (WHZ<-2).
While the sex ratio and age distribution look at the selection bias, the rest of the tests look at
the measurement bias.
A score is generated for each category of test based on pre-set criteria and an overall score is
calculated. Based on the overall score, the survey data is classified as excellent, good,
acceptable, and problematic. Some categories of tests are considered to be more important
than the others. For example, missing/flagged data has the highest penalty points.
The plausibility check report should be included in the final survey report as an annex.
85
7. DATA ANALYSIS
7.1 Data Analysis: Nutrition Survey Data
Once indices have been calculated and data checked for quality, the analysis of the data can
be conducted.
7.1.1 Classification of malnutrition
7.1.1.1 Definitions of acute malnutrition
WFH and MUAC are the two indicators used to assess moderate and severe wasting, monitor
changes in the nutrition status of the population, and make decisions on admission and
discharge of individuals to and from feeding programs.
Oedema
Pitting oedema on both feet (bilateral oedema) is the sign of kwashiorkor. In an emergency
context, any person with bilateral oedema has severe malnutrition14
and is classified as
severely malnourished even if the WFH z-score is normal. ENA does not provide WHZ and
WAZ for children with oedema.
Moderate, severe, and global acute malnutrition
The classification of acute malnutrition is traditionally based on the presence/absence of
oedema and on WFH index, as detailed in Table 7.1.
Table 7.1: Classification of acute malnutrition based on WFH index and oedema
Category Degree of malnutrition Definition using
z-score
Acute Malnutrition Moderate - 3.0 and <-2.0
Severe <-3.0 or oedema
Global Acute Malnutrition (GAM) Moderate and Severe <-2.0 and/or oedema
Severe Acute Malnutrition (SAM) Severe <-3.0 and/or oedema
Acute malnutrition can also be descried based on MUAC and oedema, as described in table
7.2
Table 7.2: Classification of acute malnutrition based on MUAC and oedema
Category Degree of malnutrition Definition using
MUAC
Acute Malnutrition Moderate 115 mm and <125 mm
Severe <115 mm or oedema
Global Acute Malnutrition (GAM) Moderate and Severe <125 mm and/or oedema
Severe Acute Malnutrition (SAM) Severe <115mm and/or oedema
14
There are other causes of bilateral oedema such as heart failure, kidney disease (nephrotic
syndrome), thiamine deficiency, and pre-eclampsia in pregnant women. However, in an emergency
context, most bilateral oedema, especially in children, is due to kwashiorkor.
86
Severe Acute Malnutrition (SAM) is the term used to include all children with severe wasting
or children who have oedema.
Global acute malnutrition (GAM) is the term used to include all children with moderate
wasting, severe wasting or oedema, or any combination of these conditions.
Note that the terms "severe wasting" and "severe acute malnutrition" are not synonyms. A
child with severe acute malnutrition is either severely wasted, oedematous, or both. Severe
acute malnutrition is the sum of severe wasting and oedema.
The user of this manual will not have to make these calculations: they are done automatically
using ENA software. GAM and SAM should be presented as prevalence expressed as a
percentage of the population.
7.1.1.2 Definition of chronic malnutrition
The long time scale over which HFA changes makes it less useful for deciding when to
intervene in an emergency. It is useful, however, for long-term planning and policy
development. Although at an individual level stunting develops slowly, the degree of stunting
can change within a few months when averaged over an entire population. Table 7.3 details
the cut-offs used to classify chronic malnutrition, based on z-score.
Incorrect age data makes HFA information misleading, and reliable age data can be difficult
and time-consuming to obtain.
Table 7.3: Classification of chronic malnutrition
Category Height-for-age z-scores
Severe stunting <-3 Z scores
Moderate stunting - 3.0 and <-2.0
Total stunting (moderate + severe) <-2 Z score
7.1.1.3 Definition of underweight
Although underweight is not widely used in nutrition assessment for intervention it is useful
to report. Table 7.4 details the cut-offs used to classify underweight, based on z-score and %
of median.
Table 7.4: Classification of underweight
Category Weight-for-age z-scores
Severe Underweight <-3 Z scores
Moderate Underweight - 3.0 and <-2.0
Total Underweight <-2 Z score
87
7.1.2 Nutrition survey results
ENA should be used in the analysis of nutrition survey data. ENA automatically calculates
nutrition indices of each individual and nutrition status of the surveyed population, along with
95% confidence interval. The analysis should be done based on the WHO standards.
Box 7.1: Using ENA to analyse nutrition survey data
Nutrition survey data can be analysed in ENA (in the Results Anthropometry screen) based
on nutrition indices, MUAC cut-offs, theoretical distributions, and different exclusion
criteria. The results are displayed in diagrams and graphs (which can be copied onto the
clipboard and pasted onto reports and other documents) as well as actual numbers. Nutrition
indices and MUAC data can be further disaggregated into various categories such as sex,
cluster, age, etc. The results can also be obtained based on WHO standards or NCHS
reference by selecting the appropriate options. The results anthropometry also helps generate
a sample report.
[Note that parameters should be set accordingly in the Options screen before any analysis is
carried out in the Results Anthropometry screen.]
Go to Options screen
Click on the “Reset” button
Select age between 6-59.99
Select observed mean SMART flags
Go to the Results Anthropometry screen
Select WHO standards 2006 (no. 1)
Select SMART flags (No. 4)
Select the nutrition index to be analysed (e.g. Weight/Height, MUAC, etc.) (No. 5)
Look at the results for the selected index by
o Selecting the appropriate categories such as All, Sex, Cluster, and Age (No. 6)
o Selecting the type of graph from the drop down menu (if cluster is selected above,
distribution of cases can also be analysed) (No. 3)
Generate sample survey report (No. 2)
Copy graph to clipboard and use it in report as appropriate (No. 2)
1 2
3 4
8 7
6 5
88
Data can be disaggregated by age groups. The age group can be set in the "Option" screen in
ENA. For analysis by other variables, the filter function on the "Data Entry Anthropometry"
screen should be used.
The results from all clusters should be combined to give an estimate for the whole population
from which the sample was taken. The results from each cluster should NOT be used as an
estimate of the prevalence of malnutrition in those individual locations because the sample
size in each location is too small to be representative of that location.
7.2 Data Analysis: Death Rate Survey Data
ENA should be used to analyse death rate survey data and calculate death rates.
7.2.1 Calculating death rates
ENA automatically calculates the crude as well as 0-5 death rate and provides the rates with
95% confidence interval. Sex and other age specific death rates can be obtained and causes
and location of death can be determined if data are entered into Data Entry Individual level
panel of the Death Rates screen (see section 6.1.2 for details).
It should be noted that the Panning screen and the Death Rates screen are linked in ENA. In
order for ENA to calculate the death rates correctly, either the same ENA file used for the
planning of the particular death rate survey (i.e. sample size calculation) should be used for
data entry or planning screen should be set accordingly if a new ENA file is used to enter
death survey data. This is because the
The formula used by ENA to calculate the crude death rate as follows:
(
)
7.2.1.1 Total population (the population at risk)
The "total population" is estimated by assuming that those who were not present in the
household for the whole of the recall period (those who left and those who joined the
household, those who died, and those were born) were present on average for half the recall
period15
. Thus, the "Total Population" ENA uses is the sum of:
+ all those present in the household at the time of the survey
+ half the deaths
+ half those present at the beginning of the recall period but who had left by the
time of the survey
− half those present at the time of the survey but who joined the household during
the recall period.
− half the births
15
Note that the denominator is actually person x days. It is mathematically equivalent to count half a
person as it is to count half the recall period for that person.
89
The 0-5 population uses the same formula, but for children under five only. Similarly, for sex
specific and other age group specific death rates, the respective population will be used.
Infants that were born and died within the recall period should be counted as deaths (in the
numerator) but should not be included in the total population (denominator)16
.
Similarly, a person who entered the household and subsequently died during the recall period
should be counted as deaths (in the numerator) but should not be included in the calculation
of the total population.
7.2.1.2 Number of recall days
To calculate the exact number of days comprising the recall period, please refer to section
3.8.1.
7.2.2 Using ENA to analyse death rate survey data
As described in section 6.1.2, death rate survey data can be entered and analysed by
individual or by household. The sections below describe both individual level data analysis as
well as household level data analysis of death rate survey data.
Box 7.2: Using ENA to analyse death rate survey data – individual level
Death rate survey data entered in the Data Entry Individual level panel under the Death
Rates screen is automatically analysed and displayed in Results Individual Level panel. The
individual level data analysis in ENA not only generates the crude and 0-5 death rates but
also helps analyse death rates by sex and other age groups. Additionally, cause and location
of death can also be analysed and population pyramid can also be obtained when using
Results Individual Level panel.
16
If a birth and death were entered for the same person, the two events would cancel each other out in
terms of contributions of “person-time” of exposure.
90
1: Age groups can be defined in this section; flexible as well as fixed intervals (5 or 10 year
intervals) used can be
2: This section provides a summary of death rate survey data by age
3: This section provides a summary of death rate survey data by cluster
4: Crude death rate as well as death rates by sex and age groups (as specified in box 1) are
displayed in this section.
5: Cause of death and location of death are gives as percentage in this section.
6: Population pyramid can be obtained using the Pop. pyramid button and the entire survey
data can be transferred to Microsoft Excel using the Transfer tables to Excel button in this
section
1
3
2
4
5
6
91
Box3 7.3: Using ENA to analyse death rate survey data – household level
Death rate survey data entered in the Data Entry Household level panel (i.e. the summary
information on the death rate form) under the Death Rates screen is automatically analysed
and displayed in Results household Level panel.
1: The entire death rate survey data is summarised and given by cluster in this section.
2: The crude and 0-5 death rates are displayed in this section along with their 95% confidence
interval and design effects.
7.3 Data analysis: Other data
Other data such as health data, WASH data, etc. can be analysed using the Statistical
calculator in ENA. Most of the data recorded will need basic descriptive analysis.
Categorical data should be presented as proportions in frequency tables (number of subject
presented the condition / total population). Continuous data should be presented with a mean,
a standard deviation, and a range.
Statistical calculator in ENA can also be used to cross tabulate variables and calculate 95%
confidence intervals and design effect. However, statistics such as p-values cannot be
calculated. See box 7.4 for details on how to use statistical Calculator in ENA.
1 2
92
Box 7.4: Using Statistical calculator in ENA
The Statistical calculator can be accessed from the Extras menu or from the last shortcut
button on the Data Entry Anthropometry screen.
1: The variable to be analysed (e.g. MUAC of mother) is specified in this section and the type
of variable, i.e. either case or cont, should also be specified (note case refers to categorical
variables and cont refers to continuous variable). The variable can be further disaggregated
using the Broken down option and specified in the boxes, form (low) and to (high) – e.g.
(mother MUAC between 175 and 209 – note: since the range includes border values, for
MUAC <210 mm, 209 should be entered).
2: The variable selected above (1) can be cross tabulated using the Crosstable with. E.g. if
you want to disaggregate BCG vaccination status between girls and boys, you need to select
BCG under Variable (1) and sex under Crosstable with. Multiple variables can be analysed
using Crosstable with at the same time.
3: This section allows calculating prevalence estimates for a single variable. E.g. if you want
to calculate reported illness among children in a survey, it can be obtained by simply
selecting the variable (e.g. illness) and clicking on calculate. 95% confidence intervals and
design effect can also be calculated using this.
4: These buttons allows you to calculate the results, transfer the results to Microsoft Excel,
and reset the calculator. Note that results are better viewed exporting results to Excel and
adjusting columns; it is a good practise to reset the calculator after each analysis.
7.4 Data Analysis: Qualitative data
The data from focus group discussions, key informant questionnaire, and direct observation
should first be reviewed and summarised removing all the non-essential information. The
main themes arising from the raw data should then be identified and summarised.
It is highly recommended that this summary information, along with information from other
sources, be discussed with the survey team members to confirm that various people got the
1
2 3
4
93
same impression about the situation and explore reasons if they are different. The key
findings from the qualitative data are then finalised and used to triangulate the survey results.
94
8. INTERPRETATION OF RESULTS
8.1 Interpreting the results
Once data are analysed, the survey results should be put in context to explain the findings and
make recommendations for interventions. In order to fulfil these challenges, the following
questions need to be answered:
How critical are the level of malnutrition and death rate for the population in the
current season and within the context of the area?
How can the nutrition levels (and death rates, if death rate survey was also
conducted) be explained?
The interpretation of the results is probably the most difficult part of a nutrition survey
because there is no standard method for interpreting nutrition data, and there are many
different factors to consider at the same time. However, a proper interpretation of the results
is crucial in order to design the right intervention. The following sections include information
on areas that should be considered when interpreting a nutrition survey results with
established thresholds.
8.1.1 Comparing the results with establish thresholds
8.1.1.1 Thresholds for malnutrition levels
A severity index for malnutrition based on prevalence of wasting among children less than 5
years of age is generally used to classify malnutrition (see table 8.1). For stunting, a WHO
classification of worldwide prevalence ranges of low height-for-age among children under 5
years of age is also available (see table 8.2). These thresholds help interpret the seriousness of
a situation.
Table 8.1: Severity index for malnutrition in emergency situations based on prevalence of
wasting for children less than 5 years of age
Classification of severity Prevalence of wasting
(% of children below –2 Z-scores)
Acceptable <5
Poor 5-9
Serious 10-14
Critical 15
Table 8.2: WHO classification of worldwide prevalence ranges of low height-for-age among
children under 5 years of age
Classification of severity Prevalence ranges (% of children below –2
Z-scores)
Low <20
Medium 20-29
High 30-39
Very high 40
95
Table 8.3: WHO classification of worldwide prevalence ranges of low weight-for-age among
children under 5 years of age
Classification of severity Prevalence ranges (% of children below –2
Z-scores)
Low <10
Medium 10-20
High 20-30
Very high 30
Although this classification was made for NCHS references, it should also be used with
WHO standards during the transition phase until thresholds relevant to WHO standards are
adopted.
Additionally, the malnutrition levels obtained from the survey can also be compared with the
CDC one survey calculator17
and probability of exceeding a certain threshold can be
obtained. This can be used as an additional piece of information to assess the severity of the
situation.
It should be noted that a threshold based classification is a supportive tool: it should not be
strictly followed as a set of rules. How a situation will be classified varies greatly according
to the context and must be adapted accordingly. The co-existing aggravating factors should
also be taken into consideration when interpreting the results using a threshold based
classification. For example, the interpretation of malnutrition prevalence of 10% will vary
depending on whether or not there is a measles outbreak in the survey area at the time of the
survey.
Potential aggravating factors include:
Poor household food availability and accessibility (due to a poor harvest, poor
pasture conditions, high market prices, insecurity, or inadequate general distribution
in a camp setting, etc.).
Epidemics of measles, cholera, or other communicable diseases; high level of
malaria.
Low levels of measles vaccination and vitamin A supplementation.
Inadequate shelter and severe cold.
Inadequate safe water supplies (quality and quantity) and sanitation.
Gender inequalities manifested in specific community social norms around feeding
practices that are different for girls and for boys
Consideration of aggravating factors is an absolutely essential part of a good interpretation of
anthropometric data. If more than one aggravating factor is present then the situation may be
worse than if there is just one.
17
CDC calculator and instructions as to how to use the calculator can be downloaded from this link:
http://www.cdc.gov/globalhealth/gdder/ierh/ResearchandSurvey/calculators.htm
96
Note the secondary data review that is done prior to the survey and the qualitative
information collected during the survey through focus group discussions and key informant
interviews would provide information on most of the aggravating factors such as food
security, epidemics of diseases, vaccination levels, etc. in the survey area.
8.1.1.2 Thresholds for death rate levels
In the case of death rates, the following thresholds should be used (see table 8.3), and level of
death rates should be analysed along with anthropometric results.
Table 8.4: Thresholds for death rate survey results
Agency Emergency threshold for
CDR
Emergency threshold for 0-
5DR
WHO (WHO 2005) ≥2/10,000/day ≥4/10,000/day
Sphere (Sphere 2011) Sub-Saharan Africa: 0.8 Sub-Saharan Africa: 2.1
8.1.2 Comparing results with previous survey results
Wherever possible, the survey results should be compared to the previous survey(s)
conducted in the survey area and findings should be compared and discussed. As malnutrition
prevalence is affected by seasonality, only surveys conducted in the same season should be
compared. If results are different, two surveys conducted in the same season can also be
compared and statistical significance can be tested using the CDC two survey calculator18
.
If multiple surveys have been conducted in the area in the past, the malnutrition trend over
time should be included in the report and discussed. For example, if the current malnutrition
levels are higher than the previous surveys, factors that might have contributed to this
difference should be explored, presented, and discussed.
8.1.3 Analysing the context
In order to be able to interpret correctly the malnutrition levels from a survey, it is necessary
to consider the following factors:
Determine seasonal variations.
Compare the results to previous surveys in the same area or livelihood zone at the
same time of year.
Using the livelihood information obtained from the food security cluster, explore
how the levels of malnutrition can be explained by a change in the livelihood of the
population compared to the baseline livelihood zone profile.
Interpret all results in their cultural, socio-economic and agro-ecological context,
together with other supporting data such as indicators on health, food supply,
markets, etc. Especial attention should be paid on equity, empowerment, and gender
equality.
Analyse death rates in the survey area.
Look at what intervention are already being implemented in the survey area.
18
CDC calculator and instructions as to how to use the calculator can be downloaded from this link:
http://www.cdc.gov/globalhealth/gdder/ierh/ResearchandSurvey/calculators.htm
97
8.1.4 Using UNICEF conceptual framework
It is highly recommended that the UNICEF conceptual framework is used in the
interpretation of survey results. Both secondary information as well as information collected
during the survey can be summarised under immediate as well as underlying causes of
malnutrition and appropriate inferences can be made.
Example 8.1: Use of UNICEF conceptual framework in interpreting a survey results
A nutrition and death rate survey showed the following results:
Indicator N Results with 95% CI
GAM 866 13.5% [11.2 – 16.2]
SAM 866 1.3% [0.7 - 2.4]
MUAC <125mm 885 6.6% [5.1-8.4]
Crude Death Rate 0.53 [0.11-2.59]
0-5 Death rate 0.23 [0.11-0.45]
It is useful to apply the UNICEF conceptual framework when explaining these single
estimates. Some of the questions that may be asked in interpreting the results include the
following:
Immediate causes
What is the health status of the population surveyed? What is the status of common childhood
illnesses (ARI, diarrhoea, Measles, etc.) in the survey area? Has there been any disease
outbreak?
Underlying causes
How is the food security situation in the survey area?
What are the feeding practices? Exclusive breastfeeding? Feeding of young children?
What is the status of WASH? General hygiene practices in general? Use of safe drinking
water?
How are the health systems functioning? Access to health services? Utilisation of health
facilities?
Note that most of this information would be collected as secondary data review and used here
to explain the malnutrition prevalence obtained from the survey. Some information such as
WASH, health may also be collected during the survey.
8.2 Presenting the results, writing the report
The nutrition survey report should provide an accurate account of the nutrition situation in a
given area for intervention planning, decision-making, and advocacy. It should contain all the
information that allows the reader to understand why the survey was conducted, the methods
used and decisions made, the population to which the results apply, the results themselves,
and a summary of problems encountered.
98
A preliminary survey report should be prepared within 2 weeks of completing the data
collection. The results should be discussed with the district/county level health authorities in
the field before the preliminary survey report is submitted to the National Nutrition Forum.
The survey report should include the following information (see annex 2 and 13 for the
sample preliminary and final survey report format):
Main topics Information to be included
Introduction Geographic description of survey area
Description of the population
Justification to conduct the survey
Survey Objectives
Methodology Sample size
Sampling methodology
Sampling procedure: selecting clusters
Sampling procedure: selecting households and children
Case definitions
Questionnaire
Survey teams and supervision
Training
Data analysis
Results Anthropometric results
Death rate survey results
Children’s morbidity
Vaccination coverage
WASH results
Food security results
Other results
Discussion
(Critical Analysis)
Nutritional status
Death rates
IYCF results
Possible causes of malnutrition
Conclusions and
recommendation
Future nutrition monitoring
8.3 Making recommendations
Clear, specific, and time-bound (immediate, medium and long-term) recommendations
should be made from the survey results and included in the survey report. The
recommendations should always be linked to the survey findings and should not be made
arbitrarily. These proposed recommendations should be discussed with the district/county
health authorities and other stakeholders and finalised.
The individual sectorial plans should be reviewed before recommendations are made so that
recommendations can be linked to sector priorities.
In areas where repeated surveys are carried out, the recommendations made from the last
survey should be revisited and actions that have been taken since the previous survey
99
recommendations were made should be reviewed before new recommendations are made.
This should be part of the dissemination meeting that is organised to discuss the survey
findings with the district/county authorities and other stakeholders.
8.4 Planning the response
The selection of appropriate nutrition interventions and strategies largely depends on the
context. Consequently, a fixed intervention blueprint does not exist. However, it is important
that there are relatively equal responses to nutrition emergencies in all parts of the country,
i.e., nutrition interventions must be fair.
To choose the right intervention, the following should be considered:
The prevalence of global and severe acute malnutrition, death rates, MIYCN
(Maternal, Infant and Young Child Nutrition) coping mechanisms, seasonality and
other aggravating factors.
The analysis of the context and the interpretation of the situation. An informed
decision can be made about what interventions to prioritise.
The population’s future needs, including immediate food prospects, potential
disease outbreaks and potential changes in caring practices.
What other on-going interventions already exist.
What resources are available and what constraints exist.
100
Annex 1: Survey proposal format
1. Background information
Geographical area to be covered in the survey
Background information on survey population, health and food security situation
Nutrition status of the area – from past nutrition survey(s)
2. Survey justification
Rationale for conducting a survey
Objectives of the survey
Timing of the survey (including seasonal calendar)
Area to be surveyed (including discussion of homogeneity and Livelihood zone).
3. Survey methodology
Calculation of sample size for nutrition and death rate surveys (including rationale for
the selected prevalence, precision, design effect (if cluster survey), household size, %
of under 5 population, and recall period (if mortality)
Description of sampling frame (including source of population data)
Description of sampling methods (including number of clusters and number of
households per clusters for cluster surveys)
List of selected clusters (for cluster surveys)
Information of household selection methods
Data to be collect, and data collection methods and tools
Data analysis plan
4. Organization of the survey
Description of main activities and timeframe
101
Annex 2: Preliminary survey report format
Implementing agency:
Survey date:
Survey area:
Survey objectives:
Methodology:
Sample size (for anthropometry, mortality, and IYCF as relevant)
Household selection
General comments on the survey (i.e. survey population, survey limitations, etc.)
Survey results
Anthropometric results (based on WHO standards 2006):
Death rate survey results (retrospective over x months/days prior to interview)
IYCF results
Children’s morbidity
Vaccination coverage
WASH results
Food security results
Other results
Conclusions
Recommendations (both short term and long term)
102
Annex 3: Random number table
67594 63100 37579 30635 41209 73080 82555 78577 74647 81058 87062 37659 30145 75645 28051 37618 28754 71462 65290 94121 50440 83974 19419 98412 79181 39377 71243 73176 49173 39997 76624 46346 40733 78182 88592 87066 26995 24143 88447 80534 24984 15722 16463 10934 87176 50553 44567 14192 42128 33584 65823 24755 85272 25425 98057 33131 13468 99502 81493 13394 49417 48474 92008 42379 14513 12884 39783 74789 21243 67523 85976 30926 28714 63460 11157 66265 37420 56220 67564 14598 21817 53066 42114 78958 71826 84874 43611 97049 66842 10542 33704 40385 28342 14425 36525 18886 69695 79758 87665 65117 54264 73528 90426 84913 85389 30772 39183 23594 94351 68772 12224 49502 54907 14103 78879 39059 35493 18019 18316 10090 42681 38133 29820 22610 82000 46868 36912 68800 74694 59638 70157 46392 11525 88244 95984 22185 27213 57436 21388 24900 11602 15118 86837 69104 16146 89168 82240 54415 36817 26337 73313 16712 27019 61197 38188 60561 26602 25601 66613 44585 45584 21639 96583 13990 83650 63542 75745 56966 59049 76512 17421 84190 72959 42946 73599 53134 17933 19016 49726 11418 81501 37089 58650 75902 28545 21933 73563 36761 28514 51204 32275 98238 56094 53157 97674 60316 46420 37070 71709 28009 38415 84342 42741 87501 12368 70727 48613 10854 50325 12685 70270 73489 30403 63314 73281 41181 68607 15825 17107 65764 64258 85039 44456 51285 57610 79869 95569 14808 70770 42261 14784 28598 15486 50549 69212 62905 93928 57713 21888 71056 71038 27493 54214 51081 49537 23836 15066 20598 91207 21635 66385 44157 10511 39247 57615 24785 59174 45735 23810 73934 92793 18327 84782 46550 53092 98036 25104 90503 31897 93937 27337 32064 95440 39040 93303 31679 70074 13257 30770 16735 53004 81409 15373 10555 94110 46752 50121 79328 37483 92994 87348 81194 83738 80261 80424 54213 46721 52990 65094 32427 70686 40212 32782 81734 16557 41205 10691 19796 58341 31961 66068 16705 13312 30471 64448 55608 82045 11259 24249 35034 26892 22168 42539 67119 50010 68840 49335 98465 84515 74875 30265 72841 84865 68135 29950 77451 55072 46150 80938 26982 15821 76116 45537 82153 44105 37430 10398 51995 91463 57255 96179 11555 29411 12059 97146 69271 33170 90619 18046 30715 87275 26442 78105 87941 29160 30082 34475 86135 39324 84320 40009 71812 69153 47666 52664 79254 50008 64174 56414 14426 49667 80006 45997 68075 14707 79751 10336 42244 19936 67936 81997 46906 78456 13718 52509 95952 28452 89211 85897 24233 35307 24437 75275 89896 20133 24342 59838 38715 38307 45872 92095 99644 54118 15560 37696 23309 10103 80608 82686 37662 32181 98910 45532 57509 94170 26013 23780 77132 17778 89462 67661 17726 76673 23509 15515 23875 65713 79652 18358 65774 28942 70975 53445 66421 12431 20749 79176 85501 10578 68278 76175 24182 36936 97441 51901 47529 93186 25920 18625 63769 12334 95554 67121 42125 74729 76821 50914 93420 78001 12887 38428 70200 54508 21216 12876 85562 92379 23183 57384 67594 11525 88244 95984 22185 27213 57436 21388 24900 11602 15118 86837 69104 16146 89168 82240 54415 36817 26337 73313 16712 27019 61197 38188 60561 26602 25601 66613 44585 45584 21639 96583 13990 83650 63542 75745 56966 59049 76512 17421 84190 72959 42946 73599 53134 17933 19016 49726 11418 81501 37089 58650 75902 28545 21933 73563 36761 28514 51204 32275 98238 56094 53157 97674 60316 46420 37070 71709 28009 38415 84342 42741 87501 12368 70727 48613 10854 50325 12685 70270 73489 30403 63314 73281 41181 68607 15825 17107 65764 64258 85039 44456 51285 57610 79869 95569 14808
Annex 4: Decision tree for selecting households at the last stage of cluster sampling
104
Annex 5: Local events calendar
MONTH Seasons 2007 2008 2009 2010 2011 2012
January
Paush-Magh
Cold
season
58
Yomari
Purnima
46 Yomari
Purnima
34 Yomari
Purnima
22 Yomari
Purnima
10 Yomari
Purnima
FEBRUARY
Magh-Falgun
Cold
season
57
45 33 21 9
MARCH
Falgun-
Chaitra
Dry
season
56 Holi 44Holi 32 Holi 20 Holi 8 Holi
APRIL
Chaitra-
Baishak
Dry
season
55
43 31 19 7
MAY
Baishak-
Jesdha
Dry
season
54
42 30 18 6
JUNE
Jesdha Ashad
Rainy
season
53 41 29 17 5
JULY
Ashad-
Shrwan
Rainy
season
52 40 28 16 4
AUGUST
Shrwan-
Bhadra
Rainy
season 51
Gaijatra
39 Gaijatra 27 Gaijatra 15 Gaijatra 3 Gaijatra
SEPTEMBER
Shrwan
Dry
season
50 38 26 14 2
OCTOBER
Bhadra
Dry
season 49 Dashin
37 Dashin 25 Dashin 13 Dashin 1 Dashin
NOVEMBER
Ashbin
Cold
season
60 Diwali 48 Diwali 36 Diwali 24 Diwali 12 Diwali 0 Diwali
DECEMBER
Kartik
Cold
season
59 47 35 23 11
105
Annex 6: Weight-for-height z-score table, WHO 2006 Child Growth Standards
Weight-for-Length Look-Up Table, Children 6-23 Months, WHO 2006 Child Growth Standards
Boys' Weight (kg) Length a Girls' Weight (kg)
-3 SD -2 SD -1 SD Median (cm) Median -1 SD -2 SD -3 SD
1.9 2.0 2.2 2.4 45 2.5 2.3 2.1 1.9
2.0 2.2 2.4 2.6 46 2.6 2.4 2.2 2.0
2.1 2.3 2.5 2.8 47 2.8 2.6 2.4 2.2
2.3 2.5 2.7 2.9 48 3.0 2.7 2.5 2.3
2.4 2.6 2.9 3.1 49 3.2 2.9 2.6 2.4
2.6 2.8 3.0 3.3 50 3.4 3.1 2.8 2.6
2.7 3.0 3.2 3.5 51 3.6 3.3 3.0 2.8
2.9 3.2 3.5 3.8 52 3.8 3.5 3.2 2.9
3.1 3.4 3.7 4.0 53 4.0 3.7 3.4 3.1
3.3 3.6 3.9 4.3 54 4.3 3.9 3.6 3.3
3.6 3.8 4.2 4.5 55 4.5 4.2 3.8 3.5
3.8 4.1 4.4 4.8 56 4.8 4.4 4.0 3.7
4.0 4.3 4.7 5.1 57 5.1 4.6 4.3 3.9
4.3 4.6 5.0 5.4 58 5.4 4.9 4.5 4.1
4.5 4.8 5.3 5.7 59 5.6 5.1 4.7 4.3
4.7 5.1 5.5 6.0 60 5.9 5.4 4.9 4.5
4.9 5.3 5.8 6.3 61 6.1 5.6 5.1 4.7
5.1 5.6 6.0 6.5 62 6.4 5.8 5.3 4.9
5.3 5.8 6.2 6.8 63 6.6 6.0 5.5 5.1
5.5 6.0 6.5 7.0 64 6.9 6.3 5.7 5.3
5.7 6.2 6.7 7.3 65 7.1 6.5 5.9 5.5
5.9 6.4 6.9 7.5 66 7.3 6.7 6.1 5.6
6.1 6.6 7.1 7.7 67 7.5 6.9 6.3 5.8
6.3 6.8 7.3 8.0 68 7.7 7.1 6.5 6.0
6.5 7.0 7.6 8.2 69 8.0 7.3 6.7 6.1
6.6 7.2 7.8 8.4 70 8.2 7.5 6.9 6.3
6.8 7.4 8.0 8.6 71 8.4 7.7 7.0 6.5
7.0 7.6 8.2 8.9 72 8.6 7.8 7.2 6.6
7.2 7.7 8.4 9.1 73 8.8 8.0 7.4 6.8
7.3 7.9 8.6 9.3 74 9.0 8.2 7.5 6.9
7.5 8.1 8.8 9.5 75 9.1 8.4 7.7 7.1
7.6 8.3 8.9 9.7 76 9.3 8.5 7.8 7.2
7.8 8.4 9.1 9.9 77 9.5 8.7 8.0 7.4
7.9 8.6 9.3 10.1 78 9.7 8.9 8.2 7.5
8.1 8.7 9.5 10.3 79 9.9 9.1 8.3 7.7
8.2 8.9 9.6 10.4 80 10.1 9.2 8.5 7.8
8.4 9.1 9.8 10.6 81 10.3 9.4 8.7 8.0
8.5 9.2 10.0 10.8 82 10.5 9.6 8.8 8.1
8.7 9.4 10.2 11.0 83 10.7 9.8 9.0 8.3
8.9 9.6 10.4 11.3 84 11.0 10.1 9.2 8.5
9.1 9.8 10.6 11.5 85 11.2 10.3 9.4 8.7
9.3 10.0 10.8 11.7 86 11.5 10.5 9.7 8.9
9.5 10.2 11.1 12.0 87 11.7 10.7 9.9 9.1
9.7 10.5 11.3 12.2 88 12.0 11.0 10.1 9.3
9.9 10.7 11.5 12.5 89 12.2 11.2 10.3 9.5
10.1 10.9 11.8 12.7 90 12.5 11.4 10.5 9.7
10.3 11.1 12.0 13.0 91 12.7 11.7 10.7 9.9
10.5 11.3 12.2 13.2 92 13.0 11.9 10.9 10.1
10.7 11.5 12.4 13.4 93 13.2 12.1 11.1 10.2
10.8 11.7 12.6 13.7 94 13.5 12.3 11.3 10.4
11.0 11.9 12.8 13.9 95 13.7 12.6 11.5 10.6
11.2 12.1 13.1 14.1 96 14.0 12.8 11.7 10.8
11.4 12.3 13.3 14.4 97 14.2 13.0 12.0 11.0
11.6 12.5 13.5 14.6 98 14.5 13.3 12.2 11.2
11.8 12.7 13.7 14.9 99 14.8 13.5 12.4 11.4
12.0 12.9 14.0 15.2 100 15.0 13.7 12.6 11.6 a Length is measured for children under 2 years or less than 87 cm height. For children 2 years or older or 87 cm height or greater, height is
measured. Recumbent length is, on average, 0.7 cm greater than standing height; although the difference is of no importance to individual
children, a correction may be made by subtracting 0.7 cm from all lengths above 86.9 cm if standing height cannot be measured.
106
Weight-for-Height Look-Up Table, Children 24-59 Months, WHO 2006 Child Growth Standards
Boys' Weight (kg) Height a Girls' Weight (kg)
-3 SD -2 SD -1 SD Median (cm) Median -1 SD -2 SD -3 SD
5.9 6.3 6.9 7.4 65 7.2 6.6 6.1 5.6
6.1 6.5 7.1 7.7 66 7.5 6.8 6.3 5.8
6.2 6.7 7.3 7.9 67 7.7 7.0 6.4 5.9
6.4 6.9 7.5 8.1 68 7.9 7.2 6.6 6.1
6.6 7.1 7.7 8.4 69 8.1 7.4 6.8 6.3
6.8 7.3 7.9 8.6 70 8.3 7.6 7.0 6.4
6.9 7.5 8.1 8.8 71 8.5 7.8 7.1 6.6
7.1 7.7 8.3 9.0 72 8.7 8.0 7.3 6.7
7.3 7.9 8.5 9.2 73 8.9 8.1 7.5 6.9
7.4 8.0 8.7 9.4 74 9.1 8.3 7.6 7.0
7.6 8.2 8.9 9.6 75 9.3 8.5 7.8 7.2
7.7 8.4 9.1 9.8 76 9.5 8.7 8.0 7.3
7.9 8.5 9.2 10.0 77 9.6 8.8 8.1 7.5
8.0 8.7 9.4 10.2 78 9.8 9.0 8.3 7.6
8.2 8.8 9.6 10.4 79 10.0 9.2 8.4 7.8
8.3 9.0 9.7 10.6 80 10.2 9.4 8.6 7.9
8.5 9.2 9.9 10.8 81 10.4 9.6 8.8 8.1
8.7 9.3 10.1 11.0 82 10.7 9.8 9.0 8.3
8.8 9.5 10.3 11.2 83 10.9 10.0 9.2 8.5
9.0 9.7 10.5 11.4 84 11.1 10.2 9.4 8.6
9.2 10.0 10.8 11.7 85 11.4 10.4 9.6 8.8
9.4 10.2 11.0 11.9 86 11.6 10.7 9.8 9.0
9.6 10.4 11.2 12.2 87 11.9 10.9 10.0 9.2
9.8 10.6 11.5 12.4 88 12.1 11.1 10.2 9.4
10.0 10.8 11.7 12.6 89 12.4 11.4 10.4 9.6
10.2 11.0 11.9 12.9 90 12.6 11.6 10.6 9.8
10.4 11.2 12.1 13.1 91 12.9 11.8 10.9 10.0
10.6 11.4 12.3 13.4 92 13.1 12.0 11.1 10.2
10.8 11.6 12.6 13.6 93 13.4 12.3 11.3 10.4
11.0 11.8 12.8 13.8 94 13.6 12.5 11.5 10.6
11.1 12.0 13.0 14.1 95 13.9 12.7 11.7 10.8
11.3 12.2 13.2 14.3 96 14.1 12.9 11.9 10.9
11.5 12.4 13.4 14.6 97 14.4 13.2 12.1 11.1
11.7 12.6 13.7 14.8 98 14.7 13.4 12.3 11.3
11.9 12.9 13.9 15.1 99 14.9 13.7 12.5 11.5
12.1 13.1 14.2 15.4 100 15.2 13.9 12.8 11.7
12.3 13.3 14.4 15.6 101 15.5 14.2 13.0 12.0
12.5 13.6 14.7 15.9 102 15.8 14.5 13.3 12.2
12.8 13.8 14.9 16.2 103 16.1 14.7 13.5 12.4
13.0 14.0 15.2 16.5 104 16.4 15.0 13.8 12.6
13.2 14.3 15.5 16.8 105 16.8 15.3 14.0 12.9
13.4 14.5 15.8 17.2 106 17.1 15.6 14.3 13.1
13.7 14.8 16.1 17.5 107 17.5 15.9 14.6 13.4
13.9 15.1 16.4 17.8 108 17.8 16.3 14.9 13.7
14.1 15.3 16.7 18.2 109 18.2 16.6 15.2 13.9
14.4 15.6 17.0 18.5 110 18.6 17.0 15.5 14.2
14.6 15.9 17.3 18.9 111 19.0 17.3 15.8 14.5
14.9 16.2 17.6 19.2 112 19.4 17.7 16.2 14.8
15.2 16.5 18.0 19.6 113 19.8 18.0 16.5 15.1
15.4 16.8 18.3 20.0 114 20.2 18.4 16.8 15.4
15.7 17.1 18.6 20.4 115 20.7 18.8 17.2 15.7
16.0 17.4 19.0 20.8 116 21.1 19.2 17.5 16.0
16.2 17.7 19.3 21.2 117 21.5 19.6 17.8 16.3
16.5 18.0 19.7 21.6 118 22.0 19.9 18.2 16.6
16.8 18.3 20.0 22.0 119 22.4 20.3 18.5 16.9
17.1 18.6 20.4 22.4 120 22.8 20.7 18.9 17.3 a Length is measured for children under 2 years or less than 87 cm height. For children 2 years or older or 87 cm height or more, height is
measured. Recumbent length is, on average, 0.7 cm greater than standing height; although the difference is of no importance to individual
children, a correction may be made by subtracting 0.7 cm from all lengths greater than 86.9 cm if standing height cannot be measured.
107
Annex 7: Nutrition and death rate survey sample questionnaire
NUTRITION SURVEY QUESTIONNAIRE
MORTALITY FORM
*This page must be filled in for every household
COUNTY:______________ CLUSTER NO. [ ][ ] HOUSEHOLD NO. [ ][ ]
DISTRICT:______________ TEAM NO. [ ][ ] NAME OF INTERVIEWER:____________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/ [ Y ][ Y ]
DR01 DR02 DR03 DR04 DR05 DR06 DR07 DR08
No. Name Sex
Male=1
Female=2
Age
(years)
Born since
_____ (insert
the start of the
recall period)
Joined since _____
(insert the start of
the recall period)
Cause of
death
Location of
death
a) Starting with the youngest, how many members are present in this household19 now? List them.
b) Starting with the youngest, how many members have left this household (out migrants) since _____ (insert the start of the recall
period)? List them
c) Do you have any member of the household who has died since _____ (insert the start of the recall period)? List them
Summary
Details U5 Total
Current household Members
Number of members who have joined during the Recall period
Number of members who have left during Recall period
Births during recall
Deaths during recall period
19 Household definition: a group of people who live together and share a common cooking pot
108
NUTRITION SURVEY QUESTIONNAIRE
HOUSEHOLD FORM
*This page must be filled for every households.
DISTRICT:__________________ CLUSTER NO. [ ][ ] SETTLEMENT:__________________
VDC:_______________________ TEAM NO. [ ][ ] TEAM LEADER:________________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/[ Y ][ Y ]
HOUSEHOLD CHARACTERISTICS
HH1 Who is the head of the household?
Circle only one (1) response.
Father…………………………………………... 1 Mother…………………………… 2
Grandfather…………………………………… 3 Grandmother…………………….. 4 Other (Specify)………………………………… 5
HH2 What is your caste?
Circle only one (1) response.
Dalit Hill/Teraj.........................……………… 1 Disadvantaged Janajati/ Hill/ Tera.. 2
Disadvantage non Dalit Terai caste group…… 3 Religious minority……………….. 4
Relatively advantaged Janajati group………… 5 Other (Specify)…………………... 6
WASH
WS1 What is the MAIN source of drinking
water for members of your household?
Circle only one (1) response.
Piped/tap water……………...…… 1 Tube well/hand or rower pump……. 2
Dug well………………….............. 3 Water from spring….………….. 4
Rainwater collection……………… 5 Tanker-truck……………………. 6
Cart with small tank/drum 7 Surface water (river, dam, canal). 8
Other (specify)…………………..... 8
WS2 Do you do anything to the drinking water to make it safer to drink?
Circle only one (1) response.
Boil…………. ................................ 1 Add bleach/chlorine................... 2
Strain it through a cloth.................... 3 Use water filter….. ................... 4
Solar disinfectant.............................
Nothing..................………………..
5
7
Let it stand and settle………......
Other (specify)………………...
6
8
WS3 What kind of toilet facility do members of your household usually use?
Circle only one (1) response.
Flush/pour flush (water seal)........... 1 Pit latrine…............................... 2
Composting toilet…….................... 3 Tin/bucket toilet......................... 4
No facility, bush, field, etc.……… 5 Other (specify) ............................ 9
WS4 Please mention all the occasions when
it is important to wash your hands.
Circle all that apply.
Before eating……………............... A After eating…………………….. B
Before praying…………………..... C Before breastfeeding/feeding a child D
Before cooking/preparing food….. E After defecation/urination………... F
After cleaning child’s defecation… G When the hands are dirty…………... H
After cleaning toilet or potty…… I Other (specify)…………………….. J
Don’t know………………………. K
WS5 What do you use to clean (wash) your
hands? Circle only one (1) response.
Water only………...................... 1 Water and soap……………...…… 2
Water and ash……….………….. 3 Water and soil………....................... 4
Other (specify) .......................... 5
FOOD SECURITY
FS1 What is the MAIN source of income for this household?
Circle only one (1) response.
Crops farming .................................. 1 Livestock farming.................... 2
Fishing………….............................. 3 Casual labour……… ............... 4
Remittance ........................................ 5 Trade/business. ........................ 6
Social assistance (pension, etc.)……. 7 Regular employment………… 8
Forest product collection…………… 9 Other (specify)……………….
FS2 What is the main source of the DOMINANT food item consumed in the past month?
Circle only one (1) response.
Own production.............................. 1 Borrowed................................. 5
Purchases......................................... 2 Gathering /wild ....................... 6
Gifts from friends/ family................. 3 Traded or bartered.................... 7
Food aid........................................... 4 Other (specify)......................... 8
FS3 From this time yesterday until now, what did
your household members consume? Include any snacks consumed as well.
Select all that apply.
Cereals.............................................. A Fish and seafood....................... G
Roots and tubers.............................. B Pulses/legumes/nuts…………. H
Vegetables........................................ C Milk and milk products............ I
Fruits ............................................... D Oils/fats .................................. J
Meat, poultry, offal.......................... E Sugar/honey………………… K
Eggs................................................. F Miscellaneous……………… L
OTHER
OT1 Do you use iodised salt (with
2 child logo) for cooking? Circle only one (1) response.
Yes…………………………………………..... 1 No………………………………... 2
Unknown……………………………………… 3 Other (specify)…………………… 4
OT2 Do you have a homegarden?
Circle only one (1) response.
Yes…………………………………………....... 1 No………………………………... 2
Other (specify)…………………………………. 3
109
NUTRITION SURVEY QUESTIONNAIRE
CHILD 6-59 FORM
*This page must be filled in for every household with a child aged 6-59 months; every child in this range should be
included.
DISTRICT:__________________ CLUSTER NO. [ ][ ] SETTLEMENT:__________________
VDC:_______________________ TEAM NO. [ ][ ] TEAM LEADER:________________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/[ Y ][ Y ]
CH01 CH02 CH03 CH04 CH05 CH06 CH07 CH08 CH9 CH10 CH11 CH12
HH
No.
Child
No.
DOB
Or
Age
(months)
If DOB
available,
record it as
DD/MM/Y and
do not calculate
age
Age
Verification
1 = Birth
certificate
2 =
Vaccination
card
3 = Recall
4 = Other
(Specify)
Sex
1 =
Male
2 =
Female
Weight
(kg)
Measure
to nearest
0.1kg
Write
down
the
decimal
and DO
NOT
round off
Height
(cm)
Measure
to nearest
0.1cm
Write
down
the
decimal
and DO
NOT
round off
Oedema
0 = No
1 = Yes
Refer the
child to
OTP if
there’s
oedema
MUAC
(cm)
Measure to
nearest
0.1cm
Write down
the decimal
and DO NOT
round off
Refer the child
to OTP if
MUAC <115
Z-score
of the
child
1 = >-3
2 = < -3
Refer the
child to
OTP if z-
score <-
3
How many
capsules of
vitamin A has
the child
received in the
past 6 months?
Show sample
capsules
0 = None
8 = DNK
Has the child
taken any drug
for intestinal
worms in the
past year?
0 = No
1 = Yes, Card
2 = Yes,
Recall
8 = DNK
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
26
27
110
CH12 CH13 CH14 CH15 CH16 CH17 CH18 CH19 CH20 CH21 CH22 CH23
Chil
d
No.
Was the child
sick in the
last TWO (2)
WEEKS?
0 = Not sick
1 =
Fever/Malari
a
2 = ARI
/Cough
3 = Watery
diarrhoea
4 = Bloody
diarrhoea
5 = Other
(specify)
8 = DNK
See case
definitions
below
If the
child was
sick in
the last 2
weeks,
what did
you do?
0 =
Nothing
1 = Took
the child
to
hospital
2 = Took
the child
to
traditiona
l healer
3 = Took
the child
to CHV
4 = Other
(specify)
If the child
was sick in
the last 2
weeks, did
you give
the child
any
food/drink
(including
breastmilk)
?
0 = No
1 = Yes
If the
child was
sick with
diarrhoea
, was the
child
given
zinc?
0 = No
1 = Yes
Has child
received
DPT-
HepB, (an
injection
given in
the left
thigh,
sometime
s given
during
polio
drops)?
0 = No
1 = Yes,
Card
2 = Yes,
Recall
8 = DNK
(Look at
card or
the SCAR
on the
LEFT
LOWER
ARM)
If the
child
has
receive
d DPT-
HepB,
how
many
timed?
(enter
the
number
)
Has child
received
measles
vaccinatio
n (a shot
given in
the arm at
the age of
9 months
or older, to
prevent
him/her
from
getting
measles)?
0 = No
1 = Yes,
Card
2 = Yes,
Recall
8 = DNK
Age of
mothe
r in
years
MUAC
of
mother
(cm)
Measur
e to
nearest
0.1cm
Enter
only
one
value
Write
down
the
decimal
and do
not
round
off
What is the
mother’s
physiologica
l status
1= Pregnant
2= BF with
child <6m
3=None of
the above
8=DNK
Did the
mother
consume
iron tablets
during the
last
pregnancy
?
0 = No
1 = Yes
8 = DNK
Is this child
6-23
months of
age?
0 = No
1 = Yes
IF YES,
PROCEE
D TO
NEXT
MODULE.
IF NO, GO
TO NEXT
CHILD.
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Fever: High temperature with shivering Cough/ARI: Any episode with severe,
persistent cough or difficulty breathing
Watery diarrhoea: Any episode of three
or more watery stools per day
Bloody diarrhoea: Any episode of three
or more stools with blood per day
111
NUTRITION SURVEY QUESTIONNAIRE
CHILD 6-23 FORM
*This page must be filled in for every household with a child aged 6-23 months; every child in this range should be
included.
DISTRICT:__________________ CLUSTER NO. [ ][ ] SETTLEMENT:__________________
VDC:_______________________ TEAM NO. [ ][ ] TEAM LEADER:________________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/[ Y ][ Y ]
CH21 CH22 CH23 CH24 CH25 CH26 CH27 CH28 CH29
HH
No.
Child
No.
Sex
1 =
Male
2 =
Female
Age of
the
child
Months
or
DOB
(this
can be
taken
from
the
child
6-59
form)
When did you start
breastfeeding this
child after the
delivery?
0 = Never
1 = Less than1 hr
2 = More than 1
but Less than24hrs
3 = More than one
day days
From this
time
yesterday
until
now, was
the child
given
breast
milk?
0 = No
1 = Yes
From this
time
yesterday
until now,
did the child
receive
solid, semi-
solid or soft
foods?
0 = No
1 = Yes
From this time
yesterday until
now, how
many meals or
snacks was the
child fed?
Has the child
received
micronutrient
powder
(MNP)?
0 = No
1 = Yes
From this time yesterday until now, what did
the child eat?
0 = Did not consume
1 = Consumed
Do not leave any blank
Gra
ins,
roots
, an
d t
uber
s
Leg
um
es a
nd n
uts
Dai
ry p
roduct
s
Fle
sh f
oods
Eggs
Vit
amin
A-r
ich f
oods
and v
eget
able
s
Oth
er f
ruit
s an
d v
eget
able
s
Grains, roots, and
tubers: Bread, noodles,
biscuits, chivada, rice,
porridge, maize, wheat
Legumes and nuts:
Beans, peas, lentils,
nuts, seeds or food made
from these
Dairy products:
milk, curd, cheese or
other milk products
Flesh foods: Pork,
lamb, goat, rabbit, wild
game, chicken, duck or
other birds.
Fresh or dried fish.
organ meats or blood
based food
Eggs Vitamin A-rich foods:
Carrot, dark green leafy
vegetables, pumpkin,
mango, red palm oil
Other fruits and
vegetables: Ripe mangoes,
dried amla, bananas, apples,
seasonal fruit.
Pumpkin, carrots, squash, or
sweet potatoes that are
orange inside
112
NUTRITION SURVEY QUESTIONNAIRE
FOCUS GROUP DISCUSSION GUIDE
Guidelines for the facilitator:
1. Find a quiet shaded area suitable for a group of 8-10 people to sit.
2. Invite 8-12 people of mixed ages and sex, with high representation from households with
a child under 5. Do not force people to join the group. Try to generate interest and
willingness.
3. Explain the objectives of the FGD. We are trying to get their experience to learn more
how to improve the programme and as a result, improve the health and nutrition status of
children in their community.
4. Explain that the information is confidential and no names are taken so they can openly
explain their real experience/opinions on the topics discussed.
5. The FGD will last 45mins-1hr. Explain this at the start.
6. Get permission from the local government authorities but it is not necessary to include the
local government authorities in the discussion. They can often take over the whole
meeting! People may be reluctant to talk openly in their presence.
7. Facilitator leads the groups through the 5 main themes, promoting responses and making
sure that the main topics are covered. Keep an idea of time spent and keep the discussion
on the theme.
8. At the end of each point, the facilitator summarises what the group has agreed as a
response. For example, “So, maize is the main staple in this area this season”. “So, most
of you say you have enough maize stored to last for 3 months”. “So only 4 members of
the group have access to protected spring water”. Do this for every issue discussed.
9. Note taker writes the summary information for every issue discussed. When possible, use
numbers to show how many people in the group agreed on the issue. For example, 8 out
of 10 group members believed that ‘healthy child’ = strong, happy, not sick. The other 2
did not comment.
10. Note taker writes direct examples given as experiences shared. For example, ‘one young
mother sold her cattle last month because she ran out of food and she was worried she
would not have enough resources to buy food for the coming months’. ‘3 male group
members plan to take factory jobs outside the area for the next 3 months, to make ends
meet’. Include the approximate age and sex of the speaker who you quote with an
interesting response.
11. Get the group to give concluding remarks on overall food security in their community.
12. Thank the group.
Follow the themes below and take notes in the space provided. If other health and food
related issues come up, report back to the supervisor.
113
Theme 1: Food Availability
What are the main food crops on the ground this season?
What are the main cash crops in this area this season?
What condition do you expect the next harvest to be (good, poor)? Explain and give
reasons why. For example, poor because of pest infestation, lack of rain, late rain, not
sufficient seeds planted, etc.). Good because the rains were good this season, fertiliser
was used, no pests experienced....etc.
How much does it cost to buy 1kg of the staple from the market? How does this compare
to the same month last year (more, less, how much more?)
How much does it cost to buy 1kg of fruits and vegetables from the market? How does
this compare to the same month last year (more, less, how much more?)
How much does it cost to buy 1kg of animal products (meat, dairy, eggs. etc.) from the
market? How does this compare to the same month last year (more, less, how much
more?)
114
Theme 2: Food Access
Explain the main income sources of this community this season.
How many of you have more than one income source (e.g., work in the fields during
harvest time, have daily labour work in the town in other seasons, other HH member
contributes earnings too).
Tell me about stored food in your home: what staple do you store, how long does it last?
What do you do when you finish the stored food?
How many of you keep livestock? What do you keep? How do you use these animals (sell
the milk, breed and sell after xx months, fatten and sell after x months).
115
Theme 3: Vulnerability and Coping Mechanisms during Food Insecurity
Explain what food insecurity means to you?
Have you faced food insecurity in the last 3 months?
What months are you most likely to face food insecurity in this area?
What do you do when you don’t have enough food for three meals a day (base this on
experience if possible using the most recent examples).
In your opinion, which households face food shortage in this community?
WHY?
116
Theme 4: Perceptions of Child Health and Child care Practices
What makes a child healthy (describe aspects showing ‘health’)
What actions do you take to keep the child healthy?
What are the main childhood illnesses children (U5) experiences in this community?
What are the main ways of caring for the child when experiencing each illness (BEST TO
ASK MOTHERS WITH U5 IN THEIR EXPERIENCE if possible for accuracy, note use
of healers, or health services, and reasons for healers or reasons for health service
utilisation)?
What are the feeding practices during pregnancy and early childhood? What is the role of
grandmothers and fathers in the feeding practices?
Which seasons are children most vulnerable to sickness in this area?
Which children are most vulnerable? (Young <2, poorest, big family, live near river??
Etc.)
117
Theme 5: Multi-Sectorial Programming
How is the multi-sectorial collaboration functioning?
Are the ward health management committees in this community? Are the VHMC or
FCHVs engaged with Farmer's Group or Lead Farmer - to promote in tandem both supply
(production) and consumption of nutrient dense foods (animal-source foods - milk or
dairy products, eggs, etc.) especially during adolescence, pregnancy and young children
(6-23 months of age)?
What proportion of disadvantaged groups is accessing child grant (Rs. 200/month/child
for up to two children) in this community? To what extent the FCHVs are encouraging
mothers to use the grant to enhance child health and nutrition (e.g. buy relatively more
expensive fruits, vegetables, animal-source foods - milk, eggs, etc.)
Are there ECD facilitators in this community? To what extend ECD facilitators and
FCHVs working in tandem to encourage early child stimulation and optimal feeding
practices?
118
NUTRITION SURVEY QUESTIONNAIRE
KEY INFORMANT QUESTIONNAIRE FOR THE MINISTRY OF AGRICULTURE
DISTRICT:__________________ CLUSTER NO. [ ][ ] SETTLEMENT:__________________
VDC:_______________________ TEAM NO. [ ][ ] TEAM LEADER:________________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/[ Y ][ Y ]
Name of Key Informant:_____________ Key Informant’s Position:_______________
1. What is the staff availability and capacity compared to the national standards to provide
services in the survey area?
2. What are the main crops cultivated and what is the yield in the last season? Has there any
increase/decrease in the production compared to previous seasons/years?
3. If there’s increase/decrease in production, what is the reason?
4. How is the prospect for harvest in the upcoming season?
5. How is the overall rainfall pattern – poor, adequate, good, etc.?
6. Has there been any shock such as flooding, drought, storm, in the area?
119
7. How is the livestock situation in the area? Has there any increase/decrease in livestock?
8. If there’s increase decrease, what is the main reason?
9. How is the overall livestock situation in the area – poor, good, etc.?
10. Has there been increase/decrease in food prices in the market? If so, why?
11. How is the availability of protected water sources?
12. What is the overall food security situation in the area at present and in the near future?
120
NUTRITION SURVEY QUESTIONNAIRE
KEY INFORMANT QUESTIONNAIRE FOR THE MINISTRY OF HEALTH STAFF
DISTRICT:__________________ CLUSTER NO. [ ][ ] SETTLEMENT:__________________
VDC:_______________________ TEAM NO. [ ][ ] TEAM LEADER:________________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/[ Y ][ Y ]
Name of Key Informant:______________ Key Informant’s Position:_____________
1. What is the staff availability and capacity compared to the national standards to provide
services in the survey area?
2. Are there health facilities not functioning in the area? If yes, how many and why?
3. How is the availability of essential medicine in the health facilities?
4. Has there been any disease outbreak in the area recently?
5. What are the major illnesses reported among children under 5 years of age?
121
6. What is the immunization coverage in the area?
7. Are there routine immunization services available at the health facility?
8. Are there antenatal or postnatal care activities in the area? If no, why?
9. How is the overall health situation in the survey population?
122
NUTRITION SURVEY QUESTIONNAIRE
DIRECT OBSERVATION CHECKLIST
DISTRICT:__________________ CLUSTER NO. [ ][ ] SETTLEMENT:__________________
VDC:_______________________ TEAM NO. [ ][ ] TEAM LEADER:________________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/[ Y ][ Y ]
1. What is the food availability and access at the time of your visit?
[Direct observations are made through observation of commodities available, livestock and
pasture condition, checking warehouses/storage rooms, visiting food distribution sites,
observing food prices (price tags), cleared pastures, etc.]
2. What is the WASH situation at the time of your visit?
[Direct observations are made during visits to water catchments areas and noting both the
time taken for a return trip, flow and quality of water, storage facilities of the water and the
sanitation system.]
3. What is the health situation at the time of your visit?
[The general health condition for example in terms of skin diseases, eye problems and runny
nose/ARI are observed.]
4. How is the overall situation in the area?
123
Annex 11: Example of standardization test data collection forms
Measure 1 Measure 2
Enumerator’s name:_____________________
Enumerator’s ID No.:______
Enumerator’s name:_____________________
Enumerator’s ID No.:______
Child Weight
(Kg)
Height
(cm)
MUAC
(cm)
Child Weight
(Kg)
Height
(cm)
MUAC
(cm)
1 1
2 2
3 3
4 4
5 5
6 6
7 7
8 8
9 9
10 10
124
Annex 12: Cluster control form
CLUSTER CONTROL FORM
COUNTY:____________ CLUSTER NO. [ ][ ] NAME OF TEAM LEADER:_______________
DISTRICT:___________ TEAM NO. [ ][ ] TEAM LEADER’S PHONE NO.:___________
DATE OF INTERVIEW [ D ][ D ]/[ M ][ M ]/ [ Y ][ Y ]
HH
No.
Name of the head of household Outcome of the visit
0 = completed
1 = partially completed
2 = abandoned HH
3 = refused
4 = absent
No. of eligible
children
No. of eligible
children
measured
HH need to be
re-visited?
0 = No
1 = Yes
Comments
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
125
Annex 13: Final survey report format
Executive summary (one to two pages only)
Geographic area surveyed and population type, dates of survey
Methodology used (sampling, sample size, main indicators)
Main anthropometric and death rate results
Other important results – vaccination coverage, morbidity, etc
Brief interpretation of the results
Recommendations
1. Introduction
Geographic description of survey area
Description of the population
Justification to conduct the survey
1.1 Survey Objectives
2. Methodology
2.1 Sample size
Sampling methodology
Sample size calculation for all major indicators used (i.e. nutrition and death rate) with
assumptions for expected prevalence, expected design effect (if cluster sampling),
precision; if number of children was converted into the number of households, describe
how this was done
Information on how sample sizes we reconciled?
If cluster sampling, how the no. of clusters and households per cluster decided
2.2 Sampling procedure: selecting clusters
Population figures used and the source
Information on the no. of clusters visited and, if some clusters were not visited why?
2.3 Sampling procedure: selecting households and children
Sampling technique used and why
Details of segmentation
How the sampling technique was applied
If different sampling methods are used in different clusters, description of the methods
2.4 Case definitions
Household definition
Definition for GAM and SAM
Length or recall period in the death rate survey; how it was set?
Case definitions for IYCF indicators, morbidity, and immunization coverage
2.5 Questionnaire, training and supervision
Questionnaire
Language of the questionnaire
Language of interviews
Was the questionnaire pre-tested (piloted) before the survey?
126
Survey teams and supervision
Survey team composition
No. of teams trained and used in the survey
Average qualification of survey team members
If field supervisors were used, provide details (number, experience, etc.)
Details of supervision (frequency, duration, etc.)
Training
Details of trainers, duration of the training
Topics covered during the training
Details of standardisation test for anthropometry
Details of pre-test; no. of children measured; no. of households visited
2.6 Data analysis
Where and by whom the data was entered?
Details of quality check put in place for data entry
Details of computer software used
3. Results
3.1 Anthropometric results (based on WHO standards 2006):
Demography of the sample
Distribution of age and sex of sample
Prevalence of acute malnutrition by sex
Prevalence of acute malnutrition by age
Distribution of acute malnutrition and oedema based on weight-for-height z-scores
Prevalence of acute malnutrition based on MUAC cut offs (and/or oedema) and by sex
Prevalence of acute malnutrition by age, based on MUAC cut offs and/or oedema
Prevalence of underweight by sex
Prevalence of stunting by sex
Mean z-scores, Design Effects and excluded subjects
3.2 Death rate survey results (retrospective over x months/days prior to interview)
CMR (total deaths/10,000 people / day): (95% CI)
U5MR (deaths in children under five/10,000 children under five / day): (95% CI)
The main causes of death
3.3 Other results
3.4 Children’s morbidity
Prevalence (and 95% CI) of reported illness in children in the two weeks prior to the
survey
3.5 Vaccination coverage
Vaccination coverage (with 95% CI) – note: BCG for 6-59 months and measles for 9-59
months
127
3.6 WASH results
3.7 Food security results
3.8 Other results
4. Discussion
4.1 Nutritional status
Discuss sample sex ratio, age distribution, etc. and the possibility for bias
Prevalence of acute malnutrition
If previous survey results are available, how do these results compare to before?
How does the prevalence compare to established thresholds of malnutrition?
Is the prevalence of malnutrition typical for the area?
4.2 Death rates
Potential for any bias in the data
Death rates
If previous survey results are available, how do these results compare to before?
How do the rates compare to benchmarks?
Is the death rate typical or not?
4.3 Other results
If previous survey results are available, how do these results compare to before?
What are the possible contributing factors?
Is a detailed IYCF assessment required?
4.4 Causes of malnutrition
Possible causes of malnutrition – immediate, underlying and basic causes
What are the prospects for the coming months? Will the situation get better or worse?
Who is worst affected?
What are the chronic causes of malnutrition?
What does the community recommend?
A diagram to show the causal framework of malnutrition may be useful.
5. Conclusions
Overall conclusions on the severity of the situation and the urgency of the response
required
6. Recommendations and priorities
Prioritise recommendations and try to give a time when action would be appropriate (e.g,
immediate, medium term or longer term).
Future nutrition monitoring
Is it necessary to carry out another nutrition survey in this area in the near future? Who
should do it? Should there be any changes to the survey methodology? When should the
survey take place?
Should there be food security indicator monitoring in this area? Who should do it?
128
7. References
List all secondary sources to which you have referred in the text.
8. Acknowledgements
9. Appendices
Appendix 1: Plausibility report
Copy of plausibility report from the ENA software
Appendix 2: Assignment of clusters
Copy of the cluster assignment sheet from the ENA software
Appendix 3: Evaluation of enumerators
Copy of the report from the ENA software
Appendix 4: Maps of the survey area
Appendix 5: Survey questionnaire
129
REFERENCES
Cogill, Bruce (2003). Anthropometric Indicators Measurement Guide. Food and Nutrition
Technical Assistance Project, Academy for Educational Development, Washington, D.C.
ENA for SMART 2011 website. Software for Emergency Nutrition Assessment. Available at:
http://www.nutrisurvey.de/ena2011/ - accessed on 28 July 2012.
CDC Atlanta. Epi Info software. Available at: http://www.cdc.gov/epiinfo/
ENCU/EWD/MoARD (2008). Guideline for Emergency Nutrition Surveys in Ethiopia.
Interim new version. September 2008.
FSNAU (2011). Guidelines for Emergency Nutrition and Mortality Surveys in Somalia. June
2011.
Global Nutrition Cluster / Assessment Working Group (2008). Meeting on Standardized
Monitoring and Assessment in Relief and Transition (SMART). April 7-9, 2008. Rome,
Italy. Final meeting report, 30 May 2008.
CARE USA (2010). Infant and Young Child Feeding practices. Collecting and Using Data: A
Step-by-Step Guide. January 2010.
Sampling Paper (2012). Sampling Methods and Sample Size Calculation for the SMART
Methodology. June 2012. Available at: http://www.smartmethodology.org/ - accessed on
28 July 2012
SMART Indicators (2006). Measuring Mortality, Nutrition Status, and food Security in Crisis
Situations: SMART methodology version 1. April 2006. Available at:
http://www.smartmethodology.org/ - accessed on 28 July 2012.
Standardised Training Package Modules. Available at: http://www.smartmethodology.org/ -
accessed on 28 July 2012.
The Save the Children Fund (2004). Emergency Nutrition Assessment. Guidelines for Field
Workers. 2004.
The Sphere Project (2011). Humanitarian Charter and Minimum Standards in Disaster
Response. Edition 2011.
WFP (2009). Comprehensive Food Security and Vulnerability Analysis Guidelines. 1st
edition. January 2009.