1

Short interpregnancy intervals and child survival

Ph. D. Thesis

Frank Mugisha Kaharuza

Department of Epidemiology and Social Medicine, University of Aarhus

Faculty of Health Sciences, University of Aarhus 2001

2

Preface

The work presented in this PhD thesis was carried out both at the Department of

Epidemiology and Social Medicine, Aarhus University and in Uganda at the Child

Health and Development Centre.

I would like to thank my supervisors, Associate Professor Svend Sabroe, Senior

Associate Professor Flemming Scheutz, for their guidance and constructive criticism

throughout the study period. Special thanks to Professor Jorn Olsen and Associate

Professor Olga Basso and all the academic staff of the Department of Epidemiology

and Social medicine for their moral support, inspiring discussions, and for introducing

me to the worlds of social and reproductive epidemiology. Antonio Sylvester

Vethanayagam, Kirsten Overgaard, Helle Trolle Birkegaard, Dorit Lindblad, and

Lillian Enemaerke, at the Department of Epidemiology and Social Medicine, and

Jorgen Pedersen, Research Secretary at the Department of Anthropology, Copenhagen

University are thanked for their practical support that made my stay in Denmark

pleasurable. I am also grateful to Janet Mikkelsen for carefully and patiently ‘dotting

the i’s and crossing the t’s’, and my colleagues and friends, Dr Yuan Wei, Dr. Weijin

Zhou, for their moral support.

I am very grateful to the TORCH project, especially Professor Susan Reynolds

Whyte and Dr. Jessica Jitta for giving me this opportunity to further develop my

interest and skills in studying and understanding maternal health issues. The practical

support given to me by CHDC staff especially Jennipher Twebaze, Peter Nyango and

Augustine Mutumba is appreciated.

Special thanks to all the mothers who generously participated in the study. I

would like to acknowledge Dr. John Wanyama and the 16 research assistants, and for

their support and patience during the data collection process.

I am very grateful to my friend and mentor, Dr Tom Barton, for kindling and

nurturing the spirit of research in me and supporting my family in many ways during

my absence.

Finally my deepest gratitude to my wife Jolly, and my sons Philip, Mark and

Jesse for their patience, support, and encouragement during the long periods of

absence both in the field and in Denmark.

Financial support for this work was obtained from DANIDA through the

ENRECA programme.

Aarhus, September 2001

3

The thesis is based on the following papers and a brief communication:

I. Kaharuza FM, Sabroe, S, Scheutz, F. Determinants of short interpregnancy

intervals in rural Uganda. Submitted to the International Journal of Obstetrics

and Gynecology.

II. Kaharuza FM, Basso O, Sabroe S. Choice or Chance: Determinants of short

interpregnancy interval in Denmark. Acta Obstetrica et Gynecologica

Scandinavica 80: 532-538.

III. Kaharuza FM, Sabroe, S, Scheutz, F. Determinants of child mortality in rural

Eastern Uganda. Submitted to the East African Medical Journal.

IV. Kaharuza FM, Sabroe, S, Scheutz, F. An Adaptable pregnancy interval

calculator (letter) Submitted to Studies in Family Planning.

4

Abbreviations

BMI Body Mass Index

CHDC Child Health and Development Centre

DANIDA Danish International Development Agency

DHS Demographic and Health Survey

ENRECA Enhancement of Research Capacity

IMR Infant Mortality Rate

IUGR Intrauterine Growth Retardation

LBW Low Birth Weight

MCH Maternal and Child Health

NNM Neonatal Mortality

PTB Preterm Birth

SES Socio-economic status

SGA Small-for-gestation age

TORCH Tororo Community Health

WFS World Fertility Survey

5

Contents

Preface............................................................................................................................2

Contents .........................................................................................................................5

1. Introduction............................................................................................................7

1.1. Definition of birth interval .....................................................................7

1.2. Components of birth interval .................................................................8

1.3. Magnitude of the problem of short birth intervals ...............................10

1.4. Short birth interval as an outcome .......................................................18

1.5. Social-economic status and birth interval ............................................19

2. Aims of the thesis.................................................................................................20

3. Materials and Methods.........................................................................................21

3.1. Uganda .................................................................................................21

3.2. Denmark...............................................................................................25

4. Statistical analysis................................................................................................28

4.1. Overall methodology ...........................................................................28

4.2. Data management.................................................................................28

5. Results..................................................................................................................31

5.1. Short interpregnancy intervals in Uganda............................................31

5.2. Short interpregnancy intervals in Denmark .........................................33

5.3. Child mortality in Uganda ...................................................................35

6. Discussion............................................................................................................38

6.1. Main results..........................................................................................38

6.2. Source of bias.......................................................................................38

6.3. Main findings compared to other studies.............................................40

7. Conclusions..........................................................................................................43

8. Perspectives..........................................................................................................44

9. Summary..............................................................................................................45

10. Danske Resumé................................................................................................47

11. References........................................................................................................49

6

Map showing location, population density, and health facilities of Bunyole, Tororo District 1999.

Location of Tororo in Uganda

Sour ce: Digita l Map by NIEC, 1995

Source. 1991 Uganda Population and housing census, MoFEP 1995

TORORO#

Study area

Population density (persons/sq.km)

Non Study area

Kilometers3020100

N

NN

101-150151-200201-250

< 100

> 251

ÆP

×

Ñ

Ñ

Ñ

Ñ

Ñ

#

Kangalaba SDButaleija HC

Busolwe Hospital

Bunawale DMU

Busaba CCF SD

Bugalo SD

Nabiganda SD

Tororo

Pallisa

Iganga

Mbale

Busia

7

1. Introduction

The timing of births involves both biological and socio-cultural processes. The

biological processes include fecundity, ovulation and spermatogenesis, gestation and

delivery of a normal child, and lactational infecundability. Socio-cultural processes

include behaviour that limits ovulation, those that avoid pregnancy, or those that

promote childbearing. Although some variation in the biological processes exists,

socio-cultural factors play a major role in the pace of childbearing. Birth spacing is

an important determinant of the rates of population growth and has a strong bearing

on maternal and infant health. A different mix of mechanisms causing either short or

long intervals may represent women or families with different bio-demographic and

social risk profiles for short and long intervals. The distribution of and reasons for

short and long intervals varies considerably among populations and health effects of

short intervals are not the same for all populations.1 Specific reproductive health

policies and programming are needed to address these differences. It is therefore

important to know the specific factors and circumstances that determine, and the

outcomes that follow, short birth intervals in any given population.

1.1. Definition of birth interval

Researchers determining factors that affect fertility and fertility outcomes have used

various definitions of birth interval. Definitions range from the interval between two

consecutive live births (interbirth interval), to the interval between the outcome of one

pregnancy and the conception of the next (birth-to-conception interval), to the interval

between two consecutive pregnancies (interpregnancy interval), and the number of

births within a given time frame (average birth interval).1,2 Different definitions are

used because they are appropriate in different situations. For example, surveys done

among populations unsure of conception dates use interbirth intervals, register-based

and cohort studies use either pregnancy or birth-to-conception intervals. Thus, lack of

a uniform definition of pregnancy intervals presents a problem in interpreting and

comparing studies.3

The definition of a short birth interval has varied among studies principally due to

the interval definition (birth-to-conception or interpregnancy interval), and the

pregnancy outcomes being studied. An “optimum” birth interval is defined as the

interval associated with the greatest probability of giving birth to a normal full term

infant and with the lowest risks of adverse outcomes to the mother and the preceding

child.4 Short birth-to-conception intervals have ranged from three months5 to 18

months6 while studies using Demographic and Health Survey (DHS) data consider an

interbirth interval less than 24 months short.7 On the other hand, long intervals,

greater than 100 months are also associated with an increased risk of adverse

pregnancy outcomes.8

8

1.2. Components of birth interval

To understand the variation and effects of birth spacing, one needs to understand the

determinants of human fertility and reproduction. Women will produce a certain

number of children by the end of their reproductive lifetimes because of the way in

which they time the various reproductive events. Figure 1 depicts the female

reproductive life course as a series of time intervals. Menarche signals the beginning

of fecundity (the biological capacity to reproduce) while marriage represents all

women living in sexual union or are having regular sexual intercourse. Following

pregnancy, a woman will remain infecundable until the normal pattern of ovulation

and menstruation is restored.

Figure 1. A female reproductive life course viewed as a series of time intervals.9

The time spent in different reproductive states will cause interpopulation and

intrapopulation variation in fertility. However, a number of factors determine the

time spent in each of these intervals.

Initially, Davis and Blake10 described a framework that mainly considered the

socio-cultural factors that affected the three necessary steps of human reproduction:

intercourse, conception, and pregnancy (gestation and parturition). This was later

modified and quantified by Bongaarts.11 The frameworks consider that if any factor is

to affect fertility, it must do so through its effect on one or more of the “proximate

determinants” or “intermediate fertility variables”. Figure 2 shows another

framework that considers biological determinants as well. The exposure factors

determine the probability of conception while the susceptibility factors govern the

conditional probability of successful reproduction, given that the exposure occurs.9

Thus, after menarche and before menopause, fertility will depend on factors that

affect fecundability, defined as the probability that a fecund couple will conceive

during the month of exposure to unprotected intercourse.

Conception

Inter-birth interval

Conception Return to ovulation Intrauterine

death Time added by intrauterine death

Third Birth

Last birth

Onset of permanent sterility

Menarche Marriage First Birth Second

Birth Menopause Age 0

Birth Conception

Return to ovulation

Birth

9

Figure 2. The proximate determinants of natural fertility 9

Proximate determinants exert a direct effect on each of time intervals because

they determine the length of these intervals. Figure 3 shows that a birth interval is

composed of three periods that can be modified by biological and behavioural factors.

The postpartum infecundable period is a period between the delivery and first

postpartum ovulation. This period is lengthened by lactational behaviour and

nutritional status of the mother.12-16 The fecund waiting time to conception, also

called the fecundable period, occurs from the first postpartum ovulation to

conception. Fecundity depends on ovum viability and sperm capacity to fertilise.

Sexual activity patterns including low frequency of sexual intercourse and postpartum

abstinence, and use of contraception would prolong birth intervals.17-20 Although

gestational length is an important determinant for foetal viability, variation in the

gestational length is unlikely to be a major cause of variation in interval length except

for the very short intervals. In some cases prolonged birth intervals are due to time

added by a pregnancy loss.

I Exposure Factors 1. Age at menarche 2. Age at marriage or entry into sexual union 3. Age at menopause 4. Age at onset of pathological sterility (if earlier than menopause) II Susceptibility factors

5. Duration of lactational infecundability 6. Duration of the fecund waiting time to conception (determined by the

following fecundability factors): a. frequency of insemination b. length of ovarian cycles c. proportion of cycles ovulatory d. duration of fertile time given ovulation e. probability of conception from a single insemination in the fertile period

7. Probability of foetal loss 8. Length of the nonsusceptible period associated with each foetal loss 9. Length of gestation resulting in live birth

10

Figure 3. Components of a birth interval and possible modifiers of each interval.

It is also known that factors that affect fertility will also affect child survival.

Mosley and Chen 21 proposed an analytical framework to study the determinants of

child survival in developing countries that suggests that all social and economic

determinants must operate through a set of intermediate determinants, which will

directly influence the risk of morbidity and mortality. These determinants have been

divided into five categories of which birth interval is an important determinant of the

‘maternal factors’ category.

1.3. Magnitude of the problem of short birth intervals

The overall public health importance of short interpregnancy interval is determined

not only by the risks for mortality and morbidity of the preceding child, subsequent

child, and the mother, but also by the prevalence of short intervals in the population.

1.3.1. Prevalence of short birth intervals

Since most research in developed countries has focused on the association between

short birth intervals and adverse perinatal outcomes, different cutoff points for short

birth intervals have been used. The prevalence of short birth intervals range from 5-

30%.5,6,8,22-24

World Fertility Survey and later the Demographic and Health Survey (DHS) are

nationally representative cross-sectional surveys, mainly carried out in developing

countries, to study fertility and demographic changes. In these surveys, an interval

less than 24 months is considered short. However, the prevalence of short interbirth

intervals of less than 18 months, which is comparable to a birth-to-conception interval

of 9 months, also ranges from 6-24%.7

1.3.2. Short birth intervals and risk for adverse pregnancy outcome

Abortion

Studies from developed countries suggest that both short and long birth intervals are

Birth Return to ovulation

Birth Conception

Postpartum infecundable period

Time to conception

Gestation period

Lactation Postpartum abstinence Frequency of intercourse Contraceptive use Sperm and ovum viability Ability to fertilise

Period

Period

Modifiers Pregnancy loss

11

associated with spontaneous abortion in the subsequent pregnancy, especially if the

interval is less than six months or longer than 60 months.25,26 The risk of spontaneous

abortion following an induced abortion was increased if the mother became pregnant

within three months of the procedure.27

Low birth weight, preterm birth and small-for-gestational age

Meta-analysis of studies from 1970-1984 on low birth weight (LBW) has shown that

the effect of short birth intervals on low birth weight is inconclusive, but also that the

short birth intervals may not be an important cause of intrauterine growth retardation

(IUGR) in the United States, where most of the reviewed studies were carried out.28

Recent studies are still inconclusive, but more studies show a positive association

between short birth intervals and low birth weight,22,29 preterm birth,5,6,8 and small-

for-gestational age,30-33 while others show little or no association with some of these

outcomes.34,35 The risk for these adverse outcomes was increased and ranged from

1.2 to more than 3.0. The risk was high in some developing countries36 and slightly

lower in the developed countries.22 The LBW outcomes were curvilinear with a high

risk for the short and very long intervals in United States37 and India.38 Table 1

presents more recent studies that show the large variation in prevalence and effects of

short interpregnancy intervals on adverse pregnancy outcomes in developed

countries.

12

Table 1. Summary of selected major cohort studies of association between short birth intervals and adverse pregnancy outcomes

Key results: Adverse pregnancy

outcome (OR or RR)

Author

Publication Year

Country

Population Study year

Study

Design

Exposure

Birth interval (months)

Confounder control Ref.

Interval (months)

Prevalence %

[Cut-off point in months]

LBW PTB SGA Other

Basso, O 22 1998

Denmark 10,187 1980-92

Historical cohort

BCI (<4) (>8)

MA Parity SES GA

24-36 5..5 [8] 1.0 ns 0.9 ns

3.6 2.3

Fuentes-Afflick E 6 2000

USA 289,842 1991

Cohort records review

BCI (< 18) (<6)

(7-11) (12-17)

MA Parity Education Obstetric history Ethnic group

18-59 36.9 [18] 23.3 [12]

1.2 1.2 1.1

Very preterm

1.47 1.39 1.26

Miller JE33 1989

Sweden 92,084 1973

Cohort IPI <12 MA Previous live births Gestational age PBI

18-59 1.3

Rawlings JS 35 1995

USA 1922 1983-93

Cohort BCI (< 6) BW BCI (<9)

WW BCI (<3)

Race

≥6

19.5 [6] 2.7 4.2

1.0 ns

Schults RA5 1999

USA 34,569 1988-94

Cohort Records review

BCI ( ≤ 3 ) Marital status Prenatal care M Ed Race

13-24 3.2 [12] 1.2

1.6

Wohlfahrt J25 2000

Denmark 181,483 1988-92

Cohort Register based

IPI (< 6) (>60) (>120)

MA Parity Spontaneous abortion

18-23 Abortion 1.1 1.3 1.1

Zhou, W27 2000

Denmark 61,763 1980-94

Cohort Register based

BCI ≤ 3 induced abortion

MA Residence Parity

3.4 [6] abortion 4.1

Zhu, BP 1999 8

U SA 173,205 1989-96

Cohort records review

BCI (< 18 ) 16 factors 18-23 5.4 [6] 19.1 [12]

1.4

1.4 1.3

BCI: Birth-to-conception interval BI: Birth interval BMI: Body Mass Index GA: Gestational age IPI: Inter-pregnancy interval LBW: Low Birth Weight MA: Maternal age M Ed: Maternal

Education NR: Not reported OR: Adjusted Odds ratio ns: not significant PTB: Preterm Birth RR Relative Risk SES: Socio-economic status SGA: Small-for-gestational age

13

Child mortality

The relationship between birth interval and neonatal, infant and childhood mortality

has been extensively studied using data from World Fertility Surveys, Demographic

Health Surveys, and household survey data obtained from a number of developed and

developing countries.2,7,39-49 Table 2 presents selected studies from developed

countries that consistently show that short birth intervals are associated with an

increased risk of neonatal, post neonatal, and early childhood mortality. The

association between short birth intervals and perinatal mortality is strong even after

controlling for length of gestation, previous child mortality, and other potential

confounders.3 Child mortality trend studies show that despite reduction of child

mortality and some associated risk factors decreasing,50,51 short birth interval levels

have remained unchanged.52 Given the consistent association between short birth

intervals and child mortality, reduction in short birth intervals would have a great

impact on reducing child mortality.

Other adverse childhood outcomes

Child mortality and adverse pregnancy outcomes aside, few studies show long-term

adverse effects of short birth intervals on child growth and development. The effects

on the preceding or index child’s development vary amongst populations. In Nepal,

the linear growth of the preceding child reduced by 28% following short

interpregnancy intervals,53 while there was no effect on child growth in a Kenyan

population.54

Another study identified short interpregnancy interval as a consistent prenatal

factor that increased the risk for cerebral palsy.55 In Singapore, children born after

short birth intervals scored lower verbal and perceptive tests than those born after a

long birth interval. Short birth intervals accounted for 39% of the variation in the Mill

Hill vocabulary test. The same study showed a positive linear effect of gradual

improvements with increase in birth interval.56

14

Table 2. Summary of selected major studies from developing countries of association between short birth intervals and child mortality

Key results: Child mortality (RR)

Author

Publication Year

Country

Population

Study year

Study

Design

Exposure

Birth interval

(months)

Confounder

control

Ref.

Interval

(months)

NNM PNM IM

Koenig 2 1990

Bangladesh 11,454 births 1973-78

Cohort (DSS)

PBI<24

SBI <24

Child Gender MA Parity SES M Ed

24-36 1.5 1.2 ns 1.3 ns

3.2

Miller 3 1989

Hungary and Sweden 128,563

Cohort BCI<12

BCI>60

GA MA Child gender

24-35 1.8

1.3

Boerma 7 1992

17 Developing countries 1986-89

Cross-sectional survey (DHS)

PBI <18 Twins MA M Ed Residence SES

24-35 1.4-4.8

0.8-2.8

0.9-2.4 ns

Hobcraft 57 1983

26 countries 140,100 1974-80

Cross-sectional survey (WFS)

PBI<24 M Ed MA

1.5-3.4 1.1-7.2 0.8-3.9

Miller 45 1992

Bangladesh 1755 1975-80 Philippines 3029 1983-89

DSS Cohort

BCI <15 MA M Ed SES Familial mortality

1.8

1.6

Rutstein 52 56 developing countries 1986-1998

Repeated Cross-sectional

PBI<24 17 factors

0.8/1000* 1.1/1000*

BCI: Birth-to-conception interval DHS: Demographic and health survey DSS : Demographic Surveillance System GA: Gestational age IMR : Infant Mortality MA: Maternal age NNM : Neonatal mortality ns: not significant PNM : Post Neonatal Mortality PBI: Preceding Birth Interval RR: Relative risk SES: Socio-economic status sig. significant WFS: world fertility survey U5MR Under-5 mortality

15

1.3.3. Short birth intervals and maternal outcome

The few studies carried out to determine the effect of short intervals on the mother are

inconclusive. Table 3. presents results of some studies that consider the effects of

short intervals on maternal morbidity and mortality. It has been postulated that short

birth intervals and prolonged lactation worsen maternal nutritional status.58,59. Some

studies demonstrate an increase in maternal weight with parity,33,60 but a cohort study

in Pakistan suggested that maternal depletion occurs and may cause poor reproductive

outcomes.61

Short interpregnancy intervals were associated with maternal postpartum

depression62 and an increased risk of uterine dehiscence or rupture during trials of

labour following a previous Caesarean delivery.63,64

The effect of birth intervals on maternal mortality is inconclusive with higher risk

of death in some studies based on hospital records24 and no effect in others

surveys.65,66

16

Table 3. Studies of association between short birth intervals and adverse maternal outcomes

Author

Publication

Year

Country

Population Study

Year

Study Design Exposures

Birth interval

(in months)

Outcomes Confounder control Key results (OR)

Conde-Agudelo24 2000

Latin America 456,889 1985-1997

Cross-sectional study

IPI (<6) Maternal morbidity and mortality

16 major obstetric confounders

Increased risk of maternal death (OR 2.5); Anaemia (OR1.3) Third trimester haemorrhage (OR1.7); puerperal endometritis(1.3)

Esposito MA64 2000

USA 170 [43 cases 127 controls] 1990-1999

Case-referent

BCI ( <6 ) Uterine scar failure / Uterine rupture

MA Parity Use of cervical ripening agents in labour

Increased risk for uterine scar failure (OR 3.9)

Fortney JA 66 1998

Multiple 377 cases, 1420 controls 1977-1993

Case-referent Matched case-referent

BI (<24) Maternal mortality Identified but not reported

Non significant association, Significant trends of maternal death with shortening IPI (1-2.4 for BI <12 months)

Gurel SA, 62 2000

Turkey 85 Not reported

Case-referent IPI (< 24)

Depression BDI > 17

M Ed Parity Congenital Abnormality Unwanted pregnancy

Associated with moderate to severe depression (OR 6.8)

Khan KS 61 1992

Pakistan 278 1986-1992

Cohort Continuous variable

Post partum weight BMI Haemoglobin status

MA Parity Maternal weight

Short IPI interval associated with lower postpartum weight and lower BMI

Miller JE 60 1989

Lesotho 873 1982

Cross-sectional study

IBI (<12)

Maternal nutrition [BMI, Triceps thickness Arm muscle area]

MA Parity Pregnancy status

Short birth interval associated with short term maternal depletion Nutritional Indices lowest at 18-23 months Arm muscle area significantly reduced at 12 months

Ronsmans 65 1998

Bangladesh 390 cases 1170 controls 1982-1993

nested case-referent study

BCI (<9) BCI (<14)

Long BCI (>39)

Maternal mortality MA M Ed Residence Religion

Short birth interval not associated with maternal death. Risk of maternal death highest in extremes of age and parity

Shipp 63 2001

USA 2409 1984-1996

Cross-sectional

IBI (<18) Symptomatic Uterine Rupture

Maternal age Time in labour Gestational age Oxytocin use

Increased risk for uterine scar failure OR (3.0)

Smits67 1999

Canada 2062 1850-99

Historical cohort

BCI (<6) Fertility among daughters

Husband’s age and occupation Daughters age Preceding child survival

Higher risk for conception failure if daughter delivered after short interval. (OR 1.1) Highest fertility among daughters born in winter

BCI: Birth-to-conception interval BDI: Beck’s Depression Inventory BI: Birth interval BMI: Body Mass Index IBI: Inter Birth Interval IPI: Inter-pregnancy interval MA: Maternal age M Ed: Maternal Education OR: Adjusted Odds ratio

17

1.3.4. Preceding and subsequent birth interval

Two time periods have to be considered in birth interval research. The preceding

birth interval is the period between the previous pregnancy and the index childbirth,

while the subsequent interval is the period between the index childbirth and the next

pregnancy.

Intervening processes through which preceding or subsequent birth intervals

operate to influence the survival chances of the child are still unclear. Most studies

consider the effect of preceding birth interval on child survival. Fewer studies

consider the effect of the subsequent birth interval on child survival because of the

difficulty of determining the cause-effect relationship, since accurate dates on

weaning and death of the index child are usually not available.2,39,68

1.3.5. Theoretical concepts of birth interval and adverse pregnancy outcomes

and child survival

Five hypotheses have been postulated to explain the association between birth

intervals and adverse pregnancy outcomes.

a) Maternal depletion hypothesis: Short birth intervals and increased periods

of lactation result in the worsening of maternal nutritional status, because

the short time does not allow the mother’s body full physiological

recovery.40,58,59 This causes a hostile intrauterine environment, which

increases the likelihood of poor reproductive outcomes (low birth weight,

prematurity, intrauterine growth retardation).

b) Sibling competition: Closely spaced births lead to two children of roughly

the same size and age, and consequently to reduced family care and

support for both or one of the children. They will share the meagre family

resources which, in turn, may lead to increased morbidity and mortality

among the children.57

c) Disease transmission hypothesis: Increase in the number of children in a

home will increase the likelihood of acquiring transmissible diseases,

which may lead to increased morbidity and mortality among children.69

d) Breast feeding hypothesis: shortened breast feeding periods, early

weaning, and less maternal support to the preceding child may lead to

increased morbidity and mortality in the preceding child.46,70

e) Preovulatory overripeness of the oocyte: It is hypothesised that females,

conceived in a condition of maternal endocrine irregularities, face an

increased risk of ovarian dysfunction through overripeness-induced

gonadal maldevelopment.71

Figure 4 depicts the intervals where some of the postulated mechanisms may act.

Biological determinants which include prematurity, intrauterine growth retardation,

and impaired lactation, are more likely to influence the preceding interval while

behavioural factors are more likely to influence the subsequent interval.

18

Preceding pregnancy

Subsequent pregnancy

Index child Preceding interval Subsequent interval

• Biological determinants

• Maternal depletion

• Sibling competition for resources

• Disease transmission

• Sibling competition for resources

• Less maternal care

• Disease transmission

Figure 4. Postulated mechanisms for the association of short birth

intervals and poor pregnancy outcomes and the birth intervals they may

influence.

The evidence for maternal depletion and maternal nutritional status on fertility

and reproductive outcomes is inconclusive.33,39,60,61,72 Short interpregnancy intervals

may have an impact on the nutritional status of the mother and on the mechanical

properties on hysterotomy wound healing. Although the time required for peak

myometrial strength to return after hysterotomy is not well known, a short inter-

delivery interval, less than 12 months, is an important risk factor for uterine rupture or

scar dehiscence in the subsequent pregnancy.63,64

Studies show the increased risk of preceding child mortality, support the sibling

competition hypothesis. However, some studies show that the risk of death of the

index child was still high despite previous child death with death clustering observed

in families which may support the disease transmission hypothesis.47,73,74

Overripeness ovopathy may be supported in part by the increased risk of

menstrual disturbances among daughters born to women conceived after the short

interpregnancy interval, increased risk of spontaneous abortion and increased adverse

pregnancy outcomes following short intervals.25,75-77

1.4. Short birth interval as an outcome

One suggested hypothesis explaining the association between birth intervals and poor

perinatal health is the selection hypothesis. The selection hypothesis suggests that

“the increasing health risks observed among foetuses conceived shortly after the

preceding birth are not attributable to short spacing per se, but to the over

representation of this group of women, who are at high risk of bearing unhealthy

infants for reasons other than short conception intervals”.68

Thus, a given maternal characteristic/variable should be strongly associated with

both short birth intervals and adverse perinatal outcomes. Some studies have shown

that previous reproductive loss, maternal age, race and maternal social status are

important predictors of both short birth intervals and adverse perinatal

outcome.7,22,37,40,68 Similarly, one study has shown that there is a tendency to repeat

19

gestational age and birth weight in successive pregnancies even after controlling for

the interpregnancy interval, foetal loss, and onset of labour.78 However, the

contributions of these factors to “causing” short birth intervals or adverse pregnancy

outcomes differ from population to population. For example, Park79developed a two

tier causal model, which shows that birth order, survival status of the preceding child,

and short birth intervals had direct roles in increasing infant mortality risk. The

effects of maternal age, maternal education, and birth order were mediated through

the preceding birth interval.

1.5. Social-economic status and birth interval

An important risk factor for short birth intervals is the social-economic status of the

mother. Social class was measured either as a unitary concept with all indicators

measuring the same concept or multidimensional with different social class indicators

measuring multiple aspects of social class. Social economic status (SES) has been

defined as “a composite measure that typically incorporates economic status,

measured by income, social status, measured by education and work status measured

by occupation.”80 Measures of social class can be grouped according to each of the

three single indicators commonly used (occupation, education, and income) or can be

used as composite index measures. Reviews on social class, socio-economic status or

socio-economic position in epidemiological and public health literature show that

many studies include some measure of social class.81,82 SES was considered either as

a confounder, a risk factor, or a descriptive variable of the study sample. Education

was the most frequently used measure followed by occupation, and composite

measures.82,83 Of all the SES measures, education has the advantage of being the

simplest to collect, is stable over ones lifetime, reasonably accurate, and is associated

with lifestyle characteristics and health behaviour. Unfortunately, the bulk of

research on defining and measuring SES derives from studies in developed countries.

Some of the measures emphasise occupational standing, which is difficult to

determine in developing countries, as most people are self-employed in agriculture or

unemployed. A number of studies have developed amenity scores used for their

specific studies but these are usually population and study specific.84,85

Despite different measures of SES, the role of SES in increasing the risk of short

birth intervals differs among the different populations studied. Most studies show a

negative association between SES and adverse pregnancy outcome and short birth

intervals, but the association between SES and birth intervals ranges from negative to

no association.6,22,23,35,86-88

20

2. Aims of the thesis

Given the large variation in risk factors associated with short interpregnancy intervals

among populations, the overall aim of this thesis was to identify which women are

more likely to have short birth intervals, under which circumstances, and for what

reasons.

The thesis is based upon two study populations; one from rural Uganda and the

other from Denmark.

The specific aims of the thesis are:

a) To determine risk factors associated with short interpregnancy intervals in a

rural population in Uganda.

b) To determine short interpregnancy interval and other risk factors associated

with child mortality in a rural Ugandan population.

c) To determine risk factors associated with short interpregnancy intervals in

Denmark.

21

3. Materials and Methods

3.1. Uganda

3.1.1. Country description

Uganda is a landlocked country in East Africa, and covers an area of 241,039 square

kilometres. Administratively, Uganda is currently divided into 56 districts, each of

which is further divided into counties, subcounties, parishes, and villages. The

majority (86%) of the estimated 20.9 million people live in rural areas. Uganda’s

average population density is 105 persons per sq km, but this varies widely from 38 to

301 persons per sq. km among rural districts. The average population growth rate is

2.5% per year, but this also varies from –0.93 to 5.9 in the different districts.89

Uganda is one of the poorest nations in the world and is ranked 141st on the

human development index. This index, developed by UNDP, considers three basic

dimensions of human development: life expectancy, adult literacy and per capita gross

domestic product.90 In 1999, about 37% of the population lived on less than a dollar a

day, the annual per capita gross national product (GNP) is estimated at 320 US $ and

35% of the adults were illiterate.91

Participatory poverty assessment studies in Uganda show that poverty is dynamic

and changes over time. The most frequent causes of poverty include poor health,

excessive alcohol consumption, large families, lack of education and employable

skills, and limited access to financial services, capital or markets.92 Table 4 shows

some socio demographic and health indicators for Uganda with high child and

maternal mortality, high fertility, low life expectancy, and poor access to health

services.

22

Table 4. Selected demographic, socio-economic and health indicators for Uganda

Indicator Uganda

Demographic

Projected total population, 1999 (millions) 21.6

Annual population growth rate(1980-1991) 2.5

Population density (per sq. km) 105

Percent female 15-49 years (1991) 43.5

Percent children 0-14 years (1991) 47.3

Male life expectancy at birth (years) 1999 43

Female life expectancy at birth (years) 1999 42

Mortality (1995)

Infant mortality rate (per 1000 live births) 97

Under five mortality rate (per 1000 live births) 147

Maternal mortality ratio (per 100 000 live births) 506

Socio-economic

Per Capita GNP US $ (1998) 320

Prevalence of child malnutrition (1995) 26

Radio (per 1000 people) 123

Percent of male adults literate (1997) 73

Percent of female adults literate (1997) 47

Reproductive health

Total fertility rate per woman (2000) 6.9

Contraceptive prevalence rate (2000) 20

Percent births within 24 months of a previous birth(1995) 27.8

Percent deliveries supervised by skilled personnel (2000) 38

Access to health services (1997)

Population: doctor ratio 20,228

Population: midwife ratio 7,431

Percent population with access to health units 49

Sources.89-91,93-95

In trying to address some of these problems, the Ugandan government over the

last five years introduced universal free primary education for children to improve

literacy rates, designed a poverty eradication action plan to address some causes of

poverty, and developed a health policy and strategic plan with the aim of reducing

morbidity and fertility.95,96 The 2001-2005 health sector strategic plan sets out to

deliver services through implementing a minimum health care package that addresses

the major causes of morbidity and mortality in Uganda and strengthening the health

23

support systems mainly through decentralisation of the health services.94,97

3.1.2. Study area

Bunyole health sub-district in Tororo district is about 260 km east of the Ugandan

capital, Kampala. It is administratively divided into six subcounties and 21 parishes.

The estimated size of Bunyole is 644 sq. km and the estimated population in 1999 was

134,000. The population density varies between the six subcounties (Map1).

Subsistence agriculture is the main economic activity for 83% of the working

population. Food crops grown include cereals and legumes, while coffee and cotton

are grown as cash crops. Rice is grown in some of the reclaimed swamp areas. Non-

commercial livestock management is the second most important activity in Bunyole

and cows, goats, sheep, and poultry are reared as a source of income, as food, and for

prestige. The average family land holdings are about 2.2 hectares.98

Bunyole has one hospital and six health units. In 1999, the staff establishment

that were directly involved in maternal health services included three doctors, 13

midwives (nine of who work at the hospital), 71 trained traditional birth attendants, 30

Community Reproductive Health Workers and many other untrained traditional birth

attendants. The district health team is involved in training more traditional birth

attendants and community health workers.

3.1.3. Study design

Figure 5 presents the study design which included two phases: an initial cross-

sectional survey used to identify all pregnant women and a follow-up study to

determine the factors associated with short interpregnancy intervals.

The sample size was calculated using the largest estimated sample for the

smallest outcome measure for the study, a relative risk worth detecting of 2, a 1:2

ratio of the exposed to unexposed, and a study power of 80% and 95% confidence

level, and allowing for a 10% dropout from the study. The required sample size was

700 mothers with short interpregnancy intervals (exposure) and a control group of

1400 unexposed mothers.99

24

Figure 5. Design for recruitment and follow up of mothers in Bunyole health

sub district, Uganda 1999

Cross-sectional study

The cross-sectional survey was carried out in April and May 1999 in all parishes in

the project area. Data were collected through house-to-house interviews on a

predetermined date in all parishes. All pregnant multigravidae and those who had

delivered during the data collection month and were resident in the survey area were

eligible for recruitment. Twenty local female health workers were trained for one

week to collect the data using a pretested questionnaire. Training included translating

the questionnaire into the local language, back-translation of the questionnaire to

English, and using a gestation age calculator designed for the study. The

questionnaire was designed to collect information on child mortality using the

preceding birth technique.100-102 The date and outcome of the last pregnancy, the

status of the child if born alive, and the date of death, had the child died, were also

obtained. The interviewers used a field manual to maintain consistency of data

collection, and a birth interval calculator to reduce arithmetic errors (Appendix 4:

Letter 1). Two supervisors monitored the data collection process by spot checks and

monitoring the interviews.

Cross -sectional Survey

Pregnant women identified (3708)

Eligible (para2+) and willing to participate (3253)

Short Interpregnancy Interval <24 months

746 (31%)

B. Follow up study

Follow up Bimonthly interviews for child morbidity and mortality Neonatal and delivery information at time of delivery or soon after

Child anthropometry weight at visits

Recording Assessments of the maternal and child health outcomes

Termination of follow up Death of mother Withdrawal of participation Loss to followup End of study (after 2years)

Not interviewed (617)

Incomplete data (228)

Prime gravida (455)

Exclusion Criteria

Short Interpregnancy Interval ? 24 months

1662 (69%)

25

Cohort study

From June to December 1999, six teams, each comprising of a midwife and

community health worker, identified eligible mothers from the cross-sectional study.

Each team was assigned to specific subcounties and interviewed participants either at

home or at antenatal sessions held at the six health units in the study area. The

baseline questionnaire, which was administered in the initial three months of the

study, included core questions on fertility, socio-economic status, and maternal

history also used in the DHS.

Follow-up interviews were carried out at the participants’ homes. Mothers and

traditional birth attendants who attended to them at delivery were asked to inform the

research team on delivery, so that the first postnatal interview could be carried out

soon after delivery. Follow up interviews collected information on the delivery and

current status of the mother and child.

3.1.4. Outcome measures

The outcome variable was the preceding interpregnancy interval. The preceding

interpregnancy interval was defined as the period from date of birth of the previous

pregnancy, irrespective of the outcome, to the date of birth of the recently concluded

pregnancy. For comparability to the DHS and other studies in developing countries,

an interval less than 24 months was considered as a short interval. While this

corresponds to a 15-month birth-to-conception interval, birth-to-conception interval

estimation is unreliable in developing countries because most mothers are either

unsure of their menstrual dates, may have lactational amenorrhoea, or the gestational

age is not usually known.

Child mortality was defined as the proportion of children born alive who were

dead at the time of interview.

3.1.5. Risk factors

The risk factors studied were maternal age, parity, marital status, education, socio-

economic status measured by the domestic animal score and the amenity score,

husband’s education, lactation patterns, previous pregnancy outcome, antenatal

attendance, residence, and ethnic group.

3.1.6. Ethical considerations

In Uganda, ethical clearance was obtained from the Uganda National Council of

Science and Technology. District health authorities gave permission to carry out the

study in their area, and mothers gave their informed consent to participate in the

study. Sick mothers were referred to hospital for treatment, and mothers could stop

participating in the study when they so desired.

3.2. Denmark

To study the determinants of short interpregnancy intervals in Denmark, data from a

population-based study that studied the social risk factors and pregnancy outcomes

were used.

26

3.2.1. Data sources

The data used were collected in a population based cohort study of pregnant women

in two counties in Denmark who were recruited for a health behaviour community

trial.103 The study obtained information about potential risk factors for adverse

pregnancy outcome from pregnant women residing in two cities in Denmark. From

April 1984 to April 1987, a self-administered questionnaire about the mothers current

lifestyle was given to all pregnant women within the geographical area of Odense and

Aalborg cities attending a routine antenatal visit in the 36th gestational week. In 1987,

Odense and Aalborg had a total population of 250,036.104 More than 95% of

pregnant women (n=13,815) received the questionnaire, and 11,980 (87%) completed

the questionnaire. Obstetric information was extracted from the medical birth records

after delivery.

Since 1968, all residents in Denmark have received a unique 10-digit central

personal registration number (CPR) from the Civil Registration System. All children

born alive in Denmark are given a CPR number which is linked to their mothers. We

obtained CPR numbers for the pregnant women in the cohort. Valid CPR numbers

were linked at the Civil Population Registry to obtain a complete list of children born

to these mothers. Figure 6 shows the linkage process used to obtain information

about the preceding birth and develop the cohort for the analysis. This linkage

process produced information about all previous live births that mothers had up to the

time of birth of the index child. Children born during the study period and matched

the date of birth in the questionnaire were identified as the “index” child. The date of

birth of the preceding child was also recorded. All key variables had less than 1 %

missing data except planned pregnancy, of which 5% of the responses was missing.

Mothers with twin pregnancy and those who could not be matched in the CPR register

were excluded from the study population. Of the 11,980 mothers, only 2904 mothers

met the inclusion criteria for the study.

27

Figure 6. Study design for social determinants of short interpregnancy intervals

in Denmark.

3.2.2. Outcome measure

The outcome variable was the preceding interpregnancy interval defined as the period

from the date of birth of the previous child to the estimated conception date of the

index child. The estimated conception date was determined by subtracting the

number of completed weeks of gestation from the date of birth of the index child. An

interpregnancy interval was considered short if it was less than nine months. This

cut-off point was used because a previous study in Denmark showed adverse perinatal

outcomes occurring in mothers with an interpregnancy interval of less than eight

months.22

3.2.3. Risk factors

Risk factors studied included maternal age, parity, maternal education, maternal social

status, husband’s education, housing, smoking and alcohol consumption in pregnancy.

3.2.4. Ethical considerations

The Regional Committee for Ethics and Science approved the initial study. The

National Board of Health, Denmark granted permission to use the data and link it to

the CPR registry. Record linkage procedure was carried out in accordance with

Mothers with valid information (11,155)

Respndents to questionnaires (11,980)

CPR registry Children ever born (22,260)

Primipara (5446)

Eligible Para1+ (5709)

Short Intervals ≤ 9 months 269 (4.7%)

Long Intervals >9months 5440 (95.3%)

Pregnant women identified (13,815)

Record linkage

Incorrect CPR (206) No CPR (342) Twins (14) Unlisted (133)

Exclusion criteria

28

guidelines set by the National Board of Health Denmark. The original data file with

mothers’ CPR was used. After linkage, each mother and children were given unique

identification numbers and all CPR numbers were extracted to a new separate

identification file. Analysis was performed using a new file without CPR numbers.

4. Statistical analysis

4.1. Overall methodology

Parametric and non-parametric analyses were used to compare the mean birth interval

between risk groups. Stratified analysis and Mantel-Haenszel tests were carried out to

control for confounding.105 The association between the explanatory risk factors and

the study outcomes was expressed as crude and adjusted odds ratios. Multivariate

logistic regression was used to control for confounding in the two datasets.

Continuous variables were recoded into categorical variables and categories

considered a priori as low risk were used as reference. Stata 6.0106 and PEPI 107were

used in the analyses.

Model building strategy

Full models included all variables considered a priori associated with the outcome of

interest and statistically significant at p <0.25 in the univariate analysis. Unimportant

variables were identified by examining the Wald statistic for each variable and

comparing the estimated coefficients in the model with the coefficient from the

univariate model. Unimportant variables were excluded, one by one, if the Wald

statistic was not significant at p < 0.05 level and new models fitted. The significance

of the variable to the model was determined by comparing the new model to the older

larger model using the log-likelihood ratio test. Confounding was also evaluated by

considering 10% or more change in estimate in the coefficient. Variables not initially

selected in univariate analysis were then added to the model, one at a time, and

retained if significant. Plausible interactions were identified, and included in the

model, one at a time, and tested for significance using the log-likelihood ratio test.

Regression diagnostic plots were drawn to examine for influential observations.

Influential covariate patterns were then excluded and models fitted without them and

any changes in the estimates noted. Overall assessment of fit of the final model was

carried out using Hosmer-Lemeshow statistic.108

4.2. Data management

4.2.1. Analysis of the Ugandan dataset

Double entry for the collected data was carried out using EPIIFO 6.04.99 Consistency

and missing data checks were carried out. In the cross-sectional data, we imputed

missing data on maternal age (n=43) using the mean age of the mothers of same

parity. Missing data on husband’s age was included as a separate category in the

analysis. Graphical assessment for digit preference for maternal age was carried out.

29

Child mortality was estimated as a proportion of deaths of children born alive. This

proportion of immediately preceding births dying before the second birth date

approximates well with the probability of dying before the second birthday on a

standard life table, if the birth interval is approximately 30 months. The study used the

preceding birth technique, a robust indirect method for estimating child mortality

where vital registration does not exist.100,101,109,110 Descriptive analysis for seasonal

variation of child mortality was done by plotting proportion of deaths and monthly

rainfall patterns for the two years before prior to the survey date. Freedman's non-

parametric test for seasonal variation was carried out.111

Figure 7 depicts the variables and modelling strategy used in studying the risk

factors for short interpregnancy interval. Of the distal determinants considered,

maternal age and parity potentially affect fecundity, marital status and social status

are factors that affect exposure to intercourse and cultural value for children. Maternal

age was categorised in five-year age groups and maternal education into four groups.

A 10-point weighted “amenity score” was developed based on the relative cost and

importance of the type of roofing material (3 points), and on ownership of a bicycle (3

points), radio (2 points), or a food storage facility (2 points). A “domestic animal

score” was based on the size of the cattle herd and presence of goats, pigs, sheep, and

poultry. A “contraceptive knowledge score” considered the number of contraceptive

methods spontaneously mentioned by respondents. All three scores and household

population size were grouped using tertiles. Antenatal care attendance and gestational

age at first visit were used as proxies for health service utilisation.

Figure 7. Strategy used for building models to study risk factors of

interpregnancy intervals in Bunyole , Uganda , 1999

The first model included previous pregnancy outcome and all distant determinant

Proximate determinants Distal determinants Available resources

Household amenities Domestic animals Husband’s education Household size

Maternal individual factors Age Education Marital status

Utilisation of health services Contraceptive knowledge Antenatal attendance

Previous child death

Abortion and stillbirth

Lactation

Interpregnancy interval

Model 1

Model 2

30

variables that were significant at p <0.25. The second model fitted excluded previous

pregnancy outcome to assess any direct association between the distal determinants

and short interpregnancy interval. A third model was fitted with selected

dichotomised distant determinants to determine their adjusted attributable risk fraction

to short interpregnancy intervals.112 A sensitivity analysis was performed by fitting

two models, one with all mothers, and another without mothers unsure of their

menstrual dates. This analysis did not change the estimates for maternal age, or

previous pregnancy outcome. Crude and adjusted odds ratios for the association

between lactation and short interpregnancy intervals were determined.

We compared previous pregnancy outcome, maternal age, and education

characteristics of the eligible mothers not interviewed to those mothers interviewed.

4.2.2. Analysis of the Danish dataset

Analysis of the Danish data was restricted to 2904 mothers, whose preceding

interpregnancy interval was less than 37 months. Univariate analysis and the crude

odds ratios (with 90% CI) were determined. Explanatory variables were grouped into

three major groups: social factors, fertility related factors, and lifestyle factors.

Stratified analysis was carried out in each of the three major groups to assess

confounding within the groups and to select all variables that showed statistical

association with short interpregnancy intervals. The full model included all variables

that were statistically significant in all groups and those considered a priori to be

associated with poor reproductive outcomes such as maternal education. A variable

with change in the OR estimate of more than 10% was considered in the reduced

model.113

31

5. Results

5.1. Short interpregnancy intervals in Uganda

The mean interpregnancy interval was 28 months. About 31% of the mothers had an

interpregnancy interval of less than 24 months and 10.6% of the mothers had an

interval less than 18 months. Mothers above 35 years of age, of high parity, of

secondary education level, and mothers whose preceding child was still alive had

longer mean birth intervals. Short intervals were more common among mothers of

the Nyole tribe, and those who breastfed for less than one year (Appendix 1: Table 1).

Of the mothers whose children died, 80% became pregnant after the death of the

child.

Table 2: Appendix 1, shows that factors strongly associated with short

interpregnancy intervals were adverse pregnancy outcome, previous child death, and

short lactation periods. Distal determinants strongly associated with short intervals

were adolescent pregnancy and high parity. Low maternal education was not

significantly associated with short interpregnancy intervals.

We explored the association between the distal determinants and lactation and

adverse pregnancy outcome. Uneducated mothers, and mothers of the Nyole tribe

were more likely to have short lactation periods. Uneducated mothers, low domestic

animal score, and young maternal age, was associated with child mortality.

Another cut-off point for very short intervals was used to determine risk factors

associated with very short interpregnancy intervals. Table 5 shows risk factors

associated with very short intervals of less than 18 months. Adverse pregnancy

outcome in the previous pregnancy and high parity were strongly associated with

short interpregnancy intervals. Mothers who had not started attending antenatal care

in the index pregnancy were more likely to have very short intervals. The low

domestic animal score was still associated with short intervals (OR 1.5, 95% CI: 0.9-

2.7) but this was not statistically significant in the multivariate analysis.

32

Table 5. The association of selected maternal characteristics (expressed as an odds ratio (OR and 95 % Confidence interval (CI)) and short

interpregnancy intervals less than 18 months in Bunyole (n= 2408).

Characteristic Short interpregnancy Interval Univariate analysis Full model Reduced Model

N <18 % OR 95% CI OR 95% CI OR 95% CI

Previous pregnancy outcome alive 2032 98 4.8 1 1 1 born alive and died 248 72 29.0 8.1 5.7-11.4 8.2 5.7-11.7 8.1 5.7-11.6 adverse pregnancy outcome 120 84 70.0 46.0 29.7-71.5 49.4 30.8-79.4 50.3 31.3-80.7

Current maternal age (in years) <19yrs 248 50 20.2 2.7 1.8-4.0 2.9 1.5-5.5 2.9 1.5-5.6 20-24 825 85 10.3 1.2 0.9-1.7 1.4 0.9-2.3 1.4 0.9-2.3 25-29 694 60 8.6 1 1 1 30-34 376 36 9.6 1.1 0.7-1.7 0.7 0.4-1.2 0.7 0.4-1.2 35-39 265 25 9.4 1.1 0.7-1.8 0.6 0.3-1.1 0.6 0.3-1.1

Parity 2-3 819 90 11.0 1 1 1 4-5 732 75 10.2 0.9 0.7-1.3 1.7 1.1-2.8 1.7 1.1-2.8 6 or more 857 91 10.6 1.0 0.7-1.3 2.6 1.4-4.6 2.6 1.4-4.6

Gestational age at first attendance 1-6months 718 72 10.0 1 1 1 7-9 months 853 80 9.4 0.9 0.7-1.3 1.1 0.8-1.7 1.1 0.8-1.7 not started yet 794 100 12.6 1.3 0.9-1.8 1.5 1.0-2.2 1.5 1.0-2.2

Maternal education none 808 91 11.3 1.5 0.8-2.6 1.2 0.6-2.4 1.2 0.6-2.4 primary 1-4 637 68 10.7 1.4 0.7-2.5 1.1 0.5-2.2 1.1 0.5-2.3 primary 5-7 769 82 10.7 1.4 0.8-2.5 1.0 0.5-2.0 1.0 0.5-2.0 secondary 174 14 8.0 1 1 1

Marital status married monogamous 1626 187 11.5 1 1 married polygamous 744 64 8.6 0.7 0.5-1.0 0.8 0.6-1.2 single 38 5 13.2 1.2 0.4-3.0 1.1 0.4-3.7

Pearson chi2(263) =250.53 p= 0.6996

Hosmer-Lemeshow chi2(8) =3.83 p= 0.8718

33

5.2. Short interpregnancy intervals in Denmark

The median interpregnancy interval was 37 months and only 4.8% of mothers had an

interval less than 9 months. Mothers with short intervals were slightly older (mean

26.5 ± 3.7 years) than mothers with longer intervals (mean 25.0 ±4.5 years). Other

than women aged 35-49, first parity mothers had consistently shorter intervals within

their age groups (Appendix 2: Table 1).

Table 3, Appendix 2, shows that the statistically significant risk factors associated

with short intervals were unplanned pregnancy, high parity, being unemployed, and

living in a flat. There was a significant trend in increasing risk for short birth

intervals with increasing maternal age, maternal smoking (Appendix 2: Figure 2).

Table 6 shows a broad overview of the results of the two studies. For

comparative purposes, Ugandan data presented consider interpregnancy intervals less

than 18 months, which is comparable to a 9-month birth-to-conception interval used

in the Danish dataset. Younger mothers in Uganda were more likely to have short

intervals while older mothers in Denmark were more likely to have short intervals.

While there is a significant inverse relationship between maternal education and short

interpregnancy intervals in Denmark, this is not evident from the Ugandan data. Poor

social status is associated with short interpregnancy intervals in both countries.

34

Table 6. Overall description and summary of key results for both studies from Denmark and Uganda.

Study Characteristics Denmark Uganda

Type of study Population based cohort with register linkage Population based cohort study Study period 1984-1987 1999 Study base All pregnant women in Odense and Aalborg

(n=11850) All pregnant women in Bunyole at given time (n= 3708)

Sampled population 85 % of all pregnant women in the study area 75% pregnant women Eligible respondents 5709 2623 Sample used in analysis 2904 2408 Outcome variable Preceding interpregnancy interval <9 months Preceding interpregnancy interval <18 month Proportion of mothers with birth interval < 9 months

4.8% 10.6%

Median interpregnancy interval 37 months 27 months Mean maternal age 26.5(s.d. 3.7) 25.1 (s.d. 6.5) High Parity >3 4.3% 66%

Risk factors associated with short

interpregnancy intervals

OR 95% CI OR 95% CI

Previous pregnancy outcome Abortion and stillbirth Child dead at time of interview

50.3 8.1

31.3 – 80.7 5.7- 11.6

Short lactation periods (mothers with children alive)

1.7 1.1-2.6

Menstrual irregularity 1.7 1.1-2.5 Unplanned pregnancy 2.9 2.2-3.9 High parity 2.6 1.4-4.6 Maternal age <25 0.7 0.4-1.3 2.3 1.5-3.4 Maternal age 31-49 1.7 1.1-3-1 0.7 0.6-1.0 Low/no maternal education 1.2 0.8-1.9 1.0 0.7-1.5 Heavy smoking 1.6 1.0 -2.5 - Living in flat 1.7 1.1-2.6 Amenity score 0.9 0.6-1.4 Domestic animal score 1.9 0.7-4.7

35

5.3. Child mortality in Uganda

Using the preceding birth technique, the child mortality rate was estimated at 108 per

1000 live births. Approximately 93% of child mortality occurred before the age of

two. The estimated annual under-two child mortality rate was 82 per 1000 child years

of risk (Appendix 3: Table 2). Deaths were three times more likely to occur in the

neonatal period than in the post neonatal period. Risk factors associated with

increased risk for child mortality were no maternal education, young maternal age,

and mothers above 35.

Uganda experienced increased and prolonged rains caused by the El Niño

Southern Oscillation weather pattern in 1997 and 1998. There was a seasonal

correlation of child death, which closely followed the El Niño weather pattern

(Appendix 3: Figure1). There was a one-month time lag between high monthly

rainfall and child mortality. A high proportion of child deaths also followed the

unusual peak rainfall in November and December 1997. Although there were few

recorded deaths in 1997 (n=47), Table 7 presents results of parametric and non-

parametric tests for seasonal variation107 and shows that the peak child mortality

occurs between June and August both in both 1997 and 1998. This corresponds to the

dry season that occurs immediately after the peak rainy season in Tororo.98

Mortality levels were different at parish level. Figure 8 shows higher mortality

occurring in parishes that had the lowest population density, and those in the

Southwest of the county.

36

Table 7. Results of parametric and non-parametric tests for seasonal variation of child mortality among in children born alive in Bunyole, Uganda.

Number of child deaths

Months 1997 1998

January 2 12 February 1 6 March 1 11 April 1 27 May 4 18 June 8 19 July 6 22 August 5 25 September 3 26 October 6 16 November 2 18 December 8 16 Total 47 216

Tests for seasonal variation Peak Months p Peak Months p Edward’s test August 0.05* August 0.001 Freedman’s test deviation from uniform incidence < 0.05 0.01 Hewitt’s rank-sum test: 4 month peak May-August > 0.09 June-September >0.09 Hewitt’s rank-sum test: 6 month peak May-October > 0.13 April-September 0.03 * Conservative p values (n<50)

37

Figure 8. Map of study area showing mortality and population density differences in Bunyole.

Sources: 1991 Uganda Population and Housing census . MoFEP 1995. Kaharuza FM Short interpegnancy intervals and child survivial 2001

Child Mortality (per 1000 livebirths)<8081-100101-120121-140141+Missing

Population density (persons/sq. km)<100101-150151-200201-250251 +

3 0 3 6 9 12 15 Kilometers

N

ÆP

×

Ñ

Ñ

Ñ

Ñ

Ñ

38

6. Discussion

6.1. Main results

In Uganda, interpregnancy intervals, less than 24 months, occurred among one third

of the mothers. Very short intervals, less than 18 months, occurred in 10.6 % of the

sampled population. Risk factors associated with short interpregnancy intervals were

adverse pregnancy outcomes, previous child death, short lactation periods, young

maternal age, and low social status. Maternal education was not associated with short

interpregnancy intervals. The under-2 child mortality is still high and may be

considered an important cause of short interpregnancy intervals.

Child replacement is probably the most important determinant of short intervals

in this community. The findings that 80% of mothers whose children died became

pregnant after the death of the child, and that mothers whose children died after one

year of age were more likely to have longer intervals than mothers with living

children, imply that child replacement may be the most important determinant of short

intervals in this rural population.

In the Danish cohort, only 4.8% of the mothers had short birth-to-conception

interval of less than nine months, which is comparable to an 18-month interpregnancy

interval if the subsequent delivery is a full term delivery. Short interpregnancy

intervals were more likely to occur among mothers who did not plan their pregnancy,

mothers with irregular menstruation, and in older mothers. Low social status, living

in a flat, poor education and heavy smoking were all significant social factors

associated with short interpregnancy interval.

6.2. Source of bias

6.2.1. Selection bias

In the Ugandan study, approximately 75% of the households in the geographical area

were visited and 19% of all women were pregnant at the time of the visit. This is

similar to the estimated prevalence of pregnant women in a given time in a developing

country.114 Previous pregnancy outcomes, age, and education of the eligible mothers

not interviewed were similar to the study participants. Withdrawals and loss to

follow-up was reduced because mothers were actively identified at their homes by the

local teams. By December 1999, only 58 mothers who were lost to follow-up had

either divorced or migrated from the study area. Crude estimates for previous

pregnancy outcome, maternal education, and social status score for mothers

interviewed were similar to estimates obtained with 2408 mothers, who had all data

available included in the multivariable analysis.

By selecting only pregnant mothers, we excluded mothers that may have had long

intervals because they are subfecund or are using a postpartum contraceptive method.

39

Although this would seem to overrepresent mothers with short intervals, the

prevalence of short intervals between live births in this study was similar to the 1995

Demographic and Health Survey which includes include both pregnant and non-

pregnant respondents.93 We therefore believe that selection bias was minimised in

this study and the results are largely representative for this population.

In the Danish cohort information was obtained from 87% of all eligible women in

a well-defined geographical area. For less than 6% of the mothers information could

not be linked to the CPR register and these were excluded from the analysis. We

restricted our analysis to mothers with interpregnancy intervals less than 37 months.

This was done because information was collected at the time of the second birth of the

pair and the longer the interval, the higher the probability that the social determinants

under study may have changed. This also reduces the bias introduced by other factors

that potentially modify and increase interpregnancy intervals such as subfecundity and

postpartum contraceptive use.

6.2.2. Information bias

In the Ugandan study, the reporting age, birth dates, and pregnancy outcomes are

subject to recall bias. In maternity history data collection, mothers may not accurately

report the dates of birth of the preceding child or other pregnancy outcomes. This was

minimised by asking about the recently ended pregnancy and the current pregnancy,

and by training interviewers to specifically ask about abortion and stillbirths.

Furthermore, sensitivity analysis showed minor differences in the risk estimates for

previous pregnancy outcome, maternal age and parity. The respondents' age structure

was similar to that in the DHS study.

Lactation data were imprecise and showed digital preference for 12, 18, and 24

months. Self-reported lactation periods are usually overestimated, given that mothers

may want to tell you what you would like to hear. This may lead to non-differential

misclassification of lactation patterns. However, this information was adequate for

this study, given that we dichotomised lactation to a short period less than 12 months,

and long periods. Analysis of the lactation patterns was restricted to mothers with

living children, since mothers stop breastfeeding when their children die.

Two limitations of the study were that information on pregnancy planning was

not collected, and data on use of contraception before the current pregnancy were not

used. Pregnant women are not expected to give you accurate information about

planning of the current pregnancy and the data would thus be subject to recall bias.

Information on contraceptive use before index pregnancy suggests that 7% of mothers

used a method of contraception. However, withdrawal, abstinence, and barrier

methods were the most commonly used methods, which are not very effective.

A common problem that exists in many studies is the inability to distinguish the

temporal relationship between child mortality and short birth intervals. By

subtracting 280 days or 9 months from the date of birth, we estimated the conception

date and compared this date to the date of death of the preceding child. This study

40

shows that short intervals usually occur after the death of the previous child.

In the Danish dataset, information on contraceptive use, breastfeeding patterns,

and spontaneous or induced abortion was not collected. Mothers with spontaneous

abortion or induced abortion are more likely to be erroneously classified as having

long intervals.

6.2.3. Confounding

Stratified analysis indicated that the association between parity and short

interpregnancy interval was confounded by age in both datasets. This is expected

given that fecundity is affected by age and parity. Lactation pattern analyses were

restricted to mothers with living children. In the Ugandan data set, estimates from

univariate analysis and the full models show that the association between previous

pregnancy outcome and short intepregnancy intervals is not confounded by social

factors studied.

In the Danish dataset, although stratified analysis among identified risk factor

groups did not show confounding among the groups, crude and adjusted risk estimates

in the final multivariate model indicates some confounding from the social risk

factors.

6.3. Main findings compared to other studies

6.3.1. Prevalence of short birth intervals

The prevalence of short interpregnancy intervals in this study is comparable to other

studies in sub-Saharan Africa.7,43,44,86,115 A more recent review of DHS data from 20

countries shows that on average, 17 % of birth intervals are less than 24 months, and

this prevalence varies from 9% to 27%.116 Birth interval trend analysis shows that

birth intervals have been increasing over time but this increase is small in Uganda,

about 2 months from 1970s to 1990s.117

The DHS data also includes information on mothers preferred birth intervals.

Although there are many limitations in measurement of preferred birth intervals from

DHS data, it has been shown that most women in many countries prefer to have

longer intervals than the actual interval they had. If the women’s preferred birth

intervals prevailed, the prevalence of interpregnancy intervals less than two years

would decline from the average 17% to 11%. On average, this decrease would

translate into a 7% decrease in child mortality.117 From the point of view of the

potential reduction in neonatal and infant mortality, the prevalence of short

interpregnancy intervals should be reduced.

Prevalence studies from developed countries are difficult to compare with

developing countries given that the different cut-off points are used, when studying

different adverse pregnancy outcomes. This emphasises that there lacks a definition

of an "optimum interval"4 below which adverse pregnancy outcomes may occur. One

reason is that the definition of this "optimum" interval is "population-specific". What

is clear however, is that most interpregnancy intervals below 12 months are associated

41

with an increased risk of adverse pregnancy outcomes. Different populations need to

determine a birth interval below which adverse pregnancy outcomes are most likely to

occur.

6.3.2. Determinants of short birth intervals

Child mortality and adverse pregnancy outcome

Previous child loss and adverse pregnancy outcome are strongly associated with short

interpregnancy intervals. It is usually difficult to sort out the direction of causation

between pregnancy intervals and infant mortality.118 Efforts have been made to

determine whether increased fertility after child mortality is merely a replacement

effect or whether additional parental concern may lead to extra insurance births.119

Child replacement strategy has been observed in Senegal47 and some of the results of

this study are consistent with the child replacement strategy.

Lactation

The greater fertility following child mortality has both biological-lactational and

behavioural components and the relative balance probably depends on the local

situational determinants. This relationship is difficult to determine when you consider

that lactation is not purely biological since its duration is itself behaviourally

determined.12,119 Early child death would remove the protective effect of lactation.

Lactation durations are behaviourally determined as mothers may opt to breastfeed for

shorter periods. This leads to fertility differences amongst populations.17 The fact

that 30 % of mothers with living children breastfed for less than a year, and Nyole

mothers were more likely to have short breastfeeding patterns implies that lactation is

behaviourally determined in this population.

Maternal age and parity

The association between maternal age and parity in the Ugandan study is consistent

with other studies that show young mothers in developing countries are more likely to

have short interpregnancy intervals. It has been shown that recovery of ovarian

function was faster among young mothers than older mothers and may be an

important determinant of short birth intervals among young mothers.15 However, in

Denmark, where the mean age at first birth is 26.1 years, older mothers were more

likely to have a shorter interval so that they can quickly complete their family size.

Thus, changing the maternal age at first birth will change the age risk profile of

mothers having short interpregnancy intervals.

Socio-economic risk factors

The observed relationship between maternal education and fertility is not uniform

amongst developing countries.117,120 Most studies show that low maternal education

is significantly associated with short intervals, while other studies show no

relationship with maternal education and short birth intervals. Comparative studies

show that maternal education in Uganda is not associated with short interpregnancy

intervals.121 This study shows that low maternal education is not associated with an

42

increased risk of short interpregnancy intervals.

Studies from both developing and developed countries consistently show the

association between low social status and short interpregnancy intervals.86-88 Social

status indicators are difficult to define in developing countries. In an anthropological

study in a predominantly pastoral society in Kenya, livestock wealth was correlated

with reproductive success, measured by the number of children. This reproductive

success may be a result of early marriages and polygamy practiced by the wealthier

respondents.122 Although short interpregnancy intervals were not associated with

short birth intervals, mothers in polygamous union were more likely to be of high

parity and to have a high domestic animal score than those in monogamous

marriages. Of the two socio-economic scores used in the present study, low domestic

animal score is consistently associated with an increased risk of short interpregnancy

intervals. This relationship between the two SES scores and short intervals seem to act

in different proximate determinants. Low animal score seems to act through reduced

child survival while the low amenity score seems to act through short lactation

periods.

6.3.3. Determinants of child mortality

Study findings on determinants of child mortality in Uganda are consistent with

earlier studies on child mortality.52,123,124The study also shows the importance of

parental literacy, especially husband education, and is consistent with other studies

that have considered husband education as a risk factor independent of SES.125

The preceding technique has been used in community and antenatal

settings126,127and is a robust method that can be used by health management teams to

monitor child mortality trends. A study on the effects of length of birth interval on

infant and child mortality, concluded that short preceding birth intervals are

associated with a 58% higher risk of dying before the age of 5 years, while long birth

intervals are associated with a 28% lower risk of dying, compared with intervals 24 to

27 months in length.128

The seasonal patterns of child mortality with El Niño weather phenomenon is

consistent with other studies that show seasonal differences in morbidity and

mortality in other areas of the country.129,130

43

7. Conclusions

The results of these studies on short interpregnancy intervals show that in this rural

Ugandan population, child replacement is probably the predominant strategy used by

mothers. Short lactation patterns seen in this community are also an important

proximate determinant of short interpregnancy interval. Interventions for improving

obstetric outcome and child survival would reduce the occurrence of short

interpregnancy intervals.

In Denmark, strong predictors for short interpregnancy intervals are mainly

biological - menstrual irregularity and unplanned pregnancy. Thus most short

intervals are unplanned and women, especially toward the end of their reproductive

career may consciously choose to have short intervals.

Although the level of maternal education is not strongly associated with short

interpregnancy interval, it is strongly associated with child mortality. Maternal

education may indirectly lengthen birth intervals interpregnancy through improving

child survival, postponing the age at first birth, and increasing contraceptive use.

Regardless of the definition of social status, both studies show that low social

status is strongly associated with short intervals; this implies that short interpregnancy

interval is indeed a marker of women who may have a poor reproductive outcome.

The preceding births technique offers an opportunity to study mortality trends in

the age group usually targeted for child survival programmes. It would also give

information for planning health service delivery at community level.

44

8. Perspectives

One of the fundamental and important population policy guidelines is that individuals

and couples should be enabled to realise their reproductive intentions and preferences,

that is, to have the number and spacing of children that they desire. Given the

emphasis on promotion of family planning for purposes of birth spacing, the results of

this research are directly relevant to public health and population policy for it is useful

to know the characteristics of mothers with short interpregnancy intervals or who

prefer short interpregnancy intervals. It has been shown that fertility preferences

differ among different communities.131 Although education seems to have a role to

play in most countries, it seems to have a limited role in Uganda. It is therefore

important to determine other factors that mask the effect of maternal education on

short interpregnancy intervals.

Assessing birth intervals allows one to generate information about intentional

birth spacing, rates of adverse pregnancy outcome, and information that helps identify

patterns of high-risk maternal groups. The preceding birth technique is robust and

flexible tool for this purpose. It can be applied to routinely collected information

during antenatal sessions or during community surveys that interview reproductive

age women such as nutrition surveys or immunisation surveys. Assessing birth

intervals should be introduced as an indicator for fertility and mortality trend research.

Using the preceding birth technique in routine data collection allows optimum use of

routine data collected rather than waiting for expensive cross-sectional surveys to

determine trends in child mortality.

45

9. Summary

Background

Short interpregnancy intervals still constitute a major reproductive health problem in

both developed and developing countries. The distribution of, reasons for, and effects

of short interpregnancy intervals vary considerably amongst populations. Given the

large variation of risk factors associated with short interpregnancy intervals, it is

important to know and compare the specific factors that determine, and outcomes that

follow short interpregnancy intervals in any given population. The present studies

aimed at identifying key determinants of short interpregnancy intervals among women

from two populations and beyond short interpregnancy intervals, identify other

determinants of child mortality in rural Uganda.

This thesis is based on three papers and a brief communication on two

population-base cohorts; one from Uganda and another from Denmark.

Methods

The Ugandan study was carried out in two phases. An initial cross-sectional house-

to-house survey identified and interviewed all women in a geographically defined

rural area in Eastern Uganda pregnant at the time of the survey. The follow-up phase

was carried out with 2635 eligible pregnant women who were interviewed during

pregnancy, to determine maternal risk factors for short interpregnancy intervals, and

soon after delivery for information on delivery and pregnancy outcome.

In Denmark, determinants of short interpregnancy intervals were studied, in an

existing dataset, from a cohort of 2904 mothers with interpregnancy intervals less

than 37 months . These mothers were identified from a population-based cohort study

from a geographically two defined areas, Odense and Aalborg.

Results

In Uganda, 31.0% of mothers had short interpregnancy intervals less than 24 months

and 10.6 % mothers had intervals less than 18 months. Adverse pregnancy outcomes

(spontaneous abortion and stillbirth), previous child mortality and short lactation

periods were strongly associated with short interpregnancy intervals. Being a young

mother, or of high parity, or of low social economic status was associated with an

increased risk of having short interpregnancy intervals less than 24 months. Maternal

education was not associated with short interpregnancy intervals.

The preceding birth technique was used to estimate mortality rate among 2888

children born alive. The under-2 mortality rate was 108 per 1000 live births. Mothers

under 25 or above 35 years of age, and couples with no formal education were more

likely to have lost the preceding child. Child mortality was strongly correlated with

prevalent weather patterns.

In Denmark, 4.8% of the mothers had a short birth-to-conception interval

interpregnancy interval less than 9 months. Older mothers, mothers of low social

46

status, and low education, and those with irregular menstruation or those who had

planned their pregnancies were more likely to have short interpregnancy intervals

Conclusion

Child replacement plays an important role in determining short birth intervals in rural

Uganda while short intervals are by choice among women in Denmark. Regardless of

the definition of social status, both studies show that low social status is associated

with short interpregnancy intervals. Biological factors, such as irregular menstruation,

play a significant role in determining very short interpregnancy intervals.

The preceding births technique offers an opportunity to study trends in child

mortality in the under-2, an age group that is usually targeted for child survival

programmes.

47

10. Danske Resumé

Baggrund

Kort afstand mellem to fødsler er et sundhedsproblem for børn og mødre, som

forekommer både i industrialiserede lande og i udviklingslande. Hyppigheden af,

årsagerne til og konsekvenserne af kort afstand mellem to fødsler varierer fra land til

land. Da disse variationer kan skyldes, at der er forskellige risikofaktorer, der er

dominerer i det enkelte land, kan en kortlægning og sammenligning af forskellige

landes risikomønster give viden, der kan anvendes til at bedre børns og mødres

sundhed.

Formålet med det aktuelle studie er, i to forskellige populationer, at identificere

nogle centrale determinanter for kort afstand mellem fødslerne, samt at identificere

andre risikofaktorer, der ud over kort interval mellem fødslerne, er af betydning for

børnedødelighed i et landdistrikt i Uganda.

PhD-afhandlingen består af en oversigt samt fire videnskabelige artikler som alle

bygger på data fra to populationsbaserede kohorteundersøgelser fra henholdsvis

Uganda og Danmark.

Metode

I Uganda blev de kvinder, der indgik i studiet, kontaktet to gange. I det indledende

survey blev alle gravide i et geografisk afgrænset område spurgt om en række

risikofaktorer, og i et follow-up, der foregik efter fødslen, blev de samme 2.635

kvinder spurgt om fødslens forløb og om barnets tilstand.

I studiet i Danmark indgik 2.904 kvinder, som havde født to børn inden for en

periode af 37 måneder. Undersøgelsen baseredes på tidligere indsamlede data fra to

kohortestudier fra henholdsvis Odense og Aalborg

Resultater

Inden for en periode på 24 måneder havde 31% af kvinderne i kohorten fra Uganda

født to børn og 10,6% havde født to gange inden for 18 måneder. Kortvarigt

ammeforløb, dødsfald af et nyfødt barn samt et utilsigtet forløb af graviditeten ( f.eks.

spontan abort, dødsfødsel) var tæt associeret med kort afstand mellem fødslerne.

Andre risikofaktorer for kort interval mellem to fødsler var: ung alder, mange fødsler

og lav social status. Varigheden af moderens skoleuddannelse var derimod ikke en

risikofaktor.

De 2.888 mødre, der inden den aktuelle fødsel havde født et levende barn, indgik

i et studie, hvor formålet var at estimere børnedødeligheden. Det viste sig at 10, 8%

døde inden for de første to leveår. Øget risiko for død blev fundet blandt børn af

forældre uden uddannelse og blandt børn, hvis mødre var under 25 eller over 35 år.

Børnedødelighed varierede en del hen over året, og svingningerne viste sig at være tæt

korreleret til klimaet, idet dødeligheden var højest én måned efter regntiden.

48

I Danmark blev 4,8% af mødrene gravide igen, inden der var gået 9 måneder

efter fødslen. Lav social status, kort uddannelse, uregelmæssige menstruationer, og

planlagt graviditet var faktorer, der alle var associeret til kort afstand mellem

graviditeterne, desuden havde også ældre kvinder en øget risiko for at blive gravid to

gange inden for så kort en periode.

Konklusion

I landdistrikterne i Uganda tydede risikofaktorerne på, at et kort interval mellem

fødslerne var udløst af et ønske om at erstatte tabet af det forrige barn, medens det i

Danmark så ud til, at de, der fik to børn hurtigt efter hinanden, bevidst valgte dette.

Uanset hvordan social status blev defineret, så var lav social status i begge lande

associeret med kort afstand mellem to fødsler. I Danmark spillede en biologisk faktor

som uregelmæssige menstruationer også en signifikant rolle, for at kvinder hurtigt

blev gravide igen.

I udviklingslande bliver dødeligheden inden for de første to leveår hyppigt

anvendt som monitoreringsmål i børneprogrammer. Dette dødelighedsmål kan man

estimere ved i surveys at spørge om forløbet af den forrige graviditet.

49

11. References

1. Winikoff B. The effects of birth spacing on child and maternal health. Stud

Fam Plann 1983; 14:231-245.

2. Koenig MA, Phillips JF, Campbell OM, D'Souza S. Birth intervals and child

mortality in rural Bangladesh. Demography 1990; 27:251-265.

3. Miller JE. Is the relationship between birth intervals and perinatal mortality

spurious? Evidence from Hungary and Sweden. Popul studies 1989; 43:479-

495.

4. Klebanoff MA. The interval between pregnancies and the outcome of

subsequent births. N Engl J Med 1999; 340:643-644.

5. Shults RA. Effects of short interpregnancy intervals on small-for-gestational

age and preterm births. Epidemiology 1999; 10:250-254.

6. Fuentes-Afflick E, Hessol NA. Interpregnancy interval and the risk of

premature infants. Obstet Gynecol 2000; 95:383-390.

7. Boerma JT, Bicego GT. Preceding birth intervals and child survival: searching

for pathways of influence. Stud Fam Plann 1992; 23:243-256.

8. Zhu BP. Effect of the interval between pregnancies on perinatal outcomes. N

Engl J Med 1999; 340:589-594.

9. Wood JW. Dynamics of human reproduction: biology, biometry, demography.

New York: Walter de Gruyter, 1994.

10. Davis K, Blake J. Social structure and fertility: an analytical framework. Econ

Dev Cult Change 1956; 4:211-235.

11. Bongaarts J. A framework of analysing the proximate determinants of fertility.

Popul Develop Rev 1978; 4:105-132.

12. Howie PW, McNeilly AS. Effect of breastfeeding patterns on human birth

intervals. J Reprod Fertil 1982; 65:545-557.

13. Habicht JP, Da Vanzo J, Butz WP, Myers L. The contraceptive role of

breastfeeding. Popul studies 1985; 39:213-232.

14. Glasier A, Van Look PFA, von Hertzen H, Kennedy KI, Ayeni O, Pinol AP.

The World Health Organization Multinational Study of Breast-feeding and

50

Lactational Amenorrhea. II. Factors associated with the length of amenorrhea.

World Health Organization Task Force on Methods for the Natural Regulation

of Fertility. Fertil Steril 1998; 70:461-471.

15. Moran C, Alcazar LS, Carranza-Lira S, Merino G, Bailon R. Recovery of

ovarian function after childbirth, lactation and sexual activity with relation to

age of women. Contraception 1994; 50:401-407.

16. Tracer DP. Socioeconomic mediation of birth interval duration in a Papua

New Guinea population. Ann N Y Acad Sci 1994; 709:231-213.

17. Bongaarts J, Frank O, Lesthaeghe R. The proximate determinants of fertility in

Sub-Saharan Africa. Popul Develop Rev 1984; 10:511-537.

18. Benefo KD. The determinants of the duration of postpartum sexual abstinence

in West Africa: a multilevel analysis. Demography 1995; 32:139-157.

19. Brown MS. Coitus, the proximate determinant of conception: inter-country

variance in sub-Saharan Africa. J Biosoc Sci 2000; 32:145-159.

20. van de Walle E, van de Walle F. Birth spacing and abstinence in sub-Saharan

Africa. Int Fam Plann Perspect 1988; 14:25-26.

21. Mosley WH, Chen LC. An analytical framework for the study of child

survival in developing countries. Popul Develop Rev 1984; 10:25-48.

22. Basso O, Olsen J, Knudsen LB, Christensen K. Low birth weight and preterm

birth after short interpregnancy intervals. Am J Obstet Gynecol 1998; 178:259-

263.

23. Khoshnood B. Short interpregnancy intervals and the risk of adverse birth

outcomes among five racial/ethnic groups in the United States. Am J

Epidemiol 1998; 148:798-805.

24. Conde-Agudelo A, Belizan JM. Maternal morbidity and mortality associated

with interpregnancy interval: cross sectional study. BMJ 2000; 321:1255-

1259.

25. Wohlfahrt J, Andersen AM, Melbye M. Interval between pregnancies and risk

of spontaneous abortion. Epidemiology 2000; 11:92-93.

26. Crenn Hubert C, Bouyer J, Collin D, Menger I. Spontaneous abortion and

interpregnancy interval. Eur J Obstet Gynecol Reprod Biol 1986; 22:125-132.

27. Zhou W, Olsen J, Nielsen GL, Sabroe S. Risk of spontaneous abortion

51

following induced abortion is only increased with short interpregnancy

interval. J Obstet Gynaecol 2000; 20:49-54.

28. Kramer MS. Determinants of low birth weight: methodological assessment

and meta-analysis. Bull World Health Organ 1987; 65:663-737.

29. Brody DJ. Short interpregnancy interval: a risk factor for low birthweight. Am

J Perinatol 1987; 4:50-54.

30. Lang JM, Lieberman E, Ryan KJ, Monson RR. Interpregnancy interval and

risk of preterm labor. Am J Epidemiol 1990; 132:304-309.

31. Lieberman E, Lang JM, Ryan KJ, Monson RR, Schoenbaum SC. The

association of inter-pregnancy interval with small for gestational age births.

Obstet Gynecol 1989; 74:1-5.

32. Kallan JE. Effects of interpregnancy intervals on preterm birth, intrauterine

growth retardation, and fetal loss. Soc Biol 1992; 39:231-245.

33. Miller JE. Determinants of intrauterine growth retardation: evidence against

maternal depletion. J Biosoc Sci 1989; 21:235-243.

34. Klebanoff MA. Short interpregnancy interval and the risk of low birthweight.

Am J Public Health 1988; 78:667-670.

35. Rawlings JS. Prevalence of low birth weight and preterm delivery in relation

to the interval between pregnancies among white and black women. N Engl J

Med 1995; 332:69-74.

36. Ferraz EM, Gray RH, Fleming PL, Maia TM. Interpregnancy interval and low

birth weight: findings from a case-control study. Am J Epidemiol 1988;

128:1111-1116.

37. Spratley, E., Taffel, S. Interval between births: United States 1970-1977. US

Department of Health and Human Services;Data from the national vital

statistics system. Series 21 No. 39. 1981.

38. Mavalankar DV, Gray RH. Interpregnancy interval and risk of preterm labor.

Am J Epidemiol 1991; 133:958-959.

39. Cleland JG, Sathar ZA. The effect of birth spacing on childhood mortatlity in

Pakistan. Popul studies 1984; 38:401-418.

40. Hobcraft JN, McDonald JW, Rutstein SO. Demographic determinants of

infant and early child mortality: a comparative analysis. Popul studies 1985;

52

39:363-385.

41. Agha S. The determinants of infant mortality in Pakistan. Soc Sci Med 2000;

51:199-208.

42. Gubhaju BB. Effect of birth spacing on infant and child mortality in rural

Nepal. J Biosoc Sci 1986; 18:435-447.

43. Ikamari L. Birth intervals and child survival in Kenya. Afr J Health Sci 1998;

5:15-24.

44. Manda SOM. Birth intervals, breast feeding and determinants of childhood

mortality in Malawi. Soc Sci Med 1999; 48:301-312.

45. Miller JE, Trussell J, Pebley AR, Vaughan B. Birth spacing and child

mortality in Bangladesh and the Philippines. Demography 1992; 29:305-318.

46. Retherford RD, Choe MK, Thapa S, Gubhaju BB. To what extent does

breastfeeding explain birth-interval effects on early childhood mortality?

Demography 1989; 26:439-450.

47. Ronsmans C. Birth spacing and child survival in rural Senegal. Int J Epidemiol

1996; 25:989-997.

48. St George D, Everson PM, Stevenson JC, Tedrow L. Birth intervals and early

childhood mortality in a migrating Mennonite Community. Am J Human Biol

2000; 12:50-63.

49. Zenger E. Siblings' neonatal mortality risks and birth spacing in Bangladesh.

Demography 1993; 30:477-488.

50. Hill K, Pebley AR. Child mortality in the developing world. Popul Develop

Rev 1989; 15:657-687.

51. Ahmad OB, Lopez AD, Inoue M. The decline in child mortality : a

reappraisal. Bull World Health Organ 2000; 78:1175-1191.

52. Rutstein SO. Factors associated with trends in infant and child mortality in

developing countries during the 1990's. Bull World Health Organ 2000;

78:1256-1270.

53. Bohler E, Singey J, Bergstrom S. Subsequent pregnancy affects nutritional

status of previous child: a study from Bhutan. Acta Pædiatr 1995; 84:478-483.

54. Boerma JT, Vianen HAW. Birth interval, mortality and growth of children in a

rural area in Kenya. J Biosoc Sci 1984; 16:475-486.

53

55. Torfs CP, van den Berg B, Oechsii FW, Cummins S. Prenatal and perinatal

factors in the etiology of cerebral palsy. J Pediatr 1990; 116:615-619.

56. Martin EC. A study on the effect of birth interval on the development of 9-

year old schoolchildren in Singapore. J Trop Pediatr Environ Child Health

1979; 25:46-76.

57. Hobcraft JN, McDonald JW, Rutstein SO. Child spacing effects on infant and

early child mortality. Popul index 1983; 49:585-618.

58. Jelliffe, D. B. The assessment of nutritional status of the community. World

Health Organisation;WHO Monograph Series. Report No. 53. 1966.

59. Winkvist A, Rasmussen KM, Habicht JP. A new definition of maternal

depletion sydrome. Am J Public Health 1992; 82:691-694.

60. Miller JE, Huss-Ashmore R. Do reproductive patterns affect maternal

nutritional status? An analysis of maternal depletion in Lesotho. Am J Human

Biol 1989; 1:409-419.

61. Khan KS, Chein PFW, Khan NB. Nutritional stress of reproduction. Acta

Obstet Gynecol Scand 1998; 77:395-401.

62. Gurel SA, Gurel H. The evaluation of determinants of early postpartum low

mood: the importance of parity and interpregnancy interval. Eur J Obstet

Gynecol Reprod Biol 2000; 91:21-24.

63. Shipp TD, Zelop CM, Repke JT, Cohen A, Lieberman E. Interdelivery interval

and risk of symptomatic uterine rupture. Obstet Gynecol 2001; 97:175-177.

64. Esposito MA, Menihan CA, Malee MP. Association of interpregnancy interval

with uterine scar failure in labor: a case-control study. Am J Obstet Gynecol

2000; 183:1180-1183.

65. Ronsmans C, Campbell OM. Short birth intervals do not kill women:

Evidence from Matlab, Bangladesh. Stud Fam Plann 1998; 29:282-290.

66. Fortney JA, Zhang J. Maternal death and birth spacing. Stud Fam Plann 1998;

29:436.

67. Smits L, Zielhuis G, Jongbloet P, Bouchard G. The association of birth

interval, maternal age and season of birth with the fertility of daughters: a

retrospective cohort study based on family reconstitutions from nineteenth and

early twentieth century Quebec. Paediatr Perinat Epidemiol 1999; 13:408-

420.

54

68. Miller JE. Birth intervals and perinatal health: an investigation of three

hypotheses. Fam Plann Perspect 1991; 23:62-70.

69. Aaby P, Bukh J, Lisse IM, Smits AJ. Measles mortality, state of nutrition, and

family structure: A community study from Guinea-Bissau. J Infect Dis 1983;

147:693-701.

70. Bohler E, Bregstrom S. Child growth and weaning depends on whether mother

is pregnant again. J Trop Pediatr 1996; 42:104-109.

71. Smits LJ, Jongbloet PH, Zielhuis GA. Preovulatory overripeness of the oocyte

as a cause of ovarian dysfunction in the human female. Med Hypotheses 1995;

45:441-448.

72. Bongaarts J. Does malnutrition affect fecundity? A summary of evidence.

Science 1980; 208:564-569.

73. Madise NJ, Diamond I. Determinants of infant mortality in Malawi: an

analysis to control for death clustering within families. J Biosoc Sci 1995;

27:95-106.

74. Ikamari L. Sibling mortality correlation in Kenya. J Biosoc Sci 2000; 32:265-

278.

75. Basso O, Olsen J, Christensen K. Risk of preterm delivery, low birthweight

and growth retardation following spontaneous abortion: a registry-based study

in Denmark. Int J Epidemiol 1998; 27:642-646.

76. Smits LJ, Willemsen WN, Zielhuis GA, Jongbloet PH. Conditions at

conception and risk of menstrual disorders. Epidemiology 1997; 8:524-529.

77. Jongbloet P, Zielhuis G, Pasker de Jong PM. Interval between pregnancies and

risk of spontaneous abortion. Epidemiology 2000; 11:738-739.

78. Bakketeig LS, Hoffman HJ, Harley EE. The tendency to repeat gestational age

and birth weight in successive births. Am J Obstet Gynecol 1979; 135:1086-

1103.

79. Park CB. The place of child-spacing as a factor in infant mortality: a recursive

model. Am J Public Health 1986; 76:995-999.

80. Alder NE, Boyce T, Chesney MA, Cohen S, Folkman S, Kahn RL et al.

Socioecomic status and health. Am Psychol 1994; 49:15-24.

81. Krieger N, Williams DR, Moss NE. Measuring social class in US public

55

health research: concepts, methodologies, and guidelines. Annu Rev Public

Health 1997; 18:341-378.

82. Liberatos P, Link BG, Kelsey JL. The measurement of social class in

epidemiology. Epidemiol Rev 1988; 10:87-121.

83. Bollen, K. A., Glanville, J. L., Stetclov, G. Socioeconomic status and class in

studies of fertility and child health in developing countries. Chapel hill, North

Carolina: MEASURE Evaluation, Carolina Population Center USAID Grant

No.: HRN-A-00-97-0018-00. Sponsored by USAID. 1999.

84. Cortinovis I, Vella V, Ndiku J. Construction of a socio-economic index to

facilitate analysis of health data in developing countries. Soc Sci Med 1993;

36:1087-1097.

85. Abramson JH, Gofin R, Habib J, Pridan H, Gofin J. Indicators of social class.

A comparative appraisal of measures for use in epidemiological studies. Soc

Sci Med 1982; 16:1739-1746.

86. Abebe GM, Yohannis A. Birth interval and pregnancy outcome. East Afr Med

J 1996; 73:552-555.

87. Klerman LV, Cliver SP, Goldenberg RL. The impact of short interpregnancy

intervals on pregnancy outcomes in a low income population. Am J Public

Health 1998; 88:1182-1185.

88. Duncan J, Nangle B, Streeter N, Bloebaum L, Tingey DC, Olson JA. Risk

factors for short inter pregnancy interval- Utah, June 1996-June 1997. MMWR

1998; 47:930-934.

89. Statistics Department, Ministry of Finance and Economic Planning. The 1991

Population and Housing Census Analytical report. Vol1. Entebbe, Uganda:

Ministry of Finance and Economic Planning 1995.

90. UNDP(United Nations Development Programme). Human Development

Report 2001. New York: Oxford University Press, 2001.

91. UNDP(United Nations Development Programme). Human Development

Report 2000. New York: Oxford University Press, 2000.

92. Uganda Ministry of Finance, Planning and Economic Development. Learning

from the poor. Kampala: Ministry of Finance Planning and Economic

Development 2000.

93. Statistics Department [Uganda], Macro International Inc. Uganda

56

Demographic and Health Survey 1995. Calverton, Maryland, USA: Statistics

Department [Uganda] and Macro International Inc. 1996.

94. Hutchinson, P. Health care in Uganda: Selected issues. Washington: World

Bank Publications; 1999. World Bank discussion paper no. 404.

95. Uganda Ministry of Finance, Planning and Economic Development. Uganda

poverty status report, 1999. Kampala, Uganda: Ministry of Finance, Planning

and Economic Development 2000.

96. Uganda Ministry of Health. National health policy. Kampala: Ministry of

Health 1999.

97. Uganda Ministry of Health. Health sector strategic plan 2000/01-2004/05.

Kampala: Ministry of health 2001.

98. National Environment Management Authority. District environmental profile

report, Tororo. Kampala,Uganda: National Environment Management

Authority 1995.

99. Dean AG, Dean JA, Coulombier D, Brendel KA, Smith DC, Burton AH et al.

Epi Info Version 6: a word processing, database and statistical system.

Atlanta, GA: Centers for Disease Control. 1994.

100. David PH, Bisharat L, Hill AG. Measuring Childhood Mortality: A guide for

simple surveys. New York: UNICEF Regional Office of the Middle East and

North Africa, 1990.

101. Hill AG, Aguirre A. Childhood mortality estimates using the preceding birth

technique: Applications and extensions. Popul studies 1990; 44:317-340.

102. Bicego GT, Augustin A, Musgrave S, Allman J, Kelly P. Evaluation of a

simplified method for estimation of early childhood mortality in small

populations. Int J Epidemiol 1989; 18:S20-S32.

103. Olsen J, Frische G, Poulsen AO, Kirchheiner H. Changing smoking, drinking.

and eating behaviour among pregnant women in Denmark. Scand J Soc Med

1989; 17:277-280.

104. Danmark Statistik. Statistisk aarbog 1989 [Statistical yearbook]. 93.

Kobenhavn: Danmarks Statistik, 1989.

105. Mantel N, Haenszel W. Statistical aspects of the analysis of data from

retrospective studies of disease. J Natl Can Inst 1959; 22:719-748.

57

106. StataCorp. Stata statistical software: Release 6.0. College Station, Texas: Stata

Corporation. 1999.

107. Abramson JH, Gahlinger PM. Computer programs for Epidemiologists: PEPI

Version 3.01. dos/windows. Llanidoes, Powys, Wales: Brixton books. 1999.

108. Hosmer DW, Lemeshow S. Applied Logistic Regression.2 ed. NewYork: John

Wiley & Sons Inc, 2000.

109. Materia E, Mele A, Mehari W, Rosmini F, Stazi MA, Damen HM et al.

Estimation of early childhood mortality using preceding birth technique in a

community-based setting. Ann Ist Super Sanita 1993; 29:465-467.

110. Bairagi R, Shuaib M, Hill AG. Estimating childhood mortality trends from

routine data: a simulation using the preceding birth technique in Bangladesh.

Demography 1997; 34:411-420.

111. Freedman LS. The use of a Kolmogorov--Smirnov type statistic in testing

hypotheses about seasonal variation. J Epidemiol Community Health 1979;

33:223-228.

112. Greenland S, Drescher K. Maximum likelihood estimation of the attributable

fraction from logistic models. Biometrics 1993; 49:865-872.

113. Maldonado G, Greenland S. Simulation study of confounder-selection

strategies. Am J Epidemiol 1993; 138:923-926.

114. Vaughan JP, Morrow RH. Manual of epidemiology for district health

management. Geneva: World Health Organisation, 1989.

115. Oryem-Ebanyat, F. The patterns of birth interval in eastern Uganda.

Baltimore, MD, USA: John Hopkins University 1990.

116. Rafalimanana H, Westoff CF. Potential effects on fertility and child health and

survival of birth- spacing preferences in sub-Saharan Africa. Stud Fam Plann

2000; 31:99-110.

117. Rafalimanana H, Westoff CF. Gap between preferred and actual birth intervals

in sub-Saharan Africa. DHS Analytic studies 2. Calverton, Maryland: ORC

Macro, 2001.

118. Preston SH. Mortality in childhood: lessons from WFS. In: Cleland JG,

Hobcraft JN, (eds). Reproductive change in developing countries: Insights

from the World Fertility Survey. New York: Oxford University Press, 1985. p.

253-272.

58

119. Taylor CE, Newman JS, Kelly NU. The child survival hypothesis. Popul

studies 1976; 30:263-278.

120. Jain AK. The effect of female education on fertility: a simple explanation.

Demography 1981; 18:577-595.

121. Namkee A, Shariff A. A comparative study of socioeconomic and

demographic determinants of fertility in Togo and Uganda. Int Fam Plann

Perspect 1994; 20:14-22.

122. Cronk L. Wealth, status, and reproductive success among the Mukogodo of

Kenya. Am Anthropol 1991; 93:345-360.

123. Bicego GT, Boerma JT. Maternal education and child survival: a comparative

study of survey data from 17 countries. Soc Sci Med 1993; 36:1207-1227.

124. Cleland JG, Van Ginneken JK. Maternal education and child survival in

developing countries: the search for pathways of influence. Soc Sci Med 1988;

27:1357-1368.

125. Rao RS, Chakladar BK, Nair NS, Kutty PR, Acharya D, Bhat V et al.

Influence of parental literacy and socio-economic status on infant mortality.

Indian J Pediatr 1996; 63:795-800.

126. Woelk GB, Arrow J, Sanders DM, Loewenson R, Ubomba JP. Estimating

child mortality in Zimbabwe: results of a pilot study using the preceding births

technique. Cent Afr J Med 1993; 39:63-70.

127. Pickering H, David PH, Hill AG. Continuous monitoring of child mortality

from clinic records. J Trop Med Hyg 1989; 92:71-74.

128. Bicego GT, Ahmad OB. Infant and child mortality. DHS Comparative studies

No. 20. Claverton, Maryland: Macro International Inc, 1996.

129. Kilian AH, Langi P, Talisuna A, Kabagambe G. Rainfall pattern, El Nino and

malaria in Uganda. Trans R Soc Trop Med Hyg 1999; 93:22-23.

130. Lindblade KA, Walker ED, Onapa AW, Katungu J, Wilson ML. Highland

malaria in Uganda: prospective analysis of an epidemic associated with El

Nino. Trans R Soc Trop Med Hyg 1999; 93:480-487.

131. Blanc AK, Wolff B, Gage AJ, Ezeh AC, Neema S, Ssekamatte-Ssebuliba J.

Negotiating reproductive outcomes in Uganda. Calverton, Maryland: Macro

International Inc. and Institute of Statistics and Applied Economics [Uganda],

1996.

59

60

Appendices

Paper I Social determinants of short interpregnancy intervals in rural Uganda

Paper II: Choice and chance: Determinants of short interpregnancy intervals in

Denmark

Paper III Child mortality in rural Uganda

Accompanying letter: A simple birth interval calculator

Abstract [Separate page]