animal disease surveys

Upload: moldovan-nicolae-andrei

Post on 15-Oct-2015

26 views

Category:

Documents


0 download

DESCRIPTION

epidemiology

TRANSCRIPT

  • 5/25/2018 Animal Disease Surveys

    1/34

    An imal disease

    surveys

  • 5/25/2018 Animal Disease Surveys

    2/34

    Veterinary surveys

    Information may be obtained from:

    Censuses

    Sample surveys

  • 5/25/2018 Animal Disease Surveys

    3/34

    To gain information on:

    Counts (e.g. prevalence)

    Measurements (e.g. milk yield)

  • 5/25/2018 Animal Disease Surveys

    4/34

    Surveys of widespread relevance:

    National picture of disease(e.g. lameness, mastitis)

    National and regional control and

    eradication campaigns(e.g. rinderpest)

    Local (farm) eradication and controlcampaigns (e.g. enzootic abortion)

  • 5/25/2018 Animal Disease Surveys

    5/34

    Bias

    Systematic(as opposed torandom)

    departures from true values

  • 5/25/2018 Animal Disease Surveys

    6/34

    Veterinary surveys:

    Definitions

    Target population

    Study population

    Elementary units

    Stratum

    Sampling frame

    Sampling unit(may, or may not, be elementary units)

    Sampling fraction

  • 5/25/2018 Animal Disease Surveys

    7/34

    Types of sampling

    1.Non-probability sampling:

    2. Probability sampling:

    - choice of investigator

    - a deliberate unbiasedprocess

    - each sampling unit has an equalprobability of selection

    - basis of random sampling

  • 5/25/2018 Animal Disease Surveys

    8/34

    Non-probability sampling methods

    Convenience sampling

    Purposive selection

  • 5/25/2018 Animal Disease Surveys

    9/34

    Probability sampling methods

    Simple random sampling

    Systematic sampling

    Stratified sampling

    Cluster sampling

  • 5/25/2018 Animal Disease Surveys

    10/34

    Random sampling

    Methods:1. Tables

    2. Pocket calculators

    3. Software

  • 5/25/2018 Animal Disease Surveys

    11/34

    Region Number of

    cows

    Number sampled

    Devon and Cornwall 302 647 302 647 x 0.05 = 15 132SW England other than

    Devon and Cornwall

    469 486 469 486 x 0.05 = 23 474

    S England 271 225 271 225 x 0.05 = 13 561

    E England 119 835 119 835 x 0.05 = 5 992

    East Midlands 189 817 189 817 x 0.05 = 9 491

    West Midlands 462 826 462 826 x 0.05 = 23 141

    Wales 342 346 342 346 x 0.05 = 17 117

    Yorkshire/Lancashire 255 626 255 626 x 0.05 = 12 781

    N England 273 838 273 838 x 0.05 = 13 692

    Scotland 260 366 260 366 x 0.05 = 13 018

    Totals 2 948 012 147 399

  • 5/25/2018 Animal Disease Surveys

    12/34

    50300 1020 40 km

    A

    B

    C

    PUNAKHA

    WANGDI

    BUMTHANG

    LHUNTSHI

    THIMPHU

    PARDHA

    PHODRANG TONSA

    SAMCHI

    CHHUKHA

    DAGA

    CHIRANGLEYLEGPRUG

    SHEMGANG

    MONGAR

    TASHIGANG

    SAMDRUP

    JONGKHAR

    PEMA-

    GATSEL

    Thimphu

    A

    B

    C

    PUNAKHA

    WANGDI

    BUMTHANG

    LHUNTSHI

    THIMPHU

    PARDHA

    PHODRANG TONSA

    SAMCHI

    CHHUKHA

    DAGA

    CHIRANGLEYLEGPRUG

    SHEMGANG

    MONGAR

    TASHIGANG

    SAMDRUP

    JONGKHAR

    PEMA-

    GATSEL

    Thimphu

  • 5/25/2018 Animal Disease Surveys

    13/34

    Cluster sampling

    Used when:

    1. Sampling frame is incomplete

    2. Random sampling from all clusters is

    impracticable

    May be:

    1. One-stage: sampleallanimals in

    selected clusters2. Two-stage: sample someanimals in

    selected clusters

    {3. Multistage}

  • 5/25/2018 Animal Disease Surveys

    14/34

    Sample size determination

    Estimation of disease prevalence

    Detecting the presence of disease

  • 5/25/2018 Animal Disease Surveys

    15/34

    Sample size determination

    Not a matter of guess-workor convenience

    For proportions, depends on:

    Expected proportion (P)

    Precision of sample estimate (d)

    Degree of confidence in estimate

    (usually 95%)

    Population size

  • 5/25/2018 Animal Disease Surveys

    16/34

    How many animals?

    Simple random sampling

    n=1.962 Pexp(1-Pexp)

    d2

    Where:

    n = required sample size

    Pexp = expected prevalence

    d = desired absolute precision

  • 5/25/2018 Animal Disease Surveys

    17/34

    n = 1.962 0.30 (1-0.30)

    0.052

    Pexp = 0.30

    d = 0.05

    Example:

    = 3.84 x 0.21

    0.0025

    = 323

  • 5/25/2018 Animal Disease Surveys

    18/34

    nadj = (Nx n) / (N+ n)

    where N= population size

    Finite correction

    N = 900Example

    nadj = (900 x 323) / (900 + 323)

    = 238

    S l i i d t tt i d i d fid

  • 5/25/2018 Animal Disease Surveys

    19/34

    Sample size required to attain desired confidenceinterval around expected percentage of 5%

    Percenta e

    6000

    4000

    2000

    1000800

    600

    400

    200

    10080

    60

    40

    20

    0 5 10 15 20 25 30 35 40 45

    95%

    95%

    99%

    6000

    4000

    2000

    1000800

    600

    400

    200

    10080

    60

    40

    20

    99%

    Re

    quiredsamplesize

  • 5/25/2018 Animal Disease Surveys

    20/34

    If an idea of estimated prevalence

    is not available ( a common event!)

    What do we do??

    Assume 0.5 (50%)

  • 5/25/2018 Animal Disease Surveys

    21/34

    Detecting the presence of disease

    (Certifying absence )

    n= {1(1-p1)1/d} {N-d/2} + 1,

    Where:

    N = population size

    d = number of affected animalsin the population

    n = required sample size

    p1 = probability of finding atleast one case in the sample

  • 5/25/2018 Animal Disease Surveys

    22/34

    Example: (Pan-African Rinderpest Campaign)

    N = 200d = 10 (5% of 200)

    N = {1- (1-0.95)1/10} {20010/2} + 1

    p1 = 0.95

    Assume at least 5% seroconvert in infected herds Set p1at 0.95

    Apply to herd size 200

    = {1- 0.051/10} x 195 + 1= {1- 0.74113} x 195 + 1

    = 0.2589 x 195 + 1

    = 50 + 1

    = 51

  • 5/25/2018 Animal Disease Surveys

    23/34

    FMD serosurveillance

    EU council directive 85/511/EC Probability 0.95

    Flock prevalence: 2%

    Within-flock prevalence: 5%

  • 5/25/2018 Animal Disease Surveys

    24/34

  • 5/25/2018 Animal Disease Surveys

    25/34

  • 5/25/2018 Animal Disease Surveys

    26/34

    Sampling frames

    Dumfries & Galloway (+ Wigton):752 premises

    Duns: 176 premises

    Jedburgh: 129 premises

  • 5/25/2018 Animal Disease Surveys

    27/34

    Samples: premises (animals)

    Dumfries & Galloway (+ Wigton):138 (18,040)

    Duns: 102 (11,187)

    Jedburgh: 91 (10,614)

  • 5/25/2018 Animal Disease Surveys

    28/34

  • 5/25/2018 Animal Disease Surveys

    29/34

    Calculation of confidence intervals:

    Simple proportion

    (e.g., simple random sampling)

    1. Exact method (binomial)

    2. Approximate methods:

    Normal approximation

    Poisson approximation(for rare diseases)

    C fid i t l f l

  • 5/25/2018 Animal Disease Surveys

    30/34

    P* = Sample estimate

    n = Number in sample

    Thus:P* = 0.40

    E.g., in a sample 100 animals, taken from a population of

    10,000, 40 are diseased.

    Normal approximation:for a 95% interval

    Where:

    95% c.i. = 0.401.96 (0.40 x 0.60 / 100)

    Confidence intervals for prevalence:simple random samples

    P*- 1.96P* (1-P*)

    n, P*+ 1.96

    P* (1-P*)n

    to 0.40 + 1.96 (0.40 x 0.60 / 100)= (0.400.096), (0.40 + 0.096)= 0.304, 0.496

    30.4% - 49.6%

  • 5/25/2018 Animal Disease Surveys

    31/34

    Finite correction:

    If the sampling fraction, f, is large (say,

    >10%):

    The numerator, P* (1-P*), should be

    multiplied by (1-f)

    E.g., if a sample of 100 animals is drawn

    from a population of total size 500, thenf= 0.20, and (1-f ) = 0.80

  • 5/25/2018 Animal Disease Surveys

    32/34

    Multipliers used in construction of

    common confidence intervals, based

    on the Normal distribution

    Confidence

    interval

    80% 90% 95% 99% 99.9%

    Multiplier 1.282 1.645 1.960 2.576 3.291

    Confidence intervals for

  • 5/25/2018 Animal Disease Surveys

    33/34

    If P* = estimated prevalence

    andn = sample size

    1. P 0.05

    Formula should only be used when:

    Restriction on asymptotic method:

    and

    Confidence intervals forsimple random samples:

    2. nPand n(1-P) 5

    If these conditions do not apply, use the exactbinomial method:

    1. from tables

    2. with appropriate software, e.g., WINPEPI

  • 5/25/2018 Animal Disease Surveys

    34/34

    Calculation of confidence intervals

    Systematic sampling

    If no periodicity:use simple proportion formula

    Stratified samplingFor proportional allocation:

    use simple proportion formula

    Cluster sampling complex!!Formula for simple proportion will produce

    too narrowa confidence interval

    (i t i )