Surveys Flashcards
Sound expertise
Hierarchy of populations (3)
- Target population: to whom results can be extrapolated
- Source population: from where the study subjects are drawn
- Study sample: individuals from source population who end up in the study
Survey designs (3)
- Prevalence survey
- Incidence rate survey
- Survey to demonstrate freedom of disease
Survey - steps
- Determine objective
a. Estimate disease prevalence
b. Estimate disease incidence
c. Detect disease or demonstrate freedom of disease - Define target/source population and units of interest
- Select survey design
a. Prevalence survey
b. Incidence rate survey
c. Survey to demonstarte freedom from disease - Build first-stage sampling frame
a. One-stage: list of individual animals (sampled by SRS or systematic sampling)
b. Two-stage: list of herds/villages (sampled by SRS)
**5. Calculate sample size (list sample size considerations) **
- Select herds
(7. Build second-stage sampling frame and select individuals) - Collect specimens
9. Analyse specimens to determine disease status (list diagnostic test implications)
- Enter data
11. Analyse data (strata, sample weights, cluster)
- Report findings
Sampling frame
List of all the sampling units in the source population, e.g. all farms in catchment area of interest, from which a sample will be drawn.
Prevalence surveys - applications (2)
Estimate proportion of population that has disease or particular status (e.g. immunity).
Applications:
- Assess burden of disease [by region] (for prioritization, development of control strategies)
- Monitor progress of control program (e.g. proportion of animals with antibodies due to vaccination)
Prevalence surveys - sampling strategies (3)
Multi-stage sampling typically used. Sample frame is usually list of villages or herds. First-stage sampling done with replacement (villages may be selected twice, but then twice as many animals are sampled from the village). Second-stage sampling done without replacement. Three designs:
- Probability proportional to size (PPS) - villages/herds selected with a probability proportional to herd size, then fixed number of animals selected from each herd using SRS (possible only if complete sampling frame is available on all herds/villages AND reliable livestock population data exists)[advantage: simplifies field work since number of animals to be sampled is known prior to visiting village/herd]
- Simple random sampling (SRS) – every village/herd has same probability of being selected, fixed proportion of animals selected from each (suitable when a complete sampling frame is available on all herds/villages BUT no reliable livestock population data exists) [disadvantage: field work more difficult since number of animals to be sampled is unknown prior to visiting village/herd]
- Random geographic coordinate sampling – RGCS used to select villages/herds, fixed proportion of animals selected at each (suitable when no sample frame is available for villages/herds)
Cluster sampling may also be possible if herd size is only a few animals (e.g. developing countries)
Prevalence survey - analysis (2)
- If sensitivity/specificity of diagnostic test is known, the prevalence can be corrected to yield the true prevalence.
- Compare 2 prevalence to see if there has been a change (e.g. evaluate progress of disease control program)
Survey to demonstrate freedom - applications (3), designs (2)
Applications:
- Accreditation schemes on farms - producers can by from farm without risk of spreading disease
- National disease control and eradication programs - decision to cease program activities
- International trade - WTO may ask exporting country to show there is not risk of spread of disease
Designs:
- Single village/herd
- Large area (e.g. nation)
Survey to demonstrate freedom - issues (2)
Issues:
- Sampling - may not be possible to test entire herd therefore may miss positive animal
- Diagnostic test - false negatives/positives
Therefore impossible to prove that a population is free from disease. Is possible to show that it is unlikely there is disease if we test enough animals and take the performance of test into account.
Antibody status lasts much longer than clinical disease and therefore has a higher prevalence (cumulative exposure). Therefore much more common to use serostatus to demonstrate freedom.
Survey to demonstrate freedom - minimum expected prevalence
Prevalence expected if a contagious disease were to enter a herd. When conducting a survey this is the lowest disease prevalence that the survey can reliably detect. For low contagious diseases, this is the maximum acceptable prevalence (non-zero). Included in sample size calculation.
Based on knowledge of the epidemic behaviour of the disease:
- FMD - 30% animals in herd affected
- Johnes - 3%
- If unknown, refer to Terrestrial Animal Health Code for disease in question (or use 2%)
Survey - data analysis considerations (4)
- Stratification - calculate stratum specific estimates
- Sampling weights - if individuals don’t have equal probability of being sampled then used weighted values to generate population values (assigned weight usually made equal to the inverse of the probability of being sampled)
- Clustering - affects variance structure, use multi-level methods
- Imperfect tests - adjustment for sensitivity/specificity less than 100%
Capture-recapture technique
Enables estimation of total population.
- Capture animals over X days and tag them, keeping track of total number of animals caught (Na). Release.
- Wait period of few days
- Capture animals over same period, keeping track of total number of animals caught (Nb) and number with tags (Nab). Release.
Total population = Na*Nb/Nab
Random geographic coordinate sampling (RGCS)
Used when no sampling frame exists and it is not practical to create one (e.g. surveys in remotes regions affacted by political instability)
- Two random numbers are selected which are the x and y coordinates of a rancome point somewhere in the study area
- Towns located within a certain distance (“selection radius”) are identified; if more than one then on eis selected at random
- Data from selected herds are weighted proportional to the total numbewr of herds within the selection radius
