Veterinary Epidemiology - Epidemiological studies Flashcards
1
Q
What is epidemiology
A
= the scientific study of the distribution (freq/pattern and determinants (i.e. causes and risk factors) of health (NOT just disease) in specified populations
- However, epidemiology has historically been concerned with disease prevention
- Epidemiology focuses on studying the health status of populations. The emphasis is on the population (more so than the individual)
2
Q
what is the role of veterinary epidemiology
A
- To describe patterns of health and disease within populations
- To interpret the differences
- To apply our results to improve animal health and welfare and protect public health
- To evaluate the effect of interventions
3
Q
two key factors in most veterinary epidemiological studies
A
- Exposure: the risk factor that can influence the outcome
(e.g. bushfire, exposure to another animal with disease)
~ Can be time variant or invariant (whether animal is male or female = not going to change over time) - Outcome: the disease or event of interest
We assume that exposures and outcomes are part of a intricate web of interactions.
4
Q
challenges determining the populations at risk
- closed and open populations
A
- Closed population
~ no additions to study population,
~ few losses.
~ Rel. easy in controlled laboratory conditions.
~ Animals lost to the study are called withdrawals
Open population –
~ animals enter and leave the study population throughout study period e.g. cases of canine allergic dermatitis in a group of veterinary practices over one year
5
Q
observational vs interventional epidemiology
A
- Observational
= patterns of health or disease in a population are described
~ no alterations are made to the factors that may influence these patterns - Two types of Observational epidemiology:
~ Descriptive : Measure the burden of disease in a population
~ Analytical : Investigate risk factors for a disease or outcome - observational - hard to interpret findings = measuring associations between exposures and outcomes
- even if demonstrate association = dont nessasarily mean exposure/risk factor caused the outcome
- Interventional epidemiology:
= randomized controlled trials RTCs or quasi-experimental designs
~ RTC - randomly allocated to treatmen/control group, animals folowed to see how many got outcome
~ Q - animals not randomly assigned to groups BUT there is experimental manipulation
(e.g. evaluate passibe immune transfer between cows that suckled from dam and separated immediately, cant comp randomise as if calf born at night unsure suckle from dam, artificially put calf in suckling group, see calf born = put into sep group)
6
Q
Common study designs in epidemiology (how linked)
A
observational
~ descriptive
~ analytic - cohort study
- cross-sectional study
- case-control study
interventional
- randomised control trial
- quasiexperimental
7
Q
cross sectional study
- about
A
- An observational study in which the measure of interest is prevalence (i.e. the number of cases that already exist in a population when the study begins)
- A sample is obtained, and the presence/absence of the outcome at the time of sampling is determined. The current or previous exposure is noted.
- For analysis, the prevalence in the exposed and non-exposed subjects is compared
8
Q
cross sectional study
- advantages
A
- Basic structure is straightforward
- Good for initial hypothesis exploration (i.e. May be a useful first step in generating hypotheses for future studies)
- Can evaluate multiple exposures and diseases
- Relatively cheap and fast to carry out
9
Q
cross sectional study
- disadvantages
A
- This study design doesn’t support causal inferences
~ Often difficult to determine if the “exposure” preceded the “outcome” when the timing of disease occurrence is unknown
(in order for exposure to cause disease = come before disease) - Inefficient if disease is rare or if there is a short duration of clinical signs or persistence of antibodies
(only 2 cases, difficult to relate to an exposure as too small subsample - Multiple biases possible (we will talk more about biases later)
- If lab tests are used, then the prevalence determined will depend upon the sensitivity and specificity of the test used
10
Q
cohort study
- about
A
- Cohort: group that has a characteristic in common
- In epidemiological studies, the characteristic of interest is the exposure status (already exposed/unexposed)
- Cohorts (i.e. groups of study subjects) are selected based upon exposure (For example, a group of exposed animals is included and a group of unexposed animals is included)
- We would then observe the cohorts for a defined follow-up period and compare the incidence of disease in the groups. So, the subjects must NOT have the disease at the beginning of the follow-up period to be included
11
Q
cohort study
- prospectively vs retrospectively
A
- In a prospective study, animals are selected based upon exposure status. Disease has not occurred in the study animals at the time the study starts
- In a retrospective study, the follow-up period has ended, and the disease event has occurred. (But cohorts are still selected based upon prior exposure and must have been healthy at beginning of follow-up)
12
Q
cohort studies
- exposures
A
= Any potential risk factor of disease e.g. infectious agent, management factors, feed ingredient etc.
- Can be exposed or non-exposed (dichotomous variable); low, medium or high dose (ordinal scale); organisms per gramme of feed (continuous scale)
- Objective: Identify consequences of a specific exposure factor in terms of e.g., developing disease
13
Q
cohort stidies
- advantages
A
- Can assess multiple outcomes (e.g., diseases) from a single exposure factor.
- Possible to assess temporality (we know that exposure has preceded disease).
- Less potential bias involved as all animals were disease-free to begin with
- If done well, prospective cohort studies provide the strongest evidence towards causality among the observational study designs (still, we would need a randomised controlled trial to prove causality)
14
Q
cohort studies
- disadvantages
A
- Impractical for rare diseases (in other words, a large number of exposed animals would nee to be followed to detect just a few cases of disease)
- Impractical for diseases with a long incubation period (individuals would have to be followed for a long period of time)
- May lose subjects to follow-up during the study (e.g., due to sale or death of an animal)
- Can be expensive and lengthy before results can be obtained
15
Q
case control study
- about
A
- Subjects are selected based upon the outcome (e.g., disease status). In other words, a group of cases and a group of non-cases (i.e. controls) are selected from a population
- Frequency of exposure in the cases is contrasted against frequency in the controls
- Case-control studies are retrospective (since the outcome has occurred when the study begins)
16
Q
case control studies
- advantages
A
- Can investigate multiple exposures and trace back to see links between animals (select based on outcome)
- Good for studying rare diseases or those with a long latent period (because you already have the cases from the start of the study so you aren’t waiting around for the cases to occur)
~ You can select a sufficient number of animals upfront (Often cases and controls are matched 1:1, but not always: for example, can have 3 or 4 controls per case) - Quicker and cheaper than cohort studies
17
Q
case control studies
- disadvantages
A
- Bias if cases and controls selected from different populations
- Accurate determination of exposure status required. However, if and when exposure took place in the past can often only be determined by owner recall - potential danger of recall bias
- Inefficient for investigating rare exposures (small subset when trying to find exposure links between animals)
18
Q
potential exposures
e.g. bovine tuberculosis
A
- infected cattle
- infected wildlife
- infection neighboring herd
- purchased from infected farm
- exposed to high/low bac loads
- false neg
- depressed immune sys
19
Q
Bias and systemic errors
A
= any form of systematic error - may arise as a result of selection or observation
= reproducible inaccuracies that are consistently in the same direction.
~ They are due to a flawed experimental design or a flawed measuring device.
20
Q
Hpw can you avoid bias
A
Bias can be reduced by careful study design
21
Q
3 common types of bias
A
- Selection bias
- Misclassification bias
- Observation bias
22
Q
selection bias
A
- Occurs when individuals or groups in a study differ systematically from the population of interest leading to a systematic error in an association or outcome
- May be due to non-response rate (i.e. potential participants who do not join the study)
- E.g., You’re trying to see whether there is an association between sheep farmers that are stressed out with work and whether or not their flocks have sheep scab.
~ stressed farmers with sheep scab less likely to participate due to time/circumstance
23
Q
Misclassification bias
A
= Some individuals are placed into the wrong groups (i.e. misclassified).
- Incorrect diagnosis due to limited knowledge
- Complex diagnostic process (e.g., sensitivity and/or specificity is poor)
- Laboratory error
- Not acknowledging subclinical disease – animal looks healthy, but may have subclinical disease
- Detection bias (e.g., animal given a more thorough exam if exposed)
- Reluctance to be truthful
- Records incorrectly coded in database
24
Q
Observation bias
- recal
A
= where cases are more likely to remember prior exposure compared to controls
e.g. food poisoning
25
Observation bias
| - interviewer
= where those carrying out the study knowingly or unwittingly treat the comparative groups differently (can possibly lead to misclassification bias)
26
Observation bias
| - loss to follow up
= where there is differential loss in different groups
- some participants may drop out of study in a particular group
- cohort studies
27
Confounding
- A possible result of complex interactions that exist between multiple exposures and disease
- the mixing together of the effects of two or more factors on the outcome
= distortion or masking of an association between an exposure and an outcome by another extraneous variable
- This other variable confounds the relationship between the exposure and the outcome = confounding variable
- Common confounders include age, sex, breed
28
Example of confounding - association between allowing dogs to roam freely among sheep and flea infestation in the dogs
- Results show that dogs allowed to roam among sheep flocks were more likely to have fleas
CONFOUNDING FACTOR:
- further investigation showed that those owners who allowed their dogs to roam among sheep flocks were also much less likely to have previously treated their dogs with flea prevention
29
How to control for confounding
- Exclusion (restricted sampling)
e.g. select only females, select only chimps in a given age range, select only dogs treated with flea prevention, etc.
- Matching
~ make the distribution of potentially confounding variables the same in both groups (assign an equal number of elderly chimps to each group)
- Adapt the statistical analysis to cope with potential confounding
