Epidemiological Concepts Causal Concepts Inductive reasoning: The process of making generalized inferences about ‘causation’ based on repeated observations. Deductive reasoning: The process of infer- ring that a general ‘law of nature’ exists and has application in a specific, or local, instance. Cause: Any factor that produces a change in severity or frequency of the outcome. Necessary cause: One without which the disease cannot occur. Sufficient cause: Produces the disease if the factor is present. Component-cause: One of a number of factors that, in combination, con Target Population: The population to which it might be possible to extrapolate re- sults from a study. Source Population: The population from which the study subjects are drawn. Study Sample/Group: Consists of the indi- Outcomes and data analysis viduals (animals or groups of animals) that Continuous // dichotomous // nominal // count // time to event end up in the study. Animals Internal validity: The study results are valid Herds causal for members of the source population. Areas inferences External validity: The study results are valid Direct for the source population, target population, Causal-web model: Consists of multi- cause and beyond. ple indirect and direct causes. The fol- Indirect Outcome lowing is an example of a Causal-web Cause model. Direct Cause Sampling (Exposure) Non-probability sampling: individual’s Stratified random sample: Prior to sampling, Sampling frame: List of all sampling probability of selection is not determined the population is divided into mutually exclu- units in the source population (Judgment, Convenience, Purposive) sive strata based on factors likely to affect the Type I (α) error: Concluding that the outcome. Probability sampling: every element has outcomes in the groups being compared a known non-zero probability of being Cluster sampling: Every study subject within are different (association exists) when they included in the sample the cluster (collection of subjects with 1 or are not. more common characteristics) is included in the Simple random sample: Every study Type II (β) error: Concluding that the sample and the primary sampling unit is larger subject in the source population has an outcomes are not different (no associa- than the unit of concern. equal probability of being included. tion) when they are Multistage sampling: After the primary sam- Systematic random sample: A complete Power: Probability that you will find a pling unit is chosen, then a sample of secondary list of the population to be sampled is not statistically significant difference when it sampling units is selected. required provided an estimate of the total exists and is of a certain magnitude (i.e. number of animals is available and all the Targeted (risk-based) sampling: Animals are power = 1-β) animals are sequentially available. assigned point values based on the probability Created by Keila Perez of them having the disease of interest and sam- pling is proportional to that estimate of risk. [email protected] Sampling Equations Types of Error: Conclusion of stat. True state of nature analysis Effect present Effect absent 1) n= total sample size Effect present (reject Type I (α)error To estimate a sample proportion with a null) Correct (power) (p-value) Type II (β) er- desired precision: Effect absent (accept null) ror Correct For adjusting the sample size (n) for clus- For continuous and binary covariates, new To estimate a sample mean with a desired tering, the size of new n(n’) depends on n (n’) (VIF= Variance Inflation Factor): precision: intra-cluster correlation (ρ) and number of individuals sampled per cluster (m): 6) General formula for the width of CI of 2) (n=sample size per group) To compare 2 proportions: a parameter 4) Sampling to confirm disease absence Where p=(p1+p2)/2 and q=1-p Parameter ± Z*SE(parameter), where for — From finite population <1000: - Estimating a mean in a single sample – To compare 2 means: From a large (infinite) population: - Comparisons of means from 2 samples – If sampling from a finite population in 5) Adjustment of sample size (n) in multi- - If expected interaction between two di- descriptive studies, the required sample variable studies: chotomous variables size (n’) can be adjusted using FPC formu- For k continuous covariates, new n (n’) la: ρce =average correlation between expo- sure and confounders Questionnaires Questionnaire: A data-collection tool that Focus Groups: Normally a group of 6-12 Quantitative: ’Structured’ questionnaires can be used in a wide variety of clinical people that provide opportunity for a designed to capture information about and epidemiological research settings. structured form of consultation with mem- study subjects and their environment bers of the intended study population, the Survey: An observational study designed Open Question: There are no re- end users and/or the interviewers. to collect descriptive information about an strictions on the types of responses ex- animal population (such as prevalence of Qualitative: ‘Explorative’ questionnaires pected. disease, level of production etc.) consisting mainly of open questions. Closed Question: The response has to be selected from a pre-set list of answers. Measures of Disease Frequency Study Period: Period of time over which Incidence (I): The number of new events Absolute rates: Number of cases of dis- the study is conducted. in a defined population within a specific ease related to the time period of observa- period of time tion Risk period: Time during which the indi- vidual could develop the disease of interest -Incidence times: Times which incident Closed Population: No additions to the cases occur population for the duration of the study Count: The number of cases of disease or (nor losses) number of animals affected with a condi- -Incidence count: Count of number of tion in a given population cases of disease observed in a population Open Population: Animals are leaving and entering the population Proportion: Ratio in which the numerator -Incidence risk: Probability an animal will is a subset of the denominator develop a disease in a defined time Prevalence (P): Cases of disease existing at a specific point in time rather than new Odds: Ratio in which the numerator is not -Incidence rate: Number of new cases of cases occurring over a period of time a subset of the denominator. disease in a population per unit of animal time during a given time period (D=mean duration of disease) Rate: Ratio in which the denominator is the number of animal-time units at risk Measures of Association Measure of association (MA): Assesses Approaches for hypothesis testing in- - Compute confidence interval (CI) for the magnitude of the relationship between clude: the point estimate. CI reflect the level of an exposure to a disease and a disease uncertainty in point estimates and indicate - Estimating standard error (SE) of the the range of values that a parameter might Attributable fraction (Afe): Proportion parameter as a measure of precision of the have (with values closer to the center being of diseases in exposed that is due to the point estimate (uncertainty) more likely than those at the ends of the exposure - Compute test statistic and from the ex- range). pected distribution of this test statistic determine p-value Interpretation of Risk ratio (RR), Rate ratio (IR), and Odds ratio (OR): <1 exposure is protective, =1 no effect, and >1 exposure is positively associated with disease Interpretation of Risk difference (RD) and Incidence difference (IR): <0 exposure is protective, =0 no effect, and >0 exposure is positively associated with disease The range for AFe: Values from 0 (risks equal regardless of exposure) to 1 (no disease in non-exposedà i.e. all disease is due to ex- posure). Vaccine efficacy is a form of AFe. Diagnostic Tests Accuracy: Average is close to true value Multiple Tests Interpretation: True prevalence: The true state of nature. Precision: The amount of variability - Series: Result is considered positive only Apparent prevalence: The result in the among test results. if both tests are positive study due to imperfections in the diagnos- Coefficient of variation (CV): Standard - Parallel: result is considered positive if tic tests. Devation/Mean (for repeat runs on same either test is positive Predictive Values: The probability that sample) Sensitivity (Se): proportion of diseased the animal has or does not have the dis- Pearson correlation coefficient (PCC): animals that test positive (TP): p(T+|D+) ease, given the test result. Ignores the scales of the 2 sets of results —PV(+) = p(D+|T+) Specificity (Sp): proportion of non- —PV(-) = p(D-|T-) Concordance correlation coefficient diseased animals that test negative (TN): (CCC): Takes into account data position p(T-|D-) Define cutoff: Sp increases, Se decreases. from equality line. See graph below. Kappa Statistic: Measure of agreement for tests with qualitative outcomes. Ranges from 0 (poor agreement) to 1 (perfect agreement. Agreement: How well 2 different tests agree on the same sample. Study Designs Descriptive Study: Describe the nature of Cross-sectional Study: Objective is to esti- Controlled trial: Planned experiment the disease. mate some sort of population parameter. The carried out on subjects in their usual - Case report: Based on individual outcome frequency of measure is prevalence environment (clinical trail in a clinical - Case series: Based on group since this study looks only a snip of time. setting) - Survey: Based on population •Phase I: (formulation trials): Trials in Explanatory Study: Objective is to identi- healthy animals to evaluate safety of fy associations between factors (exposures) the drug (dose, adverse reactions…) and disease status. (Experimental and Ob- servational Studies) •Phase II: Trials in a small number of Cohort Study: A cohort is a group of sub- animals from the target population Experimental Study: Objective is to jects with common exposure, and the objec- (e.g., sick animals) to document the identify the effect of an exposure that is tive of a cohort study is to evaluate causal activity of the drug.