BIOL 437 Week Eight P.2 (Statistical Methods in Analytical Epidemiology) Flashcards
misclassification
- occurs when either the exposure or the outcome status is inaccurately assigned
- most case-control studies have some level of it
types of misclassification
- Differential (non-random)
2. Non-differential (random)
differential (non-random)
- classification of history is accurate in 90% of of cases and 100% of controls
- level of misclassification is different
- results in under or over estimation of OR
non-differential (random)
- classification of history is accurate in 90% of cases and 90% of controls
- level of misclassification is the same
- results in understimating OR
confounding
-occurs when an extrinsic factor is associated with a disease outcome
-independant of that association, it is also associated with the exposure
Ex. coffee consumption, smoking and myocardial infraction
matching
- a strategy for controlling confounding at both the design and analysis levels of a study
- distribution of potential confounding factors is forced to be similar betweeen cases and controls
case-crossover study design
- compares exposure status of a case immediately before its occurennce with that of the same case at a prior time
- each case as their own control
case-crossover study appropriate when
- individual exposures are intermittent
- disease occurs abruptly
- incubation period for detection
- induction period are short
analytic unit for case-crossover
- time
- just before the acute event is the ‘case’ time, compared with some other time: ‘control’ times
nested case-control study design
- case-cohort study
- case-control study “nested” within a cohort study
- sample of cases and controls are selected and exposure status compared
cohort studies
- a group of persons being studied who were born in the same year or period
- generally involves study of persons who have been exposed and are followed over time
- can be prospectively (forward in time) or retrospectively (backward in time)
prospective cohort study
-predictor variable is measured before to outcome has occured
retrospective cohort study
-a historical cohort is reconstructed with data on the predictor variable and data on the outcome collected
cohort effect
- generation effect
- change in disease or health status of a study population as they move through time
- includes any exposure or influence, from environmental effets to societal changes
advantage of cohort over case-control
-incidence rate of several outcome variables can be determined and associated with the exposure variable
common measures for describing cohort data
- Cumulative incidence rate: attack rate
2. Incidence density rate: person-time rate
measures of association in cohort studies
- Ratio of attack rates
2. Ratio of person-time rates
ratio of attack rates
- risk ratio (relative risk) (RR)
- probability of health event among those exposed relative to those unexposed
ratio of person-time rates
- rate ratio
- cumulative incidence rate of disease among exposed relative to unexposed
- only able to calculate when total time the exposed and unexposed are at risk is known
person-time
- for acute: hours, days, weeks, sometimes months
- for chronic: years
RRs as %
- RR>1: % increase change (RR-1)x100
- RR<1: % decrease change (1-RR)x100
attributable risk (AR)
- cumulate incidence in exposed group minus unexposed
- difference in attack rates or in person-time rates
attributable risk intrepreted as
- excessive risk of disease among the exposed group attributed to the exposure
- absolute risk in exposed group that is attributable to the exposure
AR% intrepretation
-certain % of disease in group is attributed to the exposure
population attributable risk (PAR) intrepretation
- if exposure was eliminated, the amount the rate would drop
- assumes casual association between exposure and disease
PAR equals
-person-time rate in total population minus person time-rate in unexposed population
PAR % intrepretation
-the % decrease if the exposure was eliminated
