click below
click below
Normal Size Small Size show me how
WEEK 6:
The long way round: cohort studies
| Question | Answer |
|---|---|
| cohort study | compares risks in one group of people (or cohort) with another group (eg group exposed to a risk factor compared to a group which is not) |
| design of case control study | cases (people with disease) and controls (people without disease) both split into exposed and non exposed |
| evidence for cause and effect limited by (2) | retrospective (measure of exposure, potential for recall bias, and reverse causality (dont know when dietary habits happened relative to disease - have people changed diets since developing disease) |
| strengths of cohort studies (6) | allow measurement of incidence of disease, can study multiple disease outcomes that develop, can study effect of multiple exposures, exposure measured before development of disease, reduced selection/recall bias, strongest observaional design |
| cohort studies can study effect of multiple exposures as long as | they are measured at beginnign of study |
| exposure measured before development of disease is a strength of cohort studies because | we can be fairly confident we are not seeing 'reverse causation' |
| selection bias in cohort studies | may not be representative of people with/without outcome of interest |
| how is selection bias reduced in a cohort study | by selecting a representative group of people whose exposure status is known and following over time to see who does and does not develop the outcome |
| recall bias in cohort studies reduced where | people are followed prospectively (forward in time) |
| how is recall bias eliminated in cohort studies | exposure happens (and is measured) before development of disease so people dont know they are going to develop outcome |
| advantage of recall bias | research team can set objective, reliable measurement for exposure and outcome and defined time points and there is no chance of selective memory issues |
| limitations of cohort studies (7) | time + expensive, inefficient for rare diseases/outcomes (cohort need be large), poor estimates of incidence, not all variables controlled, difficulty with confounding, no know outcomes if dropouts (likely), attrition bias |
| attrition bias | occurs when attrition is related to either exposure or outcome |
| Retrospective cohort study | Alternative design, using a cohort where there is existing data on exposure and disease. |
| pros of retrospective cohort studies | Overcomes loss-to-follow-up Overcomes long latency periods (between exposure to risk factor and development of disease) because disease has already developed Cheaper and quicker |
| cons of retrospective cohort studies | BUT need good-quality exposure data to have already been collected with limited misclassification. If the data doesn’t exist, you can’t do it, and if the data’s bad, your conclusions may be wrong. |
| Limitations of occupational samples (for any research design) – the healthy worker effect | If population can go to work, then usually healthier than the general population Mortality lower in workers Fewer severely ill or disabled people Form of selection bias |
| survival analysis allows us to describe and compare (2) | The number of people who suffer an event of interest (not necessarily just mortality) The time at which the event occurs |