Quiz 2 Epidemiology (10/30/2012)

1. Ecological - group level, no individual data
2. Cross Sectional - prevalence, individual data
3. Case Control - disease sampling study
4. Cohort - exposure sampling study
5. Experimental - manipulation and randomization

Group level
Advantages: Quick and inexpensive
Disdvantages: No individual level, only marginal totals

Group analysis doesn't apply at individual level

Collect both exposure and disease at same time (snapshot at specific time) = prevalence
Advantage:
* Info on prevalence of disease
* Assocations to generate hypotheses
* Quick and inexpensive
Disadvantage:
* No temporal relationship (chicken or egg dilema)
* Selective survival bias (only getting info from survivors of disease)
* NOT good for rare conditions

Large groups

* Begin with cases with specific DISEASE; then
* Select controsl without disease; then
* Compare cases and controls to previous exposures

* Refered to as "retrospective" because look backward from effect to cause

Advantages:
Small sample
Inexpensive
Lack of follow-up
Good for rare diseases
Evaluate several risk factors at once
Control for confounders

Disadvantages:
* Can't compute prevalence or incidence rates (unless all PAR identified). Only have odds ratio.
* Antecedent-consequent relationship (chicken/egg)
Selection bias
* Observational bias (recall bias)
* Can't look at rare exposures

* Can't computer incidence b/c

Relationship b/w E and D is different for participants vs those who would be eligible but don't participate

Recall: Cases more likely to remember exposure than controls

Interviewer: Different interviewer styles might provoke different responses to same question.

Good estimate of relative risk in case control studies when disease is rare.

Ratio of 2 odds: odds E among cases to odds of E among controls

UNMATCHED OR = a/c / b/d = AD/BC

MATCHED OR = B/C (dont' look at concordant pairs)

Control group has same distribution as cases (e.g., cases are 60% white and 40% black, so control cases the same)

Each case is matched with one or more controls.