Cohort: Hypothesis-Exposure is associated with disease/outcome
ex: Smoking associated w/lung cancer
ex: access to Cinemax under age 12 associated w/risky sexual behavior age 21+
Case Control: Hypothesis-Disease is associated w/particular exposure.
Cohort vs Case Control
Cohort:
1. Choose population (cohort) to follow some people in cohort will be exposed to thing you think associated w/disease, others not (and so people in cohort as similar to each other as possible)
2. Track Cohort-record who develops disease and who doesn't
3. Calculate incidence of disease in exposed and not exposed groups and compare
*Prospective/longitudianl study: start now, track cohort into future
*Retrospective: start now, but w/old records to identify cohort then collect data on exposures and outcomes of cohort members
ex: Semmelweis-Cohort: women who delivered babies at kronkenhous Allemegre Retrospective Review of 2 yrs of maternal ward records. exposed=Dr. delivered babies, Nonexposed=Nurse/midwife delivered babies
Outcome=Death from puerper fever
Case Control
1. Select subjects who have a disease (cases)
2. Select subjects who dont have disease (controls) but otherwise similar
3. Collect data re: subjects' exposure to thing you think associated w/disease
4. Calculate and compare porportion of cases exposed to exposure and of controls exposure
Cohort Advantages: Track population over time and record new cases-measure incidence of outcome, which is a measurement of risk-case control since one group has disease and other group doesnt, you cant measure incidence/prevalence-cant measure risk
Case Control Study Advantages: Handier at looking at diseases that take a long time to develop or diseases that are rare.
***IF THERE IS NO COMPARISON GROUP, NO ASSOCIATION BETWEEN EXPOSURE AND DISEASE CAN BE MADE.-thats why cohorts/case controls are reliable and advantagous over case studies where there are no control/comparison group.
Cohort:
Framingham (pg 171): 30-62 yr olds in framingham alabama, MA 1948 follow to see whether heart disease develops.
Exposures defined: smoking, lack of physical activity, sugar diabetes, sweetie, i cut yo 2 fingaz into a million pieces, obesity.
Problems w/cohort Studies
*prospective-drop-outs;expenses;hypothesis can change while study is being conducted (b/c of new discoveries made btw beginning of study and conducting of study) so you may not have collected the right data
*Retrospective: Since records collected by other ppl, for other purposes, in past, may not include all reliable (relevant) info..
Set up Case Control Study
Selection of Cases: Issues
1. Incident-Disadvantages=hard to study rare diseases. Adv=More accurately looks at rish and can establish before and after relationship between exposure and disease
2. Prevalent Cases-Disadvantages=may uncover association between living w/disease and exposure, not between disease itself and exposure. Adv=easier to study rare diseases because you dont have to wait for disease to pop up before study.
Selection of Controls
-healthy controls: 1. neighborhood-pick close to cases. 2. Best Friend-relatives help control for genetic factors that contribute to disease. 3. Hospitalized Controls-ones who are sick, just not with your disease of study (easy to obtain but not necessarily representative of underlying population)
Control Match to Cases
1. Group Matching-Have casesfigure out characteristics you think its important to match on (ex: demographics) and calculate % cases w/characteristic-pick controls so % of people w/different characteristics match cases
ex: Cases: 75% male and 25% female so out of 100 controls pick 75 males and 25 females.
2. Matched Pairs-enroll case, enroll control matched for significant characteristics.
ex: Case 1: 45 yo, divorced, white female (disease)
Control 1: 45 yo, divorced, white female (no disease)
difficult to do and traits that you match cases and controls on cant be studied.
Problems w/finding Exposures w/interview/survery…1. people doent accurately know info 2. Recall Bias: people who have bad outcome more likely to remember prior events than people w/out bad outcome
Multiple Controls-1. greater than 1 control/case 2. losts of subjects which increases statistical power 3. cant increase # of cases much so can increase controls to increase subject #
Cross Sectional Study
-look at overal population and divide into people with disease and exposure, people without disease with exposure, people without disease and with exposure, people without disease without exposure.
