(PH 101) Midterm 1 Review

Lecture 1-7

Cohort

(Cohort or Case-Control) Starting point is exposure status

Case-Control

(Cohort or Case-Control) Starting point is disease status

Cohort

(Cohort or Case-Control) Direction is forward (exposure to disease)

Case-Control

(Cohort or Case-Control) Direction is backward (disease to exposure)

Cohort

(Cohort or Case-Control) Can estimate incidence and relative risk

Case-control

(Cohort or Case-Control) Can only estimate odds ratio

Cohort

(Cohort or Case-Control) Best suited for a rare exposure

Case-Control

(Cohort or Case-Control) Best suited for a rare disease

Cohort

(Cohort or Case-Control) Efficient for one exposure, many outcomes

Case-Control

(Cohort or Case-Control) Best suited for many exposures, one outcome

Cohort

(Cohort or Case-Control) Longer, costlier

Case-Control

(Cohort or Case-Control) Faster, cheaper

Cohort

(Cohort or Case-Control) Main bias risk is loss to follow-up exposure misclassification

Case-Control

(Cohort or Case-Control) Main bias risk is selection bias and recall bias

Case-Control

(Cohort or Case-Control) A suspected environmental chemical may cause a rare pediatric cancer — about 1 in 100,000 children. You have one year and a modest budget for a first answer.

Cohort

(Cohort or Case-Control) A new diabetes drug is now taken by thousands of patients. Regulators want to know whether it raises the risk of ten different possible side effects — heart, liver, kidney, and more.

Cohort

(Cohort or Case-Control) A solvent used only at one small manufacturing plant (200 workers) is suspected of causing a rare blood disorder. Every worker ever exposed can be identified.

Case-control

(Cohort or Case-Control) A cluster of a rare birth defect appears in a single city. Researchers want to test many possible explanations at once — maternal diet, medications, workplace exposures, water supply.

Cohort

(Cohort or Case-Control) A national health agency wants precise 20-year incidence and relative-risk estimates for heart disease among smokers vs. non-smokers, and has funding for a large, long-term study.

experimental study

investigator assigns the exposure

observational studies

Include…

Observational Studies

• case reports/case series

Comparative/Analytical Studies

• ecological study

• Case-control study

• cross-sectional study

• retrospective cohort study

• prospective cohort study

case reports

• no assigned exposure, no comparison group

  • Descriptive study

ecological study

• no assigned exposure, no comparison group, level of analysis is PLACE

case-control study

• no assigned exposure, no comparison group, level of analysis is INDIVIDUAL, exposure and disease NOT measured together

  • Disease first, then look back

  • Ask about past exposure

cross-sectional study

• no assigned exposure, no comparison group, level of analysis is INDIVIDUAL, exposure and disease measured together

  • Both measured at once

retrospective cohort study

• no assigned exposure, no comparison group, level of analysis is INDIVIDUAL, study does not include people w. Disease, data already collected (historical)

  • Use data that is already collected

prospective cohort study

• no assigned exposure, no comparison group, level of analysis is INDIVIDUAL, study does not include people w. Disease, data collected as the study goes

  • Recruit people, follow them in time

hypothesis generating

What are ecological studies good for?

ecological fallacy

ascribing to individuals a characteristic that really only describes the group they belong to

Pros of ecological study

Pros of _____

• The data is often easy to obtain

• It’s useful for hypothesis-generating

• It shows population level impact — e.g. of a policy change

cross-sectional study

 Exposure and disease are measured simultaneously for each participant — a “snapshot in time” that identifies prevalent cases of disease

reverse causality

Because a cross-sectional design measures exposure and disease at the same moment, we don’t know which came first; temporal bias

• Found in cross-sectional study

survivorship bias

If the exposure shortens survival (exposed cases die sooner than unexposed cases) they’re less likely to still be alive and sampled.

• Found in cross-sectional study

a/(a+b) vs c/(c+d)

Cross-Sectional Study
Calculate prevalence of disease in exposed compared to non-exposed persons.
(Show the formula)

____ vs ____

a/(a+c) vs b/(b+d)

Cross-Sectional Study
Calculate prevalence of exposure in diseased compared to non-diseased persons.
(Show the formula)

____ vs ____

cases

people with disease in case-control study

controls

people WITHOUT disease in case-control study

prevalence

Case-control study cannot be used to estimate _____ of disease!

selection bias

____ bias

• Source of cases

• Selection of controls

• Source of controls

(in case-control studies)

information bias

____ bias

• Problems with recall

• Recall bias

(in case-control studies)

referral bias

occurs when the procedures used to select cases or controls result in a study population that is not representative of the source population from which the cases originated (AKA selection bias)

• (in case-control studies)

non-differential misclassification of exposure

• When exposure is misclassified to the same extent for cases and controls, the true association gets diluted (biased toward the null)

recall bias

____ bias

• Cases may ruminate more on past exposures, searching for answers

• differential recall between cases and controls

matching

Choosing controls to resemble cases on key characteristics

• (in case-control studies)

multiple controls

Using more than one control per case, of one type or several

• (in case-control studies)

group matching

The proportion of controls with a given characteristic matches the proportion of cases with that same characteristic.

• Type of matching

• (in case-control studies)

individual matching

Matched pairs: for each case, a control is selected who is similar on the specific variable(s) of concern

• Type of matching

• (in case-control studies)

case-control

What is the study for these strengths and limitations?

Strengths Efficient — less time, money, and fewer subjects than other designs No need to wait for long-latency diseases to develop Easy to explore many exposures’ effect on one outcome Useful when little is known about a disease’s etiology Important for infectious disease outbreak investigations

Limitations Often limited to studying a single outcome Inefficient for rare exposures More opportunity for recall and selection bias May be unsure of the temporal sequence between exposure and disease Cannot calculate an absolute measure of association Cannot know what prevalence/incidence of disease is in this study

case-crossover design

____ design

Each case serves as their own control: exposure right before the outcome is compared to that same person’s exposure during earlier “control” windows

• useful for brief exposures and acute outcomes

cohort

a defined group of people followed over time

cohort study

specific research design built around following a cohort