lecture+21+Confounding+2018.pdf - ReMA | Quantitative...

• Notes
• 56
• 100% (1) 1 out of 1 people found this document helpful

This preview shows page 1 - 10 out of 56 pages.

P6031—Research Methods and Applications | Quantitative Foundations 1 ReMA | Quantitative Foundations Lecture 21 | November 8, 2018 Confounding
P6031—Research Methods and Applications | Quantitative Foundations 2 Last time Described the utility of case-control studies for studying rare diseases Described the selection of controls in a case- control study: challenges and best approaches Evaluated causal inference under the case- control design Differentiated case-control from RCT, cohort designs Described the odds ratio and its estimability under the case-control design Computed a CI for OR
P6031—Research Methods and Applications | Quantitative Foundations 3 PRESENT (time your study begins) FUTURE PAST PROSPECTIVE COHORT EXPOSED, UNEXPOSED DISEASED, NOT DISEASED RETROSPECTIVE COHORT EXPOSED, UNEXPOSED DISEASED, NOT DISEASED AMBIDIRECTIONAL COHORT EXPOSED, UNEXPOSED DISEASED, NOT DISEASED Time and Study Designs CASE-CONTROL EXPOSED, UNEXPOSED DISEASED, NOT DISEASED
P6031—Research Methods and Applications | Quantitative Foundations 4 Confidence interval for the OR Recall our 2x2 table above. We know our estimator for OR is: It can be shown that: So our (1-α)x100% CI for the true ln(OR) is: Finally, we exponentiate the endpoints (L, U) to get the CI for true OR : (e L , e U ) D+ D- E+ a b N E+ E- c d N E- N D+ N D- N=Total bc ad R O = ˆ d c b a R O SE 1 1 1 1 )) ˆ (ln( + + + = l n( \$ %&) ± ) *+,- ./ 0 l n( \$ %& = 23 45 67 ± ) *+,- 1 4 + 1 6 + 1 7 + 1 5 NOTE: MARGINS DON’T MATTER!
P6031—Research Methods and Applications | Quantitative Foundations 5 Summary: Table 1 in a Case-Control Study Table 1: Patient Characteristics Characteristic Cases (N=100) Controls (N=100) Total (N=200) Age Sex
P6031—Research Methods and Applications | Quantitative Foundations 6 Advantages of case-control studies Typically less expensive than cohort studies Usually require smaller sample sizes Shorter duration than prospective studies Can study multiple risk factors for a single disease Easily reproduced in different populations by different investigators
P6031—Research Methods and Applications | Quantitative Foundations 7 Disadvantages of case-control studies Information about exposure in cases is often obtained after the disease is diagnosed Example: Diet; physical activity Dependent on subject’s memory, which may be affected by disease. The population of origin for cases is difficult to define precisely. Difficult to identify appropriate control group Case-control studies do not provide estimates of risk and risk difference. Like any observational study, confounding is a concern….
P6031—Research Methods and Applications | Quantitative Foundations 8 Questions?
P6031—Research Methods and Applications | Quantitative Foundations 9 Learning objectives for today Explain and define confounding Describe the need for adjustment procedures in observational studies in the design and analysis stages Apply simple techniques for identifying and removing the effects of confounders in observational studies