“Extraneous variable that wholly or partially accounts for the observed effect of an exposure on an outcome.”
61 What is a confounder? More descriptions. “A variable that can be used to decrease confounding when properly adjusted for.” “The identification of confounders requires expert or substantive knowledge about the causal network of which exposure and outcome are part . Attempts to select confounders solely based on observed statistical associations may lead to bias.”
62 Plato’s cave
63 Overview of approaches to identify potential confounders “Knowledge-based” approaches Public health or clinical subject knowledge Literature Conceptualization (e.g. DAGs) Data-driven approaches Associations within own study Adjusted versus crude estimates (collapsibility)
64 “Knowledge-based” approach Either you or your public health or clinical collaborators have the needed expertise “Table 1 approach” to utilize expertise might be efficient Instead, draft a “Table 1” List factors that cause the outcome Identify factors that should be “similarly” distributed between groups In addition, use outcome prediction models if possible
65 Table 1 from registry study (observational) NEJM 2008: 348(4):342
66 Table 1 from randomized trial NEJM 2004:350:221
67 “Table 1 approach” to identify confounders Group 1 Group 2 Smoking Blood pressure LDL cholesterol Physical activity Aspirin Outcome risk (e.g. Framingham risk score) Vitamin supplements Coronary heart disease
70 Literature Learn from studies that addressed the same research question - Which confounders did they consider? - What was the impact of adjusting for those confounders? - Are there comparisons from RCTs and observational studies? Learn from studies that assessed the causal relationship between each confounder and the outcome - How strong were the effects? - How consistent were the effects across studies? - Biological plausibility?
71 Tips on how to use the literature for identifying confounders Try to find systematic reviews (comprehensive, not selective) - Clinical queries on PubMed identifies systematic reviews - Look for systematic reviews of interest - Use related articles function Systematic reviews of interest - Address same research question as your own study (which confounders were considered in included studies?) - Look at association of potential confounders and outcome (strength and consistency of association)
72 One definition of a systematic review A systematic review is a summary of research, that addresses a specific research question, that uses explicit and rigorous methods to identify, critically appraise, and synthesize relevant studies
73 Steps in systematic reviews – brief overview Framing the question Population : General population Exposure : Vitamins Control : No vitamins Outcome : Coronary heart disease Framing in- and exclusion criteria for studies Randomized clinical trials Prospective cohort studies Defining the search strategy Database selection Grey and unpublished literature
You've reached the end of your free preview.
Want to read all 115 pages?
- Fall '18
- Epidemiology, Causality, Confounding, Case-control study