Abstract:
|
Healthcare researchers who investigate health disparities are often faced with disparate results: trends across groups improve overall but disparities remain or widen. If disparity is evaluated using relative measures (e.g. risk ratio, odds ratio), the disparate results can be due in part to the influence of low outcome prevalence in the advantaged group. Guidelines (eg. CONSORT, STROBE) recommend reporting both relative and absolute measures (e.g. risk differences) of disparity; however, there hasn’t been a thorough investigation of the influence of prevalence on these common measures of health disparities. Utilizing electronic health record data from a network of community health centers, we present a simulation study that could guide researchers/stakeholders in distinguishing the most entrenched disparities from those whose magnitude is due to the mathematical idiosyncrasies of relative measures. We found that the odds ratio is a fairly contained measure of disparity between 25-50% prevalence of outcome among the advantaged population. Below 10%, the odds ratio is dramatically inflated and using only ratios to measure the disparity should be strongly reconsidered.
|