National Healthcare Disparities Report Measurement Issues
(Nov. 5, 2013; rev. March 15, 2014))
This is a draft of an explanation that in 2010 the National Healthcare Disparities Report (NHDR) changed its method of analyzing health disparities by adopting a regression approach that determined directions of changes in health and healthcare disparities on the basis of which of two groups being compared experienced the larger change in its outcome rates. See page 6 of the 2010 NHDR. The change in approach was apparently intended to be consistent with the approach to measuring disparities in Healthy People 2010/2010, which measures all health and healthcare disparities in terms of relative differences in outcome rates. And the new NHDR approach would have been consistent with the Health People approach if the changes to be compared were relative changes in adverse outcomes. But apparently the changes in rates were measured in terms of absolute (percentage) differences. This would not be consistent with Healthy People 2010. In fact, as explained on many of pages of this site and perhaps most comprehensively in the October 9, 2012 Harvard University Measurement Letter, in the common situation where a relatively common favorable healthcare outcome increase in overall prevalence (and where relative differences in favorable outcomes would tend to decrease while relative differences in adverse outcomes would tend to increase, as explained on many pages of this site), absolute differences would tend to change in the same direction as the relative difference in the favorable outcome and the opposite direction of the relative difference in the adverse outcome. Some examples from the 2012 NHDR are presented in Table 5 of my 2013 Federal Committee on Statistical Methodology Research Conference presentation “Measuring Health and Healthcare Disparities.” The table shows situations the 2012 highlights as some of the fastest improving healthcare disparities where the National Center for Health Statistics (and Health People 2020) would have found increasing disparities. This matter is also the subject of the conference paper (at 27-28).