In their recent article recommending that relative measures of health disparities always be presented in terms of adverse outcomes,1 Keppel and Pearcy cite two articles of mine as observing "that a decrease in the relative disparity in rates of infant survival (a favorable outcome) is associated with an increase in the relative disparity in rates of infant mortality (an adverse outcome)."2,3 This characterization, while technically accurate, obscures the central point of those articles and the issues that point raises about the measurement of health disparities.

The referenced articles described the statistical tendency whereby, whenever two groups differ in their susceptibility to an outcome, the rarer the outcome the greater the relative disparity in experiencing it and the smaller the relative disparity in avoiding it. One implication of that tendency is that as mortality declines, demographic disparities in mortality rates will tend to increase and demographic disparities in survival rates will tend to decline. The association of declining disparities in infant survival rates with increasing disparities in infant mortality rates-which occurred during periods of overall declines in infant mortality-is just one example of that tendency.

But it is a systematic tendency. It can be observed in virtually any data set that allows one to examine the situation at various points in a continuum where two groups have different distributions of factors related to experiencing or avoiding some outcome (eg, income data, test score distributions). But it is no less present when the distributions cannot be directly observed. Until the health disparities research community recognizes the tendency and attempts to take it into account, we will derive little of value from the substantial resources committed to that research. The recommendation of Keppel and Pearcy that relative disparities always be measured in terms of adverse outcomes by no means addresses this issue. Regardless of whether one examines adverse or favorable outcomes, one still has to take the referenced tendency into account in order to determine whether changes in relative disparities are other than the natural consequence of changes in the prevalence of the outcome.

Keppel and Pearcy also note that the absolute differences between the rates at which two groups experience an adverse outcome are the same as the absolute differences in the rates avoiding the outcome. That might be seen as a reason to favor absolute differences over relative differences in measuring health disparities. Furthermore, absolute differences are useful for indicating how large a proportion of the disadvantaged group is affected by its greater susceptibility to an adverse outcome. Like relative differences, however, absolute differences tend to change when there occurs an overall change in the prevalence of an outcome. Hence, like changes in relative differences, changes in absolute difference are useful for determining whether there has been a true change in the relative well-being of two groups only if such changes are interpreted with an understanding of the types of changes that occur simply because of a change in the prevalence of the outcome.4

REFERENCES