The article by Hosseinpoor et al.[1] draws from a 2013 World Health Organization health inequalities measurement guide (which has the same first two authors as the instant article).[2] Like the WHO handbook, the article suffers critically from the failure to recognize patterns by which the measures it discusses tend to be affected by the frequency of an outcome.

The rarer an outcome, the greater tends to be the relative difference in experiencing it and the smaller tends to be the relative difference in avoiding it. Thus, as mortality declines relative differences in mortality tend to increase, while relative differences in survival tend to decrease; as rates of appropriate healthcare (or insurance coverage) increase, relative differences in receipt of such care (or coverage) tend to decrease, while relative differences in non-receipt of such care (or coverage) tend to increase.[3-6]

The WHO handbook (at 55) cited the 2005 measurement guide of the US National Center for Health Statistics (NCHS),[7] in which the agency recognized that determinations of whether health and healthcare inequalities were increasing or decreasing would commonly turn on whether one examined relative differences in the favorable outcome or relative differences in the corresponding adverse outcome and therefore determined that all health and healthcare inequalities should be measured in terms of relative differences in adverse outcomes.[7] Yet the WHO handbook showed no recognition that it was even possible for the two relative differences to yield opposite conclusions about the comparative size of inequalities, or whether they were increasing or decreasing over time, much less that they tended to do so systematically. And it showed no recognition that NCHS had determined that it would always rely on relative differences in adverse outcomes. As a result, the WHO handbook reached conclusions about the comparative size of inequalities that were essentially the opposite of those NCHS would reach (as illustrated in Table 5 (at 19) of reference 4) and without recognizing that it was doing so. The same failure of understanding is found in the instant article, even though one of its authors co-authored the NCHS guide.

This is not to say that the NCHS approach is sound. On the contrary, as discussed in reference 3 (at 332), the view of NCHS that it could arbitrarily choose to rely on the relative difference in the adverse outcome reflected a fundamental misunderstanding of the reasons society examines health inequalities and why neither relative difference was providing sound information about the strength of the forces causing outcome rates of advantaged and disadvantaged groups differ and whether the strength of those forces was increasing or decreasing over time.

The absolute difference, which is the measure preferred in the instant article, is unaffected by whether one examines the favorable or the adverse outcome. But for a measure to effectively quantify the strength of the forces causing outcome rates of advantaged and disadvantaged groups to differ the measures must remain unchanged when there occurs a simple change in the overall frequency of an outcome. And absolute differences tend also to change as the frequency of an outcome changes, though in a more complicated way than the two relative differences. Roughly, as uncommon outcomes (less than 50% for both groups being compared) become more common, absolute differences tend to increase; as common outcomes (greater than 50% for both groups being compared) become even more common, absolute differences tend to decrease. The pattern is more difficult to predict when the pattern is neither common nor uncommon or either group’s rates crosses 50% during the period examined.

As the prevalence of an outcome change, the absolute difference tends to track the direction of change of the smaller relative difference. Thus, when uncommon healthcare outcomes increase, the absolute difference will tend to show the same direction of change as the relative difference in non-receipt of care (i.e., an increase in disparity) and the opposite direction of change as the relative difference in receipt of care; as common healthcare outcomes increase, the absolute difference will tend to show the same direction of change as the relative difference in receipt of care (i.e., a decrease in disparity) and the opposite direction of the relative difference in non-receipt of care. Useful illustrations may be found in Table 5 of reference 3 (at 335) and Table 1 of reference 4 (at 13), if movement from the first to the second row is deemed to reflect an increase in an uncommon type of care and movement from the third to the fourth row is deemed to reflect an increases in a common type of care. See also Table 4 (at 334) of reference 3 and Table 26 (at 112) of reference 6. See reference 3 (at 227-39) regarding the way the failure to understand these patterns has led to unfounded perceptions about the effects of incentive programs on healthcare inequalities and misguided responses to those perceptions.

Neither the WHO handbook nor the instant article reflects an understanding of the patterns by which absolute differences tend to change as the frequency of an outcome changes or the implications of those patterns. That failure of understanding raises particular problems with respect to the article’s recommendation of goals of percentage decreases in absolute differences between care or coverage rates of advantaged and disadvantaged groups without recognition that (a) at certain levels of overall frequency improvements in care will tend to increase absolute differences (and can do so even when a sound measure would indicate a decrease in inequality) while (b) at other levels of overall frequency improvements in care will tend to reduce absolute differences (and can do so even when sound measure would indicate an increase in inequality).

More generally, it is not possible to provide useful guidance on the monitoring of health and healthcare inequalities without understanding the patterns by which measures tend to change as the frequency of an outcome changes and without addressing the implications of those patterns.

References:

1. Hosseinpoor AR, Bergen N, Koller T, Prasad A, Schlotheuber A, et al. (2014) Equity-oriented monitoring in the context of universal health coverage. PLoS Med 11(9): e1001727. doi:10.1371/journal.pmed.1001727

2. World Health Organization (2013) Handbook on health inequality monitoring: with a special focus on low- and middle-income countries. Geneva: World Health Organization.

3. Scanlan JP (2014) Race and mortality revisited. Society 51(3):327-346
http://jpscanlan.com/imag...

4. Scanlan JP (2014) Measuring health and healthcare disparities. Proceedings of the Federal Committee on Statistical Methodology 2013 Research Conference.
http://jpscanlan.com/imag...

5. Scanlan JP (2014) “Rethinking the measurement of demographic differences in outcome rates. Methods Workshop, Maryland Population Research Center of the University of Maryland.
http://jpscanlan.com/imag...

6. Scanlan JP (2006) Can we actually measure health disparities? Chance, 19(2):47-51 http://www.jpscanlan.com/...

7. Keppel KG, Pamuk E, Lynch J, Carter-Pokras O, Kim I, et al. (2005) Methodological issues in measuring health disparities. Vital Health Stat 2005;2 (141) http://www.cdc.gov/nchs/d...