Re: Importance of monitoring health inequalities
A premise of the editorial by Marmot and Goldblatt  is that health inequalities monitoring has been sound. In fact, it has been manifestly unsound. The need to know what is going stressed by authors supports an argument for suspending health inequalities research until measurement issues have been addressed.
Most health inequalities research relies on various standard measures of differences between outcome rates without recognizing patterns by which, for reasons inherent in underlying risk distributions, those measures tend to be systematically affected by the prevalence 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, for example, as mortality declines, relative difference in mortality tend to increase while relative differences in survival tend to decrease; as appropriate healthcare rates increase, relative differences in receipt of care tend to decrease, while relative differences in non-receipt of care tend to increase. Relative differences in adverse outcomes tend to be larger, while relative differences in corresponding favorable outcomes tend to be smaller, within comparatively advantaged subpopulations (where adverse outcomes are less common) than comparatively disadvantaged subpopulations.[2-6]
Absolute (percentage point) differences between rates also tend to change as the prevalence of an outcome changes. As uncommon outcomes become more common absolute differences tend to increase; as common outcomes become more even common absolute differences tend to decrease.[2,4-6]
There will be many departures from these patterns. Observed patterns are functions of (a) the strength of forces causing rates to differ and (b) the prevalence-related forces described above. Society’s interest is in understanding (a). But one must recognize (b) in order to understand (a).
Few understand these issues, however. Researchers commonly employ a measure without consideration of whether other measures would yield different conclusion as to such things as direction of changes over time. Few seem to know that it is possible for relative differences in favorable outcomes and corresponding adverse outcomes to change in opposite directions, much less that they tend to do so systematically. Researchers discussing racial inequalities in cancer outcomes commonly refer to relative differences in survival and mortality interchangeably, often stating they are analyzing one while in fact analyzing the other. Invariably they fail to recognize that the two relative differences tend to change in opposite directions as survival rates change generally or that more survivable cancers tend to show larger relative differences in mortality, but smaller relative differences in survival, than less survivable cancers.
Observers commonly attach mistaken significance to large relative differences in adverse outcomes among advantaged subpopulations. The Whitehall studies’ findings of steeper social gradients in adverse outcomes among British civil servants than in the general UK population have prompted varied theories about the nature of health inequalities. But such theories have failed to recognize the reason to expect large relative differences in adverse outcomes among Whitehall subjects simply because they tend to be healthier than the general UK population or to consider whether relative differences in the corresponding favorable outcomes are smaller among Whitehall subjects than the general population.[2-5] (See Table 3 of reference 4 and Table 5 of reference 8 regarding the larger relative differences in mortality, but smaller relative differences in survival, among younger Whitehall subjects (the advantaged subpopulation within Whitehall) than among older subjects.)
Eight years ago the US National Center for Health Statistics (NCHS) recognized that relative differences in favorable and adverse health outcomes would commonly change in different directions. But rather than addressing that neither relative difference was an effective measure of the forces causing rates of advantaged and disadvantage groups to differ, NCHS simply recommended that all inequalities be measured in terms of relative differences in adverse outcomes.[9,10] See Table 6 of reference 6 and Table 4 of reference 7 regarding situations where researchers found substantial decreases in vaccination inequalities in the US or cancer screening inequalities in the UK in circumstances where NCHS would have found substantial increases in inequalities
Understanding of patterns by which absolute differences are affected by the prevalence of an outcome is entirely missing from research that increasingly relies on absolute differences to measure healthcare inequalities. Reliance on absolute differences to measure inequalities in healthcare outcomes that were increasing – but without consideration of whether the outcome examined was uncommon or common – has led to a perception in the US that pay-for-performance (P4P) will tend to increase healthcare inequalities and a perception of in the UK that (P4P) will tend to reduce healthcare inequalities. (See pages 21-24 of reference 5 regarding Massachusetts’ use of an inequalities measure in its Medicaid P4P program that may well increase inequalities.)
A recent BMJ article, while purporting to clarify measurement issues, evidenced no understanding that the measures it discussed tended to be affected by the prevalence of an outcome or even that there existed two relative differences. The same holds for the measurement guide of the UK National Health Service, (though two members of the drafting committee have recognized the same correlations between the prevalence of an outcome and relative and absolute differences described above.)
The life expectancy differences cited in the editorial raise somewhat different issues. While seeming to be continuous measures to which the above points would not necessarily apply, life expectancy differences are affected by general changes in mortality, though not in a very predictable way. See pages 6-7 and Table 2 of reference 4. But one cannot simply rely on life expectancy differentials to measure inequalities without considering effects of those changes.
Reference 5-7 give a fair picture of the disarray in health inequalities research in the US as a result of the failure to understand or address the issues described here. The situation is not materially different in the UK or elsewhere in the world. For far too long, mishandling of measurement issues has led to wasted resources and the misleading of policy makers and a public who wrongly assume that research emanating from governments and institutions of higher learning is invariably sound.
1. Marmot M. Goldblatt P. Importance of monitoring health inequalities. BMJ 2013;347:f6576.
2. Scanlan JP. Can we actually measure health disparities? Chance 2006:19(2):47-51: http://www.jpscanlan.com/images/Can_We_Actually_Measure_Health_Dispariti... (Accessed Nov. 7, 2013)
3. Scanlan JP. Race and mortality. Society 2000;37(2):19-35:
http://www.jpscanlan.com/images/Race_and_Mortality.pdf (Accessed Nov. 7, 2013)
4. Scanlan JP. The Misinterpretation of health inequalities in the United Kingdom. British Society for Populations Studies Conference 2006, Southampton, England, Sept. 18-20, 2006:
http://www.jpscanlan.com/images/BSPS_2006_Complete_Paper.pdf (Accessed Nov. 7, 2013)
5. Scanlan JP. Letter to Harvard University, Oct. 9, 2013: http://jpscanlan.com/images/Harvard_University_Measurement_Letter.pdf (Accessed Nov. 7, 2013)
6. Scanlan JP. Measuring health and healthcare inequalities. Federal Committee on Statistical Methodology 2013 Research Conference, Washington, DC, Nov. 4-6, 2013: http://jpscanlan.com/images/2013_FCSM_Presentation.ppt (Accessed Nov. 7, 2013)
7. Mortality and Survival page of jpscanlan.com: http://www.jpscanlan.com/images/Mortality_and_Survival.pdf (Accessed Nov. 7, 2013)
8. Scanlan JP. Measuring health inequalities by an approach unaffected by the overall prevalence of the outcomes at issue. Royal Statistical Society Conference 2009, Edinburgh, Scotland, Sept. 7-11, 2009: http://www.jpscanlan.com/images/Scanlan_RSS_2009_Presentation.ppt (Accessed Nov. 7, 2013)
9. Keppel KG, Pearcy JN. Measuring relative disparities in terms of adverse events. J Public Health Manag Pract 2005;11:479–83.
10. Keppel K, Pamuk E, Lynch J, et al. Methodological issues in measuring health disparities. Vital Health Stat 2005;2(121):1–16: http://www.cdc.gov/nchs/data/series/sr_02/sr02_141.pdf (Accessed Nov. 7, 2013)
11. Pay for Performance sub-page of Measuring Health Disparities page of jpscanlan.com:
http://www.jpscanlan.com/measuringhealthdisp/payforperformance.html (Accessed Nov. 7, 2013)
12. King NB, Harper S, Young ME. Use of relative and absolute effect measure in reporting health inequalities: structured review. BMJ 2012;345:e544 doi: 10.1136/BMJ.e5774).
13. Kelly PJ, Antony M, Bonnefoy J, et al. The social determinants of health: Developing an evidence base for political action: Final Report to World Health Organization Commission on the Social Determinants of Health from the Measurement and Evidence Knowledge Network (Oct. 2007): http://www.who.int/social_determinants/resources/mekn_report_10oct07.pdf (Accessed Nov. 7, 2013)
14. Houweling TAJ, Kunst AE, Huisman M, Mackenbach JP. Using relative and absolute measures for monitoring health inequalities: experiences from cross-national analyses on maternal and child health. International Journal for Equity in Health 2007;6:15: http://www.equityhealthj.com/content/6/1/15 (Accessed Nov. 7, 2013)
Competing interests: No competing interests