Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Recognizing why dichotomous and continuous measures may yield contrary results
- James P. Scanlan (11 June 2007)
|
|
|||
|
James P. Scanlan, Attorney 1529 Wisconsin Ave., NW, Washington, DC 20007, USA
Send response to journal:
|
The study by Chandola et al.1 found social inequalities in self reported health within the Whitehall cohort to increase with age. As the authors point out, other studies of the Whitehall workforce have found inequalities in mortality to decline with age and studies of other populations have usually found inequalities in both mortality and self- assessed health to decline with age. Some part of the reason for apparent difference between the results of the Chandola study and so much other work is clear enough. Studies of mortality differences necessarily rely on a dichotomous variable. Studies of self-assessed health have usually relied on categorical appraisals of health dichotomized to “health less than good” and “health good or better.” Research into inequalities concerning these dichotomous variables has typically relied on relative differences in experiencing the adverse outcome – i.e., mortality or “health less than good” – and has usually found that, as these outcomes increased with age, relative differences in experiencing them have declined. In regarding these patterns to reflect meaningful decreases in health inequality, however, such research has failed to consider the statistical tendency whereby as an outcome becomes more common, relative differences in experiencing it tend to decline, while relative differences in avoiding it tend to increase.2-7. In cases where Whitehall data allow the calculation (as in Marmot and Shipley 8), one does in fact see that, while relative differences in mortality are smaller among the older Whitehall population, the relative differences in survival rates are larger among that population. Very likely, the SF-36 responses underlying the Chandola study would show that if the response to the question about perception of general health were dichotomized in the manner of most self-assessed health analyses, there would be a smaller relative difference in “health less than good” among the older population, simply because health is poorer among the older population. But one would probably find a larger relative difference in rates of “health good or better” in that older population as well. Both in the case or mortality and dichotomized self-assessed health, however, neither the smaller relative difference among the older population as to one outcome nor the larger relative difference as to the opposite outcome can by itself indicate whether health inequalities are greater or smaller among the older population in any meaningful sense – that is, in the sense that distributions of higher- and lower-graded employees are more similar or dissimilar among the old than the young. Because all measures of difference between rates of experiencing or avoiding an outcome tend to change when the prevalence of an outcome changes, it is not clear that dichotomous measures can be effectively employed to appraise the size of health inequalities in different settings.2,6. Continuous measures of health inequalities may offer greater promise for determining whether health inequalities are larger in one population than another or are changing over time, with the size of the difference based on the effect size of the difference between averages.6 But is not clear that SF-36 score are the types of continuous measures that are not subject to some of the same tendencies that make reliance on dichotomous measures of health inequality problematic. An SF- 36 score is a composite of points allocated to categorical responses. And in situations where the prevalence of an outcome changes thereby causing various categorical response rates to change solely because of that change in prevalence – and in which case one ought not to consider there to have occurred a meaningful changes in the relative situation of two groups ¬– the SF36 score are likely to change in some manner as well. The matter might be compared to the situation of longevity, which seems a continuous variable, but which is a function of mortality. And overall declines in mortality that do not reflect meaningful changes in inequalities tend to cause differences in longevity also to change (as shown in Table 2 of reference 6). Hence, longevity does not provide a means of identifying changes in the relative situation of two groups that are not solely the result of changes in mortality. The same may hold for SF-36 scores. So any further reliance on those scores to measure inequality should be preceded by an exploration of the extent to which the patterns of changes in score differences are affected by overall changes in the outcomes underlying the categorical responses in the questionnaires. What seem more clearly to be truly continuous variables are those such as blood pressure and serum cholesterol levels that were used by Ferrie et al. 9 to measure changes in Whitehall inequalities over time. But it is not clear that even these measures can effectively identify changes in inequalities that are not solely a result of changes in the overall prevalence of an outcome. Consider the implications of smoking, for example. For reasons explained in references 2-7, as smoking declines, one should expect larger relative decreases among groups that smoke less (advantaged groups) than among those that smoke more (disadvantaged groups), though a larger relative increase in rates of not smoking among the latter than the former. But to the extent that changes in relative rates of smoking (or not smoking) are simply the standard consequences of a decline in smoking, according to the reasoning in references 2-6, such changes should not be regarded as a meaningful worsening or improvement of the relative well-being of disadvantaged groups. Yet, given the importance of smoking as a risk factor, even changes in differential smoking patterns that would not in themselves be regarded as changing health inequality would likely cause some change in the effect size of the difference between average blood pressure levels or other continuous measures of health. In such circumstances, whether the changes in effect size on such measures should be regarded as meaningful changes in health inequalities is a difficult issue. In any case, a priority area of health inequalities research ought to be the identification of measures that can distinguish meaningful changes in health inequalities from those that are solely functions of changes in overall prevalence of outcomes. References: 1. Chandola T, Ferrie J, Sacker A, Marmot M. Social inequalities in self reported health in early old age: follow-up of prospective cohort study. BMJ 2007:334:9900996. 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_Disparities.pdf 3. Scanlan JP. Measuring health disparities. J Public Health Manag Pract 2006;12(3):294 [Lttr]: http://www.nursingcenter.com/library/JournalArticle.asp?Article_ID=641470 4. Scanlan JP. Race and mortality. Society 2000;37(2):19-35: http://www.jpscanlan.com/images/Race_and_Mortality.pdf 5. Scanlan JP. Divining difference. Chance 1994;7:38-39,48: http://jpscanlan.com/images/Divining_Difference.pdf 6. Scanlan JP. The misinterpretation of health inequalities in the United Kingdom. Paper presented at: British Society for Population Studies Annual Conference 2006, Southampton, England, Sept. 18-20, 2006: http://www.jpscanlan.com/images/BSPS_2006_Complete_Paper.pdf 7. Carr-Hill R, Chalmers-Dixon P. The Public Health Observatory Handbook of Health Inequalities Measurement. Oxford: SEPHO; 2005:http://www.sepho.org.uk/extras/rch_handbook.aspx 8. Marmot MG, Shipley MJ. Do socioeconomic mortality differences decrease wit retirement? 25 year follow up of civil servants from the first Whitehall study. BMJ 1996;313:1177-80. 9. Ferrie JE, Shipley MJ, Davey Smith GD. Change in health inequalities among British civil servants: the Whitehall II study. J Epidemiol Community Health 2002:56:922-926. Competing interests: None declared |
|||