Jump to comment:

• Confounding?
  David E Walker
  Published on: 4 December 2006
• Difficulties in comparing relative differences across subgroups
  James Scanlan
  Published on: 22 August 2006

• Published on: 4 December 2006

  Confounding?

  • David E Walker, Computer programmer

  Dear Editor

  Does the article "Marital Status and Longevity in the United States Population" correct for possible confounding? The subjects could have different degrees of mental and emotional stability, leading to better or worse lifestyle choices, risk-taking behaviour, and attention to health. Maybe people who are more sedate, emotionally stable, successful, avoid drugs and heavy use of alcohol, etc. are more l...

  Show More

  Dear Editor

  Does the article "Marital Status and Longevity in the United States Population" correct for possible confounding? The subjects could have different degrees of mental and emotional stability, leading to better or worse lifestyle choices, risk-taking behaviour, and attention to health. Maybe people who are more sedate, emotionally stable, successful, avoid drugs and heavy use of alcohol, etc. are more likely to get married, less likely to get divorced, and also more likely to take care of themselves, engage in fewer dangerous hobbies, and take better care of their health.

  In order to correct for such possible factors, the authors of the study would need access to the results of psychological tests on the subjects, and to apply sophisticated adjustments to account for personality differences. The results of the study might not be valid after such a heavy degree of required adjustment, even if the data were available.

  David Walker

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 22 August 2006

  Difficulties in comparing relative differences across subgroups

  • James Scanlan, Attorney

  Dear Editor,

  In endeavoring to understand the mechanisms underlying the greater mortality of persons never married, Kaplan and Kronick draw inferences from the sizes of the never married penalties in various subgroups.1 For example, questioning whether there is a cumulative effect of years spent unmarried, the authors note that they find a greater penalty for never marrying among the young than the old. In ques...

  Show More

  Dear Editor,

  In endeavoring to understand the mechanisms underlying the greater mortality of persons never married, Kaplan and Kronick draw inferences from the sizes of the never married penalties in various subgroups.1 For example, questioning whether there is a cumulative effect of years spent unmarried, the authors note that they find a greater penalty for never marrying among the young than the old. In questioning whether it may be poor health that leads to never marrying rather than the reverse, they note that the penalty for never marrying is greatest among those who report themselves to be in excellent health.

  In giving interpretive weight to these comparisons of the sizes of the penalty for never marrying, the authors overlook the tendency whereby the rarer an outcome, the greater the relative difference in experiencing it (though the smaller the relative difference in avoiding it). Such tendency can be observed in virtually any data set that allows one to examine the relationships of two groups with respect to falling above or below various points on a continuum of factors associated with experiencing or avoiding some outcome.2-6.

  Thus, there would be a tendency for the relative difference in mortality between the never married and the rest of the population to be greater among the young (where mortality rates are lower) than among the old (where mortality rates are higher) even if the risk distributions of the never married and the remainder the population are more similar among the young than the among the old (in consequence of the cumulative impact of the longer term social isolation of the older never married or any other factor). It warrants note that we also observe that relative socioeconomic differences in mortality tend to be greater among the young than among the old (though, as with the instant study, the absolute differences are greater than among the old than among the young).7,8. Similarly, the fact that mortality is lower among those in excellent health than those in poorer health will tend to cause the never married penalty to be greater among those in excellent health than among those in poorer health, just as, for example, racial differences in mortality tend to be greater among low risk than high risk groups.4

  These are merely tendencies, of course. Other factors can counteract (or enhance) the tendencies to some degree and sometimes may overwhelm them. Note, for example, that even though mortality is lower among women than among men, until age 65 the never married penalty is greater among men than among women. We also observe generally that socioeconomic differences in mortality are usually greater among women than among men.8. Note, too, that Kaplan and Kronick’s Table 2 shows that, while the pattern whereby the never married penalty decreases with age is present for women as well as men, the differences are small and nonsignificant for women. That these situations seem contrary to the described tendency does not mean that the tendency is not operating, but merely that other factors may be sufficient to outweigh it. In any case, the existence of the tendency renders problematic efforts to draw inferences on the basis of the sizes of the difference between the rates of two groups observed within different segments of the population. Certainly it is futile to attempt to draw such inferences without attempting to take the tendency into account. At the same time, it is difficult to estimate the role of the tendency given that ordinarily we cannot view the actual risk distribution of the groups being examined.1,6.

  The expectation that the sizes of differences in experiencing an outcome will vary depending on the prevalence of an outcome does not apply only to relative differences, but, so far as I can tell, to all measures that are at times employed to appraise the size of differences in the rates at which two groups experience some outcome. In that regard, it warrants note, that while it is difficult enough to meaningfully compare ratios of rates of experiencing an outcome given that the ratios will tend to be larger where the outcome is rarer, such comparisons are even more difficult in the case of odds ratios (which were used in the instant study and which are typically used in studies of socioeconomic differences in morbidity). For while changes in prevalence tend to cause patterns of changes in odds ratios that are similar to the patterns observed for rate ratios in circumstances where an outcome is sufficiently rare that odds ratios approximate rate ratios, where the outcome is more common, patterns of odds ratio changes do not exhibit the same consistency as patterns of rate ratio changes. Thus, in some settings, though not necessarily the instant one, comparisons of odds ratios can be even more problematic than comparisons or rate ratios.1,6.

  Finally, an additional difficulty with the instant study involves the role of AIDS-related deaths. The never married penalty that warrants study mainly involves the effects of social isolation. Gay men tend to be disproportionately never married and to comprise a much higher proportion of AIDS-related deaths than they comprise of the population. And AIDS deaths comprise a substantial proportion of deaths among young men. Arguably some AIDS deaths involve an element of risk-taking that may be an effect of social isolation related to being unmarried. But in general the unmarried status of gay men implicates different considerations vis-à- vis social isolation from the unmarried status of other men. Thus, the inclusion of the high number of AIDS-related deaths among young men compromises the study’s ability to identify the common effects of social isolation.

  References

  1. Kaplan RM, Kronick RG. Marital status and longevity in the United States Population. J Epidemiol Community Health 2006:60:760-765.

  2. Scanlan JP. Can we actually measure health disparities? Chance 2006;19(2):47-51.

  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.

  6. Scanlan JP. Measuring health inequalities. Paper presented at: 5th International Conference on Health Economics, Management and Policy, June 5-7, 2006, Athens, Greece: http://www.jpscanlan.com/images/Measuring_Health_Inequalities.pdf.

  7. Mackenbach JP, Bakker MJ, for the European Network on Interventions and Policy to Reduce Inequalities in Health. Tackling socioeconomic inequalities in health: analysis of European experiences. Lancet 2003;362:1409-1414.

  8. Huisman M., Kunst A.E. Bopp M., et al. Educational inequalities in cause specific mortality in middle-aged and older men and women in eight western European populations. Lancet 2005;36:493-500.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top