Lee and Schwarz Respond

Abstract

We agree with Kawada that the ideal data set for our study would include a confounder-free experiment that asks the self-rated health (SRH) question in different contexts (or order) and follows study participants over time to assess subsequent morbidity and mortality. This data set does not exist. As an approximation with minimal methodological noncomparability, we used the National Health Interview Survey linked with the National Death Index and the Health and Retirement Study (HRS) to examine SRH context effects on health outcome predictions. The main focus of stratified logit models is the interaction between SRH contexts and interview language on mortality prediction examined indirectly because of data limitations. The purpose of these models is not to compare odds ratios between surveys but to examine whether SRH is a significant predictor. If this proven utility of SRH holds universally, SRH should be significant for all groups in both surveys. This was not the case for Spanish-interviewed Hispanics in HRS which asked SRH without a health context. Figure 1 may clarify this point because it includes subsequent mortality rates calculated separately for baseline-year respondents in each response category of SRH. If SRH is a good predictor, one would expect the rates to increase from the “excellent” to “poor” SRH category. Echoing the logit models, all groups in both surveys show this increasing pattern except for Spanish-interviewed Hispanics in HRS, for whom mortality rates did not differ by SRH report. We analyzed models suggested by Kawada and found that Hispanic ethnicity was associated with lower mortality rates in both surveys consistent with the literature.1 FIGURE 1— Percentage of subsequent deaths for each response category of self-rated health from base year by race, ethnicity, and interview language: Health and Retirement Study 1992-2008 and National Health Interview Survey 1997-2004. In a related experimental study, we examined how variations of SRH context affect the relationship between SRH and current comorbidity.2 As in our previous experimental study, the responses of English-speaking respondents were not affected by context.3 By contrast, Spanish-speaking respondents reported higher SRH within a health context than without a health context, and more so the higher their comorbidity status. We believe that the mechanism behind this culture-specific context effect reflects cultural differences in health conceptualization4,5 and cognition6 manifested through interview language.7,8 Establishing a conceptual framework for how respondents cognitively process SRH will be a meaningful step for improving measurement of this important item.9,10