Self-reported health and adult mortality risk: An analysis of cause-specific mortality

Abstract

The relationship between self-reported health and mortality is well documented, but less well understood. This study uses the National Health Interview Survey linked to mortality data from the National Death Index to examine the association between self-reported health and a comprehensive set of underlying cause of death and multiple cause of death categories. We also examined whether gender moderates the relationship between self-reported health and cause-specific mortality risk. Results show that the relationship between self-rated health and mortality differs by cause of death and by number of causes. Deaths due to diabetes, infectious and respiratory diseases, and a higher number of causes are most strongly associated with subjective health. Self-reported health also exhibits a moderately strong association with deaths due to heart disease, stroke, and cancer. In contrast, self-rated health is only weakly or not associated with deaths due to accident, homicide, and suicide. The relationship between self-reported health and mortality risk is also found to be stronger among men for several causes, although not for all. These findings should help researchers and policy-makers to better understand the specific predictive power of this important global measure of health.

Introduction

The relationship between self-reported health status and subsequent risk of mortality has become the topic of a substantial body of research in recent years (Idler & Angel, 1990; Idler & Kasl, 1991). While the studies differ by question wording, location of the sample, length of the follow-up period, and extent of other health indicators and covariates included in the models, a great deal of commonality in the results of the studies remains. Overall, self-reported health is a strong predictor of mortality risk, even after accounting for known demographic, social, and medical risk factors (Idler & Angel, 1990; Idler & Kasl, 1991; Hays, Schoenfeld, Blazer, & Gold, 1996; Idler & Benyamini, 1997; Benyamini & Idler, 1999; Strawbridge & Wallhagen, 1999; Idler, Russell, & Davis, 2000; Franks, Gold, & Fiscella, 2003). In fact, the odds of mortality in follow-up studies for people who report poor health are generally 1.5–3.0 times higher than for people who report excellent health at baseline (Benyamini & Idler, 1999). As a result of the strong predictive power of self-reported health, it is important to arrive at a more thorough understanding of the relationship between self-reported health and mortality. Idler et al. (2000), for example, urge researchers that “studies of self-ratings of health must move to a new stage in which health outcomes for study samples are more fully explored and potential sex differences in effects are systematically addressed” (p. 882).

One potentially illuminating line of research involves the specification of the relationship between self-reported health and specific causes of death. Subjective health measures should, in fact, predict death from certain causes better than others. For instance, if individuals base such ratings on specific known illnesses, on health behaviors, and on family history, self-reported health should be more strongly predictive of deaths from diabetes, for example, in comparison to deaths from homicides. Similarly, it is also possible that subjective health predicts death from reported multiple causes of death better than for death from a reported single cause (Israel, Rosenberg, & Curtin, 1986). That is, poor ratings of health may coincide with death due to more complex combinations of causes in comparison to deaths due to single underlying causes of death. Unfortunately, research in this area is scarce.

Research involving more specific causes of death may also be useful in determining which mechanisms drive the relationship between self-reported health and mortality risk, as alluded to above. Previous research has suggested that the relationship may exist because individuals have access to information about their current and future health status, survival probabilities, or changes in future risk behaviors that is not obtainable by other means (Idler et al., 2000). Individuals may also take into account mental and emotional health, in addition to physical (Benyamini & Idler, 1999).

In addition to these possible mechanisms, it is important to note that self-rated health may also be influenced by numerous factors that could confound the relationship between self-rated health and mortality. In particular, various demographic and socioeconomic factors that are associated with mortality differentials may also influence how an individual rates his or her health status. Age, gender, race, ethnicity, education, and income, among other factors, have all been found to be significantly associated with subjective ratings of health. These factors may confound the relationship between self-rated health and mortality because they influence both the independent and dependent variables. For instance, individuals with higher levels of socioeconomic status (SES) report better self-rated health (Franks et al., 2003) and are characterized by lower levels of mortality than their lower SES counterparts (Sorlie et al., 1992; Pappas et al., 1993; Elo & Preston, 1996; Rogers, Hummer, & Nam, 2000). Health behaviors may also confound the relationship between self-rated health and mortality. Poor health behaviors, such as cigarette smoking or lack of exercise, may cause individuals to rate their health more negatively and also have been shown to lead to increased mortality risks (Rogers et al., 2000). Not controlling for these confounding variables would lead to biased estimates of the self-reported health–mortality relationship and, therefore, demographic, socioeconomic, and health behavior factors are included in our analyses.

We found only one previous study that examined the relationship between self-rated health and cause-specific mortality (Rogers et al., 2000). This study found that subjective health predicts subsequent mortality risk quite strongly for each of four broad categories of underlying causes: circulatory diseases, cancers, respiratory diseases, and other causes. However, self-reported health was not associated with the category of deaths labeled social pathologies, which included accidents, suicide, homicides, and cirrhosis of the liver. While this study took the first step to a better understanding of how subjective health influences mortality, the study did not include the examination of more specific causes of death, nor did it include the specification of multiple causes of death. Furthermore, it did not address an important issue that is currently being debated in this literature: the role of gender.

Indeed, several studies have found that the association between self-rated health and mortality risk is stronger (or only exists) among men (Hays et al., 1996; Idler et al., 2000). In fact, a recent review by Idler and Benyamini in 1997 (and updated in Benyamini & Idler, 1999) concluded that self-rated health predicts subsequent mortality risk more strongly among men than among women. However, reports of similar associations among both genders (Jylha, Guralnik, Ferrucci, Jokela, & Heikkinen, 1998; Strawbridge & Wallhagen, 1999), or a stronger association for women, have also been found (Grant, Piotrowski, & Chappell, 1995). If gender differences are found, it could be an indication that subjective health reflects serious health problems differently for men and women (Jylha et al., 1998).

The purpose of this paper is to address the role of self-rated health in predicting cause-specific mortality and mortality by multiple causes of death and to determine if gender differences exist within these relationships. In pursuing these goals, we use a large, nationally representative, prospective data set (with more than 50,000 deaths during the follow-up period) to conduct an in-depth examination of self-rated health and mortality risk among adults in the United States. Specifically, we address three questions: (1) Does the relationship between self-rated health and mortality differ by specific cause of death? (2) Does the relationship between self-rated health and mortality differ by the number of causes of death listed on the death certificate? (3) Does gender moderate the association of self-rated health with cause-specific mortality and mortality by number of causes of death?