This study reports significant association between health behaviors and health care charges, generally consistent with previous evidence suggesting that physical inactivity, overweight, and obesity account for an important portion of health care expenditures. The results of this study indicate that physical inactivity, overweight, and obesity are associated with 23% (95% CI, 10%–34%) of all health care charges at a moderately sized Minnesota health plan and 27% (95% CI, 10%–37%) of national health care charges.
The distribution of these charges across defined groups of patients is not uniform. Per- person charges associated with physical inactivity and overweight or obesity are disproportionately higher for individuals who are old, male, and have heart disease, diabetes, or both. Yet despite higher per-person charges among these groups, nearly half of associated charges are incurred in the demographic segments of the health plan population with the largest number of members — members aged 40 to 64 with no heart disease or diabetes. This suggests that a broad, population-wide approach to managing physical inactivity and obesity would be better than a strategy that limits interventions to people with the highest health care charges.
One interesting finding is the observation that after adjusting for other factors in the model, physical activity level was not related to charges among women. There are several possible explanations for this finding. First, it may be that the types of health care services that men of this age use are more related to physical activity than the types of services that women use. Second, it is possible that women and men are differentially misclassified by our measure of physical activity. For example, if the types of physical activity that women typically engage in are less likely to be reported in response to our simple query about physical activity, then we would be more likely to misclassify women than men. In either case, the exploration of sex-related differences in the health impact of physical activity should receive more attention.
The finding of higher health care charges among former smokers (compared with current smokers and people who have never smoked) has been reported elsewhere (17
) and is most likely explained by the tendency for smokers who develop health problems to be more likely to quit compared to smokers who remain free of health problems (24
Current discussions of how to control rapidly rising health care expenditures often focus on modifying payment mechanisms, increasing patient cost-sharing, or carefully managing access to certain types of care. Our data emphasize the potential importance of primary prevention and effective management of overweight and physical inactivity as a cost-control strategy. Effective interventions to ameliorate these problems are available and can be delivered at relatively low cost across a wide range of settings, from physicians' offices to telephone-based case management to community health education using a variety of communication methods (25
). Despite convincing evidence that adverse health risks increase charges and that improving risk factors lowers subsequent charges, health plans, employers, and other payers have been slow to invest resources in initiatives.
Improvements in weight and physical activity may reduce health care expenditures within 2 to 3 years in some groups of patients (17
). Yet payers are reluctant to assume new costs in an age of accelerating medical care cost inflation, and leaders of care-delivery systems may believe that their expertise is more related to high-technology procedural care than to interventions that target lifestyle or behavior (27
). The high proportion of overall charges associated with inactivity and overweight may prompt both payers and delivery-system leaders to rethink priorities.
A number of factors limit the interpretation of these data. First, the study was conducted at a single health plan, and our analysis is limited to a white population aged 40 and older, excluding pharmacy charges. Second, the variables considered in our models were derived from survey data and automated administrative databases; they did not include sophisticated physiologic measures. Third, we adjusted our predicted cell-charge estimates for self-reported heart disease or diabetes. Other chronic debilitating diseases might have also been used as adjustors, but we selected heart disease and diabetes because of their prevalence and their established contribution to health care expenditures. Fourth, our national estimates of physical inactivity and BMI reflect 2001 population behaviors.
Fifth, we were unable to report a significant interaction between the two main effects, physical inactivity and BMI. To detect an r2 change of 0.01 from an interaction term, with two-tailed power of 80% (α = .05), a sample would need 700 observations per interaction cell. Our weighted sample was too small to achieve that level of power; our sample had two-tailed power of 13% to detect an interaction between physical inactivity and BMI. Therefore, our results do not rule out a significant interaction between physical inactivity and BMI; the question of interaction effects remains open for further study.
Finally, one might argue that physically inactive subjects experience higher health care charges because impairment or health-related problems limit their ability to be active. However, we found that the proportion of charges associated with physical inactivity and overweight or obesity was similar for subjects without self-reported chronic disease compared with subjects reporting chronic disease. Moreover, we tested this hypothesis (data not presented) by estimating a model after excluding subjects whose activities are limited because of impairment or a health-related problem. The model estimated after excluding subjects with limited activity was similar to our model including all subjects; our estimate of health care charges associated with physical inactivity and overweight or obesity did not change. This evidence increases our confidence that reverse causality is not a substantial driver of our study results.
One difficult issue in the interpretation of our study results is the extent to which the associations that we measure are the result of causation. The data allow us to estimate the difference in health care charges between individuals with selected characteristics and individuals without these characteristics. We do not know how changing characteristics will impact future health care expenditures, so we remain uncertain about the extent to which our counterfactual is a good measure of the effect of improved health behaviors on health care charges. Generally, our estimates provide a theoretical upper bound on the amount of resources that may be invested in programs to improve risk-factor profiles.
Our results are interesting and important. We jointly estimate the effects of physical inactivity and overweight or obesity associated with health care charges and find they constitute a significant portion of total health care charges. From a clinical and public health point of view, benefit may be derived by addressing physical inactivity, overweight, and obesity in all segments of the population. However, from a cost-effectiveness point of view, those with behavioral risk factors and the highest health care charges may be the strongest candidates for interventions designed to improve risk profiles and reduce expenditures. More work is needed to explore the potential of behavior interventions as a strategy to improve health and to control accelerating health care charges.