The Relationship Between Body Mass Index and Long-Term Medical Costs Among Medicare Beneficiaries

The relationship between body mass index (BMI) and health care spending has been well documented using cross-sectional data. However, that contemporaneous relationship may not accurately predict spending over a longer time period for several reasons. First, a person’s weight can change over time and the associated effects on spending are uncertain.  For example, higher spending may accompany unintentional weight loss due to illness, which is particularly common among the elderly, and intentional weight loss may not eliminate the health effects and higher spending associated with obesity. Second, differential mortality by BMI may affect the number of years a person lives to incur medical expenses. Finally, estimates of the relationship between BMI and spending typically represent an average over a range of ages, and that relationship may change as people age. Our analysis avoids such complexities by directly studying the relationship between a point-in-time BMI measure and a person’s subsequent long-term Medicare spending. 

We use a unique combination of data to study the relationship between BMI and 10-year Fee-For-Service Medicare spending and mortality. Using a crosswalk file, we match respondents from the 1999-2004 Medicare Current Beneficiary Surveys (MCBS) to administrative data from the 1999-2013 Medicare Beneficiary Summary Files (MBSF). The MCBS includes self-reported height and weight, demographic data, and administrative spending measures. However, the rotating panel design of the MCBS only allows respondents to be followed for three years at most. The MBSF allows us to continue to track Medicare spending, mortality, and disease diagnoses of all MCBS respondents through 2013.

Preliminary results show differences in the relationship between BMI and healthcare spending over time. Cross-sectional estimates using spending in the year a person is first surveyed by the MCBS imply that, on average, a beneficiary with a BMI of 18 (in the underweight range) costs 40 percent more than a person with a BMI of 25 (at the lower end of the overweight range). However, over a 10-year period the average person with a BMI of 18 at the beginning of the period costs 5 percent less than a person with an initial BMI of 25, largely due to differential mortality and mean reversion. Conversely, beneficiaries initially classified as obese have higher relative spending over 10 years than cross-sectional estimates from the survey year indicate. Our estimates imply that a person with an initial BMI of 35 (in the obese range) costs 14 percent more than a person with a BMI of 25 in the survey year but 22 percent more over a 10-year period. Our results suggest this difference is driven by Medicare spending growing more rapidly for beneficiaries with higher BMIs in the survey year (with little increase in mortality below a BMI of 40). Additionally, analysis of 10-year Medicare spending by type of service suggests that, while spending of all types rise with overweight and obesity, spending on home health care, skilled nursing facilities, dialysis, and durable medical equipment is notably higher.