What Can We Learn from Medicare's Least Costly Beneficiaries?
Posted: 13 Jun 2007
Health policy makers concerned with explosive growth in US Medicare expenditures generally belong to the Willy Sutton school of managementgo where the money is. But focusing on costly patients and expensive services may divert attention from potentially significant savings at the opposite end of the spectrum. The aim of this paper is to identify mutable factors that are predictive of low spending without compromising receipt of needed services. Data for the study come from the 2002 Medicare Current Beneficiary Survey. The sample N=8,455 after exclusion of respondents with missing values for key variables. We divide the sample into 5 equal size quintiles based on cumulative medical spending and characterize beneficiaries in each decile by the following domains: (1) health indicators (self reported health, ADL limitations, presence of common chronic diseases), (2) demographic factors (age, gender, race, marital status, education), (3) access indicators (income, rural residence, Medicare supplemental health insurance, prescription coverage, usual source of care), (4) health habits (smoking, preventive health measures, obesity), (5) health care seeking/avoidance behaviors (attitudes toward providers), and (6) health shocks (health significantly worse or better than a year ago). As expected, the health indicators reflect few health problems and high levels of self-reported health in the lowest spending quintile with steadily worsening health status thereafter. The opposite is seen in health insurance and drug coverage, with the lowest rates in quintile 1 (84% and 60%, respectively) rising to 95% and 78%, respectively, in quintile 5. However, the most startling findings relate to a cluster of sociodemographic factors (never married, nonwhite and/or Hispanic, poverty level income, and less than high school education) that are disproportionately represented in the lowest and highest deciles. Never married beneficiaries represent 22% of quintile 1, 14% of quintile 3, and 17% of quintile 5. Nonwhites are 25% of quintile 1, 16% in quintile 3, and 21% in quintile 5. Beneficiaries with no high school education are 17% of quintile 1, 13% of quintile 3, and 18% of quintile 5. These U shaped distributions suggest the presence of strong mitigating factors that initially predict low spending but after a tipping point work in the opposite direction. To better understand these relationships we subset two socioeconomic status groups defined, respectively, as low SES (under 150% of FPL and less than a high school education), and high SES (above 300% FPL and some post-graduate education). We then estimated piecewise regression models of medical spending by quintile to accommodate sign switches among the sociodemographic variables. The high SES group had a much higher level of positive health investments (flu shot, usual source of care) and lower levels of negative health investments (smoking, obesity, care avoidance behaviors) compared to the low SES group throughout the spending range. The regression results showed that positive health investments were associated with higher spending in the lowest quintile but then reduced spending in the upper quintiles. The opposite was true for negative health investments. These findings suggest that improving positive health behaviors will generate future savings in Medicare, but longitudinal analysis is necessary to prove the point.
Suggested Citation: Suggested Citation