Food Stamps, Blood Sugar Control, and Medicare Costs of Older Diabetics

Year: 2009

Research Center: The Harris School of Public Policy Studies, University of Chicago

Investigator: Nicholas, Lauren Hersch

Institution: University of Michigan

Project Contact:
Lauren Hersch Nicholas
University of Michigan
Institute for Social Research
426 Thompson Street, Room 3005
Ann Arbor, MI 48106
Phone: 734-764-2562


Incidence of diabetes is rapidly escalating in the United States despite a Healthy People 2010 goal to decrease its disease and economic burden. This problem is particularly pronounced among low-income and older Americans. Nearly a quarter of the U.S. elderly population has been diagnosed with diabetes, facing increased risk of premature death, disability, heart disease, kidney failure, and other complications. The illness burden not only reduces patient quality of life but imposes significant economic costs. Medicare spent 75 percent more on elderly diabetics than nondiabetics in 2005. Despite a well-documented role of diet in diabetes management and evidence of a socioeconomic gradient in diabetes, little is known about the relationship between participation in USDA’s Food Stamp Program (FSP)—now the Supplemental Nutrition Assistance Program (SNAP)—and diabetes management.

Food stamp receipt may impact health through several mechanisms. Food stamp receipt may directly affect health by altering the amount or quality of food purchased, or if benefits function as an income transfer, by increasing household budgets for spending on food and other health inputs. Food stamp recipients may have difficulty in smoothing consumption across the benefit cycle, purchasing larger amounts of food early in the month and running out of food prior to receiving the next month’s benefit. Periods of food deprivation, such as at the end of the benefit cycle, can impede blood sugar control if households are unable to purchase appropriate foods for medical nutrition therapy. Because theoretical predictions about food stamp effects on diabetic health outcomes are ambiguous, this study empirically tests the relationship between food stamp receipt and multiple health outcomes for diabetics.

This study analyzes data from the Health and Retirement Study (HRS), a nationally representative longitudinal survey of older Americans. The HRS reinterviews respondents and participating spouses biennially. In each wave, households report FSP participation by month, income, assets, and participation in means-tested programs. Respondent-level data detail health measures, including diabetes and other chronic conditions, as well as sociodemographic variables, including age, race, education level, household composition, and employment history. Respondents also report two health behaviors, whether they smoke (counter-indicated for diabetics) and whether they engage in physical exercise (recommended for diabetics), which can indicate positive (negative) selection into the FSP.

HRS survey data is linked to Medicare administrative claims data for respondents who have previously consented to its release and measured hemoglobin A1c levels. The claims data include several health outcomes of interest. This study estimates fixed-effect regressions of (logged) total annual Medicare spending for each respondent, as well as counts of outpatient medical visits, and indicators for whether a beneficiary is hospitalized during the year for any reason, hospitalized for diabetes, or has end-stage renal disease (ESRD). The HRS Medicare data allow several controls for nonrandom selection into the FSP. Rates of food stamp take-up (participation) are particularly low among eligible elderly adults, raising concerns that participants differ from nonparticipants in typically unobservable ways that may affect program take-up and health. Medicare claims data allow controls for clinical compliance with diabetes treatment guidelines, providing information about positive or negative selection into program participation. Individual fixed effects absorb time-invariant sources of individual heterogeneity.

A biomarker collection module provides measured A1c levels for some HRS respondents, a measure of blood sugar control over the past 2-3 months. Fractional logit regressions and propensity score matching techniques are used to assess the relationship between A1c level and food stamp receipt in cross-sectional data.

This work documents the high and rapidly growing prevalence of diabetes among elderly Americans. This trend is particularly salient for elderly food stamp recipients, 32 percent of whom were diabetic in 2006. In unadjusted data, food stamp recipients have slightly higher Medicare use and spending and worse glycemic control than eligible nonparticipants. In fixed-effect regression analysis, the difference in Medicare spending, outpatient medical visits, and blood sugar control is not statistically significant between recipients and nonrecipients. Food stamp recipients are 7 percentage points more likely to experience an inpatient hospitalization, although heart disease is the primary diagnosis for the additional hospitalizations rather than hospitalizations for diabetes. Food stamp receipt is associated with a 3-percentage-point decrease in the probability of end-stage renal disease for non-Whites, who are disproportionately affected by ESRD.

Biomarker data are used to consider the relationship between food stamp receipt and glycemic control. This relationship is one pathway through which food stamp receipt could influence use of and spending on health care. Using both fractional logit regression and propensity score matching in cross-sectional data, we find an insignificant relationship between food stamp receipt and health. Food stamp receipt is associated with an economically and statistically insignificant 0.01- to 0.02-percentage-point increase in hemoglobin A1c level. Food stamp recipients are insignificantly less likely to have HbA1c levels that are in compliance with American Diabetes Association treatment guidelines but also less likely to have high levels, which indicates severe diabetes management problems.

This study contributes to an underdeveloped literature on the health effects of food stamp receipt. Overall, the study finds little evidence that food stamps hurt or improve the health of elderly diabetics. The findings are consistent with the related food stamp and obesity literature, which also fails to find causal evidence that food stamp receipt impacts obesity. While many studies have considered the high rates of obesity among food stamp recipients, the growing diabetes epidemic among food stamp recipients and eligible nonparticipants has received little attention. An important contribution of this research is the documentation of the high and growing prevalence of diabetes in the food-stamp-eligible and recipient population. With nearly a third of elderly food stamp recipients currently diabetic, policy intervention may be appropriate. A third of food stamp recipient diabetics also report food insufficiency. Changes to program design, such as larger benefits or more frequent benefit disbursements, or additional guidance on food choices, may benefit older diabetics.

Direct inquiries about this study to the Project Contact listed above.