Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Does the Supplemental Nutrition Assistance Program Affect Hospital Utilization Among Older Adults? The Case of Maryland

Does the Supplemental Nutrition Assistance Program Affect Hospital Utilization Among Older... This study sought to examine whether Supplemental Nutrition Assistance Program (SNAP) participation and benefit levels are associated with reduced subsequent hospital and emergency department utilization in low-income older adults. Study participants were 68,956 Maryland residents aged ‡65 years who were dually enrolled in Medicare and Medicaid (2009–2012). Annual inpatient hospital days and costs and emergency department visits were modeled as a function of either 1-year lagged SNAP participation or lagged SNAP benefit amounts, controlling for sociodemographic characteristics, autoregressive effects, year, health status, and Medicaid participation. SNAP participation (adjusted odds ratio [aOR]= 0.96, 95% confidence interval [CI]: 0.93, 0.99), and, among participants, each $10 increase in monthly benefits (aOR= 0.99, 95% CI: 0.99–0.99) are associated with a reduced likelihood of hospitalization, but not emergency department use. The authors estimate that enrolling the 47% of the 2012 population who were eligible nonparticipants in SNAP could have been associated with $19 million in hospital cost savings. Accounting for the strong effects of health care access, this study finds that SNAP is associated with reduced hospitalization in dually eligible older adults. Policies to increase SNAP participation and benefit amounts in eligible older adults may reduce hospitalizations and health care costs for older dual eligible adults living in the community. Keywords: food assistance, health care utilization, hospitalization, older adults, socioeconomic status Introduction found that non–health sector resources are associated with a reduced risk of hospital readmission in low-income older ne third of US older adults, comprising 13 million adults, suggesting that social determinants of health affect Oolder adults, currently live on incomes less than 200% of hospital utilization. the poverty level, according to Census data. It has long been Social service programs exist to help low-income seniors known that adults living under or near the poverty line rely meet their basic needs. Specifically, food assistance pro- more heavily on emergency department (ED)-based health grams, such as the Supplemental Nutrition Assistance Pro- care and are hospitalized more often than their higher income gram (SNAP), provide supplemental household income for 1,2 peers. Excess hospital utilization in this population was food, and may therefore improve health outcomes for lower once believed to be preventable by improving health care income older adults. SNAP provided, on average, $129 in access. However, disparities exist among older adults who supplemental monthly income for an average of 1.3 people in 1 6 have health insurance through Medicare, and are not at- an older adult household in 2014. This transfer comprises a 3,4 tributable to access to primary care providers. One study relatively large supplemental income source for these adults, Department of Acute and Chronic Care, Johns Hopkins School of Nursing, Baltimore, Maryland. Department of Community-Public Health, Johns Hopkins School of Nursing, Baltimore, Maryland. Benefits Data Trust, Philadelphia, Pennsylvania. Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. Northwestern University, Department of Economics, Northwestern University, Evanston, Illinois. Benefits Data Trust, Philadelphia, Pennsylvania. The Hilltop Institute, University of Maryland Baltimore County, Baltimore, Maryland. ª Laura J. Samuel et al. 2018; Published by Mary Ann Liebert, Inc. This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 88 FOOD ASSISTANCE AND HOSPITAL UTILIZATION 89 whose average monthly gross income was $876. Also, by tar- nursing home for 9 months or more of the prior year, because geting financial support toward food needs, SNAP can improve they likely would not have had sufficient opportunity to enroll access to a higher quality diet for food insecure adults. There is in SNAP during the prior calendar year. Individuals who were evidence of reduced caloric intake, poorer dietary quality, and enrolled in fee-for-service Medicare and enrolled in Medicaid greater risk of hypoglycemia for low-income adults at the end of for at least 1 month during the calendar year were included for the month when funds run low. Therefore, greater SNAP benefits that year. Individuals who enrolled in Medicare or died during may facilitate chronic disease management for nutrition-sensitive the calendar year were excluded if they had less than 6 months conditions, which may account for evidence of reduced chronic of claims data for the year, as in prior work. Medicare disease hospital utilization. Also, because adults with food in- Advantage enrollees were excluded because full claims data security also often report cost-related medication nonadherence for hospital ascertainment were not available. 11,12 and difficulty paying bills, food assistance may allow them to Medicaid claims and sociodemographic data were merged use their limited money for medications and other health-related with Medicare claims and SNAP program utilization data. necessities. The study was approved by the Institutional Review Boards Several studies have tested whether SNAP is associated at Johns Hopkins Medicine and University of Maryland, with reduced hospital utilization, but results are inconsistent. Baltimore County. A crosswalk between Medicare and For example, Medicaid inpatient hospital spending growth Medicaid recipients’ IDs generated for research purposes declined in Massachusetts following an increase in SNAP enabled data merging. The dependent variables in this study benefit amounts in that state and results were most dramatic included annual inpatient hospital day count, annual inpa- among those with nutrition-sensitive chronic conditions, sug- tient hospital cost, and annual ED visit count, including gesting that improved chronic disease management contrib- visits that resulted in an inpatient admission and those re- uted to the decline. However, in a study of older patients with solved on an outpatient basis. The dependent variables were diabetes that adjusted for self-management behaviors, SNAP measured during each year, 2010–2012, while SNAP vari- participants had similar Medicare spending and a 7% greater ables were measured from 2009–2011 to capture lagged risk of hospitalization than their peers. These results con- effects. It was not possible to measure ED costs separately tradict the hypothesis that SNAP participation is associated from inpatient hospital cost for visits resulting in admission. with reduced hospitalization, but may be biased by adjusting Inpatient hospital cost was measured by the total payment for self-management, because SNAP’s effect on hospitaliza- amount in Medicare claims data because Medicare is the tion likely is partly related to improved disease management. primary payer of hospital care for dually eligible adults. Understanding how SNAP relates to health outcomes of SNAP participants were compared to nonparticipants. low-income older adults is particularly important and timely Because individuals self-select into SNAP and may quali- in light of recent interest in innovative approaches that tatively differ from nonparticipants, the study team also target the social needs of older adults. Approximately 42% measured the average benefit amount among participants of income-eligible older adults participate in SNAP. Al- (scaled in $10 increments) to evaluate potential dose re- though expanding access to SNAP and increasing SNAP sponse. Both SNAP variables were treated as time variant benefits would undoubtedly increase benefit expenditures, and modeled with 1-year lags, meaning that outcomes were program costs are modest relative to hospital and emergency regressed on the prior year’s SNAP values. Lagged models care costs. Understanding how SNAP participation and ben- minimize the threat of reverse causality, whereby indi- efit amounts affect hospital utilization could inform policy viduals in poorer health select into SNAP participation. directed at improving the value and impact of federal enti- Sociodemographic variables included sex, race/ethnicity tlement programs. Therefore, this study sought to quantify (analyzed as black, white, Hispanic, other and unknown SNAP participation rates in a population of low-income older [ref]), age, annual household income (in $1000 increments), Maryland residents dually eligible for Medicare and Medic- and dummy variables indicating partially Medicaid eligible aid and test 2 hypotheses in this population. First, the study (ie, eligible for assistance with Medicare premiums and cost team hypothesized that SNAP participation is associated with sharing, but not for full Medicaid coverage) and Medicaid reduced subsequent hospital utilization. Second, the team eligibility based on medically needy spend down, both of hypothesized that greater SNAP benefits are associated with which were considered to be socioeconomic proxies. Because reduced hospital utilization among participants. Dually eli- poor health may contribute to both SNAP participation and gible older adults are all likely income eligible for SNAP and hospital utilization, the study team measured the number of have claims data suitable for these analyses. chronic conditions defined by a modified version of the Chronic Conditions Warehouse algorithm, based on diag- nosis codes in either Medicare and Medicaid claims. Receipt Methods of the Medicaid Home and Community-Based Services The sample for this study was comprised of the population Waiver, which indicates functional limitations, was measured of Maryland residents aged 65 years and older who were as a dichotomous variable. Finally, because older adults who dually enrolled in Medicare and Medicaid at any time be- participate in SNAP also may be more continuously enrolled tween 2010 and 2012. Because Medicaid income eligibility in Medicaid, and because more continuous Medicaid partic- criteria were more stringent than those for SNAP, all were ipation is strongly associated with hospital utilization in du- 5,19 likely income eligible for SNAP. Study inclusion was deter- ally eligible older adults, the study team measured the mined for each calendar year. Hospital utilization was esti- proportion of the year enrolled in Medicaid. Covariate data, mated as a function of prior year’s SNAP participation and except for chronic condition values, were drawn exclusively benefit amount. Nursing home residents are not eligible for from Medicaid data, and all covariates were treated as time SNAP. Therefore, individuals were excluded who resided in a variant, except for sex and race/ethnicity. 90 SAMUEL ET AL. Statistical approach pared with being white or having unknown race/ethnicity. They were less likely to be partially eligible for Medicaid. SNAP participation and benefit amounts were modeled Rates for hospitalization and the sociodemographic profile with 1-year lags in separate models. Zero-inflated negative of the population remained mostly consistent across obser- binomial regression models estimated inpatient days and ED vation years (see online Supplementary Table S1; Supple- visits. The zero-inflated negative binomial fit the skewed mentary Data are available online at www.liebertpub.com/ distribution of inpatient hospital days and ED visits best pop). The average cost for inpatient admissions for partici- because of the large frequency of nonusers during the year. pants hospitalized in 2012 was $25,091. Correlated outcomes were addressed by adjusting for auto- Adjusting for sociodemographic and health characteris- regressive effects and applying robust standard error esti- 17 tics, (Table 2, Model 1), SNAP participants had, on average, mates. Time trends were addressed by adjusting for study 14% lower odds of hospitalization and 10% lower odds of year. Model 1 adjusted for sex, race/ethnicity, age, annual an ED visit in the subsequent year than nonparticipants. income, partially Medicaid eligible, Medicaid spend down These associations were attenuated after additionally ad- eligibility, study year, autoregressive effects, chronic con- justing for Medicaid participation, but SNAP participation dition count, and Medicaid community waiver status. Model continued to be statistically significantly associated with 4% 2 additionally adjusted for proportion of the year partici- reduced odds of hospitalization in the final model (Model pating in Medicaid. The inpatient hospital cost models are 2). Likewise, in models that adjusted for sociodemographic similar in concept, using a Heckman 2-step selection model and health characteristics, SNAP participants had a 10% for the same reason as the zero-inflated negative binomial lower likelihood for each additional inpatient day if hospi- model: there is a preponderance of zero utilization in any talized and a 4% lower likelihood of each additional ED given year and a skewed distribution of spending. The visit if they utilized it (Model 1), but SNAP participation Heckman model uses a probit specification in the first step to was not associated with either outcome after additional indicate a propensity for zero versus nonzero spending, and adjustment for Medicaid participation. Adjusting for socio- the second step is a weighted ordinary least squares speci- demographic and health characteristics in SNAP partici- fication of the amount of spending, conditioned on the pants, a $10 increase in monthly benefit amount was propensity to be selected into the group of persons having associated with 2% lower odds of either hospitalization hospital costs. Inpatient hospital cost models adjusted for all or ED utilization (Model 1). In models that additionally covariates. adjusted for Medicaid participation, a $10 increase in Results from the Heckman 2-stage model also were used monthly SNAP benefit continued to be statistically signifi- to estimate the potential cost implications of expanding cantly associated with 1% reduced odds of hospitalization. access to SNAP to nonparticipants in 2012. Results from the Full model results are reported in Supplementary Tables S2 first stage were used to estimate potential savings attribut- and S3. able to fewer hospital admissions and results from the sec- In fully adjusted models, SNAP participants were 1.5 ond stage were used to estimate the potential savings percentage points less likely to incur an inpatient hospital attributable to less costly stays if admitted. The fully ad- expense (ie, be hospitalized; Table 3). Among those who justed results from the first stage of the Heckman model were hospitalized, SNAP participants had 5.8% lower ex- provide the predicted difference in probability of hospital penses than nonparticipants. Therefore, the study team es- admission between SNAP participants and nonparticipants, timates that expanding SNAP benefits to the 25,018 assuming mean values for all model covariates. This value nonparticipants in 2012 could have been associated with was then multiplied by the number of SNAP nonparticipants total savings of $19 million, with approximately half of the and the average cost of inpatient hospitalization to estimate savings ($9.4 million) related to an estimated 375 averted the potential cost savings attributable to averted hospital admissions and the other half ($9.7 million) related to less admissions. The fully adjusted results from the second stage costly hospital stays (Table 4). Among SNAP participants, a of the Heckman model provide the estimated percent re- $10 increase in SNAP was associated with a 0.2 percentage duction in cost among those admitted. This was multiplied point lower probability of incurring inpatient-related hos- by the expected number of hospitalizations in SNAP non- pital costs and a 1% lower average inpatient cost for those participants to calculate the potential cost savings attribut- who were hospitalized (Table 3).This savings is not included able to less costly stays. The total potential cost savings is in the cost savings estimation to avoid double counting. the sum of the cost results from both stages. Results Discussion A total of 68,956 older adults in Maryland were dually Results from this study indicate that SNAP participation enrolled in both fee-for-service Medicare and Medicaid and and increased SNAP benefits among participants were as- were eligible for this study at some point during 2010–2012. sociated with reduced hospitalization rates, but not ED visit Of those, 53,646 individuals were eligible for the study in rates, in dually eligible older adults. Notably, in this study of 2012, the most recent year of data. In 2012, 26% of par- older adults who should be eligible, only about half partic- ticipants were hospitalized and approximately 53% were ipated in SNAP. Despite the vast body of literature doc- enrolled in SNAP (Table 1). SNAP participants had a 3 umenting poor health in low-income older adults, effective percentage point lower likelihood than SNAP nonpartici- strategies to improve health outcomes for this vulnerable pants of being hospitalized, a 1 percentage point lower group remain scant. In this study, SNAP benefits were re- likelihood of having an ED visit, and were more likely to be lated to lower hospital utilization in a population that was younger, female, and black, Hispanic, or other race com- continuously enrolled in Medicare and in regression models Table 1. Characteristics of Maryland Adults Aged ‡65 Dually Enrolled in Both Medicare and Medicaid, by Supplemental Nutrition Assistance Program Participation in 2012 (n = 53,646) SNAP participants (%) Nonparticipants (%) Total sample 28,628 (53) 25,018 (47) P Age 65–69 years 14,672 (27) 7305 (29) 7367 (26) <0.01 70–74 years 11,621 (22) 4406 (18) 7215 (25) 75–79 years 9976 (19) 4164 (17) 5812 (20) 80–84 years 8098 (15) 3693 (15) 4405 (15) ‡85 years 9279 (17) 5450 (22) 3829 (13) Sex Female 37,138 (69) 19,955 (70) 17,183 (69) 0.01 Male 16,508 (31) 8673 (30) 7835 (31) Race/Ethnicity Black 17,704 (33) 10,191 (36) 7513 (30) <0.01 White 21,034 (39) 10,560 (37) 10,474 (42) Hispanic 2869 (5) 1694 (6) 1175 (5) Other 6835 (13) 4281 (15) 2554 (10) Unknown 5204 (10) 1902 (7) 3302 (13) Medicaid community waiver No 46,732 (87) 24,903 (87) 21,829 (87) 0.36 Yes 6914 (13) 3725 (13) 3189 (13) Partially Medicaid eligibile No 31,347 (58) 16,984 (59) 14,363 (57) <0.01 Yes 22,299 (42) 11,644 (41) 10,655 (43) Medicaid eligible by spend down No 52,723 (98) 28,207 (99) 24,516 (98) <0.01 Yes 923 (2) 421 (1) 502 (2) Mean number of chronic conditions 2.8 2.6 2.9 <0.01 Admitted to hospital No 40,031 (74) 21,238 (76) 18,793 (73) <0.01 Yes 13,775 (26) 6734 (24) 7041 (27) Had emergency department visit No 31,674 (59) 16,634 (59) 15,040 (58) 0.03 Yes 22,132 (41) 11,338 (41) 10,794 (42) Limited to individuals who were enrolled in fee-for-service Medicare for ‡6 months of the year and who were not residing in a nursing home for more than 9 months of 2011. Based on chi-square test statistic. SNAP, Supplemental Nutrition Assistance Program. Table 2. Associations Between Supplemental Nutrition Assistance Program Participation (n = 68,956) and Benefit Amount (n = 26,874) with Hospitalization and Emergency Department Visits, Maryland Adults Aged ‡65 Years Enrolled in Both Medicare and Medicaid (2010–2012) a a Model 1 Model 2 Model 1 Model 2 Any hospitalization Any emergency department visits OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Previous year SNAP participation (n = 68,956) 0.86 (0.84–0.89) 0.96 (0.93–0.99) 0.90 (0.83–0.97) 0.98 (0.91–1.06) Previous year mean monthly SNAP amount 0.98 (0.97–0.98) 0.99 (0.99–0.99) 0.98 (0.98–0.99) 0.99 (0.99–1.00) in participants ($10) (n = 26,874) Number of inpatient hospital Number of emergency department days among the hospitalized visits among utilizers IRR (95% CI) IRR (95% CI) IRR (95% CI) IRR (95% CI) Previous year SNAP participation (n = 68,956) 0.90 (0.81–0.99) 0.92 (0.82–1.03) 0.96 (0.93–0.99) 0.98 (0.95–1.01) Previous year mean monthly SNAP amount 0.99 (0.98–0.99) 0.99 (0.99–1.00) 0.99 (0.99–1.00) 1.00 (1.00–1.00) in participants ($10) (n = 26,874) Associations estimated from zero-inflated negative binomial regression estimated with robust standard errors. All models adjusted for autoregressive effects, study year, age, sex, race/ethnicity, annual income, partial Medicaid eligibility, Medicaid spend down eligibility, chronic condition count, and Medicaid community waiver status. Model additionally adjusted for proportion of year participating in Medicaid. CI, confidence interval; IRR, incident rate ratio; OR, odds ratio; SNAP, Supplemental Nutrition Assistance Program. 91 92 SAMUEL ET AL. Table 3. Associations Between Supplemental Nutrition Assistance Program Participation (n = 68,956) and Benefit Amount (n = 26,874) with Inpatient Hospital Expenditures, Maryland Adults Aged ‡65 Years Enrolled in Both Medicare and Medicaid (2010–2012) Any hospitalization Marginal change in probability for any >0 hospital cost (95% CI) Previous year SNAP participation (n = 68,956) -1.5% (-2.0% to -1.1%) Previous year mean monthly SNAP amount in -0.2% (-0.2% to -0.1%) participants ($10) (n = 26,874) Inpatient hospital Medicare cost among the hospitalized Estimated elasticity for ln(cost) (95% CI) Previous year SNAP participation (n = 68,956) -5.8% (-8.4% to -3.3%) Previous year mean monthly SNAP amount -1.0% (-1.3% to -0.7%) in participants ($10) (n = 26,874) Associations estimated from Heckman regression model, adjusted for autoregressive effects, study year, age, sex, race/ethnicity, annual income, partial Medicaid eligibility, Medicaid spend-down eligibility, chronic condition count, Medicaid community waiver status, and proportion of year participating in Medicaid. Evaluated at means of all covariates. CI, confidence interval; SNAP, Supplemental Nutrition Assistance Program. that adjust for the proportion of the year enrolled in Med- icaid. It is notable that all had access to both Medicare and Medicaid because policy makers have increased access to Table 4. Steps to Obtain Cost Savings of Expanding health care for low-income groups thinking that that alone the Supplemental Nutrition Assistance Program to Nonparticipants in 2012 (n = 25,018), Based would reduce high hospital utilization in low-income groups. on Heckman Model Estimates Indeed, SNAP associations in this study were attenuated after adjusting for the proportion of the year with Medicaid cov- Cost savings from averted admissions erage, and evidence elsewhere shows that continuous Med- Change in probability of any admission 1.5 icaid coverage predicts less hospital-based care. However, in the year (%) 1,3–5 results from the present study add to evidence suggesting Multiplied by: Number of 25,018 nonparticipants that high hospital utilization is attributable to both social and Gives: Estimated number of averted 375 medical determinants of health. More specifically, these results admissions identify food assistance as a social determinant of hospital uti- Multiplied by: Average annual cost 25,091 lization, but not ED utilization. Therefore, SNAP may predict of inpatient admissions ($) fewer hospital admissions, even if it does not decrease the fre- Gives: Estimated cost savings from 9,415,900 quency of ED visits. Results from other studies suggest that averted admissions ($) improving health care access also may not reduce ED utiliza- 20,21 Cost savings from less costly hospital stays tion. Together, these results suggest that hospital and ED Number of nonparticipants admitted 7041 utilization are attributable to different mechanisms. Importantly, to hospital inpatient hospital admission is determined by health care pro- Less: Estimated number of averted 375 viders, whereas visiting the ED is generally the decision of the admissions individual. Therefore, inpatient hospital utilization is likely a Gives: Estimated no. of nonparticipants 6666 better measure of individual health status than ED utilization. still admitted to hospital These findings supplement evidence that investment in Percentage change in cost for admitted 5.8 SNAP may improve health outcomes and reduce inpatient persons (%) hospital spending. For example, the American Recovery and Multiplied by: Average annual cost 25,091 of inpatient admissions ($) Reinvestment Act of 2009, which expanded SNAP eligi- Gives: Reduction in average cost 1455 bility and increased the benefit amount, has been credited for admitted persons ($) with reducing the prevalence of food insecurity in SNAP Multiplied by: Estimated no. of 6666 income-eligible US households by 2% and may have nonparticipants still admitted 10 slowed growth in Medicaid inpatient hospital spending. to hospital These results build on prior studies by demonstrating that Gives: Estimated cost savings from 9,700,490 not only SNAP participation, but greater benefit amounts less costly hospital stays ($) among participants, is associated with reduced hospitaliza- Total cost savings tion rates. This is notable because participation is suscepti- Total estimated cost savings ($) 19,116,390 ble to self-selection, but benefit amounts are assigned based a largely on household financial need. Therefore, benefit See Table 3. amount results are less susceptible to confounding and re- See Table 1. Estimated based on participants who were hospitalized in 2012. verse causality than participation results. FOOD ASSISTANCE AND HOSPITAL UTILIZATION 93 In this study, SNAP participation also was associated with participation in food programs, and making results suscepti- less costly stays among those who were hospitalized. Based ble to survival bias. Conversely, however, these results sug- on the models, the study team estimates that expanding gest that greater SNAP benefits may confer a health benefit for SNAP access to nonparticipating dual eligible older adults low-income adults, even at older ages. In addition, lag times in Maryland could have resulted in inpatient hospital cost for potential health effects of food assistance programs savings of $19 million in 2012. Based on the average per are unknown. Lagged effects may be longer than the 1 year capita costs of the SNAP program, the team estimates that modeled in this study, so the associations between SNAP and the federal government would spend approximately $39 hospital utilization may be underestimated. Conversely, the million if they extended SNAP benefits to the 2012 income- study team could have missed key proximal effects by lagging eligible nonparticipants in the study sample. Therefore, ig- the models for a year. Also, averted hospital stays may differ noring issues of inefficiencies in taxation, approximately in cost from the typical hospital stay, which would bias the half of the cost of administering SNAP could be recouped cost savings estimate. This study is strengthened by using data by the federal government in reduced Medicare inpatient for an entire state population of dually eligible older adults, hospital spending. Further work is needed to quantify the reducing nonresponse bias in a low-income group. Also, this potential savings attributable to changes in other health care study is less susceptible to reverse causality than previous utilization outcomes. cross-sectional studies because of use of lagged exposure There are at least 2 potential reasons for the associations modeling. found in this study. First, SNAP participation reduces food 23,24 insecurity and minimizes the adverse effect of food Public Health Implications insecurity on dietary quality and obesity. This is notewor- This study found low rates of SNAP program participa- thy because food insecurity is linked to worse dietary 8 25 tion (only 53% participated in the most recent study year) in quality, increased risk of chronic diseases, and increased the population of dually eligible Maryland older adults, all risk of hospitalization. Therefore, it is plausible that im- of whom are likely income eligible for SNAP. National data proved access to SNAP and increased benefits for partici- estimates a participation rate of 42% among income eligible pants may improve food security and dietary quality in older adults, suggesting that the program is underutilized. households of low-income older adults, and this may reduce Results from this study suggest that improved access to SNAP hospital utilization. may reduce hospitalization for low-income older adults. To- An alternative potential reason for these results is that gether, these results suggest that strategies to improve access SNAP provides supplemental income that reduces financial to SNAP for eligible older adults likely will improve health strain for older adults living under or near the poverty line. outcomes, despite years of accumulated exposure to food Financial strain, or the lack of adequate money for basic 27 insecurity and financial strain in this population. These results needs, is linked with earlier mortality and greater risk for 28 29 have implications for practice and policy. both malnutrition and disability, which contribute to 5,30 hospitalization. Individuals experiencing financial strain 11,12 Practice implications struggle to afford food, heat, and health care, and may be forced to choose between these basic needs. Therefore, Efforts to increase SNAP participation may include tar- improving access to programs meeting any basic need will geted eligibility screening and enrollment assistance. Older improve an individual’s ability to afford the other basic adults tend to underutilize the program. Health care pro- needs. This idea is consistent with evidence of increased viders and health care payers can invest in efforts to screen for access to food after implementation of Medicare Part D, food insecurity and income eligibility for SNAP and facilitate which improved financial access to medications for older SNAP enrollment. Resources are available to support en- adults. Therefore, it is also plausible that any supplemental rollment and advocacy efforts nationally and state-level ad- income could be associated with reduced subsequent hospital- vocacy organizations. Community health workers, primary based care for low-income older adults because it enhances care provider practices, and not-for-profit service providers their ability to meet basic needs and reduces finance-related are well positioned to provide such screening and referral, but stress exposure. they need to be compensated and supported for services to address social and economic determinants of health. Limitations and strengths Policy implications As with other SNAP studies, SNAP participants may differ from nonparticipants on unmeasured characteristics. This Policy actions can improve access to SNAP and increase may have biased associations and the study cost savings benefit amounts. Several specific policy strategies can facil- calculations. For example, older adults who do not receive itate the SNAP enrollment process for older adults. First, SNAP may benefit from other meal assistance programs, such states can reduce enrollment requirements by implementing as Meals on Wheels or congregate meals. This likely would the Elderly Simplified Application Project. This program, bias associations toward the null because of unmeasured re- implemented in 7 states, streamlines income and expense ceipt of food assistance in the control group. Conversely, verification by matching data from existing sources, extends older adults who participate in SNAP may be more likely to certification periods to 36 months, and waives the re- enroll in other public benefits programs, which could con- certification interview. Second, states can coordinate SNAP found results if such programs collectively reduce hospital and Supplemental Security Income enrollment, as is done in utilization. Also, this study is limited to an older adult pop- the Combined Application Project. This program increased ulation, precluding measurement of the cumulative lifetime SNAP participation rates in South Carolina at a time when 94 SAMUEL ET AL. national rates were declining and is currently implemented 5. Iloabuchi TC, Mi D, Tu W, Counsell SR. Risk factors for in 18 states. Third, states can leverage administrative data early hospital readmission in low-income elderly adults. J Am Geriatr Soc 2014;62:489–494. from Medicaid, the Low Income Home Energy Assistance 6. Farson Gray K, Kochhar S. Characteristics of Supplemental Program (LIHEAP), and SNAP programs, as Maryland has, Nutrition Assistance Program households: Fiscal year 2014. to identify income eligible older adults who are not enrolled Alexandria, VA: USDA, Food and Nutrition Service, 2015. in SNAP and conduct targeted outreach to increase par- 7. Nguyen BT, Shuval K, Bertmann F, Yaroch AL. The ticipation. The Affordable Care Act incentivizes states to Supplemental Nutrition Assistance Program, food insecu- harmonize administrative data sets to facilitate Medicaid rity, dietary quality, and obesity among US adults. Am J enrollment, and these efforts can facilitate SNAP enrollment Public Health 2015;105:1453–1459. as well. 8. Tarasuk V, McIntyre L, Li J. Low-income women’s dietary Besides enhancing beneficiary access, states can enhance intakes are sensitive to the depletion of household resources benefit amounts for vulnerable older adults. For example, in one month. J Nutr 2007;137:1980–1987. the State of Maryland recently passed legislation ensuring 9. Seligman HK, Bolger AF, Guzman D, Lo´pez A, Bibbins- that all SNAP beneficiaries aged 62 and older receive a Domingo K. Exhaustion of food budgets at month’s end minimum benefit of $30 monthly by supplementing the and hospital admissions for hypoglycemia. Health Aff 2014; federal benefit with state funds as needed. Furthermore, 33:116–123. efforts to significantly cut federal spending on SNAP ben- 10. Sonik RA. Massachusetts inpatient Medicaid cost response efits through block granting or other structural changes may to increased Supplemental Nutrition Assistance Program have adverse consequences. benefits. Am J Public Health 2016;106:443–448. 11. Sattler ELP, Lee JS. Persistent food insecurity is associated with higher levels of cost-related medication nonadherence Conclusion in low-income older adults. J Nutr Gerontol Geriatr 2013; This study found that SNAP participation and greater 32:41–58. benefit amounts are associated with lower inpatient hospital 12. Berkowitz SA, Seligman HK, Choudhry NK. Treat or eat: utilization in a state population of low-income older adults. food insecurity, cost-related medication underuse, and un- These findings have public health implications because the met needs. Am J Med 2014;127:303–310.e3. majority of US older adults who are income eligible for 13. Nicholas LH. Can food stamps help to reduce Medicare SNAP do not participate. As public and private sector health spending on diabetes? Econ Hum Biol 2011;9:1–13. care partners shift to outcomes-driven, value-based care, 14. Shortell SM. Bridging the divide between health and health care. JAMA 2013;309:1121–1122. social service programs such as SNAP will be a critical tool 15. Farson Gray K, Cunnyngham K. Trends in Supplemental in improving health outcomes for low-income seniors across Nutrition Assistance Program participation rates: Fiscal the country. Year 2010 to Fiscal Year 2014. Alexandria, VA. 2016. www.fns.usda.gov/sites/default/files/ops/Trends2010-2012- Acknowledgments Summary.pdf Accessed September 19, 2016. We are grateful for the partnership of the Maryland De- 16. The Lewin Group. Picture of housing and health: Medicare and Medicaid use among older adults in HUD-assisted partment of Health and Mental Hygiene and the Maryland housing. Washington, DC: US Department of Health and Department of Human Resources for their support in pro- Human Services, 2014. viding data for these analyses. 17. Singer JD, Willett JB. Applied longitudinal data analysis: modeling change and event occurrence. New York: Oxford Author Disclosure Statement University Press, 2003. Drs. Samuel, Szanton, Wolff, and Ong, Ms. Cahill, Ms. 18. Buccaneer, A General Dynamics Company. CCW chronic Zielinskie, and Mr. Betley declared no conflicts of interest conditions: combined Medicare and Medicaid data. Wa- shington, DC: Centers for Medicare & Meidcaid Services, with respect to the research, authorship, and/or publication Federal Coordinated Healthcare Office, 2012. of this article. The authors received the following financial 19. Bindman AB, Chattopadhyay A, Auerback GM. Interrup- support: This study was supported by a grant from the Ro- tions in Medicaid coverage and risk for hospitalization for bert Wood Johnson Foundation. ambulatory care-sensitive conditions. Ann Intern Med 2008; 149:854–860. References 20. Taubman SL. Medicaid increases emergency-department 1. Blustein J, Hanson K, Shea S. Preventable hospitalizations use: evidence from Oregon’s health insurance experiment. and socioeconomic status. Health Aff (Millwood) 1998;17: Science 2014;343:263–269. 177–189. 21. Wright B, Potter AJ, Trivedi A. Federally Qualified Health 2. Oster A, Bindman AB. Emergency department visits for Center use among dual eligibles: rates of hospitalizations and ambulatory care sensitive conditions: insights into pre- emergency department visits. Health Aff 2015;34:1147–1155. ventable hospitalizations. Med Care 2003;41:198–207. 22. Nord M, Prell M. Food security improved following the 3. Ricketts TC, Randolph R, Howard HA, Pathman D, Carey 2009 ARRA increase in SNAP benefits. Washington, DC: T. Hospitalization rates as indicators of access to primary USDA Economic Research Service, 2011. care. Health Place 2001;7:27–38. 23. Mabli J, Ohls J. Supplemental Nutrition Assistance Pro- 4. Nyweide DJ, Anthony DL, Bynum JPW, et al. Continuity gram participation is associated with an increase in house- of care and the risk of preventable hospitalization in older hold food security in a national evaluation. J Nutr 2015; adults. JAMA Intern Med 2013;173:1879–1885. 145:344–351. FOOD ASSISTANCE AND HOSPITAL UTILIZATION 95 24. Ratcliffe C, McKernan SM, Zhang S. How much does the through Medicare and Medicaid. N Engl J Med 2015;374: Supplemental Nutrition Assistance Program reduce food 1–4. insecurity? Am J Agric Econ 2011;93:1082–1098. 33. Institute of Medicine. Meeting the dietary needs of older 25. Seligman HK, Laraia BA, Kushel MB. Food insecurity adults: exploring the impact of the physical, social, and is associated with chronic disease among low-income cultural environment: workshop summary. Washington, DC: NHANES participants. J Nutr 2010;140:304–310. National Academies Press, 2016. 26. Phipps EJ, Singletary SB, Cooblall CA, Hares HD, Brait- 34. Supplemental Nutrition Assistance Program. Elderly sim- man LE. Food insecurity in patients with high hospital plified application project guidance. Washington, DC: utilization. Popul Health Manag 2016;19:414–420. USDA Food and Nutrition Service, 2015. 27. Szanton SL, Allen JK, Thorpe RJ, Seeman T, Bandeen- 35. United States Department of Agriculture, Social Security Roche K, Fried LP. Effect of financial strain on mortality in Administration. Combined Application Projects. Guidance community-dwelling older women. J Gerontol B Psychol for States Developing Projects. 2005. www.fns.usda.gov/ Sci Soc Sci 2008;63:S369–S374. sites/default/files/CAPsDevelopmentGuidance.pdf Accessed 28. Samuel LJ, Szanton SL, Weiss CO, Thorpe RJ, Semba RD, July 15, 2016. Fried LP. Financial strain is associated with malnutrition 36. Kauff J, Dragoset L, Clary E, Laird E, Makowsky L, Sama- risk in community-dwelling older women. Epidemiol Res Miller E. Reaching the underserved elderly and working Int 2012;2012:696518. poor in SNAP: evaluation findings from the Fiscal Year 29. Szanton SL, Thorpe RJ, Whitfield K. Life-course financial 2009 pilots. 2014. www.fns.usda.gov/sites/default/files/ strain and health in African-Americans. Soc Sci Med 2010; SNAPUnderseved-Elderly2009.pdf Accessed August 16, 71:259–265. 2016. 30. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Address correspondence to: Med Sci 2001;56:M146–M156. Laura J. Samuel, PhD, CRNP 31. Madden JM, Graves AJ, Zhang F, et al. Cost-related medi- Johns Hopkins University School of Nursing cation nonadherence and spending on basic needs following 525 North Wolfe Street, Room 446 implementation of Medicare Part D. JAMA 2008;299: 1922–1928. Baltimore, MD 21205 32. Alley DE, Asomugha CN, Conway PH, Sanghavi DM. Accountable health communities—addressing social needs E-mail: lsamuel@jhmi.edu http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Population Health Management Pubmed Central

Does the Supplemental Nutrition Assistance Program Affect Hospital Utilization Among Older Adults? The Case of Maryland

Population Health Management , Volume 21 (2) – Apr 1, 2018

Loading next page...
 
/lp/pubmed-central/does-the-supplemental-nutrition-assistance-program-affect-hospital-eLZj7OGHjb

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Pubmed Central
Copyright
© Laura J. Samuel et al. 2018; Published by Mary Ann Liebert, Inc.
ISSN
1942-7891
eISSN
1942-7905
DOI
10.1089/pop.2017.0055
Publisher site
See Article on Publisher Site

Abstract

This study sought to examine whether Supplemental Nutrition Assistance Program (SNAP) participation and benefit levels are associated with reduced subsequent hospital and emergency department utilization in low-income older adults. Study participants were 68,956 Maryland residents aged ‡65 years who were dually enrolled in Medicare and Medicaid (2009–2012). Annual inpatient hospital days and costs and emergency department visits were modeled as a function of either 1-year lagged SNAP participation or lagged SNAP benefit amounts, controlling for sociodemographic characteristics, autoregressive effects, year, health status, and Medicaid participation. SNAP participation (adjusted odds ratio [aOR]= 0.96, 95% confidence interval [CI]: 0.93, 0.99), and, among participants, each $10 increase in monthly benefits (aOR= 0.99, 95% CI: 0.99–0.99) are associated with a reduced likelihood of hospitalization, but not emergency department use. The authors estimate that enrolling the 47% of the 2012 population who were eligible nonparticipants in SNAP could have been associated with $19 million in hospital cost savings. Accounting for the strong effects of health care access, this study finds that SNAP is associated with reduced hospitalization in dually eligible older adults. Policies to increase SNAP participation and benefit amounts in eligible older adults may reduce hospitalizations and health care costs for older dual eligible adults living in the community. Keywords: food assistance, health care utilization, hospitalization, older adults, socioeconomic status Introduction found that non–health sector resources are associated with a reduced risk of hospital readmission in low-income older ne third of US older adults, comprising 13 million adults, suggesting that social determinants of health affect Oolder adults, currently live on incomes less than 200% of hospital utilization. the poverty level, according to Census data. It has long been Social service programs exist to help low-income seniors known that adults living under or near the poverty line rely meet their basic needs. Specifically, food assistance pro- more heavily on emergency department (ED)-based health grams, such as the Supplemental Nutrition Assistance Pro- care and are hospitalized more often than their higher income gram (SNAP), provide supplemental household income for 1,2 peers. Excess hospital utilization in this population was food, and may therefore improve health outcomes for lower once believed to be preventable by improving health care income older adults. SNAP provided, on average, $129 in access. However, disparities exist among older adults who supplemental monthly income for an average of 1.3 people in 1 6 have health insurance through Medicare, and are not at- an older adult household in 2014. This transfer comprises a 3,4 tributable to access to primary care providers. One study relatively large supplemental income source for these adults, Department of Acute and Chronic Care, Johns Hopkins School of Nursing, Baltimore, Maryland. Department of Community-Public Health, Johns Hopkins School of Nursing, Baltimore, Maryland. Benefits Data Trust, Philadelphia, Pennsylvania. Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. Northwestern University, Department of Economics, Northwestern University, Evanston, Illinois. Benefits Data Trust, Philadelphia, Pennsylvania. The Hilltop Institute, University of Maryland Baltimore County, Baltimore, Maryland. ª Laura J. Samuel et al. 2018; Published by Mary Ann Liebert, Inc. This Open Access article is distributed under the terms of the Creative Commons License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 88 FOOD ASSISTANCE AND HOSPITAL UTILIZATION 89 whose average monthly gross income was $876. Also, by tar- nursing home for 9 months or more of the prior year, because geting financial support toward food needs, SNAP can improve they likely would not have had sufficient opportunity to enroll access to a higher quality diet for food insecure adults. There is in SNAP during the prior calendar year. Individuals who were evidence of reduced caloric intake, poorer dietary quality, and enrolled in fee-for-service Medicare and enrolled in Medicaid greater risk of hypoglycemia for low-income adults at the end of for at least 1 month during the calendar year were included for the month when funds run low. Therefore, greater SNAP benefits that year. Individuals who enrolled in Medicare or died during may facilitate chronic disease management for nutrition-sensitive the calendar year were excluded if they had less than 6 months conditions, which may account for evidence of reduced chronic of claims data for the year, as in prior work. Medicare disease hospital utilization. Also, because adults with food in- Advantage enrollees were excluded because full claims data security also often report cost-related medication nonadherence for hospital ascertainment were not available. 11,12 and difficulty paying bills, food assistance may allow them to Medicaid claims and sociodemographic data were merged use their limited money for medications and other health-related with Medicare claims and SNAP program utilization data. necessities. The study was approved by the Institutional Review Boards Several studies have tested whether SNAP is associated at Johns Hopkins Medicine and University of Maryland, with reduced hospital utilization, but results are inconsistent. Baltimore County. A crosswalk between Medicare and For example, Medicaid inpatient hospital spending growth Medicaid recipients’ IDs generated for research purposes declined in Massachusetts following an increase in SNAP enabled data merging. The dependent variables in this study benefit amounts in that state and results were most dramatic included annual inpatient hospital day count, annual inpa- among those with nutrition-sensitive chronic conditions, sug- tient hospital cost, and annual ED visit count, including gesting that improved chronic disease management contrib- visits that resulted in an inpatient admission and those re- uted to the decline. However, in a study of older patients with solved on an outpatient basis. The dependent variables were diabetes that adjusted for self-management behaviors, SNAP measured during each year, 2010–2012, while SNAP vari- participants had similar Medicare spending and a 7% greater ables were measured from 2009–2011 to capture lagged risk of hospitalization than their peers. These results con- effects. It was not possible to measure ED costs separately tradict the hypothesis that SNAP participation is associated from inpatient hospital cost for visits resulting in admission. with reduced hospitalization, but may be biased by adjusting Inpatient hospital cost was measured by the total payment for self-management, because SNAP’s effect on hospitaliza- amount in Medicare claims data because Medicare is the tion likely is partly related to improved disease management. primary payer of hospital care for dually eligible adults. Understanding how SNAP relates to health outcomes of SNAP participants were compared to nonparticipants. low-income older adults is particularly important and timely Because individuals self-select into SNAP and may quali- in light of recent interest in innovative approaches that tatively differ from nonparticipants, the study team also target the social needs of older adults. Approximately 42% measured the average benefit amount among participants of income-eligible older adults participate in SNAP. Al- (scaled in $10 increments) to evaluate potential dose re- though expanding access to SNAP and increasing SNAP sponse. Both SNAP variables were treated as time variant benefits would undoubtedly increase benefit expenditures, and modeled with 1-year lags, meaning that outcomes were program costs are modest relative to hospital and emergency regressed on the prior year’s SNAP values. Lagged models care costs. Understanding how SNAP participation and ben- minimize the threat of reverse causality, whereby indi- efit amounts affect hospital utilization could inform policy viduals in poorer health select into SNAP participation. directed at improving the value and impact of federal enti- Sociodemographic variables included sex, race/ethnicity tlement programs. Therefore, this study sought to quantify (analyzed as black, white, Hispanic, other and unknown SNAP participation rates in a population of low-income older [ref]), age, annual household income (in $1000 increments), Maryland residents dually eligible for Medicare and Medic- and dummy variables indicating partially Medicaid eligible aid and test 2 hypotheses in this population. First, the study (ie, eligible for assistance with Medicare premiums and cost team hypothesized that SNAP participation is associated with sharing, but not for full Medicaid coverage) and Medicaid reduced subsequent hospital utilization. Second, the team eligibility based on medically needy spend down, both of hypothesized that greater SNAP benefits are associated with which were considered to be socioeconomic proxies. Because reduced hospital utilization among participants. Dually eli- poor health may contribute to both SNAP participation and gible older adults are all likely income eligible for SNAP and hospital utilization, the study team measured the number of have claims data suitable for these analyses. chronic conditions defined by a modified version of the Chronic Conditions Warehouse algorithm, based on diag- nosis codes in either Medicare and Medicaid claims. Receipt Methods of the Medicaid Home and Community-Based Services The sample for this study was comprised of the population Waiver, which indicates functional limitations, was measured of Maryland residents aged 65 years and older who were as a dichotomous variable. Finally, because older adults who dually enrolled in Medicare and Medicaid at any time be- participate in SNAP also may be more continuously enrolled tween 2010 and 2012. Because Medicaid income eligibility in Medicaid, and because more continuous Medicaid partic- criteria were more stringent than those for SNAP, all were ipation is strongly associated with hospital utilization in du- 5,19 likely income eligible for SNAP. Study inclusion was deter- ally eligible older adults, the study team measured the mined for each calendar year. Hospital utilization was esti- proportion of the year enrolled in Medicaid. Covariate data, mated as a function of prior year’s SNAP participation and except for chronic condition values, were drawn exclusively benefit amount. Nursing home residents are not eligible for from Medicaid data, and all covariates were treated as time SNAP. Therefore, individuals were excluded who resided in a variant, except for sex and race/ethnicity. 90 SAMUEL ET AL. Statistical approach pared with being white or having unknown race/ethnicity. They were less likely to be partially eligible for Medicaid. SNAP participation and benefit amounts were modeled Rates for hospitalization and the sociodemographic profile with 1-year lags in separate models. Zero-inflated negative of the population remained mostly consistent across obser- binomial regression models estimated inpatient days and ED vation years (see online Supplementary Table S1; Supple- visits. The zero-inflated negative binomial fit the skewed mentary Data are available online at www.liebertpub.com/ distribution of inpatient hospital days and ED visits best pop). The average cost for inpatient admissions for partici- because of the large frequency of nonusers during the year. pants hospitalized in 2012 was $25,091. Correlated outcomes were addressed by adjusting for auto- Adjusting for sociodemographic and health characteris- regressive effects and applying robust standard error esti- 17 tics, (Table 2, Model 1), SNAP participants had, on average, mates. Time trends were addressed by adjusting for study 14% lower odds of hospitalization and 10% lower odds of year. Model 1 adjusted for sex, race/ethnicity, age, annual an ED visit in the subsequent year than nonparticipants. income, partially Medicaid eligible, Medicaid spend down These associations were attenuated after additionally ad- eligibility, study year, autoregressive effects, chronic con- justing for Medicaid participation, but SNAP participation dition count, and Medicaid community waiver status. Model continued to be statistically significantly associated with 4% 2 additionally adjusted for proportion of the year partici- reduced odds of hospitalization in the final model (Model pating in Medicaid. The inpatient hospital cost models are 2). Likewise, in models that adjusted for sociodemographic similar in concept, using a Heckman 2-step selection model and health characteristics, SNAP participants had a 10% for the same reason as the zero-inflated negative binomial lower likelihood for each additional inpatient day if hospi- model: there is a preponderance of zero utilization in any talized and a 4% lower likelihood of each additional ED given year and a skewed distribution of spending. The visit if they utilized it (Model 1), but SNAP participation Heckman model uses a probit specification in the first step to was not associated with either outcome after additional indicate a propensity for zero versus nonzero spending, and adjustment for Medicaid participation. Adjusting for socio- the second step is a weighted ordinary least squares speci- demographic and health characteristics in SNAP partici- fication of the amount of spending, conditioned on the pants, a $10 increase in monthly benefit amount was propensity to be selected into the group of persons having associated with 2% lower odds of either hospitalization hospital costs. Inpatient hospital cost models adjusted for all or ED utilization (Model 1). In models that additionally covariates. adjusted for Medicaid participation, a $10 increase in Results from the Heckman 2-stage model also were used monthly SNAP benefit continued to be statistically signifi- to estimate the potential cost implications of expanding cantly associated with 1% reduced odds of hospitalization. access to SNAP to nonparticipants in 2012. Results from the Full model results are reported in Supplementary Tables S2 first stage were used to estimate potential savings attribut- and S3. able to fewer hospital admissions and results from the sec- In fully adjusted models, SNAP participants were 1.5 ond stage were used to estimate the potential savings percentage points less likely to incur an inpatient hospital attributable to less costly stays if admitted. The fully ad- expense (ie, be hospitalized; Table 3). Among those who justed results from the first stage of the Heckman model were hospitalized, SNAP participants had 5.8% lower ex- provide the predicted difference in probability of hospital penses than nonparticipants. Therefore, the study team es- admission between SNAP participants and nonparticipants, timates that expanding SNAP benefits to the 25,018 assuming mean values for all model covariates. This value nonparticipants in 2012 could have been associated with was then multiplied by the number of SNAP nonparticipants total savings of $19 million, with approximately half of the and the average cost of inpatient hospitalization to estimate savings ($9.4 million) related to an estimated 375 averted the potential cost savings attributable to averted hospital admissions and the other half ($9.7 million) related to less admissions. The fully adjusted results from the second stage costly hospital stays (Table 4). Among SNAP participants, a of the Heckman model provide the estimated percent re- $10 increase in SNAP was associated with a 0.2 percentage duction in cost among those admitted. This was multiplied point lower probability of incurring inpatient-related hos- by the expected number of hospitalizations in SNAP non- pital costs and a 1% lower average inpatient cost for those participants to calculate the potential cost savings attribut- who were hospitalized (Table 3).This savings is not included able to less costly stays. The total potential cost savings is in the cost savings estimation to avoid double counting. the sum of the cost results from both stages. Results Discussion A total of 68,956 older adults in Maryland were dually Results from this study indicate that SNAP participation enrolled in both fee-for-service Medicare and Medicaid and and increased SNAP benefits among participants were as- were eligible for this study at some point during 2010–2012. sociated with reduced hospitalization rates, but not ED visit Of those, 53,646 individuals were eligible for the study in rates, in dually eligible older adults. Notably, in this study of 2012, the most recent year of data. In 2012, 26% of par- older adults who should be eligible, only about half partic- ticipants were hospitalized and approximately 53% were ipated in SNAP. Despite the vast body of literature doc- enrolled in SNAP (Table 1). SNAP participants had a 3 umenting poor health in low-income older adults, effective percentage point lower likelihood than SNAP nonpartici- strategies to improve health outcomes for this vulnerable pants of being hospitalized, a 1 percentage point lower group remain scant. In this study, SNAP benefits were re- likelihood of having an ED visit, and were more likely to be lated to lower hospital utilization in a population that was younger, female, and black, Hispanic, or other race com- continuously enrolled in Medicare and in regression models Table 1. Characteristics of Maryland Adults Aged ‡65 Dually Enrolled in Both Medicare and Medicaid, by Supplemental Nutrition Assistance Program Participation in 2012 (n = 53,646) SNAP participants (%) Nonparticipants (%) Total sample 28,628 (53) 25,018 (47) P Age 65–69 years 14,672 (27) 7305 (29) 7367 (26) <0.01 70–74 years 11,621 (22) 4406 (18) 7215 (25) 75–79 years 9976 (19) 4164 (17) 5812 (20) 80–84 years 8098 (15) 3693 (15) 4405 (15) ‡85 years 9279 (17) 5450 (22) 3829 (13) Sex Female 37,138 (69) 19,955 (70) 17,183 (69) 0.01 Male 16,508 (31) 8673 (30) 7835 (31) Race/Ethnicity Black 17,704 (33) 10,191 (36) 7513 (30) <0.01 White 21,034 (39) 10,560 (37) 10,474 (42) Hispanic 2869 (5) 1694 (6) 1175 (5) Other 6835 (13) 4281 (15) 2554 (10) Unknown 5204 (10) 1902 (7) 3302 (13) Medicaid community waiver No 46,732 (87) 24,903 (87) 21,829 (87) 0.36 Yes 6914 (13) 3725 (13) 3189 (13) Partially Medicaid eligibile No 31,347 (58) 16,984 (59) 14,363 (57) <0.01 Yes 22,299 (42) 11,644 (41) 10,655 (43) Medicaid eligible by spend down No 52,723 (98) 28,207 (99) 24,516 (98) <0.01 Yes 923 (2) 421 (1) 502 (2) Mean number of chronic conditions 2.8 2.6 2.9 <0.01 Admitted to hospital No 40,031 (74) 21,238 (76) 18,793 (73) <0.01 Yes 13,775 (26) 6734 (24) 7041 (27) Had emergency department visit No 31,674 (59) 16,634 (59) 15,040 (58) 0.03 Yes 22,132 (41) 11,338 (41) 10,794 (42) Limited to individuals who were enrolled in fee-for-service Medicare for ‡6 months of the year and who were not residing in a nursing home for more than 9 months of 2011. Based on chi-square test statistic. SNAP, Supplemental Nutrition Assistance Program. Table 2. Associations Between Supplemental Nutrition Assistance Program Participation (n = 68,956) and Benefit Amount (n = 26,874) with Hospitalization and Emergency Department Visits, Maryland Adults Aged ‡65 Years Enrolled in Both Medicare and Medicaid (2010–2012) a a Model 1 Model 2 Model 1 Model 2 Any hospitalization Any emergency department visits OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Previous year SNAP participation (n = 68,956) 0.86 (0.84–0.89) 0.96 (0.93–0.99) 0.90 (0.83–0.97) 0.98 (0.91–1.06) Previous year mean monthly SNAP amount 0.98 (0.97–0.98) 0.99 (0.99–0.99) 0.98 (0.98–0.99) 0.99 (0.99–1.00) in participants ($10) (n = 26,874) Number of inpatient hospital Number of emergency department days among the hospitalized visits among utilizers IRR (95% CI) IRR (95% CI) IRR (95% CI) IRR (95% CI) Previous year SNAP participation (n = 68,956) 0.90 (0.81–0.99) 0.92 (0.82–1.03) 0.96 (0.93–0.99) 0.98 (0.95–1.01) Previous year mean monthly SNAP amount 0.99 (0.98–0.99) 0.99 (0.99–1.00) 0.99 (0.99–1.00) 1.00 (1.00–1.00) in participants ($10) (n = 26,874) Associations estimated from zero-inflated negative binomial regression estimated with robust standard errors. All models adjusted for autoregressive effects, study year, age, sex, race/ethnicity, annual income, partial Medicaid eligibility, Medicaid spend down eligibility, chronic condition count, and Medicaid community waiver status. Model additionally adjusted for proportion of year participating in Medicaid. CI, confidence interval; IRR, incident rate ratio; OR, odds ratio; SNAP, Supplemental Nutrition Assistance Program. 91 92 SAMUEL ET AL. Table 3. Associations Between Supplemental Nutrition Assistance Program Participation (n = 68,956) and Benefit Amount (n = 26,874) with Inpatient Hospital Expenditures, Maryland Adults Aged ‡65 Years Enrolled in Both Medicare and Medicaid (2010–2012) Any hospitalization Marginal change in probability for any >0 hospital cost (95% CI) Previous year SNAP participation (n = 68,956) -1.5% (-2.0% to -1.1%) Previous year mean monthly SNAP amount in -0.2% (-0.2% to -0.1%) participants ($10) (n = 26,874) Inpatient hospital Medicare cost among the hospitalized Estimated elasticity for ln(cost) (95% CI) Previous year SNAP participation (n = 68,956) -5.8% (-8.4% to -3.3%) Previous year mean monthly SNAP amount -1.0% (-1.3% to -0.7%) in participants ($10) (n = 26,874) Associations estimated from Heckman regression model, adjusted for autoregressive effects, study year, age, sex, race/ethnicity, annual income, partial Medicaid eligibility, Medicaid spend-down eligibility, chronic condition count, Medicaid community waiver status, and proportion of year participating in Medicaid. Evaluated at means of all covariates. CI, confidence interval; SNAP, Supplemental Nutrition Assistance Program. that adjust for the proportion of the year enrolled in Med- icaid. It is notable that all had access to both Medicare and Medicaid because policy makers have increased access to Table 4. Steps to Obtain Cost Savings of Expanding health care for low-income groups thinking that that alone the Supplemental Nutrition Assistance Program to Nonparticipants in 2012 (n = 25,018), Based would reduce high hospital utilization in low-income groups. on Heckman Model Estimates Indeed, SNAP associations in this study were attenuated after adjusting for the proportion of the year with Medicaid cov- Cost savings from averted admissions erage, and evidence elsewhere shows that continuous Med- Change in probability of any admission 1.5 icaid coverage predicts less hospital-based care. However, in the year (%) 1,3–5 results from the present study add to evidence suggesting Multiplied by: Number of 25,018 nonparticipants that high hospital utilization is attributable to both social and Gives: Estimated number of averted 375 medical determinants of health. More specifically, these results admissions identify food assistance as a social determinant of hospital uti- Multiplied by: Average annual cost 25,091 lization, but not ED utilization. Therefore, SNAP may predict of inpatient admissions ($) fewer hospital admissions, even if it does not decrease the fre- Gives: Estimated cost savings from 9,415,900 quency of ED visits. Results from other studies suggest that averted admissions ($) improving health care access also may not reduce ED utiliza- 20,21 Cost savings from less costly hospital stays tion. Together, these results suggest that hospital and ED Number of nonparticipants admitted 7041 utilization are attributable to different mechanisms. Importantly, to hospital inpatient hospital admission is determined by health care pro- Less: Estimated number of averted 375 viders, whereas visiting the ED is generally the decision of the admissions individual. Therefore, inpatient hospital utilization is likely a Gives: Estimated no. of nonparticipants 6666 better measure of individual health status than ED utilization. still admitted to hospital These findings supplement evidence that investment in Percentage change in cost for admitted 5.8 SNAP may improve health outcomes and reduce inpatient persons (%) hospital spending. For example, the American Recovery and Multiplied by: Average annual cost 25,091 of inpatient admissions ($) Reinvestment Act of 2009, which expanded SNAP eligi- Gives: Reduction in average cost 1455 bility and increased the benefit amount, has been credited for admitted persons ($) with reducing the prevalence of food insecurity in SNAP Multiplied by: Estimated no. of 6666 income-eligible US households by 2% and may have nonparticipants still admitted 10 slowed growth in Medicaid inpatient hospital spending. to hospital These results build on prior studies by demonstrating that Gives: Estimated cost savings from 9,700,490 not only SNAP participation, but greater benefit amounts less costly hospital stays ($) among participants, is associated with reduced hospitaliza- Total cost savings tion rates. This is notable because participation is suscepti- Total estimated cost savings ($) 19,116,390 ble to self-selection, but benefit amounts are assigned based a largely on household financial need. Therefore, benefit See Table 3. amount results are less susceptible to confounding and re- See Table 1. Estimated based on participants who were hospitalized in 2012. verse causality than participation results. FOOD ASSISTANCE AND HOSPITAL UTILIZATION 93 In this study, SNAP participation also was associated with participation in food programs, and making results suscepti- less costly stays among those who were hospitalized. Based ble to survival bias. Conversely, however, these results sug- on the models, the study team estimates that expanding gest that greater SNAP benefits may confer a health benefit for SNAP access to nonparticipating dual eligible older adults low-income adults, even at older ages. In addition, lag times in Maryland could have resulted in inpatient hospital cost for potential health effects of food assistance programs savings of $19 million in 2012. Based on the average per are unknown. Lagged effects may be longer than the 1 year capita costs of the SNAP program, the team estimates that modeled in this study, so the associations between SNAP and the federal government would spend approximately $39 hospital utilization may be underestimated. Conversely, the million if they extended SNAP benefits to the 2012 income- study team could have missed key proximal effects by lagging eligible nonparticipants in the study sample. Therefore, ig- the models for a year. Also, averted hospital stays may differ noring issues of inefficiencies in taxation, approximately in cost from the typical hospital stay, which would bias the half of the cost of administering SNAP could be recouped cost savings estimate. This study is strengthened by using data by the federal government in reduced Medicare inpatient for an entire state population of dually eligible older adults, hospital spending. Further work is needed to quantify the reducing nonresponse bias in a low-income group. Also, this potential savings attributable to changes in other health care study is less susceptible to reverse causality than previous utilization outcomes. cross-sectional studies because of use of lagged exposure There are at least 2 potential reasons for the associations modeling. found in this study. First, SNAP participation reduces food 23,24 insecurity and minimizes the adverse effect of food Public Health Implications insecurity on dietary quality and obesity. This is notewor- This study found low rates of SNAP program participa- thy because food insecurity is linked to worse dietary 8 25 tion (only 53% participated in the most recent study year) in quality, increased risk of chronic diseases, and increased the population of dually eligible Maryland older adults, all risk of hospitalization. Therefore, it is plausible that im- of whom are likely income eligible for SNAP. National data proved access to SNAP and increased benefits for partici- estimates a participation rate of 42% among income eligible pants may improve food security and dietary quality in older adults, suggesting that the program is underutilized. households of low-income older adults, and this may reduce Results from this study suggest that improved access to SNAP hospital utilization. may reduce hospitalization for low-income older adults. To- An alternative potential reason for these results is that gether, these results suggest that strategies to improve access SNAP provides supplemental income that reduces financial to SNAP for eligible older adults likely will improve health strain for older adults living under or near the poverty line. outcomes, despite years of accumulated exposure to food Financial strain, or the lack of adequate money for basic 27 insecurity and financial strain in this population. These results needs, is linked with earlier mortality and greater risk for 28 29 have implications for practice and policy. both malnutrition and disability, which contribute to 5,30 hospitalization. Individuals experiencing financial strain 11,12 Practice implications struggle to afford food, heat, and health care, and may be forced to choose between these basic needs. Therefore, Efforts to increase SNAP participation may include tar- improving access to programs meeting any basic need will geted eligibility screening and enrollment assistance. Older improve an individual’s ability to afford the other basic adults tend to underutilize the program. Health care pro- needs. This idea is consistent with evidence of increased viders and health care payers can invest in efforts to screen for access to food after implementation of Medicare Part D, food insecurity and income eligibility for SNAP and facilitate which improved financial access to medications for older SNAP enrollment. Resources are available to support en- adults. Therefore, it is also plausible that any supplemental rollment and advocacy efforts nationally and state-level ad- income could be associated with reduced subsequent hospital- vocacy organizations. Community health workers, primary based care for low-income older adults because it enhances care provider practices, and not-for-profit service providers their ability to meet basic needs and reduces finance-related are well positioned to provide such screening and referral, but stress exposure. they need to be compensated and supported for services to address social and economic determinants of health. Limitations and strengths Policy implications As with other SNAP studies, SNAP participants may differ from nonparticipants on unmeasured characteristics. This Policy actions can improve access to SNAP and increase may have biased associations and the study cost savings benefit amounts. Several specific policy strategies can facil- calculations. For example, older adults who do not receive itate the SNAP enrollment process for older adults. First, SNAP may benefit from other meal assistance programs, such states can reduce enrollment requirements by implementing as Meals on Wheels or congregate meals. This likely would the Elderly Simplified Application Project. This program, bias associations toward the null because of unmeasured re- implemented in 7 states, streamlines income and expense ceipt of food assistance in the control group. Conversely, verification by matching data from existing sources, extends older adults who participate in SNAP may be more likely to certification periods to 36 months, and waives the re- enroll in other public benefits programs, which could con- certification interview. Second, states can coordinate SNAP found results if such programs collectively reduce hospital and Supplemental Security Income enrollment, as is done in utilization. Also, this study is limited to an older adult pop- the Combined Application Project. This program increased ulation, precluding measurement of the cumulative lifetime SNAP participation rates in South Carolina at a time when 94 SAMUEL ET AL. national rates were declining and is currently implemented 5. Iloabuchi TC, Mi D, Tu W, Counsell SR. Risk factors for in 18 states. Third, states can leverage administrative data early hospital readmission in low-income elderly adults. J Am Geriatr Soc 2014;62:489–494. from Medicaid, the Low Income Home Energy Assistance 6. Farson Gray K, Kochhar S. Characteristics of Supplemental Program (LIHEAP), and SNAP programs, as Maryland has, Nutrition Assistance Program households: Fiscal year 2014. to identify income eligible older adults who are not enrolled Alexandria, VA: USDA, Food and Nutrition Service, 2015. in SNAP and conduct targeted outreach to increase par- 7. Nguyen BT, Shuval K, Bertmann F, Yaroch AL. The ticipation. The Affordable Care Act incentivizes states to Supplemental Nutrition Assistance Program, food insecu- harmonize administrative data sets to facilitate Medicaid rity, dietary quality, and obesity among US adults. Am J enrollment, and these efforts can facilitate SNAP enrollment Public Health 2015;105:1453–1459. as well. 8. Tarasuk V, McIntyre L, Li J. Low-income women’s dietary Besides enhancing beneficiary access, states can enhance intakes are sensitive to the depletion of household resources benefit amounts for vulnerable older adults. For example, in one month. J Nutr 2007;137:1980–1987. the State of Maryland recently passed legislation ensuring 9. Seligman HK, Bolger AF, Guzman D, Lo´pez A, Bibbins- that all SNAP beneficiaries aged 62 and older receive a Domingo K. Exhaustion of food budgets at month’s end minimum benefit of $30 monthly by supplementing the and hospital admissions for hypoglycemia. Health Aff 2014; federal benefit with state funds as needed. Furthermore, 33:116–123. efforts to significantly cut federal spending on SNAP ben- 10. Sonik RA. Massachusetts inpatient Medicaid cost response efits through block granting or other structural changes may to increased Supplemental Nutrition Assistance Program have adverse consequences. benefits. Am J Public Health 2016;106:443–448. 11. Sattler ELP, Lee JS. Persistent food insecurity is associated with higher levels of cost-related medication nonadherence Conclusion in low-income older adults. J Nutr Gerontol Geriatr 2013; This study found that SNAP participation and greater 32:41–58. benefit amounts are associated with lower inpatient hospital 12. Berkowitz SA, Seligman HK, Choudhry NK. Treat or eat: utilization in a state population of low-income older adults. food insecurity, cost-related medication underuse, and un- These findings have public health implications because the met needs. Am J Med 2014;127:303–310.e3. majority of US older adults who are income eligible for 13. Nicholas LH. Can food stamps help to reduce Medicare SNAP do not participate. As public and private sector health spending on diabetes? Econ Hum Biol 2011;9:1–13. care partners shift to outcomes-driven, value-based care, 14. Shortell SM. Bridging the divide between health and health care. JAMA 2013;309:1121–1122. social service programs such as SNAP will be a critical tool 15. Farson Gray K, Cunnyngham K. Trends in Supplemental in improving health outcomes for low-income seniors across Nutrition Assistance Program participation rates: Fiscal the country. Year 2010 to Fiscal Year 2014. Alexandria, VA. 2016. www.fns.usda.gov/sites/default/files/ops/Trends2010-2012- Acknowledgments Summary.pdf Accessed September 19, 2016. We are grateful for the partnership of the Maryland De- 16. The Lewin Group. Picture of housing and health: Medicare and Medicaid use among older adults in HUD-assisted partment of Health and Mental Hygiene and the Maryland housing. Washington, DC: US Department of Health and Department of Human Resources for their support in pro- Human Services, 2014. viding data for these analyses. 17. Singer JD, Willett JB. Applied longitudinal data analysis: modeling change and event occurrence. New York: Oxford Author Disclosure Statement University Press, 2003. Drs. Samuel, Szanton, Wolff, and Ong, Ms. Cahill, Ms. 18. Buccaneer, A General Dynamics Company. CCW chronic Zielinskie, and Mr. Betley declared no conflicts of interest conditions: combined Medicare and Medicaid data. Wa- shington, DC: Centers for Medicare & Meidcaid Services, with respect to the research, authorship, and/or publication Federal Coordinated Healthcare Office, 2012. of this article. The authors received the following financial 19. Bindman AB, Chattopadhyay A, Auerback GM. Interrup- support: This study was supported by a grant from the Ro- tions in Medicaid coverage and risk for hospitalization for bert Wood Johnson Foundation. ambulatory care-sensitive conditions. Ann Intern Med 2008; 149:854–860. References 20. Taubman SL. Medicaid increases emergency-department 1. Blustein J, Hanson K, Shea S. Preventable hospitalizations use: evidence from Oregon’s health insurance experiment. and socioeconomic status. Health Aff (Millwood) 1998;17: Science 2014;343:263–269. 177–189. 21. Wright B, Potter AJ, Trivedi A. Federally Qualified Health 2. Oster A, Bindman AB. Emergency department visits for Center use among dual eligibles: rates of hospitalizations and ambulatory care sensitive conditions: insights into pre- emergency department visits. Health Aff 2015;34:1147–1155. ventable hospitalizations. Med Care 2003;41:198–207. 22. Nord M, Prell M. Food security improved following the 3. Ricketts TC, Randolph R, Howard HA, Pathman D, Carey 2009 ARRA increase in SNAP benefits. Washington, DC: T. Hospitalization rates as indicators of access to primary USDA Economic Research Service, 2011. care. Health Place 2001;7:27–38. 23. Mabli J, Ohls J. Supplemental Nutrition Assistance Pro- 4. Nyweide DJ, Anthony DL, Bynum JPW, et al. Continuity gram participation is associated with an increase in house- of care and the risk of preventable hospitalization in older hold food security in a national evaluation. J Nutr 2015; adults. JAMA Intern Med 2013;173:1879–1885. 145:344–351. FOOD ASSISTANCE AND HOSPITAL UTILIZATION 95 24. Ratcliffe C, McKernan SM, Zhang S. How much does the through Medicare and Medicaid. N Engl J Med 2015;374: Supplemental Nutrition Assistance Program reduce food 1–4. insecurity? Am J Agric Econ 2011;93:1082–1098. 33. Institute of Medicine. Meeting the dietary needs of older 25. Seligman HK, Laraia BA, Kushel MB. Food insecurity adults: exploring the impact of the physical, social, and is associated with chronic disease among low-income cultural environment: workshop summary. Washington, DC: NHANES participants. J Nutr 2010;140:304–310. National Academies Press, 2016. 26. Phipps EJ, Singletary SB, Cooblall CA, Hares HD, Brait- 34. Supplemental Nutrition Assistance Program. Elderly sim- man LE. Food insecurity in patients with high hospital plified application project guidance. Washington, DC: utilization. Popul Health Manag 2016;19:414–420. USDA Food and Nutrition Service, 2015. 27. Szanton SL, Allen JK, Thorpe RJ, Seeman T, Bandeen- 35. United States Department of Agriculture, Social Security Roche K, Fried LP. Effect of financial strain on mortality in Administration. Combined Application Projects. Guidance community-dwelling older women. J Gerontol B Psychol for States Developing Projects. 2005. www.fns.usda.gov/ Sci Soc Sci 2008;63:S369–S374. sites/default/files/CAPsDevelopmentGuidance.pdf Accessed 28. Samuel LJ, Szanton SL, Weiss CO, Thorpe RJ, Semba RD, July 15, 2016. Fried LP. Financial strain is associated with malnutrition 36. Kauff J, Dragoset L, Clary E, Laird E, Makowsky L, Sama- risk in community-dwelling older women. Epidemiol Res Miller E. Reaching the underserved elderly and working Int 2012;2012:696518. poor in SNAP: evaluation findings from the Fiscal Year 29. Szanton SL, Thorpe RJ, Whitfield K. Life-course financial 2009 pilots. 2014. www.fns.usda.gov/sites/default/files/ strain and health in African-Americans. Soc Sci Med 2010; SNAPUnderseved-Elderly2009.pdf Accessed August 16, 71:259–265. 2016. 30. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Address correspondence to: Med Sci 2001;56:M146–M156. Laura J. Samuel, PhD, CRNP 31. Madden JM, Graves AJ, Zhang F, et al. Cost-related medi- Johns Hopkins University School of Nursing cation nonadherence and spending on basic needs following 525 North Wolfe Street, Room 446 implementation of Medicare Part D. JAMA 2008;299: 1922–1928. Baltimore, MD 21205 32. Alley DE, Asomugha CN, Conway PH, Sanghavi DM. Accountable health communities—addressing social needs E-mail: lsamuel@jhmi.edu

Journal

Population Health ManagementPubmed Central

Published: Apr 1, 2018

References