Hypertension, Diabetes and Medication Adherence among the Older Supplemental Nutritional Assistance Program Population

Abstract

The burdens of chronic diseases such as hypertension and diabetes for older Americans are profound. Yet, data on the population-level prevalence of hypertension and diabetes among the older adult Supplemental Nutritional Assistance Program (SNAP) population and the associated level of medication adherence is lacking despite evidence of the “treat or eat” trade-off in the general population. We used linked administrative data from SNAP and Medicaid between 2006 and 2014 in the state of Missouri to document rates of hypertension or diabetes diagnoses and medication adherence. About 69% of the study sample were found to be diagnosed with a hypertension and 40% with diabetes. Approximately 1 in 4 of those living with hypertension and 1 and 3 of those living with diabetes were nonadherent to antihypertensive or antidiabetic medications each year, on average. Furthermore, medication non-adherence increases with age and is more common among non-White and urban residents.

INTRODUCTION

The social and economic burdens of chronic diseases such as hypertension and diabetes for older Americans are profound. Cardio-metabolic diseases such as diabetes and hypertension are among the leading cause of morbidity and mortality in the United States (Xu et al., 2016). More than two-thirds of adults ages 60 years or older were living with hypertension between 2011 and 2014, and a fourth of adults 65 or older were living with diabetes (Fryar et al., 2017a, 2017b).

Cardio-metabolic diseases such as diabetes and hypertension are of particular concern among older adults who are food insecure. Not only are diabetes and hypertension more prevalent among low-income and food insecure individuals (Bhargava & Lee, 2017; Fryar et al., 2017a, 2017b; Jih et al., 2018), food insecure individuals living with diabetes and/or hypertension have poorer disease control, such as sub-optimal glycemic or blood pressure control (Bawadi et al., 2012; Walker et al., 2021; Wang et al., 2015). This may be because food insecure individuals living with diabetes and/or hypertension have poorer diet quality (Orr et al., 2019), but also because they have higher rates of cost-related medication non-adherence (Berkowitz et al., 2014; Gundersen & Ziliak, 2015), even after controlling for socio-demographic characteristics (Irving et al., 2014).

In fact, in what has been termed the “eat or treat” hypothesis, there is growing recognition that low-income households lack adequate resources to purchase both food and medication and face tradeoffs between purchasing food and purchasing prescription drugs to treat chronic illnesses such as diabetes and hypertension. Among those who reported a chronic illness in the National Health Interview Survey (Berkowitz et al., 2014), 23.4% of adults reported cost-related medication underuse, 18.8% reported food insecurity, and 11% reported both.

The Supplemental Nutritional Assistance Program (SNAP) is the country’s largest food and nutrition assistance program in the United States and is designed to address food insecurity among all ages. Despite the clinical relevance, very little is known about the prevalence of cardio-metabolic diseases such as diabetes and hypertension or adherence to antihypertensive or antidiabetic medications among the population of older adults participating in SNAP and associated levels of medication non-adherence. This limitation stems from reliance on survey data (Afulani et al., 2015; Berkowitz et al., 2014; Srinivasan & Pooler, 2018), which contains measurement error in the self-reports of chronic diseases, medication under-use, and SNAP participation. In addition, within the medical literature the standard approach to studying medication non-adherence requires administrative data containing pharmacy claims but this data typically lacks information on patient income and SNAP participation (Calip et al., 2017; Kim et al., 2018). Finally, although older adults consume a large share of medical resources and are comprising a larger share of the total population, most of the research to date has not focused on this age group (Herman et al., 2015). In order to address this gap in the literature, we examined the prevalence of hypertension and diabetes and the observed level of medication non-adherence among a population age 60 and older on SNAP between 2006 and 2014 in one Midwestern state. We used state administrative data for SNAP linked to Medicaid pharmacy claims for those with hypertension and diabetes and calculated levels of medication non-adherence for different demographic groups.

Understanding the health of the older adult population on SNAP is of critical interest to researchers and clinicians with a focus on older adults. SNAP is the largest food and nutrition program in the United States and is the only program open to older adults that allows them to fully make their own food choices. Adults qualify based on household income and expenses (including medical expenses) and receive an electronic debit card that allows them to purchase food at participating food retailers. While there are special rules for adults age 60 and older that are designed to make it easier to qualify and remain on SNAP, the average participation level among eligible older adults is approximately 45%; the average participation level among eligible younger counterparts age 18-59 is 88% (Cunnyngham, 2018). Despite the low take-up rates, estimates calculated prior to COVID-19 indicated that 1 in 4 SNAP households contained an older adult and 13% of all individuals participating in SNAP were age 60 or older with future projections estimating even higher levels of participation (Cronquist & Lauffer, 2019).

METHODS

We performed a retrospective cohort study of older adults who received SNAP between 2006 and 2014 using administrative data from the state of Missouri. We linked this sample to Medicaid claims for the same period and identified pharmacy claims for antihypertensive and antidiabetic medications for the sample of adults with a diagnosis of hypertension or diabetes.

Our study population is SNAP participants in Missouri age 60 or older during our study period (N=154,020). Of these, 106,197 had a diagnosis of hypertension; 60,965 had a diagnosis of diabetes (see Table 1). We identified these disease conditions based on the presence of one or more Medicaid claims with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes for hypertension (401.x-405.x, 437.2), hypertensive emergencies with acute target organ damage (362.81, 377.01, 428.0, 428.1, 428.20, 428.21, 428.23, 428.30, 428.31, 428.33, 428.40, 428.41, 428.43, 428.9, 410.x, 414.12, 443.21, 443.22, 443.23, 443.24, 443.29, 441.x, 430, 431, 432.x, 434.x, 435.x, 436, 437.2), and/or diabetes (250.0-250.9). Diagnosis codes used to identify hypertension and diabetes were based on previous studies (Janke et al., n.d.; Washington et al., 2013).

Table 1.

Demographic Characteristics of Older Adults on SNAP

	Full Sample		Hypertension		Diabetes
			Overall Prevalence		Overall Prevalence
	N (Persons)	Percent	N (Persons)	Percent	N (Persons)	Percent
	154,020	100.00%	106,197	68.95%	60,965	39.58%
Age
60 to 64	52,730	34.24%	29,325	55.61%	16,636	31.55%
65 to 69	35,521	23.06%	24,635	69.35%	14,503	40.83%
70 to 79	40,782	26.48%	31,481	77.19%	18,771	46.03%
80 to 89	24,987	16.22%	20,756	83.07%	11,055	44.24%
Sex
Female	96,170	62.44%	70,837	73.66%	41,694	43.35%
Male	57,850	37.56%	35,360	61.12%	19,271	33.31%
Race and Ethnicity
White	106,847	69.37%	73,628	68.91%	41,432	38.78%
Black	32,949	21.39%	23,599	71.62%	14,395	43.69%
Hispanic	2,934	1.90%	1,775	60.50%	1,091	37.18%
Other	11,290	7.33%	7,195	63.73%	4,047	35.85%
Citizenship Status
Citizen	149,741	97.22%	103,473	69.10%	59,466	39.71%
Noncitizen	4,279	2.78%	2,724	63.66%	1,499	35.03%
Geography
Rural	63,910	41.49%	47,404	74.17%	27,109	42.42%
Urban	90,110	58.51%	58,793	65.25%	33,856	37.57%

Notes: Prevalence = ever had a diagnosis for the condition between 2006 and 2014. Individuals in the SNAP sample who were not observed in the Medicaid files are included in the denominator (16% of the total SNAP sample). Age is measured at first SNAP receipt. Urban designation based on living in an Office of Management and Budget designated Metropolitan County.

We first assessed the prevalence of hypertension and diabetes in the older adult SNAP population who received benefits between 2006 and 2014. We next document the prevalence of these two chronic diseases by age (60-64; 65-69; 70-79; and 80 and older), sex (male/female), race and ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Other), citizenship (US Citizen/noncitizen), and geography (urban/rural based on the OMB designation of metropolitan and nonmetropolitan counties).

Next, we used information from Medicaid pharmacy claims to assess levels of medication non-adherence to prescribed antihypertensives and antidiabetics for individuals who were enrolled in both SNAP and Medicaid for all 12 months of a given year (52% of the total SNAP population) and who had a diagnosis of the relevant disease condition (hypertension or diabetes). Using the HEDIS 2017 final NDC lists, we defined antidiabetics as Metformin, Alpha-glucosidase inhibitors, sulfonylureas, thiazolidinediones, dipeptidyl peptidase-4 (DPP-4) inhibitors, glucagon-like peptide-1 (GLP1) agonists, meglitinides, amylin analogs, sodium glucose cotransporter 2 (SGLT2) inhibitor, or any combination antidiabetic medication. Antihypertensives were defined as angiotensin II receptor blockers, angiotensin-converting enzyme inhibitors, beta-blockers, calcium-channel blockers, vasodilating agents or any combination antihypertensive.

We calculated medication non-adherence on an annual basis based on the proportion of days covered from the date of the first fill for a medication through December 31^st of that year. If a prescription for the same medication was filled prior to medications from the prior fill being exhausted, then we shifted the prescription start date forward to the day after the previous days’ supply was exhausted. Calculations of medication non-adherence were done using the Stata packaged medadhere (Linden, 2019). We present the average of the annual rates of medication adherence to antihypertensives and to antidiabetics for SNAP participants with diagnosis of these conditions by demographic characteristics. In line with previous studies, we define nonadherence as proportion of days covered below 0.80 (Centers for Medicare & Medicaid Services, 2018; National Quality Forum, 2015; Pharmacy Quality Alliance, 2018).

Finally, we use ordinary least squares (OLS) regression analysis to model the rate of medication adherence (proportion of days covered) for each chronic disease condition (hypertension and diabetes) as a function of demographic characteristics. We use maximum likelihood estimation with a probit link function to model the likelihood of adherence (the probability that a proportion of days covered is greater than or equal to .80) as a function of demographic characteristics.

All analyses were conducted using STATA/MP version 16.1. Statistical significance was assessed at the α=.05 level. This study was categorized as exempt from Human Subject Research oversight by the University of Missouri Institutional Review Board.

RESULTS

Approximately 154,000 adults age 60 or older received SNAP in the state of Missouri between 2006 and 2014. Table 1 shows that 69% of SNAP recipients were living with hypertension and 40% with diabetes based on health care claims received by Medicaid.

Among the SNAP population, the risk of hypertension increases with age: while 56% of SNAP recipients age 60-64 have been diagnosed with hypertension, the prevalence rises to 83% among those age 80 and older. In contrast, the risk for diabetes is similarly low in the 60-64 age group (32%) but peaks at age 70-79 (46%) before dropping slightly among those age 80 and older (44%).

Among female SNAP recipients, 74% have a diagnosis of hypertension and 43% diabetes; among male SNAP recipients 61% have a diagnosis of hypertension and 33% of diabetes.

In terms of the distribution of chronic conditions by race and ethnicity, Black SNAP recipients have prevalence rates of hypertension and diabetes of 72% and 44%, respectively, while rates for White, Hispanic or those with other racial and ethnic identities range from 61-69% for hypertension and 36-39% for diabetes. While less than 3% of older adults who receive SNAP are noncitizens, 64% have been diagnosed with hypertension and 35% with diabetes.

Finally, in terms of geography, 74% of older adults SNAP recipients who reside in counties designated as rural have a diagnosis for hypertension and 42% diabetes; 65% of those who reside in urban counties have a diagnosis for hypertension and 38% for diabetes.

Next, we present information on annual rates of adherence to antihypertensives and antidiabetics for the older adult SNAP population with diagnoses of these conditions. We also present information on the percentage nonadherent, using proportion of days covered less than .80, a commonly used threshold (Centers for Medicare & Medicaid Services, 2018; National Quality Forum, 2015; Pharmacy Quality Alliance, 2018). Results are reported in Table 2. Overall, for the older adult SNAP population with hypertension, the annual rate of medication nonadherence is 0.86 and 24% are nonadherent on average. For the older adult SNAP population with a diagnosis of diabetes, annual rates of medication adherence are lower (0.80) and a much higher percentage (36%) are nonadherent.

Table 2.

Annual Rates of Medication Adherence and Percentage Nonadherent by Demographic Characteristics for Older Adults on SNAP

	Antihypertensive Medications			Antidiabetic Medications
	Persons per Yeara	Adherence Rateb	Percentage Nonadherent (PDC<.8) b	Persons per Yearc	Adherence Rateb	Percentage Nonadherent (PDC<.8) b
	21,755	0.86	24.35%	5,298	0.80	35.81%
Age
60 to 64	5,346	0.88	22.27%	1,348	0.82	32.89%
65 to 69	5,416	0.86	26.03%	1,404	0.79	37.83%
70 to 79	7,186	0.86	24.67%	1,787	0.80	36.07%
80 to 89	3,807	0.86	24.21%	760	0.80	36.95%
Sex
Female	16,135	0.87	24.09%	3,886	0.80	36.31%
Male	5,621	0.86	25.10%	1,412	0.81	34.43%
Race and Ethnicity
White	15,542	0.87	23.18%	3,812	0.81	34.47%
Black	4,482	0.84	27.79%	1,026	0.77	41.63%
Hispanic	293	0.85	26.87%	87	0.80	35.46%
Other	1,438	0.86	25.75%	372	0.81	33.78%
Citizenship Status
Citizen	21,234	0.86	24.30%	5,147	0.80	35.92%
Noncitizen	522	0.85	26.56%	151	0.82	32.46%
Geographyd
Rural	10,972	0.88	22.28%	2,770	0.81	33.83%
Urban	10,784	0.85	26.54%	2,528	0.79	38.05%

Notes:

^a.Sample inclusion in each year is conditional on diagnosis of hypertension and 12 months of SNAP and Medicaid participation.

^b.Annual estimates averaged across the eight-year study period.

^c.Sample inclusion in each year is conditional on diagnosis of diabetes and 12 months of SNAP and Medicaid participation.

^d.Urban designation based on living in an OMB Metro County.

Among older adults on SNAP with hypertension, annual rates of medication adherence range from 0.86-0.88 across age groups while the percentage nonadherent ranges from 22% among those age 60-64 to 26% among those age 65-69. For those with diabetes, annual rates of medication adherence ranges from 0.79-0.82 across age groups and the percentage nonadherent is 33% among the lowest age group and 38% among those age 65-69.

Female SNAP recipients with hypertension have annual rates of medication adherence of 0.87 and male SNAP recipients have rates of 0.86 on average, with 24% and 25% nonadherent in each group respectively. Female SNAP recipients with diabetes have annual rates of medication adherence of 0.80 and male SNAP recipients have rates of 0.81, with 36% and 34% nonadherent in each group respectively.

Among older adult SNAP recipients with hypertension, annual rates of medication adherence are 0.87 for Whites, 0.84 for Blacks, 0.85 for Hispanics, and 0.86 for those in the “other” racial category on average. In terms of nonadherence, 23% of Whites, 28% of Blacks, 27% of Hispanics and 26% of Other identities are nonadherent. Among those living with diabetes, the annual rate of medication adherence among Blacks is 0.77, and closer to 0.81 for Whites, Hispanics, and Other identities. However, in terms of the percentage nonadherent there is more variability across racial and ethnic groups (Blacks 42%; Hispanics 34%; Whites 34%; Others 33%).

In terms of citizenship status, we find annual rates of medication adherence of 0.86 for hypertension and 0.80 for diabetes among US citizens on average; compared to annual rates of medication adherence of 0.85 for hypertension and 0.82 for diabetes among non-citizens. Approximately 24% of citizens and 27% of non-citizens with hypertension are classified as nonadherent and 36% of citizens and 32% of non-citizens with diabetes are classified as nonadherent in a given year on average.

When we look at geography, we note a consistent pattern across both disease conditions and measures of medication adherence: among SNAP participants with hypertension, rural residents have annual rates of medication adherence of 0.88 and urban residents have annual rates of 0.85. For SNAP participants with diabetes, rural residents have annual rates of medication adherence of 0.81 and urban residents have annual rates of 0.79. In addition, 22% of rural residents with hypertension and 34% with diabetes meet the definition of nonadherent. For urban residents the respective percentages are 27% and 38% in a given year on average.

Finally, we estimate multivariate models to examine the conditional effect of demographic characteristics on each chronic condition in Table 3 for both—continuous and dichotomous—measures of medication adherence. Results are consistent across continuous and dichotomous measures but vary across chronic conditions. In terms of age, we find medication adherence is lower among older SNAP recipients than those age 60-64 for both hypertension and diabetes (p<.001 for each age group and both outcomes) after controlling for other demographic characteristics. Females have higher levels of medication adherence than do males for hypertension but lower levels for diabetes after controlling for other demographic characteristics. In terms of race and ethnicity, on average Blacks (p<.001), and Hispanics (p<.01) have lower levels of medication adherence than do Whites for hypertension but the coefficient for Hispanics is not statistically significant when the dichotomous measure of medication adherence is estimated. Similarly, for diabetes, average levels of medication adherence are lower for Blacks (p<.001) relative to Whites but are not statistically different for Hispanics and Other identity groups. US citizenship has no relationship with medication adherence for hypertension after controlling for other demographic characteristics; for diabetes, non-citizens have higher rates of medication nonadherence than do U.S. citizens (p<.01). Finally, in terms of geography, older adults on SNAP living in urban areas have lower levels of medication adherence than do those living in rural areas for both hypertension and diabetes after controlling for other characteristics (p<.001).

Table 3.

Predictors of Medication Non-adherence for Older Adults on SNAP

	Hypertension PDC		Diabetes PDC
	Continuous	Adherent (PDC>.8)	Continuous	Adherent (PDC>.8)
Age (60 to 64)
65 to 69	−0.02***	−0.04***	−0.03***	−0.05***
	(0.002)	(0.003)	(0.003)	(0.006)
70 to 79	−0.01***	−0.02***	−0.01***	−0.03***
	(0.001)	(0.003)	(0.003)	(0.006)
80+	−0.01***	−0.02***	−0.02***	−0.03***
	(0.002)	(0.003)	(0.004)	(0.008)
Sex (Male)
Female	0.01***	0.01***	−0.01**	−0.02**
	(0.001)	(0.002)	(0.003)	(0.005)
Race/Ethnicity (White)
Black	−0.02***	−0.03***	−0.03***	−0.06***
	(0.001)	(0.003)	(0.003)	(0.006)
Hispanic	−0.01*	−0.02	−0.01	−0.00
	(0.005)	(0.009)	(0.010)	(0.019)
Other	−0.00	−0.01	0.01	0.02
	(0.002)	(0.004)	(0.005)	(0.010)
Citizenship Status (Citizen)
Noncitizen	−0.01	−0.01	0.02*	0.03*
	(0.004)	(0.007)	(0.008)	(0.015)
Geography (Rural)
Urban	−0.02***	−0.03***	−0.02***	−0.03***
	(0.001)	(0.002)	(0.003)	(0.005)
Constant	0.89***		0.84***
	(0.001)		(0.003)
Observations (Person-Years)	174,043	174,043	42,383	42,383
	0.004		0.007

Notes: Probit regression analysis shown with standard errors in parenthesis under the average marginal effect. Omitted reference group is shown in parenthesis next to the variable name.

^***p<0.001

^**p<0.01

^*p<0.05

DISCUSSION

In this study of older adult SNAP recipients, 69% were diagnosed with hypertension and 40% with diabetes, two chronic conditions that require daily medication to manage properly. Yet, our study documented that 1 in 4 older adults on SNAP and Medicaid with hypertension were nonadherent in a given year on average; the rate of nonadherence rises to 1 in 3 for those with diabetes. Furthermore, we find higher levels of nonadherence at older ages and among non-White racial and ethnic groups and rural residents. However, some interesting disease specific findings are present among our population of older adults on SNAP: Males with hypertension are more likely than females to be nonadherent to antihypertensives while females with diabetes are more likely than males to be nonadherent to antidiabetics. Additionally, citizenship is an important determinant of medication nonadherence for those with diabetes but not for those with hypertension.

Previous research on SNAP participation and medication underuse has relied on survey data and self-reports of key measures, focused on a wider age range and has not been capable of describing the health of the SNAP population (Afulani et al., 2015; Berkowitz et al., 2014; Herman et al., 2015). As a consequence, this study is the first to describe the population-level health of the older adult SNAP population and document levels of medication non-adherence for two common chronic conditions—hypertension and diabetes. For comparison, among the general population, more than two-thirds of adults ages 60 years or older were living with hypertension between 2011 and 2014, and a fourth of adults 65 or older were living with diabetes (Fryar et al., 2017a, 2017b). Furthermore, reported levels of cost-related medication underuse in studies using survey data range from 23.4% using an analysis of all ages and all chronic conditions to 12.8% among older SNAP participants with diabetes (Berkowitz et al., 2014; Pooler & Srinivasan, 2019). We report a similar prevalence of hypertension and much higher prevalence of diabetes among the older adult SNAP population and substantially higher levels of medication nonadherence related to these two disease conditions (although levels are more similar to those reported in the Medicare population for hypertension) (Q. Yang et al., 2017).

The observed high rates of medication non-adherence have significant implications for both patient health and state Medicaid budgets. Medication non-adherence has been associated with higher hospitalization and mortality rates among individuals living with diabetes (Ho et al., 2006). Low levels of adherence to antihypertensive medications have also been associated with a higher prevalence of acute cardiovascular events, as well as higher emergency department and hospitalization costs (Q. Yang et al., 2017; Z. Yang et al., 2016). Medication non-adherence to beta-blockers, ACE inhibitors and statins has also been associated with higher cardiovascular mortality among patients with coronary heart disease (Ho et al., 2008; Q. Yang et al., 2017). Furthermore, previous findings suggest that these individuals are at risk for poorer hypertensive or diabetic outcomes that could affect overall quality of life among these patients. Further, these poorer outcomes could exacerbate competing demands for indirect healthcare expenditure (such as transportation costs) and household food needs. Our findings therefore highlight a need for targeted interventions towards the older adult SNAP subpopulations at risk for medication non-adherence aimed at improving medication adherence. Clearly this is an area of research that deserves further attention (Gellad et al., 2011; Krousel-Wood et al., 2011).

Limitations

This study has several limitations. First, we relied upon Medicaid claims filed to identify the presence of hypertension and diabetes. While this approach is more reliable than using self-reports, this approach likely underestimates the prevalence of these two conditions. These conditions may be emergent or diagnosed while covered by other health insurance. Second, our population-level analysis of all older adults on SNAP is limited to one state and the time-period 2006-2014 and may not be generalizable to states whose demographic characteristics differ substantially, or to states who increased their Medicaid population after 2014. However, SNAP participation among eligible older adults as well as levels of food insecurity are very similar to the national level (Food Research & Action Center, 2019). Third, our measures of medication non-adherence are calculated for those that receive both SNAP and Medicaid for an entire calendar year (about 52% of our total SNAP population), and may not represent the levels for the full SNAP population in a given year. Additionally, we have included SNAP recipients in assisted living facilities, who likely have higher levels of medication adherence than those living independently. Finally, we are only able to observe medication nonadherence for two disease conditions: hypertension and diabetes. As a consequence, this study cannot provide an estimate of the total level of medication nonadherence for all disease conditions within the older adult SNAP population.

CONCLUSION

In conclusion, chronic health conditions are very common among the older adult SNAP population with 69% having a diagnosis for hypertension and 40% for diabetes. Adherence to prescribed medications is critical to maintain proper disease control and to prevent avoidable complications and hospitalizations among this subpopulation. Therefore, the prevalence of medication non-adherence to antihypertensive and antidiabetic medications that we observed among individuals living with hypertension (24%) or diabetes (35%) each year, on average, has patient-level and population-level implications.