Association of Unmet Social Needs With Metformin Use Among Patients With Type 2 Diabetes

Metformin remains the first-line pharmacotherapy for type 2 diabetes (T2D). It is affordable, widely available, and reduces all-cause mortality risk. However, adherence to metformin is suboptimal, and metformin underuse contributes to poor glycemic control (1). The terms “compliance” and “adherence” can be stigmatizing (2) and frame medication underuse as an attribute related to personal responsibility and minimize the role of structural factors. Although factors such as adverse effects and dislike of taking medications are partial explanations, it is important to consider other factors that contribute to suboptimal use. One category of such factors are individual-level social needs (e.g., food insecurity, social isolation) that result from social determinants of health (SDOH)—structural, population-level factors (e.g., income distribution) that drive health outcomes and disparities for individuals with T2D (3). Unmet social needs are associated with poor T2D outcomes (i.e., hemoglobin A_1c [HbA_1c]) (4,5) through various pathways, including medication underuse; however, the relationship between social needs and metformin use is understudied.

A 2021 systematic review and meta-analysis on the impact of SDOH on adherence to all medication types identified few studies, even fewer among patients with T2D, and no studies that examined metformin adherence (6). Other evidence suggests that food insecurity and copayment cost burden contribute to medication underuse in general among patients with T2D, without specifically examining the relationship with metformin (7–9). Given that metformin has one of the lowest adherence rates of oral antihyperglycemic medications (10), it is important to understand how social needs, or SDOH more broadly, contribute to underuse. Our objective was to examine how social needs are associated with metformin adherence among low-income patients with T2D. We hypothesized that greater burden of needs would be associated with lower metformin use.

The prospective cohort study was conducted at a Federally Qualified Health Center in Durham, North Carolina, that serves a patient population of which 81% are at or below 100% of the federal poverty level; 51% uninsured; 92% racial and/or ethnic minorities. The study period was 2018–2020. We included adults with T2D, who had completed a Protocol for Responding to and Assessing Patients’ Assets, Risks, and Experience (PRAPARE) assessment, a validated (11) social needs assessment widely used in primary care settings (Supplementary Appendix). We excluded participants who did not have an active metformin prescription 90 days before and 30 days after the PRAPARE assessment based on data availability and to ensure adherence was assessed during the time periods surrounding when social needs were reported.

To assess metformin adherence, we linked data from the electronic health record (EHR) and Surescripts (Surescripts, LLC) e-prescription network database on medication (re)fills from pharmacies and pharmacy benefit managers. Surescripts supports e-prescription information exchange between health care organizations and pharmacies. Surescripts refill data are embedded in the Federally Qualified Health Center EHR. Using the Surescripts linked data, we calculated adherence based on expected refill frequency using a proportion of days covered (PDC) approach over the 120-day study window. PDC is the preferred adherence measurement methodology since it assesses the proportion of days medication is available based on prescription refill frequencies over a period of time (12). Further, we examined adherence at an 0.8 PDC given that it is a widely used threshold for a high level of adherence. We considered the following covariates, extracted from the EHR: age, sex, race, ethnicity, insurance status, and Charlson comorbidity index, which are shown to predict mortality and cost in primary care patients (13).

We examined demographic and clinical characteristics to describe metformin adherence and associated unmet social needs. Further, as an exploratory supplementary analysis, we describe HbA_1c at an 0.8 PDC level. Next, we used linear regression models to examine the relationship between the total number of social needs and metformin adherence. Finally, we examined the association between unmet social needs by estimating the association between cumulative needs (i.e., one or two or more needs) and individual specific needs and metformin adherence. This enabled us to examine the cumulative effect of social need burden and the effect of each need controlling for other needs present. We also evaluated adherence at an 0.8 PDC threshold using logistic regression models. The Duke University Health System Institutional Review Board determined that this study was not human subjects research.

The data sets generated and/or used during the current study are not publicly available due to data stewardship and privacy regulations.

The sample consisted of 722 adults with a mean age of 54 years, 52.6% were Black, and 54.7% were uninsured. Mean HbA_1c was 8.1%. Approximately 60% of patients reported one or more social needs (mean = 1.36). Lack of access to medicine/health care (20.3%), social isolation (19.6%), and food insecurity (18.2%) were the most frequently reported social needs (Table 1). Metformin adherence ranged from 0.90 PDC among those with no social needs to 0.82 PDC among those who reported two or more social needs. Overall, 56.3% had a PDC of ≥0.8. Of patients with zero unmet social needs, 60.8% had an 0.8 PDC or greater, whereas 51.7% of patients with two or more social needs had an 0.8 PDC or greater. Those with an 0.8 PDC or greater had a mean HbA_1c of 8.38%, whereas those with PDC <0.8 had a mean of 7.94% (see Supplementary Table 1).

In adjusted analysis, those with two or more unmet needs had lower metformin adherence than those with no needs (−0.046; 95% CI −0.089, −0.003). When individual needs were modeled as independent variables of interest, being without housing (−0.069; 95% CI −0.121, −0.018), and lack of access to medicine/health care (−0.058; 95% CI −0.115, −0.000) were associated with lower adherence (Table 2) compared with not having those needs. In adjusted analysis of the 0.8 PDC threshold for metformin adherence, we found that being without housing was associated with lower adherence and age was associated with higher adherence (see Table 3).

In this study, we found that both total burden of unmet social needs and specific needs, particularly lacking housing and health care access, were associated with lower metformin use. The magnitude of these associations was clinically meaningful, representing between 2 and 4 weeks of missed medication over a 1-year period.

These findings highlight the role of specific social and economic barriers to metformin adherence that may contribute to its low adherence rates. Social needs screening and response in primary care represents an opportunity to respond to barriers to metformin adherence through social service and community-based organization referrals. Identifying and alleviating social and economic barriers to taking medication as prescribed may also have implications for provider bias in how patients are labeled as “nonadherent” (14) by providing context surrounding self-management activities. Our findings highlight that lack of housing represents a major barrier to metformin use and suggest that integrated care models are needed in primary care to respond to the intersecting medical and social needs of individuals without housing (15). However, health systems alone cannot modify broader policy- and system-level factors (e.g., social insurance programs, regulations, and labor laws) that influence the prevalence of social needs that shape T2D management and outcomes. Policies to address housing affordability are a useful exemplar. Among eligible households for the Section 8 Housing Choice Voucher Program, the nation’s largest source of rental assistance for low-income households, only 25% receive any rental assistance after an average waiting time of 2.5 years (16).

The results of this study should be interpreted in light of several limitations. This was a single-site study that included low-income patients from minority groups, which limits generalizability. Further, given that PDC captures medication refills but not use, it is possible that patients could have refilled metformin but not taken it, which could bias results. Future research should examine the impact of addressing social needs on metformin use over time.

Helping patients overcome barriers to adherence is a crucial part of T2D management. While medication underuse is multifactorial, social needs are one important factor. To improve diabetes care for low-income patients, researchers and practitioners should design strategies to overcome social and economic barriers to adherence as a mechanism to improve T2D outcomes and equity.

Acknowledgments. The authors are grateful to Dr. Howard Eisenson, Carolyn Crowder, and clinicians at the Lincoln Community Health Center for their clinical leadership and support. The authors thank the Center of Innovation to Accelerate Discovery and Practice Transformation (ADAPT) at the Durham Veterans Affairs Health Care System for its support.

Funding. This study was supported by the Blue Cross and Blue Shield Foundation of North Carolina. This study was also supported in part by the National Institutes of Health, National Heart, Lung, and Blood Institute under award number K12HL138030 (to C.D., L.Z., and H.B.). The study was also supported in part by U.S. Department of Veterans Affairs Health Services Research and Development Service Career Development Award 19-035 (IK2HX003085-01A2 to D.B.). H.B.B. reports research funding through his institution from National Heart, Lung, and Blood Institute.

The content is solely the responsibility of the authors and does not necessarily reflect the position or policy of Duke University, the U.S. Department of Veterans Affairs, or the U.S. government.

Duality Interest. C.D. reports a financial relationship with ZealCare. H.B.B. reports research funding through his institution from BeBetter Therapeutics, Boehringer Ingelheim, Esperion, Improved Patient Outcomes, Merck, Novo Nordisk, Otsuka, Sanofi, VA, Elton John Foundation, Hilton Foundation, and Pfizer. H.B.B. provides consulting services for Abbott, Esperion, Imatar, Novartis, Sanofi, Vidya, Walmart, and Webmed and was also on the board of directors of Preventric Diagnostics. D.V.B. reports a consulting relationship with the Eating Recovery Center. S.A.B. reports research grants from National Institutes of Health, North Carolina Department of Health and Human Services, Blue Cross Blue Shield of North Carolina, and Feeding America, and personal fees from the Aspen Institute, Rockefeller Foundation, Gretchen Swanson Center for Nutrition, and Kaiser Permanente, outside of the submitted work. L.L.Z. reports a consulting relationship with Novartis. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.I. oversaw clinical operations required for this analysis. C.D. and J.M.A. conducted data abstraction and analysis. C.D., J.M.A., D.V.B., and L.L.Z. conceptualized the study and analysis plan. C.D., J.M.A., B.C.B., H.B.B., S.A.B., and L.L.Z. wrote the manuscript. All authors reviewed, edited, and provided feedback on the manuscript. C.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this work were presented as an oral presentation at the AcademyHealth 2022 Annual Research Meeting, Washington, DC, 4–7 June 2022.