Associations of Community and Individual Social Determinants of Health With Medication Adherence in Patients With Atrial Fibrillation: A Retrospective Cohort Study

Abstract

Background Despite guideline-recommended use of oral anticoagulation (OAC) for stroke prevention in atrial fibrillation (AF), OAC medication adherence among patients with AF in the United States ranges from 47% to 82%. To characterize potential causes of nonadherence, we analyzed associations between community and individual social risk factors and OAC adherence for stroke prevention in AF. Methods and Results A retrospective cohort analysis of patients with AF was conducted using the IQVIA PharMetrics Plus claims data from January 2016 to June 2020, and 3-digit ZIP code-level social risk scores were calculated using American Community Survey and commercial data. Logistic regression models evaluated associations between community social determinants of health, community social risk scores for 5 domains (economic climate, food landscape, housing environment, transportation network, and health literacy), patient characteristics and comorbidities, and 2 adherence outcomes: persistence on OAC for 180 days and proportion of days covered ≥0.80 at 360 days. Of 28 779 patients with AF included in the study, 70.8% of patients were male, 94.6% were commercially insured, and the average patient age was 59.2 years. Multivariable regression found that greater health literacy risk was negatively associated with 180-day persistence (odds ratio [OR]=0.80 [95% CI, 0.76-0.83]) and 360-day proportion of days covered (OR, 0.81 [95% CI, 0.76-0.87]). Patient age and higher AF stroke risk score and AF bleeding risk scores were positively associated with both 180-day persistence and 360-day proportion of days covered. Conclusions Social risk domains, such as health literacy, may affect OAC adherence among patients with AF. Future studies should explore associations between social risk factors and nonadherence with greater geographic granularity.

Keywords: anticoagulants; atrial fibrillation; medication adherence; social determinants of health; social risk factors.

Figure 1. Patient selection criteria.

Figure 2. Associations between community‐ and patient‐level characteristics and persistence on index anticoagulant at 180 days (multivariable regression).

Risk scores are continuous variables on a 1–5 scale, where 1=little to no risk, 2=low risk, 3=moderate risk, 4=high risk, and 5=severe risk. Community characteristics are measured at the 3‐digit ZIP code (ZIP‐3) level. Proportional community‐level measures are scaled by a factor of 10. For example, a 1‐unit increase in “Prop. aged 70 and older” is equivalent to a 10% increase in the prevalence of residents aged ≥70 years in the ZIP‐3 area. Other measures represent median or mean values. DOAC indicates direct‐acting oral anticoagulant; and HAS‐BLED, atrial fibrillation bleeding risk score.

Figure 3. Associations between community‐ and patient‐level characteristics and PDC at 360 days (multivariable regression).

Risk scores are continuous variables on a 1–5 scale, where 1=little to no risk, 2=low risk, 3=moderate risk, 4=high risk, and 5=severe risk. Community characteristics are measured at the 3‐digit ZIP code (ZIP‐3) level. Proportional community‐level measures are scaled by a factor of 10. For example, a 1‐unit increase in “Prop. aged 70 and older” is equivalent to a 10% increase in the prevalence of residents aged ≥70 years in the ZIP‐3 area. Other measures represent median or mean values. DOAC indicates direct‐acting oral anticoagulant; HAS‐BLED, atrial fibrillation bleeding risk score; HMO, health maintenance organization; PDC, proportion of days covered; and PPO, preferred provider organization.