Statin adherence in primary care among patients with hypertension with and without other cardiometabolic diseases

Abstract Background Statin adherence has been reported suboptimal, although it effectively prevents cardiovascular disease and is available at a low cost. Knowledge on statin adherence among hypertensive patients in primary care is scarce. Purpose We examined statin adherence in Swedish primary care among patients with hypertension alone or combined with diabetes and/or ischemic heart disease. Also, associations between baseline characteristics and adherence were assessed. Method A retrospective register-based cohort study from a 1.8 million people Swedish region. Clinical individual data from all primary health care centres (PHCCs) were obtained from the regional quality register for primary care, linked to national and regional registers containing data on comorbidities, dispensed drugs, and socioeconomic factors. Patients diagnosed with hypertension who retrieved their first-ever statin prescription 2012–2015 were included. The index date was the first date of dispensation. Adherence was assessed over two years from statin initiation, by the proportion of days covered (PDC) i.e, the total days with statin supply divided by the total days in the observation period. Machine learning (ML) with XGBoost and SHAP values, accounting for variable interactions was used to determine patient baseline characteristics influence on adherence prediction. An adjusted multilevel regression model was used to estimate odds ratios (OR) and median odds ratio (MOR) for the associations between the ten most influential baseline characteristics and adherence (PDC 80%). Results The study included 30 497 patients (mean age 66 years, 48% women) with hypertension. The statin indications were primary prevention (39%), diabetes without cardiovascular disease (16%), or secondary prevention (45%). The most dispensed statins were simvastatin (58%) and atorvastatin (41%). Overall, the mean two-year PDC was 73%. Adherence (PDC80%) differed (p0.001) between patients with primary prevention (54%), diabetes without cardiovascular disease (61%), and secondary prevention (65%). The 10 most influential patient characteristics for adherence prediction according to ML (SHAP values) are shown in figure 1, and the result of the adjusted regression model in figure 2. The associations between the top 5 patient characteristics and adherence were type of statin (atorvastatin vs. simvastatin OR 1.37), dispensed antithrombotic therapy (OR 1.34), older age (OR 1.08 per 10 years), PHCCs (MOR 1.13), and country of birth outside of Sweden (OR 0.70–0.74). The pattern of associations was similar in unadjusted models. Conclusion Statin adherence was associated with the type of statin, patient age, socioeconomic factors, and the dispensation of other cardiometabolic drugs as a proxy for comorbidities, although not with the diagnoses themselves. Adherence varied across PHCCs, indicating variability in care. Clinicians might consider these associations when initiating statin therapy.Figure 1 Figure 2