Determinants of medication adherence in people with epilepsy: A multicenter, cross-sectional survey

OBJECTIVES

Poor medication adherence in people with epilepsy (PwE) increases mortality, hospitalization, and poor quality of life, representing a critical challenge for clinicians. Several demographic, clinical, and neuropsychological factors were singularly found associated with medication adherence in several studies, but the literature lacks a comprehensive study simultaneously assessing all these variables.

METHODS

We performed a multicenter and cross-sectional study using online questionnaires with the following clinical scales: Morisky Medication Adherence Scale (MMAS-8), Quality of Life in Epilepsy Inventory 31 (QoLIE-31), Beck Depression Inventory-II (BDI-II), Generalized Anxiety Disorder-7 (GAD-7) and 14-item Resilience scale (RES14) in a population of 200 PwE. We used the ANOVA test and Spearman's correlation to evaluate the relationship between medication adherence and demographic, clinical (seizure frequency, number of anti-seizure medications), and neuropsychological characteristics. We trained separate machine learning models (logistic regression, random forest, support vector machine) to classify patients with medium-high adherence (MMAS-8 ≥ 6) and poor adherence (MMAS-8 < 6) and to identify the main features that influence adherence.

RESULTS

Women were more adherent to medication (p-value = 0.035). Morisky Medication Adherence Scale -8 showed a direct correlation with RES14 (p-value = 0.001) and age (p-value = 0.001), while was inversely correlated with BDI-II (p-value = 0.001) and GAD-7 (p-value = 0.001). In our model, the variables mostly predicting treatment adherence were QoLIE-31 subitems, followed by age, resilience, anxiety, years of school, and disease duration.

CONCLUSION

Our study confirms that gender, age, and neuropsychological traits are relevant factors in predicting medication adherence to PwE. Furthermore, our data provided the first evidence that machine learning on multidimensional self-report questionnaires could help to develop a decisional support system in outpatient epilepsy clinics.