摘要：Background: Electronic monitoring devices (EMDs) are regarded as the “gold standard” for assessing medication adherence in research. Although EMD data provide rich longitudinal information, they are typically not used to their maximum potential. Instead, EMD data are usually combined into summary measures, which lack sufficient detail for describing complex medication-taking patterns. This paper uses recently developed methods for analyzing EMD data that capitalize more fully on their richness. Methods: Recently developed adaptive statistical modeling methods were used to analyze EMD data collected with medication event monitoring system (MEMS™) caps in a clinical trial testing the effects of motivational interviewing on adherence to antihypertensive medications in a cohort of hypertensive African-Americans followed for 12 months in primary care practices. This was a secondary analysis of EMD data for 141 of the 190 patients from this study for whom MEMS data were available. Results: Nonlinear adherence patterns for 141 patients were generated, clustered into seven adherence types, categorized into acceptable (for example, high or improving) versus unacceptable (for example, low or deteriorating) adherence, and related to adherence self-efficacy and blood pressure. Mean adherence self-efficacy was higher across all time points for patients with acceptable adherence in the intervention group than for other patients. By 12 months, there was a greater drop in mean post-baseline blood pressure for patients in the intervention group, with higher baseline blood pressure values than those in the usual care group. Conclusion: Adaptive statistical modeling methods can provide novel insights into patients’ medication-taking behavior, which can inform development of innovative approaches for tailored interventions to improve medication adherence.
关键词：adaptive statistical modeling; hypertension; medication adherence; Medication Event Monitoring System
