ASSESSING THE EFFECTS OF ANTIHYPERTENSIVE MEDICATION ADHERENCE THROUGH BAYESIAN D

Boston University Medical Campus, Boston MA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Over 78 million American adults, or about one third of the population over 20 years old, suffer from hypertension, with an even higher prevalence in priority populations such as African American (AA) adults at 44%. Left untreated, hypertension can lead to a range of serious and costly health concerns, such as cardiovascular disease, stroke, and renal disease. While 80% of adults with hypertension are aware of their condition, less than half have their blood pressure (BP) under control. Among the many factors associated with a lack of BP control, patients' non-adherence to prescribed antihypertensive medications is a major concern. While little doubt exists that patients who more poorly adhere to their prescribed antihypertensive medication are at risk for worse BP outcomes, it is less clear how to accurately measure the impact of time- varying patient adherence on BP levels, and how to accurately measure medication effectiveness and the effects of socio-demographic and risk factors on BP levels accounting for the variation in adherence. This proposed project is designed to address these questions. In this study, we will develop a novel approach to measuring BP control based on daily measurements of adherence through Bayesian dynamic linear models (DLM), and apply the approach to a pre-existing cohort of hypertensive patients. DLMs have a long history as a statistical framework for forecasting and measuring trajectories in many contexts, including real-time missile tracking as well as financial securities forecasting. The application of DLMs to measuring BP control is novel, but fits naturally into the DLM framework because BP can be tracked over time as adherence data is accumulated. The aims of our study include (1) to develop a Bayesian dynamic linear model framework for BP levels from daily antihypertensive medication adherence, and assess the fit and applicability of the models, (2) to assess the extent to which the DLM framework results in improvement over models that incorporate adherence data using more traditional approaches, and (3) to estimate the effects of risk factors and medications on BP that control for time-varying adherence. Based on the medication effect sizes estimated from the third aim, the results of this work could form the basis of a follow-up large-scale study on a more heterogeneous population to examine the comparative effectiveness of antihypertensive medications, controlling for detailed adherence. Furthermore, if successful, the DLM approach we develop for BP could be adapted eventually to other disease settings where adherence is crucial to treatment success, such as HIV, cancer, and mental illnesses such as depression.