CHARACTERIZING MEDICATION ADHERENCE TO AN ORAL GLUCOSE LOWERING DRUG, METFORMIN USING GROUP BASED TRAJECTORY MODELS

Medication adherence;Pharmacoepidemiology;group based trajectory models;type 2 diabetes mellitus;metformin;Epidemiology

Iyer, Geetha Sankar Narayanan

Johns Hopkins University

关键词: Medication adherence;    Pharmacoepidemiology;    group based trajectory models;    type 2 diabetes mellitus;    metformin;    Epidemiology;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/39475/IYER-THESIS-2016.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

BackgroundMedication non-adherence is a major contributor to suboptimal control of chronic diseases. Its consequences include inferior clinical outcomes as well as unnecessary health care costs. Group based trajectory models (GBTM), a type of finite mixture model, can be used to identify distinct trajectories of medication adherence among the patient population. It models adherence as a longitudinal parameter. ObjectiveTo characterize patterns of medication adherence among adult patients in the first year of initiating metformin using group based trajectory models, compared to the traditional summary measure of proportion of days covered.MethodsWe identified patients who initiated metformin, an oral glucose lowering drug, between 1st January to 31st December, 2011 from pharmacy prescription claims in Truven MarketScan Commercial Claims and Encounter database and followed them for a period of 360 days. We evaluated the number of days covered by metformin in 12 30 day periods and generated 12 monthly indicators to indicate whether that month was fully covered (defined as 24 or more days out of the 30 days). We modeled trajectories using group based trajectory models (2 to 6 groups) using these monthly indicators. We also calculated a traditional summary measure, proportion of days covered (PDC). We used Bayesian Information Criterion (BIC), posterior probabilities and clinical relevant interpretations in order to decide the best fit model. Additionally, we compared the accuracy of prediction of adherence of the different summary measures.ResultsAmong 77,279 patients who initiated metformin in 2011, we found that the 5 and 6 group model performed comparably. Overall the 6 group trajectory model summarized long term adherence best (C statistic 0.951) and PDC categorized as 80% or more (value of 1) and less than 80% (value of 0) summarized adherence worst (C statistic 0.767). However, keeping in the mind the relevance of clinical interpretation, we chose the 5 group model to be best fit model.ConclusionGroup based trajectory models can be used to summarize medication adherence more accurately than proportion of days covered. This newer method can be used by payers, clinicians and researchers in order to identify groups of patients with distinct adherence patterns and to identify targeted interventions for its improvement.