Background: Medication nonadherence is known to worsen glycemic control. Few studies have examined this relationship over several years.
Objective: The aim of this study was to examine the longitudinal effect of medication nonadherence on glycemic control among a large cohort of veterans.
Methods: Analysis was performed on a cohort of 11 272 veterans with type 2 diabetes followed from April 1994 to May 2006. The primary outcome measures were mean glycosylated hemoglobin A1c (A1C) and proportion in poor control (A1C > 8%) over time. The main predictor was medication nonadherence based on medication possession ratio (MPR). Other covariates included sociodemographics and ICD-9 coded medical and psychiatric comorbidities. Generalized linear mixed models (GLMMs) were used to assess the relationship between MPR and A1C after adjusting for covariates.
Results: Mean follow-up was 5.4 years. In the linear mixed model, after adjusting for baseline A1C and other confounding variables, mean A1C decreased by 0.24 (P < 0.001) for each 10% increase in MPR (95% CI = -0.27, -0.21). In the fully adjusted GLMM, each percentage increase in MPR was associated with a 48% lower likelihood of having poor glycemic control (odds ratio = 0.52; 95% CI = 0.4, 0.6). In both continuous and dichotomized A1C analyses, average A1C showed a decreasing trend over the study period (P < 0.001).
Conclusions: In patients with type 2 diabetes, glycemic control worsens over time in the presence of medication nonadherence. Future studies need to take into account the complexity of patient- and system-level factors affecting long-term medication adherence to improve diabetes-related outcomes.
Keywords: databases; epidemiology; outcomes research/analysis; type 2 diabetes.