Discontinuation of Oral Antidiabetic Drugs among Medicare Beneficiaries with Diabetes: Different Factors Predict Short Versus Long Periods Without Drug Therapy

Oral antidiabetic drugs (OADs) are mainstay medications for type 2 diabetes (T2D) patients who cannot control glucose levels through diet and exercise.  While OAD regimens are frequently modified to maintain control, most patients can expect to remain on OAD therapy throughout their lives. Understanding factors associated with inappropriate discontinuation of OAD therapy is thus critically important to long-term diabetes care management.  There is a large literature on adherence with OADs, yet little attention has focused on dynamic changes in patient behavior and healthcare utilization patterns proximal to OAD discontinuance, and there has been no formal analysis of the extent to which different factors predict shorter versus longer periods without OAD therapy.  We assessed predictors of OAD discontinuance in a nationally representative sample of Medicare beneficiaries with T2D in fee-for-service Part D drug plans between 2006 and 2008.  Following a one-year baseline period of OAD use, we identified 3 mutually exclusive cohorts with OAD discontinuation periods of 3-6 months (N=3,925), 7-11 months (N=1,205), and 12 or more months (N=5,276).  The 3 cohorts were compared on a series of variables hypothesized to influence discontinuance including demographic characteristics, low-income subsidy status, SSDI disability status, and monthly values for potential time-varying predictors of discontinuance (count of different OADs taken; insulin use; tests for FBG and HbA1C; eye exam; hospital and SNF admissions; ED visits; Medicare and Part D spending; first month of use for OADs, ACEIs/ARBs, and statins; gaps in therapy; discontinuance with ACEIs/ARBs and statins; number of different physicians seen; and diagnoses for hypoglycemia, uncontrolled diabetes, and long and short term diabetes complications).  We estimated a multinomial logit model to identify factors that were differentially predictive of OHA discontinuance among the 3-6 and 7-11 month cohorts compared to long-term discontinuers.  Long-term discontinuers were much sicker than the other cohorts, with higher monthly Medicare and Part D drug spending, higher proportion using insulin, higher probability of hospital/SNF admissions, and greater prevalence of diabetes complications.  Three sets of prognostic factors distinguished shorter- from longer-term discontinuers:  (1) longer periods of use of OADs and ACEIs/ARBs, (2) taking 2 or more OADs in the 3 months prior to discontinuation, and (3) lower discontinuation rates for ACEIs/ARBs and statins in the 3 months prior to OAD discontinuance.  Although we cannot rule out removal of OAD therapy by prescribers, we believe that monitoring these factors should help clinicians identify patients at highest risk for inappropriate long-term OAD discontinuance.