Use of an early disease-modifying drug adherence measure to predict future adherence in patients with multiple sclerosis.

BACKGROUND: Patients with multiple sclerosis (MS) who are adherent to their treatment regimens are less likely to experience relapses and the cost associated with relapse. Pharmacists whose practice involves these specialty pharmaceuticals used to treat MS are striving for ways to improve outcomes by achieving treatment adherence in their patients. Specialty pharmacies have reported higher adherence rates than traditional pharmacies, which may translate to improved outcomes. Identifying patients who warrant increased adherence intervention is critical. Models using administrative health care claims to predict adherence have typically included demographic characteristics, comorbidities, and/or previous consumption of health care resources. Addition of a measure of early adherence may improve the ability to predict future adherence outcomes.
OBJECTIVE: To evaluate early adherence with disease-modifying drugs (DMDs) as a predictor of future adherence in patients with MS.
METHODS: The first DMD claim (i.e., index event) for adult MS patients (aged ≥18 years and aged ≤ 65 years) who received self-injected DMDs between January 1, 2006, and May 31, 2010, was identified in a national U.S. managed care database. Patients were required to have continuous eligibility for 12 months pre- and 24 months post-index. Multiple regression models were used to predict future adherence as measured by the proportion of days covered (PDC). The base model included age, gender, a medication intensity measure, presence of a non-MS-related hospitalization pre-index, and markers for physical difficulty, forgetfulness, or depression/stress. Models adding early DMD adherence as a covariate were analyzed using incrementing 30-day periods predicting the subsequent 360 days.
RESULTS: There were 4,606 patients included with an average age of 46.0 (SD 9.4) years, and 78.7% were female. Average PDC in the first 360 days post-index was 80.0% (SD 26.0). Using the first 60 days of early adherence as the only predictor in the model showed an R2 of 20.6%. The base model (i.e., no early adherence measure but other covariates included) yielded an adjusted R2 of only 2.3%. As the time period of early adherence is increased (from 60 to 360 days), the explained variance as measured by adjusted R2 values increased from 20.6% to 53.5% (early adherence-only models). Addition of the covariates, other than early adherence, increased the R2 by 1% to 2%.
CONCLUSIONS: Statistical predictive models that include early adherence with DMDs were able to explain the variance in future adherence outcomes to a greater extent than models based solely on baseline characteristics. The efficiency of an adherence intervention in reaching its intended target can be improved by using models such as these with enhanced specificity and selectivity.