Presentation (CTI)

OBJECTIVES: Comparative evidence on health care utilization between managed care and fee-for-service Medicare populations remains sparse, largely because comparable utilization data are difficult to obtain. This limitation is especially consequential for cancer patients, whose care often involves intensive and complex service use. This study aimed to predict enrollment in Medicare Advantage Prescription Drug plans (MA-PD) and standalone Prescription Drug Plans (PDP) among cancer patients, and to identify key factors associated with enrollment choice using machine learning methods.
METHODS: This study from the 2020-2022 Medicare Current Beneficiary Survey (MCBS), linked with Medicare administrative claims. Medicare beneficiaries with a cancer diagnosis were identified and classified as MA-PD or PDP enrollees. Three machine learning models, including Random Forest (RF), Support Vector Machines (SVM), and Extreme Gradient Boosting (XGB), were developed. We included 52 multidimensional variables based on a framework for comprehensive evaluation. Decision curve analysis (DCA) was conducted to compare the clinical utility.
RESULTS: Weighted analyses indicated that 54.57% of cancer beneficiaries were enrolled in MA-PD and 45.42% in standalone PDP. The RF model demonstrated the strongest predictive performance, achieving an area under the curve (AUC) of 0.74, outperforming SVM (0.66) and XGB (0.67). The DCA showed that the RF model consistently yielded greater net benefit across a range of threshold probabilities. Variable importance analysis identified poverty status, residence type, census region, income, age, race and ethnicity, hypertension status, sex, smoking status, and marital status as the top 10 predictors of enrollment.
CONCLUSIONS: Machine learning methods effectively distinguished prescription drug plan enrollment among Medicare cancer beneficiaries, with RF outperforming other approaches. The prominence of socioeconomic and demographic predictors highlights structural factors that influence enrollment decisions. Efforts to improve equity in Medicare drug coverage should account for social and regional disparities when designing plan options, outreach strategies, and enrollment guidance for cancer patients.