Comparison of Methods for Predicting High-Cost Patients Captured Within the Oncology Care Model (OCM): a Simulation Study

• Activity Number: 119 - SPEED: Government and Health Policy
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Jung-Yi Lin* and Wei Zhang and Mark Liu and Mark Sanderson and Luis Isola and Madhu Mazumdar and Liangyuan Hu
• Companies: Icahn School of Medicine at Mount Sinai and UALR and Mount Sinai Health System and Mount Sinai Health System and Mount Sinai Health System and Icahn School of Medicine at Mount Sinai and Icahn School of Medicine at Mount Sinai
• Keywords: Oncology Care Model; risk-adjustment model; machine learning; quartile regression

Abstract

The Centers for Medicare & Medicaid Services (CMS) developed the Oncology Care Model (OCM) as an episode-based payment model to encourage participating practitioners to provide better care at a lower cost for Medicare beneficiaries with cancer. CMS defined 6-month episodes for OCM that begin with the receipt of outpatient non-topical chemotherapy for cancer. Usually a log-link Gamma generalized linear model is used for estimating concurrent cost. However, Advanced Supervised Machine Learning Method (ASMLM) and Partially Linear Additive Quartile Regression (PLAQR) have recently emerged as modeling approaches with higher flexibility. The CMS risk-adjustment model includes 16 variables including age, gender, disease, and treatment. Extensive simulations were conducted to examine the performance of these methods under various contexture-motivated scenarios. The methods were also applied to the data on 4205 OCM episodes collected in 2012-2015 from the Mount Sinai Health System. The methods are compared using the root mean square error (RMSE), classification accuracy (CA), and patient cost accuracy (PCA) as the performance metrics.