Balancing case mix risk: The role of cancer type mix on savings in total cost of care models.

82 Background: The Enhancing Oncology Model (EOM) and its predecessor, the Oncology Care Model (OCM), represent total cost of care (TCOC) models in oncology. Both rely on price prediction frameworks built from historical expenditure data and specific independent predictive variables. The OCM integrates cancer types into its overall prediction model, whereas the EOM adopts a cancer type-specific approach. Case mix, reflecting the diversity, complexity, and severity of patients treated at a healthcare facility, is a key concern in TCOC models. In oncology, case mix issues arise from variations in cancer types treated and differences between attributed and comparator populations. This study examines the impact of cancer type case mix on TCOC model performance. Methods: We analyzed deidentified national baseline episode data from the Center for Medicare & Medicaid Innovation (CMMI), along with coefficients from the EOM price prediction model and the OCM performance period 8 prediction model. Baseline prices were calculated for 1.5 million episodes spanning July 2016 to June 2020, and savings were determined as the difference between baseline prices and total standardized per-episode expenditures. We examined cancer type mix, total and per-episode expenditures, baseline prices, and savings for both models. Two linear regression analyses were conducted, with cancer type mix as the independent variable and savings as the dependent variable. R-squared values, indicating the proportion of variance explained by cancer type mix, were compared to assess its effect on savings. Results: The analysis revealed significant case mix differences between participating physician group practices (PGP), especially for PGPs with < 72 episodes per performance period (66% of overall PGPs). The EOM linear regression analysis had an R-squared value of 0.058, indicating that 5.8% of savings variance was explained by cancer type mix. Conversely, the OCM linear regression analysis had an R-squared value of 0.406, meaning 40.6% of savings variance was attributable to cancer type mix. This suggests that cancer type mix plays a greater role in the OCM model’s financial outcomes than in the EOM model. Conclusions: Cancer type mix significantly influences TCOC model performance in oncology. The OCM's broad integration of cancer types as independent variables increased case mix risk for participating PGPs, while the EOM’s cancer type-specific prediction model mitigates some risk by accounting for cancer type variability. We commend CMMI for making favorable refinements to the price prediction model. PGPs must navigate resource and cost management challenges in TCOC models without any control over case mix. These findings highlight the need for ongoing refinement of TCOC models to maintain predictive accuracy while minimizing unnecessary risk from case mix.