The impact of applying clinical adjusters via histology to lung cancer in the EOM.

88 Background: The Enhancing Oncology Model (EOM) incentivizes participating practices to decrease total cost of care, and compares actual episode costs to benchmark episode costs to assess performance. In the benchmark episode calculation, static clinical adjusters estimated from historical episode costs are used as multipliers to reflect the variation in the cost of care of episodes - metastatic status for solid tumors and HER2 histology for breast cancer. However, lung cancer does not have a histology-based clinical adjuster despite the fact that CMS requires practices to submit this information. Methods: We used Medicare claims data from the first performance period episodes with treatment initiated between July 2023 and December 2023 for a large community oncology practice that represents 10% of lung cancer episodes. We use the performance period reconciliation files that include actual and benchmark price cost per episode, and the histologies reported by the practice despite not being included in the benchmark price calculation. There were 445 reconciliation eligible lung cancer episodes in the performance period. 47 are missing a reported histology. We use the remaining 398 episodes for the regression model. These episodes were reported as: Mesothelioma (7), Adenocarcinoma (184), Large Cell Carcinoma (4), Squamous Cell Carcinoma (73), Small Cell Carcinoma (44), and Other histology (86). We regress the difference between the actual and the benchmark episode costs on the histology dummy variables. We also build a model using only 3 histologies - Adenocarcinoma, Squamous Cell Carcinoma, and Small Cell Carcinoma - as these are the most common histologies. All models use robust standard errors that adjust for heteroskedasticity by down-weighting observations with high leverage (HC2). Results: Regressing the difference between the actual and the benchmark episode costs on the histologies reported by the practice yields a model that is not statistically significant (p=0.08). The pruned model fitted using episodes with only 3 histologies is statistically significant (p=0.03) and explains 1.4% of the variance in the difference between episode spend and benchmark price. Conclusions: The EOM requires that practices report histology data for lung cancer, but these data are not currently used to inform benchmarks for lung cancer episodes. Our findings using a subset of histologies suggest that the amount of variance in the residuals explained by these data elements may be low. However, this analysis with a single large practice is underpowered, and an adequately powered study is required to draw conclusions about the program. Given the high reporting burden, we look forward to CMS evaluating the value of reporting these data elements. Model N R^2 R^2 Adj F-statistic p-value All histologies 398 0.024 0.012 1.950 0.085 Excluding “other” histology 312 0.030 0.018 2.387 0.051 3 biggest histologies 301 0.021 0.014 3.686 0.026