Abstract
BACKGROUND AND OBJECTIVES: The Risk Analysis Index (RAI) accurately predicts adverse postoperative outcomes but the inclusion of cancer status in the RAI has raised two key concerns about its suitability for use in surgical oncology: (1) the potential over classification of cancer patients as frail, and (2) the potential overestimation of postoperative mortality for patients with surgically curable cancers. METHODS: We performed a retrospective cohort analysis to assess the RAI's power to appropriately identify frailty and predict postoperative mortality in cancer patients. We assessed discrimination for mortality and calibration across five RAI models-the complete RAI and four variants that removed different cancer-related variables. RESULTS: We found that the presence of disseminated cancer was a key variable driving the RAI's power to predict postoperative mortality. The model including only this variable [RAI (disseminated cancer)] was similar to the complete RAI in the overall sample (c = 0.842 vs. 0.840) and outperformed the complete RAI in the cancer subgroup (c = 0.736 vs 0.704, respectively, p < 0.0001, Max R(2) = 19.3% vs. 15.1%, respectively). CONCLUSION: The RAI demonstrates somewhat less discrimination when applied exclusively to cancer patients, but remains a strong predictor of postoperative mortality, especially in the setting of disseminated cancer.