To operate or not: Prediction of 3-month postoperative mortality in geriatric cancer patients

Context: Appropriate selection of aging patient who fit for cancer surgery is an art-of-state. Objectives: This study aimed to identify predictive factors pertinent to 3-month postoperative mortality in geriatric cancer patients. Methods: A total of 8,425 patients over 70 years old with solid cancer received radical surgery between 2007 and 2012 at four affiliated hospitals of the Chang Gung Memorial Hospital were included. The clinical variables of patients who died within 3 months post-surgery were analyzed retrospectively. Recursive partitioning analysis (RPA) was performed by randomly selecting 50% of the patients (testing set) to identify specific groups of patients with the lowest and highest probability of 3-month postoperative mortality. The remaining 50% were used as validation set of the model. Results: Patients' gender, Eastern Cooperative Oncology Group performance (ECOG scale), Charlson comorbidity index (CCI), American Society of Anesthesiologist physical status, age, tumor staging, and mode of admission were independent variables that predicted 3-month postoperative mortality. The RPA model identified patients with an ECOG scale of 0-2, localized tumor stage, and a CCI of 0-2 as having the lowest probability of 3-month postoperative mortality (1.1% and 1.3% in the testing set and validation set, respectively). Conversely, an ECOG scale of 3-4 and a CCI >2 were associated with the highest probability of 3-month postoperative mortality (55.2% and 47.8% in the testing set and validation set, respectively). Conclusion: We identified ECOG scale and CCI score were the two most influencing factors that determined 3-month postoperative mortality in geriatric cancer patients.