Mortality and predictive models for 90-/180- day postoperative death for NSCLC patients aged ≥70 years

Older patients with lung cancer are particularly vulnerable to surgical risks. However, analyses of postoperative mortality and the development of predictive models for patients aged ≥70 years after lung cancer surgery remain limited. Focusing on elderly thoracic surgery patients, this study systematically assessed 90- and 180-day postoperative mortality risks through predictive model development and validation, and ultimately establishing a risk stratification framework to support clinical decision-making and patient management.

Postoperative mortality was analyzed, and prediction models were developed based on elderly patients who received lung cancer surgery at Peking University Cancer Hospital (PKUCH) (from 1995 to 2019). The models were then validated using data from the National Cancer Database (NCDB) cohort from 2004 to 2014.

Postoperative mortality rates for patients aged ≥70 years were analyzed in the PKUCH cohort (n=989) and the NCDB cohort (n=16,321). Both cohorts showed increasing mortality from 30 to 180 days after surgery (0.4% and 1.0% within 30 days; 90- and 180-day mortality rates were 1.72%/2.03% and 3.44%/4.19%, respectively. Both models shared four predictors: age, sex, tumor diameter, and surgical approach. The models demonstrated good performance in internal validation [area under the curve (AUC) =0.80 and 0.79] and moderate but clinically useful discrimination in external validation (AUC=0.71 and 0.69). The 90- and 180-day models showed good calibration in both the development cohort (Hosmer-Lemeshow test, P=0.51 and 0.08) and the validation cohorts (P=0.44 and 0.12). Decision curve analysis confirmed the clinical utility of the models. After risk stratification, the high-risk group consistently had the highest mortality rate compared with the medium- and low-risk groups across different ethnic cohorts.

Postoperative mortality increased from 0-30 to 91-180 days in patients aged ≥70 years. Prediction models for 90- and 180-day mortality and risk stratification may help guide clinical decision-making.