Elderly cancer patients are more predisposed to developing nosocomial infections during anti-neoplastic treatment, and are associated with a bleaker prognosis. This study aimed to develop a novel risk classifier to predict the in-hospital death risk of nosocomial infections in this population.
Retrospective clinical data were collected from a National Cancer Regional Center in Northwest China. The Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was utilized to filter the optimal variables for model development and avoid model overfitting. Logistic regression analysis was performed to identify the independent predictors of the in-hospital death risk. A nomogram was then developed to predict the in-hospital death risk of each participant. The performance of the nomogram was evaluated using receiver operating characteristics (ROC) curve, calibration curve, and decision curve analysis (DCA).
A total of 569 elderly cancer patients were included in this study, and the estimated in-hospital mortality rate was 13.9%. The results of multivariate logistic regression analysis showed that ECOG-PS (odds ratio [OR]: 4.41, 95% confidence interval [CI]: 1.95-9.99), surgery type (OR: 0.18, 95%CI: 0.04-0.85), septic shock (OR: 5.92, 95%CI: 2.43-14.44), length of antibiotics treatment (OR: 0.21, 95%CI: 0.09-0.50), and prognostic nutritional index (PNI) (OR: 0.14, 95%CI: 0.06-0.33) were independent predictors of the in-hospital death risk of nosocomial infections in elderly cancer patients. A nomogram was then constructed to achieve personalized in-hospital death risk prediction. ROC curves yield excellent discrimination ability in the training (area under the curve [AUC]=0.882) and validation (AUC=0.825) cohorts. Additionally, the nomogram showed good calibration ability and net clinical benefit in both cohorts.
Nosocomial infections are a common and potentially fatal complication in elderly cancer patients. Clinical characteristics and infection types can vary among different age groups. The risk classifier developed in this study could accurately predict the in-hospital death risk for these patients, providing an important tool for personalized risk assessment and clinical decision-making.