A machine learning-based predictive model for long-term complications of totally implantable venous access ports in cancer patients: a tool for risk-stratified nursing care

Background: Totally implantable venous access ports (TIVAPs) are essential for long-term oncology treatments, but complications like infection and thrombosis disrupt care. Oncology nurses, who manage TIVAPs, lack validated bedside tools to prospectively identify high-risk patients. This study aimed to develop and validate a practical, nurse-oriented machine learning predictive model for long-term TIVAP complications to facilitate risk-stratified nursing care. Methods: This retrospective cohort study, following TRIPOD+AI and STROBE guidelines, included 2,522 cancer patients receiving TIVAPs at a comprehensive cancer centre (September 2019--September 2024). Least absolute shrinkage and selection operator (LASSO) regression selected predictors from 13 clinical variables. Thirteen machine learning algorithms were trained and evaluated on an independent test set using the area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis. The optimal model was translated into a bedside nomogram. Results: Over a median follow-up of 385 days, 104 patients (4.12%) developed long-term complications. A seven-variable logistic regression model---incorporating late-stage cancer, neoadjuvant chemotherapy, diabetes mellitus, body mass index, port duration exceeding one year, and season of implantation---achieved the highest test-set AUC of 0.801 (95% CI: 0.715--0.880), outperforming complex algorithms like gradient boosting. The model showed good calibration and positive net clinical benefit. A nomogram was derived for point-of-care risk estimation by nurses. Conclusions: The interpretable seven-variable model accurately predicts long-term TIVAP complications. The derived nomogram empowers oncology nurses to perform individualised risk stratification post-implantation, guiding targeted surveillance, tailored preventive interventions, and optimised nursing resource allocation.