Predicting Severe Postoperative Complications after CRS-HIPEC: An Externally Validated Machine-Learning Tool

Current decision support tools designed to predict postoperative complications, following cytoreductive surgery with hyperthermic intraperitoneal chemotherapy (CRS-HIPEC), are limited by small sample sizes and lack of external validation. Our aim was to develop an externally validated machine-learning tool that can predict severe complications (Clavien-Dindo grade ≥ 3) after CRS-HIPEC.

The dataset consisted of adult patients who underwent CRS-HIPEC at the University of Pittsburgh Medical Center (UPMC) and was split into an 80:20 ratio for the training and internal validation datasets, respectively. For external validation, we used the US HIPEC collaborative dataset. Three different models were trained and tested, and SHAP values were calculated to assess variable importance.

A total of 37% (n = 719/1955) of cases in the UPMC cohort experienced severe postoperative complications compared to 22% (n = 134/617) in the US HIPEC collaborative dataset. Recursive feature elimination algorithm determined that the optimal performance was achieved with 15 variables. After optimization, the random forest model had the highest area under the ROC curve (AUC, 0.71) in the internal validation cohort and was chosen as the final model. The random forest model had an AUC that ranged from 0.60 to 0.73 (mean AUC, 0.65) in the external validation cohorts. Finally, the most important variables associated with severe complications were high PCI, subtotal gastrectomy, low albumin levels, small bowel resection, and high Charlson Comorbidity Index.

Using the largest cohort of CRS-HIPEC patients in the US, we developed and externally validated a machine-learning model that can aid with patient selection and help providers anticipate complications in the postoperative setting.