BACKGROUND: The gold standard treatment for urinary bladder cancer is cystectomy, which can result in postoperative complications, patient morbidities, and lengthy hospital visits post-surgery. At present, there are limited models that can predict length of stay for patients post cystectomy. Identifying risk factors associated with increased length of stay could have translational implications in mitigating such risk factors. As such, the goal of this study was to generate parsimonious models to predict length of hospital stay (LOHS) following cystectomy in bladder cancer patients.

METHODS: Data from the American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) was used to analyze patients who underwent cystectomy between 2014 and 2016. Modeling building approaches were used to identify the most informative preoperative and postoperative covariates to predict LOHS. Separate models were identified for preoperative covariates and for postoperative covariates and the most informative pre- and postoperative covariates were combined into a single model.

RESULTS: This analysis includes 6,074 patients (mean LOHS = 9.4 days). The preoperative (5 covariates; R squared= 0.245 and Adj R-squared = 0.240) and postoperative models (13 covariates; R squared= 0.237 and Adj R-squared = 0.236) were significantly better predictors compared to the ACS risk calculator model (19 covariates; R-square= 0.026 and Adj R-squared = 0.025). Likewise, the combined model (15 covariates) was a significantly better predictor (R squared= 0.241 & Adj R-squared = 0.239) when compared to the ACS risk calculator model. The combined model was subjected into CART analysis to stratify patients into five risk-groups based on ventilator over 48 hours, organ/space surgical site infection, wound disruption, and pneumonia. Patients in the high-risk group had significantly longer mean LOHS of 26.29 days compared to low-risk group of 8.4 days.

CONCLUSIONS: This study identified and validated a clinically important combined model that is to the ACS NSQIP surgical risk calculator to predict LOHS in patients who underwent cystectomy. Patients at high-risk may benefit from allocation of additional follow-up to prevent longer LOHS.

Citation Format: Rashmi Pathak, Jaileene Perez-Morales, Sephalie Y. Patel, Michael A. Poch, Matthew B. Schabath. Parsimonious model to predict length of stay following cystectomy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 883.