Predicting length of stay after robotic partial nephrectomy

Introduction: To investigate factors predictive of length of stay (LOS) after robotic partial nephrectomy (RPN) in an effort to identify patients suitable for RPN with overnight stay at outpatient surgical facilities. Materials and methods: Retrospective chart review of patients who underwent RPN at Memorial Sloan Kettering Cancer Center from January 2007 to July 2012 was conducted. Univariate and multivariate analyses were performed to identify the main predictors of LOS. The discrimination of the multivariate model was measured using the area under the curve (AUC); tenfold cross-validation was performed to correct for over-fit. Results: One hundred and eighty-six patients were included in the analysis; 84 (45 %) had LOS of ≤1 day (median LOS 2 day; interquartile range 1–2). On univariate analysis, preoperative variables associated with LOS > 1 included larger tumors (P < 0.0001), lower estimated glomerular filtration rate (P = 0.003), older age (P = 0.006), female gender (P = 0.035), and higher comorbidity score (P = 0.015); operative variables associated with LOS > 1 day included greater estimated blood loss (P < 0.0001) and longer operative (P < 0.0001) and ischemia (P < 0.0001) times. The AUC of the preoperative model was 0.61 (95 % CI 0.52–0.69) after tenfold cross-validation. Conclusions: LOS after RPN is influenced by age, gender, medical comorbidities, and tumor size. However, when analyzed retrospectively, these factors had limited ability to predict LOS after RPN with sufficient accuracy to develop a prediction tool.