Xiaochun ZHANG ^1* Cuizhong Wang ¹ Wei Cheng ¹ Guanwen Gong ² Gang Wang ² Wei Cheng ² Mingyue Shao ² Zhengming Deng ^2* Zhiwei Jiang ^2*

• ¹ First Clinical Medical College, Nanjing University of Chinese Medicine, Nanjing, China
• ² Affiliated Hospital of Nanjing University of Chinese Medicine, Nanjing, Liaoning Province, China

Background: Prolonged postoperative hospital stay following gastric cancer (GC) surgery is an important risk factor affecting patients' mood and increasing complications. We aimed to develop a nomogram to predict risk factors associated with prolonged postoperative length of stay (PLOS) in patients undergoing gastric cancer resection. Methods: Data were collected from 404 patients. The least absolute shrinkage and selection operator (LASSO) was used for variable screening, and a nomogram was designed. The nomogram performance was evaluated by the area under the receiver operating characteristic curve (AUC). The consistency between the predicted and actual values was evaluated via a calibration map, and the clinical application value was evaluated via decision curve analysis (DCA) and clinical impact curve analysis (CICA). Results: A total of 404 patients were included in this study. Among these patients, 287 were assigned to the training cohort, and 117 were assigned to the validation cohort. According to the PLOS quartile distance, 103 patients were defined as having prolonged PLOS. LASSO regression and logistic multivariate analysis revealed that 4 clinical characteristics, the neutrophil–lymphocyte ratio (NLR) on postoperative day one, the NLR on postoperative day three, the preoperative prognostic nutrition index and the first time anal exhaust was performed, were associated with the PLOS and were included in the construction of the nomogram. The AUC of the nomogram prediction model was 0.990 for the training set and 0.983 for the validation set. The calibration curve indicated good correlation between the predicted results and the actual results. The Hosmer‒Lemeshow test revealed that the P values for the training and validation sets were 0.444 and 0.607, respectively, indicating that the model had good goodness of fit. The decision curve analysis and clinical impact curve of this model showed good clinical practicability for both cohorts. Conclusion: We explored the risk factors for prolonged PLOS in GC patients via the enhanced recovery after surgery (ERAS) program and developed a predictive model. The designed nomogram is expected to be an accurate and personalized tool for predicting the risk and prognosis of PLOS in GC patients via ERAS measures.