Abstract In the era of enhanced recovery after surgery (ERAS), accurate estimation of postoperative length of stay (p-LOS) after gastrectomy is significant for patient discharge planning and hospital bed management. However, a precise and straightforward prediction tool is lacking. In this study, the clinical data of 924 patients who underwent gastrectomy between December 2018 and December 2023 were included from five tertiary care medical centers. There were 647 cases in the training set and 277 cases in the verification set. The linear regression analyses were used to determine the independent factors affecting p-LOS. The model's accuracy was assessed by the C-index, calibration curve, and decision curve analysis (DCA). The final model included age, American Society of Anesthesiologists (ASA) classification, Charlson comorbidity index (CCI), tumor size, and location, all of which had a significant effect on p-LOS. The nomogram comprising these five predictors exhibited sufficient predictive accuracy, with a C-index of 0.662 in the training group and of 0.664 in the verification group. The calibration curves and DCA of both groups had excellent calibration and clinical usefulness. In the high-risk group, the use of robot-assisted surgery (RAS) could effectively shorten p-LOS. The developed nomogram could provide personalized risk prediction for p-LOS, and help clinicians better manage patients within ERAS.
Luo et al. (Wed,) studied this question.