Development and validation of a prediction score for safe outpatient colorectal resections

Background: Avoiding unnecessary inpatient stay may decrease hospital-acquired complications and costs while increasing patient satisfaction. This study aimed to develop and validate a score to identify patients eligible for safe same-day discharge after colorectal resections. Methods: This bi-institutional retrospective cohort study included consecutive patients undergoing elective colon and rectal resections (2011–2018) for benign and malignant indications. Two multivariable logistic models were developed based on demographic and surgical risk factors to predict a combined endpoint (ileus, anastomotic leak, intra-abdominal abscess, and readmission). Development and validation datasets were randomly sampled from the entire cohort. Areas under the receiver operating characteristic curves (AUC) were evaluated, and Hosmer-Lemeshow goodness-of-fit tests were used to assess validation model fit. Results: Of 5,389 patients, 1,182 (21.9%) experienced at least one complication of the combined endpoint. Male gender, open surgery, ASA ≥3, wound class ≥3, ileostomy, surgical duration >3 hours, and perioperative IV fluids >3 L all had significantly greater odds of the combined endpoint in the parsimonious multivariable model (all P <.05). The reduced model considering only the 4 variables with the highest OR (>1.5) contained open surgery, ASA ≥3, wound class ≥3, and surgical duration ≥3 hours as predictors (all P <.05, AUC of 0.65; 95% CI 0.63, 0.68). Both the parsimonious model and the reduced model demonstrated no lack of fit in the validation cohort. Conclusion: The suggested score composed of preand intraoperative items may help physicians decide on patients’ same-day discharge after colorectal resection.