Related Experiment Video
Updated: Jan 7, 2026
An R-Based Landscape Validation of a Competing Risk Model
Published on: September 16, 2022
Development and Internal Validation of a Multivariable Prediction Model for Postoperative Bleeding in Patients
Ksawery Bieniaszewski1, Michał Szymański2, Maciej Wilczyński2
1Medical University of Gdańsk, Gdańsk, Poland. ksawery.bieniaszewski@gumed.edu.pl.
Background:
Postoperative bleeding is a potentially severe complication of bariatric surgery. reported in 1-4% of patients. Despite recognition of several risk factors. no validated multivariable model combining surgical and early postoperative laboratory data has been developed to support discharge decisions within Enhanced Recovery After Bariatric Surgery (ERABS) pathways.
Methods:
We performed a retrospective cohort study of all patients undergoing laparoscopic sleeve gastrectomy or gastric bypass (Roux-en-Y or one-anastomosis) between January 2017 and December 2023 at a high-volume bariatric center. The primary outcome was postoperative bleeding within 30 days. defined as any surgical or endoscopic intervention for suspected bleeding or transfusion requirement. Candidate predictors included demographic. surgical. and routinely available perioperative laboratory data. A multivariable logistic regression model was constructed using backward stepwise selection. Model discrimination and calibration were assessed. and internal validation was performed using bootstrap resampling (500 iterations).
Results:
Of 2.024 patients screened. 1.925 were included. with 49 bleeding events (2.5%). Bleeding occurred more often after gastric bypass. in male patients. in those with hypertension. lower preoperative BMI. and with excessive postoperative drainage. Significant perioperative laboratory changes included greater hemoglobin decline and postoperative increases in platelets and white blood cells. The final model incorporated five predictors: procedure type. excessive postoperative day 1 drainage. and perioperative changes in hemoglobin. platelets. and white blood cells. The model demonstrated excellent discrimination (AUC = 0.92) and calibration (Hosmer-Lemeshow p = 0.81). Bootstrap validation confirmed minimal optimism (corrected AUC = 0.917).
Conclusions:
The POD 1-DISCHARGE model offers a pragmatic. inexpensive. and accurate method for early risk stratification of bleeding after bariatric surgery. Its reliance on POD 1 clinical and laboratory data makes it readily applicable across diverse centers. supporting safer discharge practices within ERABS. Prospective multicenter validation will determine its broader applicability.
