Development and validation of a predictive model for transfusion in major abdominal surgery: a multicentre retrospective study.

BACKGROUND

Blood products are limited resources and preoperative transfusion often relies on protocols or clinical experience rather than objective risk assessment. Thus a systematic and accurate prediction strategy is needed for efficient blood utilisation and timely administration. We developed and validated a nomogram to predict intraoperative red blood cell (RBC) transfusion.

METHODS

We developed a nomogram that integrates preoperative variables to predict intraoperative RBC transfusion in patients undergoing major abdominal surgery at a tertiary care hospital (2012-21). The model was externally validated with datasets from other tertiary hospitals and an open dataset. Variables were selected using least absolute shrinkage or selection operator logistic regression. Discriminative ability was assessed with C-statistic, and calibration plots were used to evaluate the agreement between predicted and observed outcomes.

RESULTS

This retrospective study included 128 749 patients for model development and 71 590 and 3701 patients in two external validation sets, respectively. Intraoperative RBC transfusion occurred in 6.0% of patients in the development set, and in 4.5% and 7.3% of patients in the two validation sets, respectively. The nomogram incorporated seven preoperative variables including age, American Society of Anesthesiologists physical status, European Society of Cardiology surgical risk, haemoglobin, concentration, platelet number, international normalised ratio, and albumin concentration. The nomogram showed good predictive ability with C-statistics of 0.857 (95% confidence interval [CI]: 0.852-0.861), 0.847 (95% CI: 0.840-0.854), and 0.848 (95% CI: 0.822-0.873) in the development and external validation sets, respectively, demonstrating good calibration across the predicted probability range.

CONCLUSIONS

This nomogram, integrating seven preoperative variables, can effectively predict intraoperative transfusion in major abdominal surgery.

CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov (NCT06847490).