Development and validation of models for predicting mortality in intertrochanteric fracture surgery patients with perioperative blood transfusion: a prospective multicenter cohort study

Abstract

Background: The association between allogenic blood transfusions (ABT) and all-cause mortality in surgically treated hip fracture patients with perioperative transfusion (STHFPT) remained unknown. We aim to introduce transfusion-related factors, new variables to develop and validate models to predict mortality in these patients. Methods: A prospective multicenter cohort study was conducted with STHFPT hospitalized during Jan. 2018 and Jun. 2021. The database was divided into training cohort and validation cohort in a ratio of 70% to 30% using the randomization method. All participants received a minimum of 2-year follow-up and all participants’ overall and eight time-specific survival status were recorded. Prediction models were developed using multivariate logistic regression and Cox regression for variable selection. Model performance was measured by determining discrimination, calibration, overall model performance or precision, and utility. Sensitivity analyses were performed to test robustness of the results. Results: A total of 7074 consecutive patients were prospectively screened and assessed for eligibility to participate. Finally, 2490 patients met our inclusion and exclusion criteria and 1743 (70%) patients were randomized to the training cohort and 747 (30%) to the validation cohort. The median duration of follow-up was 38.4 months (IQR 28.0-62.0). Our novel models highlight that preoperative transfusion is of significance for short-term mortality while mid-term outcomes are predominantly determined by severe complications, pulmonary complications, and advanced age. Our models showed high discriminative power, good calibration and precision for mortality prediction in both training and validation cohorts, especially in short-term mortality prediction. Conclusions: We introduce transfusion-related factors, new variables to develop and validate models to predict mortality with STHFPT. The models can be further tested and updated with the ultimate goal of assisting in optimizing individual transfusion strategy.