Preoperative Risk Factors for Predicting Postoperative Human Serum Albumin Infusion after Hip Fracture Surgery: Development and Validation of a Nomogram

Abstract

Background: As one of the adverse events after hip fracture surgery, hypoalbuminemia is usually treated using human serum albumin infusion. However, the application of human serum albumin may cause complications such as postsurgical infection and increased mortality.

Aims: To examine the preoperative risk factors of human serum albumin infusion after hip fracture surgery, establish a nomogram prediction model, and verify its accuracy.

Study design: A retrospective cross-sectional study.

Methods: Eligible patients who underwent hip fracture surgery were divided into the infusion and non-infusion groups according to whether human serum albumin was infused or not. All patients were divided randomly into a training set and a testing set in line with the ratio of 7:3. In the training set, independent risk factors of postoperative human serum albumin infusion were determined by univariate logistic regression analysis, LASSO regression, and multivariate logistic regression analysis. Then, a nomogram model was established. Furthermore, the receiver operating characteristic curve and calibration curve were plotted, and decision curve analysis was performed for the training and testing sets to assess the predictability, discriminative ability, and clinical usefulness of the model.

Results: This study included a total of 1,339 eligible patients, 141 of whom were injected with human serum albumin postoperatively. Altogether, the training set incorporated 939 patients, and the testing set included 400 patients. Multivariate logistic analysis indicated five independent risk factors, including chronic lung disease (odds ratio, 95% confidence interval, 2.618, 1.413-4.849, p = 0.002), (albumin; odds ratio, 95% confidence interval, 0.842, 0.787-0.900, p < 0.001), prothrombin time (odds ratio, 95% confidence interval, 1.252, 1.071-1.463, p = 0.005), red blood cells (odds ratio, 95% confidence interval, 0.370, 0.228-0.602, p < 0.001), and type of anesthesia (odds ratio, 95% confidence interval, 0.553, 0.327-0.937, p = 0.028). Fracture type, a clinically significant factor, was also considered. Finally, the nomogram model was built based on these seven predictors. The areas under the curve of the nomogram were 0.854 (95% confidence interval, 0.811-0.898) and 0.767 (95% confidence interval, 0.686-0.847) in the training and testing sets separately. As shown in the calibration curve, the predicted result was consistent with the observed one. The decision curve analysis indicated that the nomogram has good clinical value.

Conclusion: Low preoperative serum albumin levels, low preoperative red blood cell counts, prolonged preoperative prothrombin time, history of chronic lung disease, and general anesthesia were independent risk factors for postoperative human serum albumin infusion. Besides, the fracture type, clinically significant factor, was also included. The nomogram that combined these six predictors could accurately predict the risk of postoperative human serum albumin infusion.

Figure 1

Workflow of this study.

Figure 2

LASSO analysis of 23 significant variables in the univariate analysis. (a) LASSO coefficient profiles. (b) LASSO deviance profiles.

Figure 3

Nomogram for predicting postoperative HSA infusion in patients who underwent hip fracture surgery.

Figure 4

Evaluating the performance of the nomogram in the training set. The receiver operating characteristic curve (a), calibration curve (b), and decision curve analysis (c) of the training set.

Figure 5

Validating the predictive ability of the nomogram in the testing set. The receiver operating characteristic curve (a), calibration curve (b), and decision curve analysis (c) of the testing set.