Background:

Perioperative hypoalbuminemia of the posterior lumbar interbody fusion (PLIF) can increase the risk of infection of the incision site, and it is challenging to accurately predict perioperative hypoproteinemia. The objective of this study was to create a clinical predictive nomogram and validate its accuracy by finding the independent risk factors for perioperative hypoalbuminemia of PLIF.

Methods:

The patients who underwent PLIF at the Affiliated Hospital of Qingdao University between January 2015 and December 2020 were selected in this study. Besides, variables such as age, gender, BMI, current and past medical history, indications for surgery, surgery-related information, and results of preoperative blood routine tests were also collected from each patient. These patients were divided into injection group and non-injection group according to whether they were injected with human albumin. And they were also divided into training group and validation group, with the ratio of 4:1. Univariate and multivariate logistic regression analyses were performed in the training group to find the independent risk factors. The nomogram was developed based on these independent predictors. In addition, the area under the curve (AUC), the calibration curve and the decision curve analysis (DCA) were drawn in the training and validation groups to evaluate the prediction, calibration and clinical validity of the model. Finally, the nomograms in the training and validation groups and the receiver operating characteristic (ROC) curves of each independent risk factor were drawn to analyze the performance of this model.

Results:

A total of 2482 patients who met our criteria were recruited in this study and 256 (10.31%) patients were injected with human albumin perioperatively. There were 1985 people in the training group and 497 in the validation group. Multivariate logistic regression analysis revealed 5 independent risk factors, including old age, accompanying T2DM, level of preoperative albumin, amount of intraoperative blood loss and fusion stage. We drew nomograms. The AUC of the nomograms in the training group and the validation group were 0.807, 95% CI 0.774-0.840 and 0.859, 95% CI 0.797-0.920, respectively. The calibration curve shows consistency between the prediction and observation results. DCA showed a high net benefit from using nomograms to predict the risk of perioperative injection of human albumin. The AUCs of nomograms in the training and the validation groups were significantly higher than those of five independent risk factors mentioned above (P < 0.001), suggesting that the model is strongly predictive.

Conclusion:

Preoperative low protein, operative stage ≥ 3, a relatively large amount of intraoperative blood loss, old age and history of diabetes were independent predictors of albumin infusion after PLIF. A predictive model for the risk of albumin injection during the perioperative period of PLIF was created using the above 5 predictors, and then validated. The model can be used to assess the risk of albumin injection in patients during the perioperative period of PLIF. The model is highly predictive, so it can be clinically applied to reduce the incidence of perioperative hypoalbuminemia.