Development and validation of a prediction model for postoperative delirium in elderly spinal patients: a...

Previous studies have predominantly focused on risk prediction models for delirium in patients undergoing cardiac or orthopedic joint surgeries, with limited reports on prediction models for postoperative delirium (POD) following spinal surgeries. This study first investigates independent risk factors for POD in elderly spinal surgery patients. Based on these factors, it subsequently constructs a nomogram prediction model, the accuracy and applicability of which are rigorously validated.  METHODS: The data of 846 patients who underwent surgical treatment in the Department of Spinal Surgery at the First Affiliated Hospital of Guilin Medical University from January 2019 to September 2023 were retrospectively collected. The cases were randomly divided into a training set (592 cases) and a validation set (254 cases) using a 7:3 randomization method, where the training set was used for model development and the validation set for model evaluation. Binary Logistic regression analysis was employed to identify independent risk factors for POD and to construct a predictive model and a score-risk prediction value reference table. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). Internal validation was conducted using the validation set.

The results of this study found that 144 patients developed POD, with an incidence rate of 17.0%. Multivariate regression analysis revealed that patient age, history of smoking, alcohol consumption, diabetes, preoperative psychiatric disorders, anesthesia method, and postoperative hemoglobin levels were independent risk factors for POD in elderly patients undergoing spinal surgery. Based on these factors, a nomogram model and a scoring-risk prediction table were developed. The total score of the model ranges from 0 to 350, corresponding to predicted risks ranging from less than 1% to over 90%. The model's performance was evaluated using the receiver operating characteristic (ROC) curve. The area under the ROC curve (AUC) for the training cohort was 0.81 (95% CI = 0.76-0.86), and for the validation cohort, it was 0.86 (95% CI = 0.80-0.93). Additionally, the Hosmer-Lemeshow goodness-of-fit test results showed good calibration, with p-values of 0.685 for the training cohort and 0.706 for the validation cohort, indicating favorable model consistency. Decision curve analysis (DCA) demonstrated that the net benefit within the intervention probability threshold range was 8% to 68% for the training set and 7% to 73% for the validation set, suggesting that the model has high practical value in clinical decision-making.

The patient's age, smoking history, alcohol use history, diabetes history, preoperative psychiatric disorders history, anesthesia method, and postoperative hemoglobin levels are identified as independent risk factors for the occurrence of postoperative delirium in patients undergoing such surgeries. Simultaneously, the nomogram prediction model established based on these factors demonstrates good risk prediction accuracy for postoperative delirium and excellent clinical applicability.