# Development and validation of a post-operative delirium prediction model for patients undergoing abdominal surgery: a retrospective, observational, single-center study

## TL;DR

This study developed a model to predict post-operative delirium in elderly patients undergoing abdominal surgery, identifying risk factors like age and medical history.

## Contribution

A novel prediction model for post-operative delirium in abdominal surgery patients, validated internally and externally.

## Key findings

- Old age, cerebrovascular disease, diazepam use, and intraoperative fluid imbalance were key predictors of delirium.
- The model achieved a sensitivity of 0.556 and specificity of 0.754 with an AUC of 0.703 in training data.
- External validation showed a lower AUC of 0.634, suggesting room for improvement in generalizability.

## Abstract

Postoperative delirium (POD) is a common complication following major surgery in elderly. The purpose of this study was to develop and evaluate a POD prediction model for patients undergoing abdominal surgery.

One thousand consecutive patients scheduled for elective abdominal surgery from July 2019 to March 2021 in Ruijin Hospital, Shanghai China, were retrospectively analysed, and their demographics, pre-operative evaluation, and intra-operative parameters were collected and cross-analysed. The primary outcome was the POD incidence. A prediction model of POD was established and internal validation was conducted with various analyses including univariate and multivariate regression. Data from another cohort of346 patients enrolled from July 2021 to December 2021 were used for model external validation.

After screening, 838 patients were included as the training cohort and 10.9% (91/838) of the patients manifested POD. Old age, cerebrovascular disease and diazepam use history and intraoperative fluid imbalance were the main contributors of the POD prediction model. The optimum cut-off point of the predicted probability that maximised the sum of sensitivity and specificity was 0.12. The fitting set AUC was 0.703 (95% Confidence interval (CI) 0.637–0.753). The sensitivity and specificity of the model were 0.556 and 0.754 respectively. The mean AUC during the cross and external validation of the model was 0.684 [Standard Deviation (SD) 0.068] and 0.634 (95%CI 0.511–0.758) respectively.

Our data indicated that improving perioperative management may reduce POD incidence in patients who are old age and have cerebrovascular disease history.

The retrospective data (ChiCTR2100047405) of this study was registered in the Chinese Clinical Trial registry (https://www.chictr.org.cn/).