Construction and validation of an early prediction model of delirium in children after congenital heart surgery

Transl Pediatr. 2022 Jun;11(6):954-964. doi: 10.21037/tp-22-187.

Abstract

Background: Delirium often occurs in children with congenital heart disease in the early postoperative period, which is not conducive to the rehabilitation and prognosis. There is little evidence to prove the effectiveness and safety of drug treatment of delirium in children, and the prevention has become an important topic. The purpose of this study is to analyze the early risk factors of delirium in children after congenital heart surgery, establish a nomogram prediction model, and explore the application efficiency of the model, so as to provide reference for early prevention of delirium.

Methods: A total of 362 children treated in the cardiac intensive care unit (CICU) of Shanghai Children's Medical Center after congenital heart surgery from February 15 to April 15, 2021 were enrolled for the construction of the model. Bedside nurses who received unified training used the Cornell Assessment of Pediatric Delirium (CAPD) to evaluate delirium and recorded sixteen preoperative- and intraoperative-related influencing factors. A nomogram prediction model was created using multivariate logistic regression. The prediction effect of the model was evaluated by C-index and Brier value, and 96 children from April 16 to May 15, 2021 were included for effect verification. The model's effectiveness was validated by comparing the occurrence of delirium in children predicted by the model with the actual occurrence.

Results: Multivariate logistic regression analysis showed that male gender [odds ratio (OR) =1.786, 95% confidence interval (CI): 1.018-3.134, P=0.043], age <6.5 months (OR =0.224, 95% CI: 0.126-0.399, P=0.000), disease severity ≥4 points (OR =6.955, 95% CI: 3.564-13.576, P=0.003), and operation time ≥148 min (OR =2.401, 95%CI: 1.336-4.315, P=0.000) were independent risk factors for delirium in children after cardiac surgery. The C-index of the nomogram prediction model was 0.808, sensitivity was 76.1%, specificity was 70%, and the Brier value was 0.142. The validation of the model showed that the model predicted 20 cases and the actual occurrence was 20 cases, of which 8 cases were false negative and 8 cases were false positive, and the sensitivity, specificity, and accuracy of the model were 60%, 89.5%, and 83.3%, respectively.

Conclusions: The prediction model constructed in this study could provide early prediction of the occurrence of delirium in children after congenital heart surgery to a certain extent.

Keywords: Congenital heart surgery; children; delirium; nomogram; risk prediction model.