Two-stage prediction model for postoperative delirium in ICU patients after cardiac surgery

Autor: Shining, Cai, Hang, Cui, Wenyan, Pan, Jingjing, Li, Xiaolei, Lin, Yuxia, Zhang
Rok vydání: 2022
Zdroj: European journal of cardio-thoracic surgery : official journal of the European Association for Cardio-thoracic Surgery.
ISSN: 1873-734X
Popis: Postoperative delirium is a common severe complication in ICU patients after cardiac surgery. We developed a two-stage prediction model and quantified the risk of developing postoperative delirium to assist in early prevention before and after surgery.We conducted a prospective cohort study and consecutively recruited adult patients after cardiac surgery. The Confusion Assessment Method for ICU was used to diagnose delirium five days postoperatively. The stage I model was constructed using patient demographics, health conditions and laboratory results obtained before the surgery, while the stage II model was built upon both pre- and postoperative predictors. The model was validated internally using bootstrap and externally using data from an external cohort.The two-stage model was developed with 654 patients and was externally validated with 214 patients undergoing cardiac surgery. The stage I model contained six predictors, while the stage II model included ten predictors. The stage I model had an area under the receiver operating characteristic curve of 0.76 (95% confidence interval: 0.68-0.81), and the stage II model's AUC increased to 0.85 (95% CI: 0.81-0.89). The external validation resulted in an area under the curve of 0.76 (95% CI: 0.67-0.86) for the stage I model and 0.78 (95% CI: 0.69-0.86) for the stage II model.The two-stage model assisted medical staff in identifying patients at high risk for postoperative delirium before and 24 hours after the cardiac surgery. This model showed good discriminative power and predictive accuracy and can be easily accessed in clinical settings.The study was registered with the US National Institutes of Health ClinicalTrials.gov (NCT03704324. Registered 11 October 2018).
Databáze: OpenAIRE