Postoperative delirium (POD) is a severe complication following cardiac surgery and is associated with increased morbidity and mortality. The impact of intra- and early postoperative factors on the occurrence of POD following cardiac surgery remains controversial. To close this gap, we investigated intra- and early postoperative factors and their predictive values for POD. We performed a prospective observational study that aimed to FIND DElirium RIsk factors (FINDERI) for patients undergoing elective cardiac surgery. POD was assessed using the Confusion Assessment Method (CAM) algorithm. Intra- and early postoperative factors were extracted from electronic medical records and reviewed by cardiac surgeons. To identify potential predictors of POD, we used univariate and multivariate logistic regression along with machine learning (ML) with tenfold cross-validation. In our study cohort of 490 patients, 106 screened positive for POD (21.6%). In the multivariate analysis, we found a positive association between POD occurrence and age (p<0.001), duration of surgery (p=0.027), combined (versus isolated) surgical procedures (p=0.024), opening of the cardiac chambers (p=0.046), and ventilation time (p<0.001). The ML-based decision tree identified a two level-algorithm including ventilation time and aortic cross-clamping time, with an AUC of 0.7116 (p=0.0002) in the validation set. In the ML-based LASSO regression analysis, we identified ventilation time, administration of erythrocyte concentrates (EC), and usage of cardiopulmonary bypass (CPB) as predictors of POD, with an AUC of 0.7407 (p<0.0001) in the validation set. The results of this analysis highlight the associations between ventilation time, aortic cross-clamping time, administration of EC, and usage of CPB and POD. Additionally, they suggest that the optimization of surgical protocols has the potential to reduce POD risk in individuals undergoing cardiac surgery.
Read full abstract