Delirium Risk Screening in Patients Undergoing Cardiac Surgery: Results from the Prospective Observational FINDERI Study

Background: Postoperative delirium (POD) is a common complication of cardiac surgery that is associated with longer hospital stay, higher morbidity, cognitive decline, and mortality. The incidence of POD varies from 8% to 55% after cardiac surgery. Preoperative assessments may help to identify patients´ POD risk and be important for prevention of POD. The aim of this study was to analyze predictive values of a delirium risk screening questionnaire and cognitive, frailty, and geriatric assessments for POD, and to use data-driven machine learning to develop a decision-tree based POD prediction tool.

Methods: The FINd DElirium RIsk factors (FINDERI) is a prospective observational study in patients aged ≥ 50 years undergoing an elective cardiac surgery. POD was assessed using Confusion Assessment Method (CAM). Predictors included the Delirium Risk Screening Questionnaire (DRSQ; adapted from the PAWEL study), frailty (assessed using the Canadian Study of Health and Aging Clinical Frailty Scale), and preoperative cognitive status (assessed using the Montreal Cognitive Assessment [MoCA], Trail Making Test [TMT] A, and TMT B). Predictive properties were analyzed using receiver operating characteristics and multivariate approaches (regularized regression and decision trees).

Results: In the analyzed sample of 504 patients, the mean age was 68.3 ± 8.2 years (21.4% women). Most participants underwent coronary artery bypass grafting (n = 327, 65%), and 87.5% received cardiopulmonary bypass. POD incidence was 21.0% (n = 106). DSQR showed an area under the curve (AUC) of 0.68 (95% CI 0.62, 0.73) in predicting POD after cardiac surgery. TMTB also showed a good performance in predicting POD after cardiac surgery with an AUC of 0.67 (95% CI 0.60, 0.72). Using a machine learning approach, a three-rule decision tree prediction model including DRSQ score > 7, TMTB time > 118 second and MoCA score ≤ 22 was identified. The accuracy rate of the decision tree was 80.3%.

Conclusion: The DRSQ performed well in predicting POD and may be implemented into the daily routine of cardiac surgery. A three-rule decision tree including the DRSQ, TMTB and MoCA was identified as a useful POD risk screening pathway.

Publikationsverlauf

Artikel online veröffentlicht:
13. Februar 2024