The Value of the Aperiodic Exponent of the Intraoperative Electroencephalogram for Predicting Postoperative Delirium in Elderly Patients

BACKGROUND: Postoperative delirium (POD) is a frequent and serious complication after surgery. Parameters of the electroencephalogram (EEG), such as the Bispectral Index and the occurrence of burst suppression, have been associated with POD. We analyzed the predictive properties of periodic and aperiodic parameters of the EEG power spectrum. METHODS: In a secondary post-hoc analysis of a prospective observational study, patients with an age of at least 60 years undergoing major cardiac or non-cardiac surgery were analyzed. The frontal intraoperative raw EEG was recorded by a BIS monitor and offline analyzed with the FOOOF toolbox, revealing the periodic and aperiodic parameters of the power spectrum. Patients were screened for POD and divided in a 2: 1 ratio into a training and a validation cohort. Predictors of POD were identified by uni- and multivariable logistic regression. RESULTS: Thirty-two out of the 120 training group patients developed POD. These showed a significantly longer median duration of surgery (286 interquartile range IQR 236–391 vs 223 127–331 min, P =. 005), lower median BIS (40. 4 IQR 38. 1–43. 4 vs 42. 7 39. 5–46. 0, P =. 038), and a higher mean aperiodic exponent (2. 09 ± 0. 19 vs 1. 99 ± 0. 17, P =. 017). Duration of surgery (odds ratio OR = 1. 01; 95% confidence interval CI, 1. 00–1. 01, P =. 005), peak center frequency (OR = 0. 79; CI, 0. 62–0. 97, P =. 039) and aperiodic exponent (OR = 23. 2; CI, 2. 1–318. 2, P =. 013) were predictors of POD according to univariable logistic regression. At a cutoff of 1. 967, the aperiodic exponent had a sensitivity of 0. 813 and a specificity of 0. 478 for the prediction of POD. A low duration of time spent in burst suppression was observed in both patients with POD (67 6–363 s) and without POD (173 4–641 s, P =. 30), and cumulative burst suppression time was not a predictor of POD. In a stepwise regression model, age, duration of surgery, peak power, and the aperiodic exponent were associated with POD (AUROC = 0. 80 (CI, 0. 71–0. 89, P <. 001). The predictive model was confirmed in the validation group (n = 60) with an AUROC = 0. 77 (CI, 0. 65–0. 90, P =. 001). CONCLUSIONS: A higher aperiodic exponent, for example, a less complex EEG signal, is associated with a greater POD risk, especially in combination with known POD predictors such as age and duration of surgery.