Postoperative delirium (POD) is a common neuropsychiatric disorder characterized by inattention, fluctuating levels of consciousness, and disorganized thinking. POD can have serious consequences, including institutionalization and death. Risk stratification may target prevention to individuals at greater risk of POD. The objective of this study was to identify all published POD risk prediction models (RPMs) and to compare them with regard to their clinical practicability and predictive and discriminative performance. PubMed and EMBASE were searched from inception to January 1, 2013, for articles describing POD RPMs. Studies were included if they presented data from a cohort study, examined one or more RPMs, examined POD as an outcome, and assessed the performance of the RPM(s). Thirty of 2,246 articles were included, and 37 RPMs were found. Sixteen and six studies described individuals who had undergone cardiovascular and orthopedic surgery, respectively. The Confusion Assessment Method (CAM) for the intensive care unit checklist was the most often used diagnostic method (65%), followed by the Diagnostic and Statistical Manual of Mental Disorders (DSM), Fourth Edition criteria (16%). Predictors most often used in RPMs were age (20), preoperative Mini-Mental State Examination score (10), and preoperative increased alcohol use (7). Thirty RPMs were not validated, three were validated internally, and four were validated externally. Size of the models was not associated with their discriminatory performance. Instead of creating steadily new RPMs, existing RPMs should be further tested, improved, and meta-analytically integrated. It may be too early to implement a particular PODRPM in clinical practice with confidence.

Background Prehospital detection of large vessel occlusion stroke of the anterior circulation (LVO-a) would enable direct transportation of these patients to an endovascular thrombectomy (EVT) capable hospital. The ongoing ELECTRA-STROKE study investigates the diagnostic accuracy of dry electrode electroencephalography (EEG) for LVO-a stroke in the prehospital setting. To determine which EEG features are most useful for this purpose and assess EEG data quality, EEG recordings are also performed in the emergency room (ER). Here, we report data of the first 100 patients included in the ER. Methods Patients presented to the ER with a suspected stroke or known LVO-a stroke underwent a single EEG prior to EVT. Diagnostic accuracy for LVO-a stroke of frequency band power, brain symmetry and phase synchronization measures were evaluated by calculating receiver operating characteristic curves. Optimal cut-offs were determined as the highest sensitivity at a specificity of ≥ 80%. Results EEG data were of sufficient quality for analysis in 65/100 included patients. Of these, 35/65 (54%) had an acute ischemic stroke, of whom 9/65 (14%) had an LVO-a stroke. Median onset-to-EEG-time was 266 min (IQR 121–655) and median EEG-recording-time was 3 min (IQR 3–5). The EEG feature with the highest diagnostic accuracy for LVO-a stroke was theta–alpha ratio (AUC 0.83; sensitivity 75%; specificity 81%). Combined, weighted phase lag index and relative theta power best identified LVO-a stroke (sensitivity 100%; specificity 84%). Conclusion Dry electrode EEG is a promising tool for LVO-a stroke detection, but data quality needs to be improved and validation in the prehospital setting is necessary. (TRN: NCT03699397, registered October 9 2018).

BackgroundPatients with a stroke who are transferred to a comprehensive stroke center for endovascular treatment (EVT) often undergo repeated neuroimaging prior to EVT.ObjectiveTo evaluate the yield of repeating imaging and its effect on treatment times.MethodsWe included adult patients with a large vessel occlusion (LVO) stroke who were referred to our hospital for EVT by primary stroke centers (2016–2019). We excluded patients who underwent repeated imaging because primary imaging was unavailable, incomplete, or of insufficient quality. Outcomes included treatment times and repeated imaging findings.ResultsOf 677 transferred LVO stroke, 551 were included. Imaging was repeated in 165/551 patients (30%), mostly because of clinical improvement (86/165 (52%)) or deterioration (40/165 (24%)). Patients who underwent repeated imaging had higher door-to-groin-times than patients without repeated imaging (median 43 vs 27 min, adjusted time difference: 20 min, 95% CI 15 to 25). Among patients who underwent repeated imaging because of clinical improvement, the LVO had resolved in 50/86 (58%). In patients with clinical deterioration, repeated imaging led to refrainment from EVT in 3/40 (8%). No symptomatic intracranial hemorrhages (sICH) were identified. Ultimately, 75/165 (45%) of patients with repeated imaging underwent EVT compared with 326/386 (84%) of patients without repeated imaging (p<0.01).ConclusionsNeuroimaging was repeated in 30% of patients with an LVO stroke and resulted in a median treatment delay of 20 minutes. In patients with clinical deterioration, no sICH were detected and repeated imaging rarely changed the indication for EVT. However, in more than half of patients with clinical improvement, the LVO had resolved, resulting in refrainment from EVT.