Diagnostic test accuracies of 4AT items are consistent across the range of baseline cognition: results from two prospective studies.

backgroundPresenting features of delirium can be difficult to distinguish from co-existing dementia phenomenology. While the 4 'A's Test (4AT) is a well validated delirium screening tool across multiple patient populations and healthcare settings, how its diagnostic performance varies across different levels of baseline cognition is unclear.

methodsThis study harmonises data from two prospective cohorts: the Delirium and Population Health Informatics Cohort (DELPHIC) and Delirium and Cognitive Impact in Dementia (DECIDE). Both cohorts were stratified into tertiles according to baseline cognition. Delirium was defined using DSM-IV (DELPHIC) and DSM-V (DECIDE). 4AT domain scores were operationalised from relevant delirium assessments within each study (MDAS, OSLA, months of the year backwards); item 4 was drawn from assessments of fluctuation. This process yielded an operationalised 4AT. Diagnostic accuracy of the operationalised 4AT was calculated using sensitivity, specificity, and receiver operating characteristics, with pooled estimates for each test domain.

resultsAmong 396 unique participants and 2468 assessments across both studies, pooled 4AT sensitivity and specificity were 0.73 and 0.89, respectively. Overall 4AT performance was consistent across cognitive tertiles (sensitivity: high 0.82, middle 0.73, low 0.71; specificity: high 0.91; middle 0.92; low 0.86).

conclusionsThe operationalised 4AT showed good diagnostic performance in all three levels of baseline cognition. Sensitivity would likely have been higher had clinical information to inform 4AT item 4 been available. Overall, however, the findings support the value of bedside 4AT items in detecting delirium in hospitalised patients across the spectrum of baseline cognition.