The delirium screening tool 4AT in routine clinical practice: prediction of mortality, sensitivity and specificity

Key summary pointsAim Investigate if 4AT score predicts 1 year mortality and explore the sensitivity and specificity of the 4AT when applied as part of a clinical routine. Findings 4AT score is one of several clinical characteristics predicting 1 year mortality. The 4AT has reasonable sensitivity and specificity to detect delirium in a clinical routine setting. Message The 4AT seems to be a useful tool for delirium screening and may predict mortality. Purpose Delirium is common and associated with poor outcomes, partly due to underdetection. We investigated if the delirium screening tool 4 A's test (4AT) score predicts 1 year mortality and explored the sensitivity and specificity of the 4AT when applied as part of a clinical routine. Methods Secondary analyses of a prospective study of 228 patients acutely admitted to a Medical Geriatric Ward. Physicians without formal training conducted the index test (the 4AT); a predefined cut-off >= 4 suggested delirium. Reference standard was delirium diagnosed by two geriatricians using the Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-5). We calculated hazard ratios (HR) using Cox regression based on the groups 4AT = 0, 1-3, 4-7 and >= 8, first unadjusted, then adjusted for the covariates age, comorbidity, and personal activities of daily living. We calculated sensitivity, specificity, and the area under the receiver operating curve (AUC). Results Mean age of patients was 86.6 years, 139 (61.0%) were female, 78 (34.2%) had DSM-5 delirium; of these, 56 had 4AT-delirium. 1 year mortality was 27.6% (63 patients). Compared to 4AT score 0, the group 4AT >= 8 had increased 1 year mortality (HR 2.86, 95% confidence interval 1.28-6.37, p = 0.010). The effect was reduced in multiadjusted analyses (HR 1.69, 95% confidence interval 0.70-4.07, p = 0.24). Sensitivity, specificity, and AUC were 0.72, 0.84, and 0.88, respectively. Conclusions 4AT >= 8 indicates increased mortality, but the effect was reduced in multiadjusted analyses. 4AT had acceptable sensitivity and specificity when applied as a clinical routine.