Evaluation of delirium screening tools in geriatric medical inpatients: a diagnostic test accuracy study.

Introduction: screening all unscheduled older adults for delirium is recommended in national guidelines, but there is no consensus on how to perform initial assessment.

Aim: to evaluate the test accuracy of five brief cognitive assessment tools for delirium diagnosis in routine clinical practice.

Methods: a consecutive cohort of non-elective, elderly care (older than 65 years) hospital inpatients admitted to a geriatric medical assessment unit of an urban teaching hospital. Reference assessments were clinical diagnosis of delirium performed by elderly care physicians. Routine screening tests were: Abbreviated Mental Test (AMT-10, AMT-4), 4 A's Test (4AT), brief Confusion Assessment Method (bCAM), months of the year backwards (MOTYB) and informant Single Question in Delirium (SQiD).

Results: we assessed 500 patients, mean age 83 years (range = 66-101). Clinical diagnoses were: 93 of 500 (18.6%) definite delirium, 104 of 500 (20.8%) possible delirium and 277 of 500 (55.4%) no delirium; 266 of 500 (53.2%) were identified as definite or possible dementia. For diagnosis of definite delirium, AMT-4 (cut-point < 3/4) had a sensitivity of 92.7% (95% confidence interval (CI): 84.8-97.3), with a specificity of 53.7% (95% CI: 48.1-59.2); AMT-10 (<4/10), MOTYB (<4/12) and SQiD showed similar performance. bCAM had a sensitivity of 70.3% (95% CI: 58.5-80.3) with a specificity of 91.4% (95% CI: 87.7-94.3). 4AT (>4/12) had a sensitivity of 86.7% (95% CI: 77.5-93.2) and specificity of 69.5% (95% CI: 64.4-74.3).

Conclusions: short screening tools such as AMT-4 or MOTYB have good sensitivity for definite delirium, but poor specificity; these tools may be reasonable as a first stage in assessment for delirium. The 4AT is feasible and appears to perform well with good sensitivity and reasonable specificity.