CAM: Confusion Assessment Method

The original two-site study developed the CAM and compared two trained interviewers with psychiatrist DSM-III-R diagnoses in 56 older adults. Sensitivity was 100% and 94% and specificity 95% and 90% for the two interviewers, with excellent inter-rater reliability in this intensively assessed research setting.

Across 797 hospitalised older people and 2,721 paired observations, nurses recognised delirium in only 19% of affected observations and 31% of affected patients compared with research CAM and cognitive assessments. Sensitivity ranged from 15% to 31% and specificity from 91% to 99%; hypoactive delirium, dementia, older age and visual impairment were associated with missed cases, although the comparator was not an independent DSM interview.

A pilot after one hour of training included 32 palliative-care patients and 33 assessments, yielding 0.50 sensitivity and 1.00 specificity. After enhanced training, 52 patients and 54 assessments yielded 0.88 sensitivity and 1.00 specificity, showing that training can alter CAM performance materially.

This diagnostic-accuracy review included 22 studies and 2,442 participants and analysed the original CAM separately from CAM-ICU. Pooled sensitivity was 82% and specificity 99%, but heterogeneity and study-quality limitations mean the high specificity should not obscure variable sensitivity.

This cardiac-surgery study followed 141 patients with repeated DSM-IV-TR reference assessments, and 78 developed delirium. CAM sensitivity was 68% and specificity 90%, with false negatives concentrated among patients with less overt delirium phenotypes.

In 507 acute admissions in Tanzania, 95 participants had delirium and the CAM achieved an area under the ROC curve of 0.94 among those assessed. About one in ten patients could not complete the CAM, many of whom had delirium, so complete-case sensitivity and specificity can overstate performance in the whole clinical population.

In this blinded multicentre comparison, 384 older acute medical inpatients were allocated to CAM assessment against an independent DSM-IV reference standard. CAM sensitivity was 40% and specificity 100%, demonstrating that a highly specific result can coexist with substantial under-detection in a single-assessment workflow.

This broad older-adult review included 38 studies and 7,378 participants across several delirium instruments. For the original CAM, pooled sensitivity was 0.80 and specificity 0.98, while moderator analyses reinforced the importance of setting and study design.