Confusion Assessment Method (CAM) for the diagnosis of delirium

Purpose and status. The CAM (Inouye et al., Ann Intern Med 1990) is a structured bedside screening instrument that operationalises the DSM criteria so that non-psychiatrists can detect delirium reliably. It is the most widely used delirium tool in the world and the parent of a family of derivatives. Note carefully: the CAM screens for delirium; the formal diagnosis still rests on DSM-5 or ICD-10.

The four features. (i) Acute onset and fluctuating course; (ii) Inattention; (iii) Disorganised thinking; (iv) Altered level of consciousness.

Scoring algorithm (must be written exactly). The screen is positive when the patient has feature (i) AND feature (ii), PLUS either feature (iii) OR feature (iv). In short: (1 AND 2) + (3 OR 4). Features 1 and 2 are therefore mandatory — a patient cannot be CAM-positive without acute fluctuating onset and inattention.

How to elicit each feature. Feature 1 — from a collateral informant: is this a change from baseline, and does it come and go during the day? Feature 2 (inattention) — months of the year backwards, days of the week backwards, or digit span 3–4 backwards; or simply observe that the patient is inappropriately distracted or cannot keep track of the conversation. Feature 3 (disorganised thinking) — test orientation to time and place; observe an illogical or rambling flow of ideas. Feature 4 (altered consciousness) — anything other than alert: vigilant, lethargic, stuporous or comatose.

Performance and derivatives. The CAM has a sensitivity of 94% and specificity of 89%, with high inter-rater reliability. Derivatives: 3D-CAM (scripted 3-minute version — sensitivity 96% / specificity 86% in dementia, 93% / 96% without dementia); CAM-ICU for intubated patients (RASS ≠ 0 for feature 4, SAVEAHAART letter task for feature 2); bCAM for the emergency department; and the UB-2 two-item screen derived from 3D-CAM validation data.

Advantages and limitations. Advantages: rapid, bedside, no equipment, validated across settings, usable by nurses and non-specialists, high sensitivity and specificity, standardises documentation, and enables the daily screening that guidelines require. Limitations: it needs training and a collateral history for feature 1; performance falls without formal cognitive testing of attention; it cannot distinguish delirium from dementia in an unknown baseline and is unreliable in established DSD; being a single point-in-time assessment it can be falsely negative during a lucid interval; it does not grade severity (use MDAS or DRS-R-98) and does not identify the cause; and it cannot be applied to a comatose patient.

Current guidance no longer treats the CAM as the only acceptable tool. NICE CG103 (2023) and SIGN recommend the 4AT for routine ward and emergency screening because it needs no training and, unlike the CAM, remains scoreable in patients who are too drowsy or untestable to complete a formal interview. The APA 2025 guideline requires systematic screening with a validated instrument — the CAM family remains fully acceptable, but the examinable point is that the CAM is now one validated option among several rather than the default.