A practical taxonomy of delirium tools
Delirium tools are difficult to classify because one label rarely says enough. A single instrument may assess for current delirium, measure severity, draw on family observations, suit a particular setting and lend itself to repeated use. These are different properties.
This catalogue therefore uses a layered taxonomy. The first layer describes the question a tool addresses. The other layers describe when and how it is used, the population and setting, and the available evidence. A tool can carry more than one label.
One shared broad role
The 4AT, CAM and 3D-CAM are structured tools used to assess for current delirium. They differ in format, administration, training requirements, setting and workflow. Their results inform clinical judgement and should be interpreted alongside history, examination and other relevant information. A service may use one tool or a locally designed pathway; this does not create an intrinsic hierarchy between instruments.
How this catalogue describes a delirium tool
What question does the tool address?
Assessment / detection of current delirium
Is delirium possible or indicated now?
Severity and course
How severe is it, and how is it changing?
Motor subtype / phenomenology
Which symptoms or motor pattern are present?
Experience and impact
What burden or distress has the episode caused?
Domain-specific or adjunctive measure
What do attention, arousal or cognition show?
Future-risk prediction
Who may develop delirium later?
Retrospective ascertainment
Was delirium likely to have been present earlier?
Cause / aetiology support
Which possible causes should be considered after delirium is recognised?
Contextual, historical or service record
What related evidence or service experience does the record describe?
When and how is it used?
- When delirium is suspected
- Routine serial observation
- Repeated measurement during an episode
- Patient interview or test
- Direct observation
- Staff observation during care
- Informant report
- Health-record review
- Digital, physiological or hybrid method
In whom, where and with what evidence?
- Population
- Care setting
- Rater and training
- Administration time
- Version or adaptation
- Evidence type and maturity
- Access and permissions
Dimensions considered in this catalogue
Every public record has a deliberate primary role or catalogue function. Other structured dimensions are recorded where the source material supports them; that coverage is still being expanded, so a blank field does not necessarily mean that a characteristic is absent.
On a small screen, swipe across the table to see every column.
| Dimension | Question | Examples | How it is used here |
|---|---|---|---|
| Primary role | What question does the tool or related record address? | Assessment / detection of current delirium; Severity and course; Motor subtype / phenomenology; Experience and impact; Domain-specific or adjunctive measure; Future-risk prediction; Retrospective ascertainment; Cause / aetiology support; Contextual, historical or service record | Each record has a deliberate primary role. Other dimensions may still give it several relevant characteristics. Future-risk and retrospective methods remain separate from current bedside assessment. |
| Temporal use | When is it applied? | When delirium is suspected; serial surveillance; repeated measurement during an episode; retrospective review | Monitoring describes a pattern of use, not a higher or lower class of instrument. |
| Information source | Where does the information come from? | Patient interview or test; direct observation; staff observation; informant; record; device; hybrid | Informant-based describes the source of information, not the clinical purpose. |
| Component characteristic | Does the record identify a particular contributing domain? | Cognition; attention; arousal; acute change or fluctuation | A component characteristic supports discovery but does not replace or change the record's primary role. |
| Population and setting | In whom and where was it designed or studied? | Adult; paediatric; intensive care; emergency department; general ward; perioperative; palliative care; care home | Evidence from one setting does not establish suitability in another. |
| Administration | What does use require? | Time; item or feature count; rater; training; certification; handling of untestable responses | Ultra-brief describes administration time. It is not a purpose or evidence rating. |
| Delivery | How is it presented? | Paper form; structured interview; observation chart; digital application; record algorithm; physiological method | Delivery format does not define what a tool measures or how well it performs. |
| Evidence | What has been studied? | Development; reliability; construct validity; diagnostic accuracy; external validation; pooled evidence; implementation | Evidence labels describe evidence type. They are not endorsement or a single quality ranking. |
| Catalogue relationship | What kind of record is this? | Instrument family; version; severity derivative; workflow; adjunct; emerging method; historical record | Related versions and workflows are linked without being counted as wholly independent families. |
| Catalogue context | Why is an additional descriptive tag retained? | Detailed research characterisation; harmonisation; historical or service context | Catalogue-context tags support navigation. They are not clinical purposes, evidence rankings or endorsements. |
Where the 4AT, CAM and 3D-CAM sit
The differences lie in construction and use, not in a ranking of diagnostic status.
On a small screen, swipe across the table to see every column.
| Attribute | 4AT | CAM | 3D-CAM |
|---|---|---|---|
| Broad role | Assessment / detection of current delirium | Assessment / detection of current delirium | Assessment / detection of current delirium |
| Structure | Four-part structured assessment combining observation, brief cognitive testing, attention and acute change or fluctuation | Feature-based classification applied to information gathered through an appropriate interview, cognitive assessment, observation and history | Structured interview and observations that operationalise CAM features |
| Typical administration | About 2 minutes | CAM scoring can be brief; total time depends on the assessment used to gather the features | Median 3 minutes in the original validation study |
| Training | No formal course or certification is required; staff should read the instructions and be competent to use it | Training is recommended; performance depends on the preceding assessment and feature ratings | Use the current manual and recommended training |
| Evidence | Pooled diagnostic-test-accuracy evidence | Pooled diagnostic-test-accuracy evidence | Pooled diagnostic-test-accuracy evidence |
| Severity measurement | The 4AT is not designed as a separate severity scale | CAM-S is a separate severity method | 3D-CAM-S is a separately derived severity method |
| Clinical interpretation | Alongside history, examination and other relevant information | Alongside history, examination and other relevant information | Alongside history, examination and other relevant information |
A two-stage pathway may suit a particular service, but it is one design choice rather than a universal requirement. A tool result contributes to clinical diagnosis; it does not identify the cause of delirium, and a negative result should not overrule continuing clinical concern.
Boundary cases
A test of attention or arousal may contribute useful evidence without assessing the full delirium syndrome. An informant tool may establish acute change or fluctuation but may not replace direct assessment. Risk models estimate the probability of future delirium, while chart methods look back for evidence of an earlier episode; neither performs the same task as a current bedside assessment. Aetiology tools support investigation after delirium is recognised rather than deciding whether it is present.
Paediatric emergence measures remain in an adjacent group because pain, agitation, arousal and delirium can overlap during post-anaesthetic recovery. Physiological, EEG, sensor and digital methods also need careful labels: some deliver an established assessment in a new way, while others measure a proposed marker and remain at an early evidence stage.
Disclosure
Professor Alasdair MacLullich, the curator of DeliriumTools.com, co-developed the 4AT. This is a relevant intellectual interest because the taxonomy and comparison include the 4AT. The site declares this openly and applies the same descriptive dimensions to each instrument. No external organisation has endorsed this taxonomy. See the full disclosures.
Selected sources
- Schuurmans MJ, et al. The measurement of delirium: review of scales. 2003. PubMed.
- Helfand BKI, et al. Detecting delirium in non-ICU settings: a systematic review of validated instruments. 2021. Full text.
- National Institute for Health and Care Excellence. Delirium: prevention, diagnosis and management. CG103, updated 2023. Recommendations.
- Tieges Z, et al. Diagnostic accuracy of the 4AT: systematic review and meta-analysis. 2021. PubMed.
- Shi Q, et al. Confusion Assessment Method: systematic review and meta-analysis. 2013. PubMed.
- Ma R, et al. Diagnostic accuracy of the 3D-CAM: systematic review and meta-analysis. 2023. PubMed.
- Vasunilashorn SM, et al. Derivation and validation of a 3D-CAM severity scoring method. 2016. PubMed.
- MacLullich A. A classification of delirium assessment tools. 2020. Delirium Words.