Concurrent validity of two instruments (the Confusion Assessment Method and the Delirium Rating Scale) in the detection of delirium among older medical inpatients

Age and Ageing 2005; 34: 72–83 Age and Ageing Vol. 34 No. 1  British Geriatrics Society 2005; all rights reserved Research Letters Concurrent validity of two instruments (the Confusion Assessment Method and the Delirium Rating Scale) in the detection of delirium among older medical inpatients SIR—Delirium is an acute fluctuating disorder of consciousness, attention and cognition with thought disturbances, delusions and hallucinations [1]. Over 20% of acutely ill older hospital inpatients have delirium and it is associated with worse clinical outcomes, increased mortality and health care costs [2–4]. Its presence is frequently not recognised [4–6] and management remains sub-optimal [7]. Possible explanations for this under-recognition include the fluctuating pattern of symptoms and the hypoactive, withdrawn state of some patients who escape notice [8, 9]. When cognitive function is assessed and impairment noted, staff may not confidently distinguish delirium from dementia [9, 10]. Delirium is a broader construct than cognitive impairment, so the Abbreviated Mental Test Score (aMTS) [11] or the Mini-Mental State Examination (MMSE) [12] are not sufficient for diagnosis. The BGS delirium guidelines [13] recommend routine use of cognitive testing along with specific diagnosis of delirium based upon the DSM-IV criteria. In clinical practice, non-psychiatric specialist doctors may find the use of simplified but validated diagnostic tools more feasible than formal application of ‘gold standard’ DSM-IV criteria [14, 15]. Two of the more commonly used tools are the Delirium Rating Scale (DRS) [16] and the Confusion Assessment Method (CAM) [17]. Description of DRS and CAM The DRS was developed for use by psychiatrically trained clinicians to measure the severity of delirium, based on DSM-III criteria. Each of 10 items is scored from 0 to a maximum of 2, 3 or 4 points, the maximum total being 34 points. Sensitivity and specificity are high [18] but are dependent on the cut-point used for case definition of delirium [16, 18–20]. The DRS has been used in a variety of settings. Among patients of geriatric medical and psychiatry services with a high prevalence of delirium, it was a significantly better discriminator than the MMSE [20]. Sensitivity was 0.82 and specificity 0.94 with a cut-point of 10. In contrast, the Confusion Assessment Method (CAM) was designed to assist the detection of delirium by nonpsychiatrists. It consists of an extensive rating system based on nine features (acute onset, inattention, disorganised thinking, altered level of consciousness, disorientation, memory impairment, perceptual disturbances, psychomotor agitation or retardation, and altered sleep–wake cycle) operationalised from the DSM-IIIR and a briefer diagnostic algorithm. The full rating enables a detailed semi-quantitative comprehensive assessment of delirium, but accurate detection of 72 delirium using the diagnostic algorithm requires judgements about only four features: (i) acute onset with a fluctuating course; (ii) inattention; (iii) disorganised thinking; and (iv) an altered level of consciousness. CAM-positive delirium requires the presence of (i) and (ii) and either (iii) or (iv) [17]. Sensitivity and specificity of over 0.90 have also been demonstrated for CAM used with elderly hospitalised patients, against the gold standard of psychiatrist diagnosis using DSMIIIR criteria [17]. Although CAM sensitivity (1.0) was superior to the MMSE in a study of post-operative delirium [21], its operating characteristics are affected by the assessors’ skills, the patient population and the nature of the interview used to inform judgements needed for the diagnostic algorithm [9, 15, 21–23]. Its performance in the UK has not been investigated. Thus, the DRS and the CAM have been derived from similar but not identical ‘gold standard’ criteria. Both can detect delirium accurately, i.e. they have criterion validity [18, 24] and should have high agreement in case detection. We report here a preliminary study of an investigation into the utility of CAM to detect delirium among elderly medical inpatients in the UK. Our aim was to investigate the concurrent agreement (validity) [25] of these two methods, the more rigorous DRS acting as a proxy for a formal DSM-based expert diagnosis. To our knowledge this has not been previously reported. Patients and methods Study design This study was secondary analysis of data collected in a prospective observational study examining the association between physical illness severity and the symptoms and natural history of delirium. Subjects and settings The study was performed in an urban teaching hospital Elderly Care Unit, which provides acute medical care and rehabilitation. One hundred and seventy-one consecutively admitted patients aged 70 years or over were approached for informed consent within 3 days of hospital admission. The study population (n = 94) consisted of those with capacity and willingness to consent (n = 91) and those without capacity when assent was available from relatives (n = 3). Patients re-admitted were included once only. The local research ethics committee approved the study including capacity assessment and consent procedures. Ethical considerations about possible bias consequent on capacity and consent methods have been discussed elsewhere [26]. Assessments i. MMSE [12] to assess cognition. ii. CAM [17] for a dichotomous classification of delirium/ not delirium. Research letters patients with severe impairment, 12 were classified as delirious on both DRS cut-points and CAM identified 11 of these. Among 37 with moderate impairment, agreement was less close: of 13 CAM-positive delirium cases, DRS cutpoint 12 agreed on 7, but the more sensitive cut-point 10 produced agreement on 12 with DRS identifying one additional delirium case, which was CAM negative. Among 22 patients with mild cognitive impairment, three cases were agreed between CAM and DRS cut-point 10, but one of these was a non-case on DRS 12. Among 19 without cognitive impairment, no delirium cases were identified by either method. iii. DRS [17] for a dichotomous classification, based on cut-points of 12, as originally described [17] and 10, as subsequently suggested [20]. These assessments were administered in the same order by the same researcher (D.A.). Analyses Data were analysed using SPSS (version 11.5). The kappa statistic was used to investigate the chance-corrected agreement between CAM and DRS [27]. MMSE scores were used to classify patients with severe (MMSE score <10), moderate (10–20), mild (21–25), or no (26–30) cognitive impairment, as previously described [28]. Discussion Results CAM and DRS demonstrated good to very good agreement in the classification of delirium among older acutely ill hospital inpatients. Agreement was strongest with the more sensitive DRS cut-point 10. Thus, the shorter CAM rating was highly concordant in this patient population with (...truncated)