This study was carried out at St. Mary's Hospital Center, Montreal, a community-based, university-affiliated hospital. All patients aged 65 years and older who were screened for eligibility had presented consecutively at the ED from Monday to Friday, between midnight and 3 pm, during a 4-month period and had been triaged by a nurse according to Wilson's system12
to the observation room on a stretcher. Wilson's triage system is based on 4 levels of severity: immediate (attention within minutes), urgent (attention within 30 minutes), semi-urgent (attention within 2 hours) and nonurgent (attention in more than 2 hours). The particular time of presentation was chosen because most patients presenting after midnight were held overnight in the ED and could be assessed in the morning. The research staff worked from 8 am to 6 pm to permit the evaluation of patients presenting until 3 pm. All patients able to communicate in French or English and for whom consent could be obtained were enrolled. Critically ill patients (unconscious, with unstable cardiorespiratory status or severe trauma) were excluded. This study was approved by the hospital research ethics committee.
First, eligible patients were approached and the Short Portable Mental Status Questionnaire (SPMSQ) was administered. The SPMSQ is a 10-item questionnaire that evaluates orientation, memory and concentration with good sensitivity and specificity for detecting cognitive impairment.13
It was used for screening patients' capacity to give informed consent. Patients with 4 or fewer errors in the SPMSQ were directly asked for consent. For patients with more than 4 errors, consent was obtained from family or from a caregiver. Demographic data were then gathered (age, sex, marital status, residence and level of education). Second, the Mini-Mental State Examination (MMSE) and the CAM were used by either of 2 research psychiatrists (M.E. and F.R.) to detect and diagnose delirium, blind to the emergency physician's assessment. Third, attending ED physicians, blind to the investigators' evaluations, were asked to complete a mental status checklist (Appendix 1
) and a clinical severity of illness scale14
for all of the delirious patients identified by the research investigators, and for a one-third random sample of the nondelirious patients also identified by the research investigators. The ED physicians were not aware of whether they were being asked to fill out these forms for delirious or nondelirious patients. In other respects, they completed their evaluation in the usual manner. Finally, the emergency medical records of patients with delirium and of nondelirious patients were systematically reviewed in the weeks following the initial evaluation. The record review was carried out by one of the researchers blind to the results of the screening. The reviewer looked at the diagnoses of the ED physician for the term delirium or an acceptable synonym (acute or new confusional state, acute mental status change, toxic psychosis, metabolic encephalopathy or acute organic brain syndrome). The discharge disposition from the ED, that is, whether patients were discharged home or admitted to hospital, the principal diagnosis and the number of medications were also recorded to monitor ED physicians' management of these patients.
As stated earlier, the two instruments used to detect and diagnose delirium were the MMSE15
and the CAM.16
The MMSE is a well-known practical instrument composed of 11 items that can be used to grade cognitive status and, in this study, complete the CAM. The test-retest reliability is reported to be 0.89.15
The CAM is a semi-structured instrument that can distinguish between delirium and other kinds of cognitive impairment. It has been validated in previous studies and has been shown to have a sensitivity and specificity for detecting delirium of more than 90% compared with the diagnosis of trained psychiatrists.16
The CAM was used as the "gold standard" for the detection of delirium. Using the same approach as Lewis and colleagues,11
patients with 5 of 5 DSM-IV1
criteria for delirium (disturbance of consciousness, reduced ability to focus or shift attention, a change in cognition or perceptual disturbances not due to dementia, the disturbance develops over a short period and tends to fluctuate during the day) were diagnosed with definite delirium, and patients who had 4 of 5 criteria were diagnosed with probable delirium. Patients diagnosed with probable delirium were also considered to have delirium in order to avoid overlooking patients with a partial, but clinically significant, syndrome. Interrater reliability, monitored before and during the study, showed an agreement of over 90% for both the MMSE and the CAM.
The clinical severity of illness scale14
and a mental status checklist (Appendix 1
) were filled out by the attending ED physician. The severity scale, with a score from 1 (not ill) to 9 (moribund), uses clinical judgement to classify a patient's severity of illness. Because the rapid evaluation required in the ED setting may lead some physicians not to record observed mental disorders, the ED physicians were asked to complete a checklist indicating any mental disorders observed.
The association of delirium with the demographic variables was assessed by the t-test for continuous variables and by the χ2 test for categorical variables. The discharge dispositions of patients with detected cases of delirium versus those with undetected delirium were compared by means of χ2 analysis. The severity of illness and the number of medications prescribed were compared for patients with and without delirium using the t-test.
The sensitivity and specificity of the conventional clinical assessment by the ED physician in detecting delirium (according to checklist, chart review and both combined) were calculated using the result from the CAM as the gold standard.