Study subjects were enrolled at a 400-bed, university-affiliated, primary acute care hospital in Montreal during the period February 1997 to January 1999. The study design was a prospective, observational study of 2 cohorts: the delirium cohort, in whom prevalent or incident delirium was detected during the first week of the hospital stay, and a cohort in whom delirium was not detected. Each cohort was further subdivided into those with and without dementia. The study was conducted at the same time as a randomized trial of the detection and treatment of delirium (unpublished data), and a subgroup of the delirium cohort also participated in that trial.
Patients 65 years of age and older who were admitted from the emergency department to the medical services were eligible for the studies. We excluded patients with a primary diagnosis of stroke, those admitted to the oncology unit, those admitted to the intensive care unit or cardiac monitoring unit unless they were transferred to a medical ward within 48 hours of admission, and those who did not speak English or French. At admission and during the first week of the hospital stay, a research nurse screened eligible patients for delirium using the Short Portable Mental Status Questionnaire (SPMSQ)11
and a review of the nursing notes. She also administered the Confusion Assessment Method (CAM)12
for those whose initial SPMSQ score was 3 or more, whose SPMSQ score at repeat screening increased by at least 1 point from the first assessment, or whose nursing notes indicated symptoms of delirium. Members of the cohort without delirium were selected from patients screened for delirium but free of this condition. To balance the distributions of age and prior cognitive impairment among patients with and without delirium, a stratified sampling method, based on the patient's age and initial SPMSQ score, was used. Thus, patients without delirium were selected from those 70 years of age and over, and only a subsample of patients with SPMSQ scores of less than 3 were included. Patients with an SPMSQ score of 4 or less gave informed consent to participate in the study; those with a score of 5 or more assented to participation and a relative provided written consent. The study was approved by the hospital's Research Ethics Committee.
A research assistant, blinded to study group, assessed the patients at enrolment and at 2, 6 and 12 months after admission; the research assistant also interviewed a family member. The 2-month follow-up took place in hospital for patients who had not been discharged or, for those who had been discharged, at home 8 weeks after discharge. The mean time from enrolment to the 2-month follow-up was 58 (standard deviation 20) days and was similar in the 4 study groups. In most cases the 6- and 12-month follow-ups were conducted at home.
Delirium was diagnosed by means of the CAM,12
a structured instrument validated against the clinical judgement of a psychiatrist, which assesses 10 symptoms of delirium specified in the revised 3rd edition of the Diagnostic and Statistical Manual of Mental Disorders
(DSM-III-R) (similar to the DSM-IV criteria) specifically acute onset, fluctuating course, inattention, disorganized thinking, altered level of consciousness, disorientation, memory impairment, perceptual disturbances, psychomotor activity and sleep–wake disturbance. We have previously reported that a nurse-administered CAM, validated against a consensus diagnosis, has sensitivity of 89% and specificity of 100%.13
Prevalent delirium was diagnosed if the patient met the criteria for delirium at admission and incident delirium if the criteria were met during the week after admission.
The presence of dementia was assessed on the basis of the 16-item Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE), an instrument with high internal consistency and test–retest reliability.14,15,16
In validation studies against clinical diagnosis, in which dementia was defined by a score of 3.38 or more, the sensitivity varied between 75% and 91% and the specificity between 65% and 82%.14,15
A French version tested in Quebec, in which dementia was defined by a score of 3.6 or more, had sensitivity of 75% and specificity of 95%.17
We used a cut-off value of 3.5 in defining dementia, midway between the cut-off values used previously.
Three measures of illness burden and severity were used. Comorbidity at admission was assessed by chart review with the Charlson Comorbidity Index (CCI), higher scores indicating greater comorbidity.18
Acute physiologic severity of illness was assessed with the Acute Physiology Score (APS), which is derived from the Acute Physiology and Chronic Health Evaluation (APACHE) II scale and was coded (on the basis of chart review) according to laboratory and clinical measures taken on or just before the date of enrolment; the APS score ranges from 0 (no impairment) to 44 (severe impairment).19
Clinical severity of illness was assessed by the research nurse at enrolment; the score ranges from 1 (minimal) to 9 (most severe).20,21
Finally, demographic variables (age, sex, marital status and living arrangement) and enrolment in the treatment arm of the randomized trial were recorded in study baseline forms.
The Mini-Mental State Examination (MMSE),22
a widely used instrument with established reliability and validity,23
was rated by the research assistant at enrolment and follow-up. The MMSE score ranges from 0 to 30, lower scores indicating greater cognitive impairment.
The Barthel Index (BI), which measures activities of daily living,24
was rated by the research assistant at enrolment and at follow-up, usually during a home visit. The BI score was based on observation, when possible; otherwise it was based on judgement. At 6 months, for example, the proportion of subjects whose BI scores were based on observation ranged from 1% for bathing to 87% for ambulation. We used the modified scoring suggested by Shah and associates;5 25
and inter-rater reliability (Pearson's r
) was 0.99.26
The Instrumental Activities of Daily Living (IADL) questionnaire from the Older American Resources and Services project,27
administered to an informant, was used to assess function; the scores for this scale range from 0 (completely dependent) to 16 (completely independent). At baseline, the IADL scale was rated for the premorbid period (i.e., before the current illness but not more than 1 month before hospital admission).8
Admission to a long-term care facility was determined by the research assistant during each follow-up contact and was defined as residence in a nursing home or long-term care hospital at 12 month follow-up or at the last follow-up before death or withdrawal from the study, among patients not resident in such a facility at enrolment.
We created a 4-category variable (delirium alone, dementia alone, both and neither) to study the individual and combined effects of delirium and dementia on the outcome variables. Pairwise comparisons of the 4 categories were used to make inferences about the effects of delirium and dementia on the various outcomes.
We analyzed changes over time in MMSE, BI and IADL scores using general linear models for longitudinal data with an autoregressive error structure.29
MMSE and BI scores were potentially available for all 4 time points, whereas IADL scores were potentially available only at enrolment and the 6- and 12-month follow-ups, because many patients were still hospitalized at 2 months. For each outcome we first fitted a model that included an interaction between subject group and time. If this interaction was statistically significant (i.e., there was a differential rate of change in the outcome in the 4 study groups), we fitted a separate linear regression model at each time point. Covariates in these analyses were age, sex, marital status, education, patient's living arrangement, comorbidity (measured by CCI), physiologic severity (measured by APS) and clinical severity. A treatment group covariate was added to the multivariate regression models to adjust for the possible effect of the intervention on any of the outcomes. This variable was not a statistically or clinically significant predictor of any of the outcomes and hence was not included in the final analyses. Our primary analyses used all available data; we also conducted secondary analyses with data for patients who had completed all potential assessments of each measure. In our primary analyses of the MMSE and BI scores, we did not control for the baseline level of each score, because these measures, determined in hospital, could be affected by the presence of delirium. We also did not control for premorbid IADL, as this measure is known to be affected by the presence of dementia, an independent variable in the study.23
However, we conducted secondary analyses in which we adjusted for the premorbid IADL score.
After excluding patients who had been in a long-term care facility at baseline and who died before hospital discharge, we used multiple logistic regression analysis to investigate the effect of the 4-category delirium–dementia variable on admission to long-term care, adjusting for the same set of covariates as for the analyses of the other outcomes.
Statistical tests were deemed significant if the 2-sided p value was less than 0.05.