Delirium in an adult acute hospital population: predictors, prevalence and detection

Background: To date, delirium prevalence and incidence in acute hospitals has been estimated from pooled findings of studies performed in distinct patient populations. Objective: To determine delirium prevalence across an acute care facility. Design: A point prevalence study. Setting: A large tertiary care, teaching hospital. Patients: 311 general hospital adult inpatients were assessed over a single day. Of those, 280 had full data collected within the study's time frame (90%). Measurements: Initial screening for inattention was performed using the spatial span forwards and months backwards tests by junior medical staff, followed by two independent formal delirium assessments: first the Confusion Assessment Method (CAM) by trained geriatric medicine consultants and registrars, and, subsequently, the Delirium Rating Scale-Revised-98 (DRS-R98) by experienced psychiatrists. The diagnosis of delirium was ultimately made using DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition) criteria. Results: Using DSM-IV criteria, 55 of 280 patients (19.6%) had delirium versus 17.6% using the CAM. Using the DRS-R98 total score for independent diagnosis, 20.7% had full delirium, and 8.6% had subsyndromal delirium. Prevalence was higher in older patients (4.7% if <50 years and 34.8% if >80 years) and particularly in those with prior dementia (OR=15.33, p<0.001), even when adjusted for potential confounders. Although 50.9% of delirious patients had pre-existing dementia, it was poorly documented in the medical notes. Delirium symptoms detected by medical notes, nurse interview and patient reports did not overlap much, with inattention noted by professional staff, and acute change and sleep-wake disturbance noted by patients. Conclusions: Our point prevalence study confirms that delirium occurs in about 1/5 of general hospital inpatients and particularly in those with prior cognitive impairment. Recognition strategies may need to be tailored to the symptoms most noticed by the detector (patient, nurse or primary physician) if formal assessments are not available.