Objective: To determine if DSM-III criteria or clinical or discharge diagnoses, reviewed retrospectively, are as accurate an indicator of the presence of delirium as prospective evaluation by a psychiatrist.
Design: Selection of delirious patients prospectively by a psychiatrist, followed by retrospective record review of the same patients.
Setting: A referral-based university hospital.
Patients: From a sample of 235 consecutive medical patients over age 70, 47 delirious patients were identified prospectively by a research psychiatrist using DSM-III criteria. The medical record of these delirious patients was reviewed after discharge for evidence of delirium.
Results: Four patients were assigned ICD-9 codes suggestive of delirium (sensitivity 0.09). Review of physicians' diagnoses correctly identified 8 of 47 (sensitivity 0.17) patients as being delirious or acutely confused. The specific diagnostic criteria necessary to meet a DSM-III diagnosis of delirium could be ascertained from 10 of 47 records (sensitivity 0.21).
Conclusion: The retrospective medical record review is very imprecise in establishing the diagnosis of delirium. As research in this field moves from descriptive epidemiology to studies of pathogenesis and treatment, prospective designs will be needed.