Validation of a medical record-based delirium risk assessment.

Abstract

To improve identification of patients at high risk for delirium, this study developed a chart abstraction tool for delirium risk and validated the tool against clinical expert diagnosis of delirium. Prospective cohort study. Tertiary Veterans Affairs hospital in New England. One hundred veterans admitted to the medical service. While admitted, each participant underwent serial assessments for delirium by a clinical expert. Using the four criteria of a validated delirium prediction rule (cognitive impairment, sensory deficit, severe illness, and dehydration), chart review terms were selected for each criterion, and delirium risk was the sum of criteria present (range: 0-4; 4 = worst). After discharge, a nurse blinded to the expert's diagnosis completed the chart tool. The participants were mostly male (94%) and older (mean age 81 ± 7), and 23% developed overall delirium (14% incident). The rate of overall delirium was 11% in participants with zero risk factors, 18% in those with one or two, and 50% in those with three or four (P = .01; c-statistic 0.65, 95% confidence interval (CI) = 0.54-0.76). For incident delirium, the rates were 11%, 13%, and 25%, respectively (P = .53; c-statistic 0.56, 95% CI = 0.42-0.74). Discharge to a rehabilitation center or nursing home increased with increasing delirium risk (0%, 18%, 60%, P = .02). A chart abstraction tool was effective at identifying overall delirium risk but not incident delirium risk. Although the tool cannot replace clinical assessment and diagnosis of delirium, the use of this tool as an educational, clinical, or quality measurement aid warrants additional study.