Can Variables From the Electronic Health Record Identify Delirium at Bedside?

Delirium, a common and serious disorder in older hospitalized patients, remains underrecognized. While several delirium predictive models have been developed, only a handful have focused on electronic health record (EHR) data. This prospective cohort study of older inpatients (≥ 65 years old) aimed to determine if variables within our health system’s EHR could be used to identify delirium among hospitalized patients at the bedside. Trained researchers screened daily for delirium using the 3-minute diagnostic Confusion Assessment Method (3D-CAM). Patient demographic and clinical variables were extracted from the EHR. Among 408 participants, mean age was 75 years, 60.8% were female, and 82.6% were Black. Overall rate of delirium was 16.7%. Patients with delirium were older and more likely to have an infection diagnosis, prior dementia, higher Charlson comorbidity severity of illness score, lower Braden Scale score, and higher Morse Fall Scale score in the EHR (P \u3c 0.01 for all). On multivariable analysis, a prior diagnosis of dementia (odds ratio: 5.0, 95% CI: 2.5–10.3) and a Braden score of \u3c 18 (odds ratio: 2.8, 95% CI: 1.5–5.1) remained significantly associated with delirium among hospitalized patients. Further research in the development of an automated delirium prediction model is needed