Khan A, Heslin K, Simpson M, Malone M. Data in the electronic health record can be used at the bedside to identify older hospitalized patients with delirium. Poster presented at: Aurora Scientific Day; May 20, 2020; virtual webinar hosted in Milwaukee, WI.
Poster presented at: Aurora Scientific Day; May 20, 2020; virtual webinar hosted in Milwaukee, WI.
Background: Delirium is common among hospitalized older adults and associated with adverse outcomes. Delirium remains underrecognized, and efforts are focused on early recognition and prediction. While several delirium predictive rules have been developed, only a handful have focused on electronic health record (EHR) data. The coupling of prediction rules with features of the EHR are in their infancy but hold promise in their ability to aid in identification of delirium.
Purpose: To determine variables within our health system’s EHR that can be used to identify older hospitalized patients with delirium.
Methods: This is a prospective study among hospitalized patients (≥65 years of age) from February 2016 to November 2017. Patients were excluded if they: 1) were non-English-speaking, comatose, ventilated, or combative; 2) were intensive care or surgical patients; or 3) had severe aphasia, severe dementia, or a critical illness. Researchers screened daily for delirium using the 3-minute diagnostic confusion assessment method (3D-CAM). Predictive variables were extracted from the EHR. Basic descriptive statistics were conducted. Chi-squared and Fisher’s exact tests were used to compare differences among those diagnosed with or without delirium. Binary logistic regression was used for multivariable modeling.
Results: Among 408 participants, mean age at admission was 75 years, 61% were female, and 83% were African American. The overall rate of delirium was 16.7% (prevalent delirium: 10.5% [n=43]; incident delirium: 6.1% [n=25]). There was no statistical difference in 30-day mortality (2.9% vs 2.7%) or 30-day readmission rates (13.2% vs 14.7%) between those with and without delirium (P>0.05 for both). Even so, patients with delirium were older, more likely to have a diagnosis of infection and/or cognitive impairment, and more likely to have increased severity of illness (P<0.05 for all). Moreover, patients with delirium had a lower Braden (pressure ulcer risk) score and higher Morse fall score (P<0.01 for both). In multivariable analysis, cognitive impairment (odds ratio: 5.49; 95% CI: 2.77–10.87) and lower Braden scores (odds ratio: 1.29; 95% CI: 1.18–1.41) remained significant predictors of delirium among hospitalized patients.
Conclusion: Our study found that cognitive impairment and lower Braden scores were associated with hospital delirium. Further research is needed to develop an automated, dynamic (daily) prediction model inclusive of these variables.