Automated palliative care trigger for admitted emergency department patients
Kunjummen E, Jayaprakash N, Zimny E, Miller J, and Holland C. Automated palliative care trigger for admitted emergency department patients. Acad Emerg Med 2019; 26:S54.
Acad Emerg Med
Background: Identifying ED patients at high-risk for mortality may allow clinicians to best direct early palliative care consultation or initiate discussion on goals of care. This study's objective was to develop an automated electronic medical record (EMR) trigger in the ED that would identify patients admitted to the hospital who are at high-risk for 30-day mortality and may benefit from early palliative care consultation. Methods: We established a prospective data collection tool within the EMR to gather data on comorbidities, demographics, and physiologic abnormalities. We included all consecutive adult ED patients requiring hospitalization at a single urban center over 6 months. Patients enrolled in hospice, psychiatric or labor and delivery admissions, and those transferred to another hospital were excluded. The primary outcome was death within 30-days. A palliative trigger score with multivariate logistic regression was derived with a sensitivity analysis excluding patients placed in observation. Model comparison used the area under the receiver operating curve (AUC). Results: A total of 10,019 consecutive patient hospitalizations were included, of whom 5,165 (51.6%) were female, 6,838 (68.3%) African American, and 1,739 (12.4%) admitted to an intensive care unit (ICU). The mean age was 58.4 years. Thirty and 60-day mortality was 3.8% and 5.7% respectively. The derived model was accurate in predicting 30-day mortality (AUC 0.89, 95% CI 0.88 - 0.91). Sensitivity analysis excluding observation patients resulted in similar results (AUC 0.87, 95% CI 0.84 - 0.89). Using this model, we constructed a palliative trigger score with a range from 0 - 10 that was based on 22 covariates (median value 3.4, IQR 2.5 - 4.3). Patients (n=888) with a score > 5.1 (90th percentile) had 23.4% mortality at 30-days compared to 1.9% among those (n=9,131) with a score ≤ 5.1 (OR 15.6, 95% CI 12.5 - 19.3). Similarly, those with a score > 5.1 had a 31.4% mortality at 60-days vs. 3.2% (OR 14.1, 95% CI 11.7 - 16.9). Conclusion: With further validation, an automated, palliative care trigger in the EMR has the potential to accurately identify patients at high risk of mortality who may benefit from early palliative care consultation. Such a tool could enhance goals of care discussions early in clinical management.