Lead/Presenter: Barbara Jones, COIN - Salt Lake City
All Authors: Jones BE (VA SLC Health System IDEAS Center), Ying J (Division of Epidemiology, University of Utah), He T (Division of Epidemiology, University of Utah) Nevers M (Division of Epidemiology, University of Utah) Alba PR (Division of Epidemiology, University of Utah) Patterson OV(Division of Epidemiology, University of Utah) Jones MM (VA SLC Health System IDEAS Center, Division of Epidemiology, University of Utah) Bostwick AD (Division of Pulmonary & Critical Care Medicine, University of Utah) Rutter E (Department of Emergency Medicine, VA SLC Health System) Stevens V(VA SLC Health System IDEAS Center, Division of Epidemiology, University of Utah) Dean NC(Pulmonary & Critical Care Medicine, University of Utah & Intermountain Medical Center) Fine M (Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System) Gundlapalli AV (VA SLC Health System IDEAS Center Greene T (Division of Epidemiology, University of Utah) Samore MH (VA SLC Health System IDEAS Center, Division of Epidemiology, University of Utah)
To examine temporal trends and facility variation in diagnosis, illness severity, hospitalizations, and mortality among a national population of patients diagnosed with community-onset pneumonia across 118 VA Emergency Departments.
We identified a retrospective cohort of all Veterans > age 18 presenting to Emergency Departments (EDs) at 118 VA facilities between 2006-2016 with chest imaging. Diagnoses of pneumonia were identified by both ED diagnostic codes and physician diagnosis extracted by natural language processing from clinical documents. For each year, we compared the percent of ED encounters with pneumonia diagnosis, hospitalization, 30-day mortality, and severity of illness, defined as the probability of 30-day mortality estimated from a logistic regression model that included patient comorbidities, vital signs, and laboratory values. We examined facility-level variation in hospitalization risk.
Among 3.98M ED encounters with chest imaging, 297,334 (7%) were initially diagnosed with pneumonia; the percent with pneumonia diagnoses remained stable during the study period. From 2006 to 2016, the percent hospitalized decreased from 64% to 60%, and 30-day mortality decreased from 7.7% to 5.7%. Estimated mortality risk also decreased across years, from an average estimated risk of death of 7.3% in 2006 to 6.3% in 2016. However, after adjustment for mortality risk, 30-day mortality and hospitalization still significantly decreased over time (0.1% per year for death and 0.6% per year, p < 0.001 for both). We found substantial variation in hospitalization: the highest decile of facilities demonstrated 80% risk of hospitalization, compared to 40% risk among the lowest decile of facilities. This variation was unchanged after adjustment for severity of illness.
Hospitalizations and 30-day mortality have decreased among patients treated for pneumonia in VA Emergency Departments, even after adjustment for a decline in illness severity. We found substantial facility-level variation in hospitalization that was unexplained by differences in patient severity.
Community-onset pneumonia outcomes at VA emergency departments have gradually improved. Further research is needed to establish sources of this improvement, including improvement of care process versus unmeasured improvements in illness severity. Examining variation at the facility level could further elucidate mechanisms and identify optimal care pathways.