Diagnostic Discordance, Uncertainty, and Treatment Ambiguity in Community-Acquired Pneumonia : A National Cohort Study of 115 U.S. Veterans Affairs Hospitals

Abstract

Background: Evidence-based practice in community-acquired pneumonia often assumes an accurate initial diagnosis.

Objective: To examine the evolution of pneumonia diagnoses among patients hospitalized from the emergency department (ED).

Design: Retrospective nationwide cohort.

Setting: 118 U.S. Veterans Affairs medical centers.

Patients: Aged 18 years or older and hospitalized from the ED between 1 January 2015 and 31 January 2022.

Measurements: Discordances between initial pneumonia diagnosis, discharge diagnosis, and radiographic diagnosis identified by natural language processing of clinician text, diagnostic coding, and antimicrobial treatment. Expressions of uncertainty in clinical notes, patient illness severity, treatments, and outcomes were compared.

Results: Among 2 383 899 hospitalizations, 13.3% received an initial or discharge diagnosis and treatment of pneumonia: 9.1% received an initial diagnosis and 10.0% received a discharge diagnosis. Discordances between initial and discharge occurred in 57%. Among patients discharged with a pneumonia diagnosis and positive initial chest image, 33% lacked an initial diagnosis. Among patients diagnosed initially, 36% lacked a discharge diagnosis and 21% lacked positive initial chest imaging. Uncertainty was frequently expressed in clinical notes (58% in ED; 48% at discharge); 27% received diuretics, 36% received corticosteroids, and 10% received antibiotics, corticosteroids, and diuretics within 24 hours. Patients with discordant diagnoses had greater uncertainty and received more additional treatments, but only patients lacking an initial pneumonia diagnosis had higher 30-day mortality than concordant patients (14.4% [95% CI, 14.1% to 14.7%] vs. 10.6% [CI, 10.4% to 10.7%]). Patients with diagnostic discordance were more likely to present to high-complexity facilities with high ED patient load and inpatient census.

Limitation: Retrospective analysis; did not examine causal relationships.

Conclusion: More than half of all patients hospitalized and treated for pneumonia had discordant diagnoses from initial presentation to discharge. Treatments for other diagnoses and expressions of uncertainty were common. These findings highlight the need to recognize diagnostic uncertainty and treatment ambiguity in research and practice of pneumonia-related care.

Primary funding source: The Gordon and Betty Moore Foundation.