JAMA Netw Open. 2026 Sep 1;9(9):e2635312. doi: 10.1001/jamanetworkopen.2026.35312.
ABSTRACT
IMPORTANCE: Overtesting with computed tomographic pulmonary angiography (CTPA) to diagnose pulmonary embolism (PE) is a quality improvement target in emergency departments (EDs).
OBJECTIVE: To describe how ED clinicians vary in their diagnostic practices for PE, specific to key decision moments, and how these practices are associated with CTPA diagnostic yield, the primary modality for definitively diagnosing PE, as well as potential missed opportunities to diagnose.
DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study included ED visits and clinicians at 29 Michigan EDs from January 1, 2023, to November 30, 2025. ED visits were stratified by clinician's diagnostic practices.
EXPOSURES: Three decision moments were evaluated: (1) the decision to test for PE (proportion of visits with either a D-dimer assay or CTPA performed); (2) the decision on initial testing modality (proportion of tested visits with CTPA performed without D-dimer assay first); and (3) the decision on interpreting the negative D-dimer result (proportion of negative age-adjusted visits that proceeded from D-dimer assay to CTPA).
MAIN OUTCOMES AND MEASURES: Diagnostic yield was defined as the percentage of CTPAs used to diagnose an acute PE. Potential missed opportunities to diagnose PE were defined using 10-day ED revisits and symptom-disease pair analysis.
RESULTS: The sample included 4 180 159 ED visits (median patient age, 52 [IQR, 34-69] years; 2 334 043 [55.8%] female) and 747 ED clinicians (463 [62.0%] male; median graduation year, 2013 [IQR, 2006-2017]). Diagnostic yield was associated with diagnostic practices at the decision to test for PE and the decision on initial testing modality. The decision to test for PE showed a pooled CTPA yield for the lowest-testing quintile of clinicians of 9.6% (95% CI, 9.0%-10.1%), while the corresponding yield for the highest-testing quintile was 6.4% (95% CI, 5.5%-7.3%; difference, -3.2 percentage points [95% CI, -4.2 to -2.1 percentage points]). A total of 189 potential missed opportunities to diagnose PE of 13 530 acute cases (1.4% of PEs) were identified, and these were associated with lower testing intensity.
CONCLUSIONS AND RELEVANCE: In this cohort study of 4 180 159 ED visits and 747 clinicians, variation in the first decision moment, the decision to test for PE, was associated with diagnostic yield. Future research should target standardizing this most upstream decision to best improve diagnostic quality.
PMID:42771345 | DOI:10.1001/jamanetworkopen.2026.35312