Abstract

Background: International Classification of Diseases, 10th Revision (ICD-10) discharge diagnosis codes are commonly used to identify patients with pulmonary embolism (PE) for research, surveillance, or quality improvement. Ethno-racial disparities exist in the presentation, treatment strategies, and outcomes of PE, but it is unknown whether disparities extend to how PE is coded in electronic health records. Methods: The PE-EHR+ study was designed to validate ICD-10 codes for PE. We used data from Mass General Brigham (MGB). The reference standard was the manual chart review by two independent investigators using predefined criteria. Sensitivity and specificity for ICD-10 codes were determined by race and ethnicity. Results: 1,712 randomly sampled patients in equal distribution with PE principal discharge codes, PE secondary discharge codes, and no discharge codes for PE were selected from MGB. The race and ethnicity of the patients were non-Hispanic White (n=1,368), non-Hispanic Black (n=160), non-Hispanic Asian (n=50), non-Hispanic other or unknown (n=84), and Hispanic or Latino (n=50). The sensitivity and specificity of ICD-10 discharge diagnosis codes for PE were similar across race and ethnicity (Table). The results for the principal discharge diagnosis codes showed low sensitivity (range: 45%-66.3%) but high specificity (range: 90.9%-96.7%) by race and ethnicity. The principal or secondary discharge diagnosis codes resulted in high sensitivity (range: 99.7%-100%) but low specificity (range: 66.2%-83.3%) by race and ethnicity. The hybrid approach encompassing principal codes or secondary codes plus diagnostic imaging codes displayed high sensitivity (range: 97%-100%) and moderately high specificity (range: 78.1%-90%) among different racial and ethnic groups. Conclusion: While the study shows the tradeoffs of various combinations of ICD-10 codes for identifying PE, reassuringly, there was no large variation by race and ethnicity.