Diagnostic accuracy of the international classification of disease “I26” code to detect acute pulmonary embolism in a surveillance network

Abstract

Background

Emergency departments (EDs) offer a unique platform for a surveillance network for acute pulmonary embolism (PE) using International Classification of Disease (ICD-10) codes extracted from electronic medical records.

Objectives

Test the diagnostic accuracy of the I26 "leader" ICD-10 code for the detection of PE in near real-time in a large, ED-based surveillance network.

Methods

Standardized structured language queries were deployed at 91 hospitals to extract data, including ICD-10 codes, on a weekly basis from electronic medical records on ED patients with acute respiratory complaints. We used 2 methods for coding computed tomography pulmonary angiogram (CTPA) reports to derive a criterion or gold standard for PE diagnosis: (1) research associates were trained to interpret the CTPA reports, and (2) a validated Regular Expression computer program was used to interpret PE on CTPA reports. These 2 methods were independently adjudicated (PE⁺ or PE⁻). The primary outcome was diagnostic accuracy of the I26 leader compared with the final adjudication.

Results

From 6448 valid CTPA scan reports, 442 (6.8%) were adjudicated as PE⁺. On a weekly basis, the I26 leader had a sensitivity of 50.9% (95% CI, 46.1%-55.6%) and a specificity of 99.7% (95% CI, 99.5%-99.8%), likelihood ratio (LR) negative of 0.49 (95% CI, 0.44-0.54) and LR positive of 191 (95% CI, 116-12). At 1 month, the I26 sensitivity was 57.5% (95% CI, 52.7%-62.1%), and specificity was 99.5% (95% CI, 99.2%-99.6%); LRnegative of 0.43 (95% CI, 0.38-0.47) and LRpositive of 111 (95% CI, 77-159).

Conclusion

For low-latency surveillance of PE diagnosed in EDs, the ICD leader code I26 affords high specificity and high LR(+) for detection of acute PE in the United States but has modest sensitivity.