Developing Validated Tools to Identify Pulmonary Embolism in Electronic Databases: Rationale and Design of the...

Area of Science:

• Medical Informatics
• Cardiovascular Research
• Health Data Science

Background:

• Electronic health records (EHRs) and International Classification of Diseases (ICD) codes are crucial for pulmonary embolism (PE) research.
• Natural language processing (NLP) offers automated chart review but its validity for PE patient identification requires validation.
• Uncertainty exists regarding the accuracy of ICD-10 codes and NLP algorithms in identifying PE patients within EHRs.

Purpose of the Study:

• To validate ICD-10 codes (principal and secondary discharge diagnoses) for identifying PE patients in EHRs.
• To validate NLP tools for identifying PE patients in EHRs, including from radiology reports.
• To establish a reliable reference standard for PE patient identification in EHRs.

Main Methods:

• The PE-EHR+ study employs manual chart review by two independent abstractors as the reference standard.
• Sensitivity, specificity, and predictive values will be calculated to assess code and NLP algorithm accuracy.
• Discriminatory function of code subgroups for intermediate- and high-risk PE will be evaluated.

Main Results:

• 1,734 patients from Mass General Brigham were identified with PE-related ICD-10 codes (principal, secondary) or without codes.
• Random sampling was used to select patients within each coding group.
• Data validation and analysis are pending; a subset of patients from Yale-New Haven Health System will also be included.

Conclusions:

• The PE-EHR+ study aims to confirm the reliability of tools for identifying PE patients in EHRs.
• Validated tools will enhance the accuracy of observational studies and clinical trials involving PE patients.
• Improved patient identification in electronic databases will bolster research on pulmonary embolism.