Development and Validation of Algorithms for Systemic Sclerosis Identification in Electronic Health Record Data

Area of Science:

• Medical Informatics
• Rheumatology
• Epidemiology

Background:

• Systemic sclerosis (SSc) identification in electronic health records (EHR) is challenging.
• Accurate case ascertainment is vital for clinical and epidemiological research.
• Existing methods for SSc identification in EHR data require validation.

Purpose of the Study:

• To develop and validate International Classification of Diseases (ICD) code-based algorithms for identifying SSc cases in EHR data.
• To evaluate the performance of these algorithms using a gold standard.
• To identify the most effective algorithm for reliable SSc case detection.

Main Methods:

• Utilized a large multicenter EHR dataset (TriNetX Research Network) with ICD-9/ICD-10 codes for SSc.
• Performed detailed medical record review on a sample of patients to confirm SSc diagnosis (gold standard).
• Compared seven prespecified algorithms against the gold standard using sensitivity, specificity, PPV, NPV, and ROC analyses.

Main Results:

• The best-performing algorithm required at least two outpatient SSc ICD codes ≥30 days apart, excluding mimics within 24 months (sensitivity 96%, specificity 71%, PPV 90%, NPV 87%).
• Algorithms based on inpatient ICD codes showed high specificity and PPV but low sensitivity and NPV.
• Incorporating inpatient codes into outpatient algorithms did not enhance performance.

Conclusions:

• An algorithm using two outpatient SSc ICD codes ≥30 days apart, excluding mimics, reliably identifies SSc cases.
• This implementable approach surpasses previously published complex algorithms.
• The validated algorithm supports robust clinical and epidemiologic research using large EHR datasets.