Validity of a claim-based algorithm for classifying disease activity levels using Systemic Lupus Erythematosus...

To validate algorithms for classifying disease activity in patients with systemic lupus erythematosus (SLE) using Korean claims data.

We used data from a prospective cohort of SLE patients enrolled at a single academic center between October 2014 and August 2020. Disease activity was assessed at each visit using the Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2K), with the annual average score serving as the gold standard. Three claims-based algorithms incorporating diagnostic codes for comorbidities and medication use were evaluated (1) a previously established model including SLE-related comorbidities, immunosuppressants, and oral glucocorticoids, (2) a modified version incorporating intravenous glucocorticoids, and (3) a version with adjusted glucocorticoid dosage criteria (5 mg) to better identify mild to moderate disease. The performance of each algorithm in classifying mild disease activity-defined as SLEDAI-2K <3-was assessed by calculating sensitivity, specificity, positive predictive value (PPV), negative predictive value, and the area under the curve.

A total of 151 patients were included. The mean age was 34.5±8.7 years, and 94.7% were female. The mean initial SLEDAI-2K score was 3.6±2.6. The PPV for identifying mild disease activity ranged from 75.9% to 77.2%, with Algorithm 3 demonstrating the highest PPV. However, incorporating intravenous glucocorticoids or adjusting dosage thresholds did not result in further improvement.

A claims-based algorithm using diagnostic and medication codes demonstrated a PPV of 77.2% for classifying mild disease activity in SLE. This approach may offer a practical method for disease activity assessment in Korean claims-based research.