Objective The heterogeneity of lupus clinical manifestations makes it difficult to precisely characterize the full spectrum of patient profiles with traditional real-life (claims) data sources. This study assesses the ability of the Realli™ solution to characterize patients with lupus attending a reference center in France.

Methods Retrospective longitudinal study with data extracted via the Realli solution, an advanced machine learning-powered platform providing near real time in-depth extraction and analyses of full content structured and unstructured hospital electronic medical records (EMRs), as well as all connected data sources and claims. Upon Ethics Committee approval, eligible adult patients were identified via presence of ICD-10-CM L93.X or M32.X codes in claims and lupus type was characterized by text strings searches in the hospital EMRs.

Results Between January 2018 and March 2023, 187 adult patients with lupus (mean age 51.4 years; 88% females) were identified. Overall, 37.4%, 81.8%, 34.2%, 25.7% and 7% of patients presented with lupus erythematous, systemic lupus erythematosus, cutaneous lupus erythematosus (CLE), lupus nephritis (LN), and concomitant CLE with LN, respectively. Biomarkers, such as antinuclear antibodies, C-reactive protein, C4 and CH50 were retrieved in 68.4%, 54.0%, 46.0%, and 45.5% of patients, while SLEDAI disease activity index was reported in 9.6% patients, with an average score of 11.7 (95%CI: 6.36 -17.04). Most prevalent comorbidities and complications were cardiovascular disorders (62.6%), metabolic disorders (39.6%), polyarthritis (39.0%) and high blood pressure (38.5%). Renal impairment was identified in 34.8% of patients (all renal ICD-10 codes + text strings). Lastly, 31.0% and 15.0% of patients presented with antiphospholipid antibody syndrome or Gougerot-Sjögren syndrome, respectively. Patients received cyclophosphamide (4.8%), immunosuppressants (35.8%), glucocorticoids (51.3%), hydroxychloroquine (47.1%), belimumab (15.0%), and/or rituximab (8.6%). Over the study period, patients were hospitalized on average 10.8 times (median 4.0).

Conclusions A digital platform connected to multidata hospital sources is more time efficient in characterizing complex pathologies like lupus than manual chart reviews. This type of solution also provides a breadth and granularity of information not available in traditional RWE claims data sources. Learning from this single-center study will be deployed to multihospital system in France to test the robustness of the approach.