An algorithm-based approach to ascertain patients with rare diseases in electronic health records using hypereosinophilic syndrome as an example.

Purpose: Improved hypereosinophilic syndrome (HES) ascertainment in electronic health record (EHR) databases may improve disease understanding and management. An algorithm to ascertain and characterize this rare condition was therefore developed and validated.

Methods: Using the UK clinical practice research datalink (CPRD)-Aurum database linked to the hospital episode statistics database (Admitted Patient Care data) from Jan 2012 to June 2019, this cross-sectional study ascertained patients with a specific HES code (index). Patients with HES were matched (age, sex and index date) 1:29 with a non-HES cohort. An algorithm was developed by identifying pre-defined variables differing between cohorts; model-fitting using Firth logistic regression and statistical determination of the top-five performing models; and internal validation using Leave-One-Out Cross Validation. Final model sensitivity and specificity were determined at an 80% probability threshold.

Results: The HES and non-HES cohorts included 88 and 2552 patients, respectively; 270 models with four variables each (treatment used for HES, asthma code, white blood cell condition code, and blood eosinophil count [BEC] code) plus age and sex variables were tested. Of the top five models, the sensitivity model performed best (sensitivity, 69% [95% CI: 59%, 79%]; specificity, >99%). The strongest predictors of HES versus non-HES cases (odds >1000 times greater) were an ICD-10 code for white blood cell disorders and a BEC ≥1500 cells/μL in the 24 months pre-index.

Conclusions: Using a combination of medical codes, prescribed treatments data and laboratory results, the algorithm can help ascertain patients with HES from EHR databases; this approach may be useful for other rare diseases.