OBJECTIVES: Before 2022, there was no specific International Classification of Diseases (ICD) code for atypical hemolytic uremic syndrome (aHUS), making it difficult to use claims data to study aHUS outcomes. This study aimed to develop and validate an algorithm to identify patients with aHUS using an insurance claims database.

METHODS: We used linked electronic health record and claims data (Optum’s de-identified Market Clarity Data; from Jan-2016 to Jan-2022) of patients with chronic kidney disease according to the following inclusion criteria: ≥2 hemolytic uremic syndrome diagnosis codes (ICD-10, D59.3) or ≥2 thrombotic microangiopathy diagnosis codes (ICD-10, M31.1), data available from ≥2 physician notes and no evidence of other complement-mediated diseases. Ground-truth aHUS was established using tokenized physician notes from the database. Previously, a rule-based algorithm was established. However, due to a lack of laboratory results to enable differential diagnosis, the algorithm did not provide a clear differential among TMA types. Three machine learning (ML) algorithms (elastic net, classification tree, and random forest) were developed using clinical features, including renal, pulmonary, cardiovascular, hematology and others, extracted from patient medical records. Overall, 70% of the data was used to train the models and 30% was reserved for testing.

RESULTS: Of the 1,992 patients included, 304 (15%) had ground-truth aHUS. In the test set, the elastic net ML algorithm had the highest positive predictive value (85%) for identifying aHUS with 73% sensitivity and 98% specificity. The most important features that the ML algorithm used to identify patients with aHUS were hospitalization, renal disease progression, cardiovascular/pulmonary complications, and the absence of bacterial infections.

CONCLUSIONS:

This ML algorithm approach may address the limitations of rule-based algorithms and has the potential to permit accurate identification of more patients with aHUS in claims data and provide an improved understanding of the aHUS treatment landscape.