Diagnostic accuracy of an algorithm identifying US veterans with Inclusion Body Myositis from the Corporate Data Warehouse

Objectives Inclusion body myositis (IBM) is the most common idiopathic inflammatory myopathy (IIM) in adults over 50 years of age and has only one International Classification of Diseases (ICD) code across ICD-9 and ICD-10. Despite this, little is known about the performance of administrative data in identifying individuals with IBM. An algorithm based on billing and administrative data to create a veteran IBM cohort was tested and its performance compared to clinical diagnosis and consensus diagnostic criteria. Methods Of 732 previously identified veterans with IBM, 107 were randomly selected for manual record review using a keyword search for “Inclusion Body Myositis” or “IBM” via the tvf_TIU_FullTextSearch function. Rheumatology and Neurology notes from January 1, 2011 to June 1, 2024 were extracted and data tokenized using R packages tidytext and tokenizers . Incomplete records were supplemented with Voogle data. Results 83 (77.5%) of 107 veterans had definite and 7 (6.5%) suspected clinical IBM. There were 4 cases with HIV-associated myopathy and 5 with inherited myopathies. Only 7 individuals were concurrently identified among a separate cohort of 1,136 veterans with a diagnosis of polymyositis based on administrative data. Detailed muscle biopsy pathology reports were available for 40/107 veterans of whom 30 (75%) noted findings consistent with IBM and 30 (75%) met ENMC 2024 diagnostic criteria for IBM. Based on ENMC 2011 criteria, 2 (5%), 12 (30%), and 25 (62.5%) veterans met clinicopathologic, clinical, and probable IBM definitions, respectively. Using Griggs criteria, 2 (5%) and 15 (37.5%) individuals met definite and probable IBM thresholds. The minimum positive predictive value (PPV) was 74.4% for an IBM clinical diagnoses and 88.0% for cases meeting diagnostic criteria, with specificity ≥ 90% for most groups. Conclusions The algorithm demonstrates robust performance and excellent specificity in identifying Veterans with IBM, comparable to approaches for systemic lupus and other IIMs.