Validity of Diagnosis Codes to Identify a Dermatomyositis Cohort from Electronic Health Records

Dermatomyositis (DM) is an uncommon idiopathic inflammatory myopathy resulting in characteristic patterns of cutaneous lesions and myositis. Observational evidence related to the disease is limited by small case cohorts. We aimed to evaluate the validity of specialist-specific diagnostic coding for DM in an outpatient clinical database.

Adults having an outpatient encounter between January 1, 2010, and June 30, 2023, at a US regional health system with ICD-9/-10 coding for DM were identified. We randomly sampled 156 patients with ≥1 code applied by a dermatologist or rheumatologist. The primary reference standard for case adjudication was a confirmed diagnosis of DM by the treating physician in the medical chart. Fulfilment of the 2017 European Alliance of Associations for Rheumatology/American College of Rheumatology (EULAR/ACR) criteria for "probable" or "definite" DM was used as a secondary, more stringent reference standard. Positive predictive values (PPVs) for several case definitions were calculated with 95% confidence intervals.

Among eligible patients, the median age was 51.5 years, and 81% were female. Using the treating physician's diagnosis as reference, PPVs of ≥1 and ≥2 codes applied by a dermatologist were 93.2% (95% CI 82.0-98.3%) and 96.4% (82.2-99.8%), respectively. The PPVs of ≥1 and ≥2 codes from a rheumatologist were 82.0% (77.1-86.9%) and 85.8% (80.6-91.1%), respectively. At least one and at least two codes from a rheumatologist or dermatologist had PPVs of 82.1% (77.3-86.8%) and 85.7% (80.7-90.8%), respectively. The rate of confirmed cases based on EULAR/ACR criteria ranged from 44.9% to 57.1%.

All tested algorithms yielded an accurate case cohort with high PPV. Studies prioritizing sensitivity may use ≥1 code by dermatology or rheumatology to identify patients with DM.