Presentation (CTI)

OBJECTIVES: No validated insurance claims-based algorithms specifically identify DM1. A simple rule-based algorithm based on ICD-10-CM diagnosis code was adopted and its performance validated using electronic health records (EHR).
METHODS: We used Optum^® Market Clarity Clinical and Claims database. The algorithm requires ≥1 inpatient or ≥2 outpatient claims records (≥30 and ≤365 days apart) with ICD-10-CM diagnosis code, G71.11. Index date was defined as the earliest claim for G71.11 identified between 01 July 2016 to 31 March 2023. The required length of claims enrollment was ≥6 months prior and ≥12 months after index date. EHR validation population included a sub-sample of individuals with relevant keywords in ≥1 provider note in their claims-linked EHR. Provider notes with explicit records of DM1 phenotype or positive DM1 genetic test results were considered as gold standard for either true DM1 or non-DM1. Positive predictive value (PPV) and sensitivity of the algorithm were estimated in the EHR validation sample. The algorithm was refined by adding claims factors that may improve the PPV based on their prevalence and standardized differences that may distinguish between true DM1 and non-DM1.
RESULTS: Of 250 individuals (173 true DM1 and 77 non-DM1) in the EHR validation sample, 224 were identified as having DM1 by the claims algorithm and 165 were confirmed as true DM1. The PPV was 0.74 (165/224), 95% CI 0.68-0.79 and sensitivity was 0.95 (165/173), 95% CI 0.92-0.99. A refined algorithm was built on the simple algorithm by further excluding those with DM2 genetic test, stress test, and aldolase test procedures. This numerically improved PPV to 0.78, 95% CI 0.72-0.84 but lowered sensitivity to 0.86, 95% CI 0.81-0.91.
CONCLUSIONS: The DM1 claims algorithm has a good PPV and can be applied consistently in US claims databases to generate insights about disease burden and costs in DM1.