VALIDATION OF A TRANSTHYRETIN AMYLOID CARDIOMYOPATHY (ATTR-CM) UNITED STATES (US) CLAIMS-BASED CODING ALGORITHM
Author(s)
Isaiah Jimenez, BS1, Garcia Fahed, MD1, Nixuan Cai, BS, M-TRAM1, Margarita Udall, MPH2, Hiroki Kitakata, MD, PhD1, Kevin Alexander, MD1;
1Stanford Amyloid Center, Stanford University School of Medicine, Stanford, CA, USA, 2BridgeBio Pharma, Inc, San Francisco, CA, USA
1Stanford Amyloid Center, Stanford University School of Medicine, Stanford, CA, USA, 2BridgeBio Pharma, Inc, San Francisco, CA, USA
OBJECTIVES: ATTR-CM remains substantially underdiagnosed. Understanding disease characteristics and burden in real-world settings is critical. Assessment of this information within large US insurance claims databases has limitations, including nonspecific coding and use of many inconsistent, unvalidated coding algorithms. Thus, we have developed and validated a US claims-based coding algorithm to identify patients with ATTR-CM.
METHODS: First, we developed an algorithm for identifying patients with ATTR-CM within US claims databases based on previously published studies. We then applied it to a large, well-established claims/electronic health record-linked database, which includes ATTR-CM cases—the Stanford Research Repository (STARR). The algorithm’s inclusion criteria included adults ≥60 years of age at diagnosis with ≥2 claims with an ATTR-CM diagnosis code (ICD-10-CM: E85.[0,1,2,4,82]) on separate days and ≥2 cardiac-related diagnosis codes, or ≥1 tafamidis claim; exclusion criteria included multiple myeloma or light chain amyloidosis (E85.81) and HSCT (Z94.84). Algorithm performance and refinements were assessed by cross-referencing patients captured in STARR to those in Standford’s well-phenotyped ATTR patient registry as follows: true positives (TP) were in STARR and the registry, true negatives (TN) were in neither, false positives (FP) were in STARR but not the registry, and false negatives (FN) were in the registry but not STARR. Definitions used were accuracy (TP+TN/total sample), sensitivity (TP/TP+FN), and specificity (TN/TN+FP).
RESULTS: After iterative refinement to reduce FPs, the algorithm added 3 additional exclusion codes when paired with E85.4 (I68.0, N18.6, and L99). Accuracy was 99.5%, sensitivity was 80.7%, and specificity was 99.7% for detecting patients with ATTR-CM (TP, n=511; TN, n=61,215; FP, n=182; FN, n=122).
CONCLUSIONS: This is the first US claims-based ATTR-CM coding algorithm that has been validated and shown strong clinical performance with a sensitivity >80%. Application of this algorithm will improve identification and understanding of real-world ATTR-CM disease prevalence, patient characteristics, and patient outcomes.
METHODS: First, we developed an algorithm for identifying patients with ATTR-CM within US claims databases based on previously published studies. We then applied it to a large, well-established claims/electronic health record-linked database, which includes ATTR-CM cases—the Stanford Research Repository (STARR). The algorithm’s inclusion criteria included adults ≥60 years of age at diagnosis with ≥2 claims with an ATTR-CM diagnosis code (ICD-10-CM: E85.[0,1,2,4,82]) on separate days and ≥2 cardiac-related diagnosis codes, or ≥1 tafamidis claim; exclusion criteria included multiple myeloma or light chain amyloidosis (E85.81) and HSCT (Z94.84). Algorithm performance and refinements were assessed by cross-referencing patients captured in STARR to those in Standford’s well-phenotyped ATTR patient registry as follows: true positives (TP) were in STARR and the registry, true negatives (TN) were in neither, false positives (FP) were in STARR but not the registry, and false negatives (FN) were in the registry but not STARR. Definitions used were accuracy (TP+TN/total sample), sensitivity (TP/TP+FN), and specificity (TN/TN+FP).
RESULTS: After iterative refinement to reduce FPs, the algorithm added 3 additional exclusion codes when paired with E85.4 (I68.0, N18.6, and L99). Accuracy was 99.5%, sensitivity was 80.7%, and specificity was 99.7% for detecting patients with ATTR-CM (TP, n=511; TN, n=61,215; FP, n=182; FN, n=122).
CONCLUSIONS: This is the first US claims-based ATTR-CM coding algorithm that has been validated and shown strong clinical performance with a sensitivity >80%. Application of this algorithm will improve identification and understanding of real-world ATTR-CM disease prevalence, patient characteristics, and patient outcomes.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
RWD109
Topic
Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory)