A Validated Transthyretin Amyloid Cardiomyopathy Claims-Based Coding Algorithm: ValidATTR-US

Lack of a validated United States (US) claims-based method for identifying transthyretin amyloid cardiomyopathy (ATTR-CM) limits the reliability and comparability of real-world evidence. We developed and validated a US claims-based coding algorithm (ValidATTR-US) to identify patients with ATTR-CM.

ValidATTR-US was developed by optimizing previously published algorithms using International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes and refining them using the STAnford Research Repository (STARR, claims/electronic health record-linked database; N = 61,908) to define the index cohort (n = 694). The Stanford ATTR-CM Patient Registry (N = 634) served as the validation cohort for cross-referencing the index cohort. Index cohort inclusion criteria were age ≥ 60 years at diagnosis and ≥ 2 ATTR-CM diagnosis codes (ICD-10-CM, E85.[0, 1, 2, 4, 82]) on separate days and ≥ 2 cardiac-related codes or ≥ 1 tafamidis claim. Exclusion criteria were multiple myeloma (C90), light chain amyloidosis (E85.81), hematopoietic stem cell transplantation (Z94.84), and organ-limited amyloidosis (E85.4) when co-occurring with cerebral amyloid angiopathy (I68.0), end-stage renal disease (N18.6), or skin disorders (L99). Performance metrics included accuracy ({true positives [TPs] + true negatives [TNs]}/total sample), specificity (TNs/{TNs + false positives [FPs]}), and sensitivity (TPs/{TPs + false negatives [FNs]}).

Accuracy was 99.51% (95% confidence interval [CI] 99.45-99.56%), specificity 99.70% (99.65-99.74%), and sensitivity 80.76% (77.47-83.76%) for detecting patients with ATTR-CM (TP, n = 512; TN, n = 61,092; FP, n = 182; FN, n = 122). Individual ICD-10-CM codes were strongly predictive of ATTR-CM case status (area under the curve [AUC] 0.84) and genotype classification (AUC 0.88).

ValidATTR-US is the first US claims-based ATTR-CM coding algorithm validated against a clinically confirmed ATTR-CM registry. Broader application may improve identification of patients with ATTR-CM in real-world datasets and help characterize prevalence and treatment patterns. Graphical abstract available for this article.