Abstract 15791: Evaluation of Medicare Claims for Heart Failure Diagnostic Development: A Multisense-linkage Feasibility Analysis

Abstract

Listen

Translate article

Introduction: The validity of utilizing claims data for the development of worsening heart failure (HF) diagnostics has not been previously reported. Therefore, we linked data from the MultiSENSE study, in which the HeartLogic diagnostic was first reported, with claims data in order to validate claims HF events and compare diagnostic performance. Methods: Identifiers from 791 United States study patients were directly linked with Medicare claims to produce 320 patients with continuous Medicare Part A/B fee-for-service (FFS) coverage during study participation. Claims HF events were defined using primary HF diagnosis codes (DRG and ICD-9-CM) and included inpatient events as well as outpatient events with intravenous decongestive therapy. We matched events by patient and date, calculated agreement between events (Cohen’s kappa coefficient κ), and evaluated HeartLogic diagnostic performance using sensitivity and false positive rate (FPR). Results: The linked cohort was older than the remaining patients and had greater disease burden. Study deaths (N=14) matched exactly with claims. In contrast, matching of 207 study hospitalizations with 197 acute inpatient events resulted in a total of 239 events of which 193 matched with claims (81%), 32 (13%) were claims-unique, and 14 (6%) were study-unique. Claims HF events had substantial agreement with study adjudicated hospitalizations (observed = 93.3%, κ = 0.823). The HeartLogic algorithm detected claims-derived events with a sensitivity of 75.6% and an FPR of 1.539 alerts/pt-year, which was not different from performance derived using study events (sensitivity = 77.6% and FPR = 1.528, P = 0.82 and 0.92). HeartLogic detected events contributed to 91% of HF event costs used for performance evaluation ($605,000 out of $663,000). Conclusions: The agreement between claims and study events, and equivalence of HeartLogic diagnostic performance suggest that claims data may have utility for future diagnostic development/enhancement. Additional data are needed to demonstrate safety, efficacy, and cost-effectiveness of HeartLogic-directed interventions.