Abstract P321: Assembly and Validation of a Heart Failure Free Cohort from the Reasons for Geographic and Racial Differences in Stroke Study

Lauren Balkan,Matthew Mefford,Ligong Chen,Madeline R Sterling,Raegan W Durant,Monika M Safford,Emily B Levitan,Parag Goyal
DOI: https://doi.org/10.1161/circ.141.suppl_1.p321
IF: 37.8
2020-01-01
Circulation
Abstract:Introduction: Studies of incident heart failure (HF) have provided insights into key risk factors for the disease but have been limited to select populations that often lack geographic, racial, and gender diversity. There is a need to assemble a HF-free cohort using a contemporary, geographically diverse sample. Aim: To develop and validate a strategy for assembling a HF-free cohort from the REasons for Geographic And Racial Differences in Stroke (REGARDS) study. Methods: To assemble a HF-free cohort, we identified and excluded REGARDS participants who were taking HF-specific medications at baseline including: digoxin without atrial fibrillation, angiotensin converting enzyme inhibitor/angiotensin receptor blocker plus beta-blocker in the absence of hypertension, carvedilol, spironolactone, loop diuretic or a combination of hydralazine and nitrates. We then examined the subgroup of REGARDS participants with at least 6 months of Medicare claims at the time of the baseline assessment; we evaluated diagnostic performance (negative predictive value, positive predictive value, sensitivity and specificity) using three Medicare claims-based definitions of HF as the referent standard: Hospitalization for HF, Principal Diagnosis of HF, and Any Diagnosis of HF. Results: Among 28,884 eligible participants, 3,125 used HF-specific medications at baseline, leaving 25,759 (89%) participants in the proposed HF-free cohort. Participants in the HF-free cohort had a lower prevalence of coronary disease, atrial fibrillation, and diabetes compared to excluded participants. Depending on the Medicare definition used as the referent, the percent of the HF-free cohort without evidence of HF based on Medicare claims (the negative predictive value) ranged from 95.0-99.2% (Table 1). Negative predictive value was stable across age, sex, and race strata. Conclusions: This medication-based strategy to assembling a HF-free cohort in REGARDS can serve as a basis for future studies to examine incident HF in REGARDS and similar studies.
What problem does this paper attempt to address?