Abstract 9844: Validation of the Seattle Heart Failure Model in the Swedish Heart Failure Registry

• Published in: Circulation (New York, N.Y.) Vol. 140; no. Suppl_1 Suppl 1; p. A9844
• Main Authors: Yang, Sushan, Lund, Lars H, Li, Song, Dahlstrom, Ulf, Sartipy, Ulrik, Levy, Wayne
• Format: Journal Article
• Language: English
• Published: by the American College of Cardiology Foundation and the American Heart Association, Inc 19-11-2019
• ISSN: 0009-7322; 1524-4539
• DOI: 10.1161/circ.140.suppl_1.9844

IntroductionHeart failure risk models provide important prognostic information to guide care. Many risk models, including the Seattle Heart Failure Model (SHFM-D), were developed and validated predominantly in patients with reduced ejection fraction (HFrEF). The validity of these risk models in mid-range (HFmrEF) and preserved EF (HFpEF) has not been assessed.HypothesisThe SHFM accurately estimates mortality in patients in each EF category.MethodsThe SHFM was applied to 35,140 patients (40% enrolled in the hospital) from the Swedish Heart Failure Registry (2000-2015). Ejection fraction was categorized as HFrEF (≤39%), HFmrEF (40-49%), and HFpEF (≥50%). Model discrimination was assessed by one and five-year area under the receiver operating characteristic (AUROC) curve. Calibration compared predicted versus observed survival at one and five years.ResultsThe HFrEF, HFmrEF, and HFpEF included 20,884, 7,590, and 6,666 patients, respectively. At one year, 13.8% of patients died (13.5%, 12.5%, and 16.4% respectively). The SHFM demonstrated excellent risk discrimination when applied to the entire study population (one and five-year AUROC 0.791 and 0.796). Risk discrimination was also accurate among each EF group, with five-year AUROC of 0.796, 0.802, and 0.785, respectively. Calibration analysis showed that the observed survival was lower than the predicted survival at one year (86.2% vs. 91.8%) and five years (55.4% vs. 65.6%). Clinic-enrolled patients showed appropriate calibration with similar observed and predicted survival at one year (92.3% vs. 93.5%) and five years (65.7% vs. 70.7%). Hospital-enrolled patients had lower survival than predicted; observed versus predicted survival at one year (77.3% vs. 89.4%) and five years (41.9% vs. 58.7%). In a Cox proportional hazard model, enrollment in the hospital had a hazard ratio (HR) of 1.48 ± 0.02, similar to 1 NYHA class higher risk (HR 1.41).ConclusionsThe SHFM has excellent discrimination in a real-world registry in all categories of EF. The SHFM accurately predicts one and five-year mortality in clinic-enrolled heart failure patients with reduced, mid-range, and preserved EF and underestimates mortality in hospital-enrolled patients.