Utility of the SENIORS elderly heart failure risk model applied to the RICA registry of acute heart failure

•The patients from clinical trials might not represent those from the “real world” with numerous co-morbidities.•Several HF risk models have been developed and validated in patients younger than 70 years with systolic dysfunction.•Our model is one of the few to focus on elderly HF patients including those with preserved ejection fraction.•This model provides a useful basis for clinical risk prediction for elderly patients with acute heart failure.

BackgroundHeart failure (HF) is predominantly a disease of the elderly. Reliable risk stratification would help in the management of this population, but no model has been well evaluated in elderly HF patients in both acute and chronic settings and not being restricted by ejection fraction. To evaluate the utility of the SENIORS risk model, developed from a clinical trial of elderly patients with chronic HF, in an independent cohort (National Spanish Registry: RICA) of elderly acute HF patients.MethodsWe applied the SENIORS risk model to 926 patients in RICA to estimate risk at one year of a) composite outcome of all-cause mortality or cardiovascular hospital admission and b) all-cause mortality.ResultsIn the RICA registry mean age was 78 years, mean ejection fraction 51% and 87% were in NYHA II and III. At one year death/CV hospitalization occurred in 31.9% and all-cause mortality in 19.5%. The risk model provided good separation of Kaplan Meier curves stratified by tertile for death/CV hospitalization and all-cause mortality. The observed versus expected rates of death/CV hospitalization in the lowest, middle and highest risk tertiles were (%) 34/24, 45/41 and 57/67, and for death 13/16, 32/38 and 44/70 respectively. C-statistic for all-cause mortality or CV hospitalization was 0.60 and for all-cause mortality 0.66.ConclusionThe SENIORS risk model was a reliable tool for relative risk stratification among acute heart failure patients in a “real world” registry, but predicted versus observed risk showed some variability. The model provides a useful basis for clinical risk prediction.