Risk prediction models for heart failure admissions in adults with congenital heart disease

Abstract

Background: Heart failure (HF) is the leading cause of death in adult patients with congenital heart disease (ACHD). No risk prediction model exists for HF hospitalization (HFH) for ACHD patients. We aimed to develop a clinically relevant one-year risk prediction system to identify ACHD patients at high risk for HFH.

Methods: Data source was the Quebec CHD Database. A retrospective cohort including all ACHD patients aged 18-64 (1995-2010) was constructed for assessing the cumulative risk of HFH adjusting for competing risk of death. To identify one-year predictors of incident HFH, multivariable logistic regressions were employed to a nested case-control sample of all ACHD patients aged 18-64 in 2009. The final model was used to create a risk score system based on adjusted odds ratios.

Results: The cohort included 29,991 ACHD patients followed for 648,457 person-years. The cumulative HFH risk by age 65 was 12.58%. The case-control sample comprised 26,420 subjects, of whom 189 had HFHs. Significant one-year predictors were age ≥ 50, male sex, CHD lesion severity, recent 12-month HFH history, pulmonary arterial hypertension, chronic kidney disease, coronary artery disease, systemic arterial hypertension, and diabetes mellitus. The created risk score ranged from 0 to 19. The corresponding HFH risk rose rapidly beyond a score of 8. The risk scoring system demonstrated excellent prediction performance.

Conclusions: One eighth of ACHD population experienced HFH before age 65. Age, sex, CHD lesion severity, recent 12-month HFH history, and comorbidities constructed a risk prediction model that successfully identified patients at high risk for HFH.

Keywords: Congenital heart disease; Heart failure; Prediction; Risk score; Risk stratification.