Predicting hospitalisation for heart failure and death in patients with, or at risk of, heart failure before first hospitalisation: a retrospective model development and external validation study.

BACKGROUND

Identifying people who are at risk of being admitted to hospital (hospitalised) for heart failure and death, and particularly those who have not previously been hospitalised for heart failure, is a priority. We aimed to develop and externally validate a prognostic model involving contemporary deep phenotyping that can be used to generate individual risk estimates of hospitalisation for heart failure or all-cause mortality in patients with, or at risk of, heart failure, but who have not previously been hospitalised for heart failure.

METHODS

Between June 1, 2016, and May 31, 2018, 3019 consecutive adult patients (aged ≥16 years) undergoing cardiac magnetic resonance (CMR) at Manchester University National Health Service Foundation Trust, Manchester, UK, were prospectively recruited into a model development cohort. Candidate predictor variables were selected according to clinical practice and literature review. Cox proportional hazards modelling was used to develop a prognostic model. The final model was validated in an external cohort of 1242 consecutive adult patients undergoing CMR at the University of Pittsburgh Medical Center Cardiovascular Magnetic Resonance Center, Pittsburgh, PA, USA, between June 1, 2010, and March 25, 2016. Exclusion criteria for both cohorts included previous hospitalisation for heart failure. Our study outcome was a composite of first hospitalisation for heart failure or all-cause mortality after CMR. Model performance was evaluated in both cohorts by discrimination (Harrell's C-index) and calibration (assessed graphically).

FINDINGS

Median follow-up durations were 1118 days (IQR 950-1324) for the development cohort and 2117 days (1685-2446) for the validation cohort. The composite outcome occurred in 225 (7·5%) of 3019 patients in the development cohort and in 219 (17·6%) of 1242 patients in the validation cohort. The final, externally validated, parsimonious, multivariable model comprised the predictors: age, diabetes, chronic obstructive pulmonary disease, N-terminal pro-B-type natriuretic peptide, and the CMR variables, global longitudinal strain, myocardial infarction, and myocardial extracellular volume. The median optimism-adjusted C-index for the externally validated model across 20 imputed model development datasets was 0·805 (95% CI 0·793-0·829) in the development cohort and 0·793 (0·766-0·820) in the external validation cohort. Model calibration was excellent across the full risk profile. A risk calculator that provides an estimated risk of hospitalisation for heart failure or all-cause mortality at 3 years after CMR for individual patients was generated.

INTERPRETATION

We developed and externally validated a risk prediction model that provides accurate, individualised estimates of the risk of hospitalisation for heart failure and all-cause mortality in patients with, or at risk of, heart failure, before first hospitalisation. It could be used to direct intensified therapy and closer follow-up to those at increased risk.

FUNDING

The UK National Institute for Health Research, Guerbet Laboratories, and Roche Diagnostics International.

Informations de copyright

Copyright © 2022 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Published by Elsevier Ltd.. All rights reserved.

Déclaration de conflit d'intérêts

Declaration of interests EBS serves as an adviser for HAYA Therapeutics and consults for PureTech Health. PFB was in receipt of a Joint Alliance Medical and University Hospital of South Manchester Fellowship Salary Support Grant. JHN has a part-time appointment at Bioxydyn. TM serves as the clinical lead for the National Heart Failure Audit and has received speaker fees from Novartis, AstraZeneca, and Vifor. CAM has served on advisory boards for Novartis, Boehringer Ingelheim and Lilly Alliance, and AstraZeneca; serves as an adviser for HAYA Therapeutics and PureTech Health; and has received research support from Amicus Therapeutics, Guerbet Laboratories, Roche, and Univar Solutions (none are relevant to the contents of this Article, except where described in the Role of the funding source). All other authors declare no competing interests.