Development and Validation of a parsimonious AI-Based Risk Score for Mortality in Heart Failure: A UK cohort study

Background: Accurate risk stratification in heart failure (HF) is crucial to guide clinical decisions and optimise therapeutic strategies. Existing risk scores such as the Meta Analysis Global Group in Chronic Heart Failure (MAGGIC) show modest discrimination and rely on specialised tests (e.g., echocardiography), while advanced Artificial Intelligence (AI) models face significant barriers to adoption due to their complexity and data access issues. Methods: Using a cohort of 373,389 patients aged 18 years and older with HF from the Clinical Practice Research Datalink (CPRD) Aurum dataset, we developed and validated an AI-based risk prediction model for all-cause mortality. A multi-layer perceptron survival model using variables from the MAGGIC score was enhanced by incorporating predictive features from an advanced AI model trained on longitudinal electronic health records (EHR) and distilled into a final optimised 11-variable model (named "SIMPLE-HF" - Simplified Intelligent Mortality Prediction for Longitudinal EHRs in HF patients). The model utilised readily available variables at point-of-care, including BMI, year of birth (as a proxy for birth cohort effects), and key comorbidities such as cancers. Discriminatory performance and clinical utility of SIMPLE-HF was compared against a MAGGIC score adapted for use on routine EHR (MAGGIC-EHR). In secondary analyses, the performance of SIMPLE-HF was evaluated for major adverse cardiovascular events and hospitalisation. Findings: The SIMPLE-HF model demonstrated significantly improved discriminatory performance (C-index: 0.801, 95% CI 0.795, 0.806) compared to the benchmark MAGGIC-EHR Cox model (0.735, 0.728, 0.741), while maintaining acceptable calibration. Clinical impact analysis further revealed that SIMPLE-HF consistently captures more true events while flagging fewer patients as high-risk. For every 1000 patients screened, at a high-risk threshold of 0.60, SIMPLE-HF identifies 199 patients who will have an event, whereas MAGGIC-EHR captures only 99 true events. Similarly, SIMPLE-HF delivered improved performance on cardiovascular events and hospitalisation prognostication as compared to the benchmark model. Interpretation: Leveraging insights from a complex EHR-trained AI model into easily collected clinical features, we enhance the accuracy and practicality of HF mortality risk prediction. This novel approach offers a pathway to clinically implementable tools that balance predictive accuracy with pragmatic utility and has the potential to improve HF risk stratification, along with offering significant cost efficiency by removing the need of specialised tests.