Predicting heart failure onset in the general population using a novel data-mining artificial intelligence method.

Area of Science:

• Cardiology
• Biostatistics
• Predictive Analytics

Background:

• Heart failure (HF) poses a significant global health burden.
• Predicting HF onset accurately remains a clinical challenge.
• Existing predictive models may not capture complex multifactorial interactions.

Purpose of the Study:

• To identify novel combinations of clinical factors predicting heart failure (HF) onset.
• To evaluate the relationship between the number of predictive factor combinations and HF development probability.
• To introduce and apply a novel limitless-arity multiple-testing procedure (LAMP) for clinical factor analysis.

Main Methods:

• Utilized a cohort of 32,547 individuals without HF for initial factor combination identification using LAMP.
• Analyzed combinations of fewer than four clinical factors.
• Validated the predictive capability in a larger cohort of 275,658 individuals, assessing HF probability based on the count of identified predictive combinations.

Main Results:

• Identified 549 distinct combinations of clinical factors associated with HF onset.
• Demonstrated a progressive increase in HF probability with an increasing number of matching predictive combinations.
• Classified individuals into six groups based on the number of predictive combinations, showing a clear dose-response relationship.

Conclusions:

• Novel combinations of clinical factors can effectively predict heart failure onset.
• The number of identified predictive factor combinations is directly correlated with the probability of developing HF.
• LAMP offers a powerful tool for uncovering complex predictive patterns in clinical data, enhancing HF risk stratification.