E1.L1

DATASET DESCRIPTION

 Disease Area

Cardiovascular Medicine — Heart Failure

 Specific Research Focus

Early-stage heart failure risk prediction and binary risk stratification before clinical decompensation

 Dataset Theme / Edition

Cardiovascular Heart Failure Risk Prediction Dataset — v1.0.0 (Initial Release, May 2026)

 Short Dataset Summary

A large-scale, clinically grounded synthetic dataset of 50,000 patient records built to support AI-driven early heart failure detection. Features span 11 clinical domains — from lab biomarkers and ECG findings to lifestyle factors and care history — with realistic inter-feature correlations, structured missing data, and hard-enforced clinical rules ensuring every high-risk record meets evidence-based HF criteria.

 Intended Research Applications

Binary classification · Survival analysis · Explainable AI (XAI/SHAP) · Risk stratification · Federated learning · Biomarker discovery · Healthcare policy simulation · AI model benchmarking · Medical education

 Total Number of Patients

50,000 patients (13,047 high-risk · 36,953 low-risk)

 Number of Features / Parameters

68 total — 55 clinical features + 3 outcome/target columns + 10 identifiers & source flags

Breakdown: Demographics (8) · Lifestyle (6) · Medical History (7) · Vital Signs (5) · Lab Biomarkers (12) · ECG (5) · Echo Imaging (4) · Symptoms (7) · Medications (5) · Longitudinal History (5)

 Data Modalities Included

Structured EHR · Laboratory values · ECG-derived parameters · Echocardiographic imaging metrics · Patient-reported symptoms · Pharmacotherapy records · Retrospective longitudinal history · Demographic & socioeconomic data

 Recommended AI Tasks

Binary classification (primary) · Regression (hospitalization risk, survival probability) · Survival/time-to-event modeling · Multi-task learning · Federated learning · Fairness & bias auditing · Imputation model development

 Dataset Size

13.6 MB (CSV, uncompressed) · 4.2 MB (gzip-compressed) · 68 columns × 50,000 rows · 3.34% overall missing rate · 0 duplicate records

 Release Date

May 2026

 Why This Dataset Exists

Real patient cardiovascular data is tightly access-controlled, siloed across institutions, and laden with regulatory barriers — making it nearly impossible to freely develop and benchmark clinical AI systems. This dataset provides a privacy-safe, regulation-free synthetic analog with clinical-grade realism, so researchers can build, test, and compare HF prediction models without waiting for data access agreements or risking patient privacy.

 Real-World Problem Addressed

Heart failure affects ~64 million people globally and is the leading cause of hospital readmission in adults over 65. The majority of patients are diagnosed only after acute decompensation — a preventable crisis if risk is detected earlier. This dataset targets the gap between population-level cardiovascular screening and timely clinical intervention, enabling AI tools that can flag high-risk patients weeks or months before they deteriorate.