PRIME-CVD: A Parametrically Rendered Informatics Medical Environment for Education in Cardiovascular Risk Modelling
Researchers create a fully synthetic cohort of 50,000 patients to train AI models on cardiovascular risk without privacy risks.
A team of researchers, including Nicholas I-Hsien Kuo, Marzia Hoque Tania, Blanca Gallego, and Louisa Jorm, has introduced PRIME-CVD (Parametrically Rendered Informatics Medical Environment for Cardiovascular Disease). This novel platform addresses a critical bottleneck in medical AI and informatics education: the lack of accessible, privacy-safe patient data for hands-on training. PRIME-CVD generates a fully synthetic cohort of 50,000 adults undergoing primary cardiovascular disease prevention. Crucially, the data is not derived from real Electronic Medical Records (EMRs) or generative AI models, but is parametrically rendered from a user-specified causal graph, using Australian population statistics and published epidemiological estimates. This method ensures the data has realistic risk gradients and subgroup imbalances while guaranteeing negligible re-identification risk.
The environment provides two complementary data assets. The first is a clean, analysis-ready dataset designed for exploratory analysis and survival modeling. The second restructures the same synthetic cohort into a relational, EMR-style database complete with realistic structural and lexical heterogeneity, mimicking the messy data challenges found in real hospitals. This dual-structure enables comprehensive instruction in the full data science pipeline—from data cleaning and harmonization to causal inference and policy-relevant risk modeling. By releasing PRIME-CVD under a Creative Commons Attribution 4.0 license, the team aims to standardize and democratize medical data science education, fostering greater reproducibility and transparency in a field traditionally constrained by sensitive data governance.
- Contains two synthetic datasets for 50,000 patient records, generated from public stats, not real EMRs or generative AI.
- Enables training in data cleaning, causal reasoning, and risk modeling with realistic data imbalances and zero privacy risk.
- Released under a Creative Commons license to support scalable, reproducible medical AI education and research.
Why It Matters
Unlocks hands-on AI and data science training for medical professionals and researchers without the legal and ethical hurdles of real patient data.