Simulating a Million Patients: Realistic Health Data Generation
Talkat RustConf 2026 · on Thursday, 2026-09-10· at 11:00· Room 510ac | Palais des Congrès de Montréal, Palais des Congrès de Montréal·
summary
Project simulates life courses for realistic synthetic health data using Rust for speed and correctness and exposes Rust bindings to Python and R. Life events are modeled as DAGs then collapsed into an adjacency matrix to sample coherent patient paths; ages are constrained with a new type wrapper and a 50/50 forward/backward sampling rule to reduce drift. Generates up to 10 million patients and ~3 million events in ~72 ms on a capable machine with worst-case prevalence error about 0.4%.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 7:59Synthetic data challenges
- 11:08Rust for life course simulation
- 16:01Fixing forward drift in models
- 23:06Scaling to millions of patients
- 27:27Creating realistic health records
- 29:26Future plans and takeaways
related talks
01h 00mRust for good: systems programming in Science and Public Health
36m"Rust at the Last Mile in Healthcare"
34mLongitudinal Phenotypes and Disease Trajectories at Population Scale
40mHealthy Data, Rusty Code: Epidemiology Meets Modern Systems Programming
25mApplication of C++ in Computational Cancer Modeling
42mSex, Money and Evolution - Simulation and Data Analysis with Ruby and R
13mData Nerding in Public Health
35mBetter Research Planning Through Simulation