AI
SynTabFall releases privacy-preserving synthetic health data for fall-risk AI

Healthcare AISynthetic dataGermany
Published: 24 July 2026. A peer-reviewed Scientific Data descriptor releases 745,380 fully synthetic rows with 44 attributes derived from an anonymised German hospital dataset. The authors report that models trained on the synthetic data can achieve fall-risk predictive performance comparable with models trained on the real source data.
Why it matters: It provides an open dataset and documented governance process for a high-friction health-data use case. Impact: It may accelerate privacy-conscious model development, but synthetic data can reproduce source bias and does not prove clinical generalisation. The parity claim is study-specific and not independently reproduced. Scientific Data




