Researchers pilot ‘model to data’ approach to developing predictive algorithms

Researchers from the University of Washington have proposed a “model to data” approach that could facilitate the development of predictive models using electronic health records (EHRs) without compromising patient privacy. The approach involves sending models to an isolated environment for training and evaluation on sensitive data. The researchers conducted a pilot study using the “model to data” framework and successfully created three models that predicted patient mortality. The use of synthetic data is also being explored as a potential solution to privacy concerns in healthcare data analysis.