Statistical Center for Translational Research in Medicine (SC-TRM)
Advancing translational medicine through innovative biostatistics, real-world data, and clinical research expertise.
SC-TRM Software Resources
SC-TRM develops open-source software that translates methodological research in risk prediction, model evaluation, and healthcare data analysis into practical and reproducible tools.
The software packages below are publicly available through GitHub and the Python Package Index (PyPI), with synthetic datasets and reproducible examples provided to support learning and application without requiring access to protected health information.
As predictive algorithms become increasingly integrated into healthcare, rigorous evaluation of their fairness across population subgroups is essential to ensure that they do not perpetuate or exacerbate existing health inequities. Conventional approaches typically assess fairness by comparing predictive performance metrics across subgroups, with observed disparities often interpreted as evidence of algorithmic unfairness. However, an important question remains: is such evaluation itself fair? We identified a ubiquitous confounding effect in fairness evaluation and proposed a principled method to account for it.
metrics-adjuster
Adjusted evaluation of binary prediction models across populations with different risk distributions.
Risk Bridge
Development of calibrated risk-prediction models using constrained maximum likelihood when richer source data may not represent the target population.