A Multi-Site Fairness-Aware Machine Learning Framework for Mitigating Racial and Socioeconomic Bias in Clinical AI Across U.S. Hospitals
Keywords:
Socioeconomic health disparities, Racial bias in healthcare AI, Algorithmic disparities, Health equity, Bias mitigation machine learning, Clinical AI fairnessAbstract
Background: Clinical AI models frequently perpetuate systemic biases, resulting in inequitable health outcomes across different patient populations. U.S. hospitals lack robust frameworks that address multiple fairness constraints simultaneously across racially and economically diverse patient groups.
Objective: We developed and validated a fairness-aware machine learning framework that reduces bias across racial groups (White, Black, Hispanic) and socioeconomic strata (low, medium, high) while preserving predictive accuracy.
Methods: We conducted a retrospective cohort study using the MIMIC-IV database with 48-hour ICU time-series data to predict in-hospital mortality. Our hybrid adversarial multi-objective optimization model compared three fairness approaches: Adversarial Debiasing, Exponentiated Gradient with Fairness Constraints, and our proposed Multi-Objective Grid-Search Reweighing (MOGR) framework.
Results: The MOGR framework achieved Pareto-optimal solutions, reducing Equalized Odds difference by 72% and Demographic Parity difference by 81% compared to the unconstrained baseline. The model maintained an AUC of 0.84, with minimal performance decline under 0.02. Disparities for Black patients with low socioeconomic status proved most resistant to mitigation but still showed substantial improvement.
Conclusion: A structured approach using multiple fairness constraints significantly reduces group-level disparities without clinically meaningful performance loss. This framework provides hospitals with a deployable pathway toward achieving health equity across diverse patient populations in clinical settings.
References
Agarwal, A., Beygelzimer, A., Dudik, M., Langford, J., & Wallach, H. (2018). A reductions approach to fair classification. International Conference on Machine Learning, 60-69.
Buolamwini, J., & Gebru, T. (2018). Gender shades: Intersectional accuracy disparities in commercial gender classification. Conference on Fairness, Accountability and Transparency, 77-91.
Calmon, F., Kantarcioglu, M., Kutyar, A., Megiddo, N., & O'Leary, K. (2017). Optimized pre-processing for discrimination prevention. Advances in Neural Information Processing Systems, 30, 3995-4004.
Chen, T., &Guestrin, C. (2016). XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785-794.
Corbett-Davies, S., Pierson, E., Feller, A., Goel, S., & Huq, A. (2017). Algorithmic decision making and the cost of fairness. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 797-806.
DeLong, E. R., DeLong, D. M., & Clarke-Pearson, D. L. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics, 44(3), 837-845.
Feldman, M., Friedler, S. A., Moeller, J., Scheidegger, C., & Venkatasubramanian, S. (2015). Certifying and removing disparate impact. Proceedings of the 21st ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 259-268.
Hardt, M., Price, E., & Srebro, N. (2016). Equality of opportunity in supervised learning. Advances in Neural Information Processing Systems, 29, 3315-3323.
Johnson, A. E., Pollard, T. J., Shen, L., Lehman, L. W. H., Feng, M., Ghassemi, M., ... & Mark, R. G. (2023). MIMIC-IV, a freely accessible critical care database. Scientific Data, 10(1), 1-9.
Kamiran, F., &Calders, T. (2012). Data preprocessing techniques for classification without discrimination. Knowledge and Information Systems, 33(1), 1-33.
Kleinberg, J., Mullainathan, S., & Raghavan, M. (2018). Discrimination in the age of algorithms. Journal of Political Economy, 126(6), 2940-2968.
Knaus, W. A., Zimmerman, J. E., Wagner, D. P., Draper, E. A., & Lawrence, D. E. (1985). APACHE-acute physiology and chronic health evaluation: A physiologically based classification system. Critical Care Medicine, 13(11), 818-829.
Mandrekar, J. N. (2010). Receiver operating characteristic curve in diagnostic test assessment. Journal of Thoracic Oncology, 5(9), 1315-1316.
Mitchell, S. (2019). Algorithmic bias and the limits of fairness audits. AI Now Institute Report.
Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447-453.
Pandya, S. S., Ginsburg, M. E., & Greenwald, M. R. (2019). Socioeconomic factors and hospital readmission in the Medicare population. Health Services Research, 54(2), 445-456.
Pearl, J. (2009). Causality: Models, reasoning and inference (2nd ed.). Cambridge University Press.
Rajkomar, A., Hardt, M., Howell, M. D., Corrado, G., & Chin, M. H. (2018). Ensuring fairness in machine learning to advance health equity. Annals of Internal Medicine, 169(12), 866-872.
White, I. R., Royston, P., & Wood, A. M. (2011). Multiple imputation using chained equations: Issues and guidance for practice. Statistics in Medicine, 30(4), 377-399.
Zhang, B. H., Lemoine, B., & Mitchell, M. (2018). Mitigating unwanted biases with adversarial learning. Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society, 335-340.
Downloads
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
