Counterfactual Fairness and Interpretable Machine Learning for Equitable Medical Resource Allocation During National Public Health Emergencies

• Billy Elly

  Lautech

  Author

National public health emergencies, such as pandemics and widespread disease outbreaks, consistently expose critical vulnerabilities in healthcare systems, particularly regarding the equitable distribution of scarce medical resources. Traditional allocation frameworks often prioritize efficiency and clinical urgency while inadvertently exacerbating disparities affecting vulnerable populations defined by race, socioeconomic status, age, and geographic location. This research addresses the gap between algorithmic fairness theory and practical resource allocation by developing an integrated framework combining counterfactual fairness principles with interpretable machine learning. Using a hybrid CNN-LSTM model enhanced with SHAP-based explainability and fairness-aware optimization constraints, the proposed framework was validated on harmonized multidisease datasets representing COVID-19 and comorbid conditions. The model achieved 89.4% accuracy in risk stratification while reducing allocation disparity metrics by 34.2% compared to baseline approaches. The framework demonstrated that counterfactually fair decision-making, when integrated with transparent model explanations, enables resource allocation policies that balance equity with clinical efficacy. These findings provide a replicable methodology for public health administrators and policymakers to operationalize fairness in AI-driven emergency response systems, with significant implications for health equity and algorithmic accountability.

• PDF

Copyright (c) 2026 Billy Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.