A Dynamic Machine Learning Framework for System- Wide Drug Shortages Prediction and Autonomous Supply Chain Rerouting in U.S. Healthcare Infrastructure

• Sunday Sunday

  Ladoke Akintola University of Technology

  Author

Drug shortages represent a persistent and escalating crisis within the U.S. healthcare system, threatening patient safety, inflating operational costs, and exposing fundamental vulnerabilities in pharmaceutical supply chains. Despite increasing awareness, current shortage management remains predominantly reactive, relying on manual tracking and static forecasting methods that fail to capture the complex, dynamic nature of supply disruptions. This study addresses this critical gap by developing and validating a dynamic machine learning framework that integrates predictive analytics with autonomous supply chain rerouting capabilities. The proposed framework employs a hybrid architecture combining Gradient Boosting, Random Forest, and a Dueling Double Deep Q-Network (D3QN) to forecast shortage probabilities up to 90 days in advance, coupled with a reinforcement learning-based rerouting agent that dynamically adjusts procurement pathways upon shortage confirmation. Retrospective analysis of FDA Drug Shortages Database records (2018-2026) and CMS Medicare Part D utilization data demonstrates that the framework achieves 89.4% prediction accuracy (AUC-ROC = 0.92), outperforming baseline static models by 22.3%. The autonomous rerouting mechanism reduces estimated shortage impact duration by an average of 18.6 days compared to manual intervention benchmarks. This research provides a validated, replicable framework for transitioning healthcare supply chain management from reactive crisis response to proactive, data-driven resilience.

• pdf

Copyright (c) 2026 Sunday Sunday (Author)

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