Prototyping Intelligent Software Solutions for Global Health Engagement.

Abstract

Introduction: Understanding the Global Health Engagement (GHE) landscape is the key to effective military planning and execution across tactical, operational, and strategic levels. Global Health Engagements are intended to both help partner nations achieve the capacity, capability, and interoperability necessary to provide care to their own and also delivering an appropriate standard of care to U.S. service members, when needed. As standards and capabilities vary globally, planning GHEs is increasingly data-driven, relying on sources focused on GHE information requirements as well as those focused on GHE-supportive information needs. Current challenges to the use of GHE data in planning and crisis management are its lack of availability, standardization, integration, and visualization. Solving these challenges requires improved data curation, analysis, and user interface design to properly optimize engagements and care delivery.

Materials and methods: Our research focused on supporting the Center for Global Health Engagement's (CGHE) ability to effectively plan, execute, and track a range of these engagements across the tactical, operational, and strategic levels. We leveraged Machine Learning (ML) capabilities including Large Language Models (LLMs) to assist users in making sense of large and disparate datasets. Our solution focused on automating the ingestion of data from unstructured text (GHE surveys), establishing a unified and standardized data format, developing interactive analytic tools, and presenting the results through user-specific visualizations to support decision-making and risk-informed course of action recommendations.

Results: To curate, analyze, and visualize GHE data, we developed a prototype ML algorithm that employs an LLM for global health tactical-level hospital data ingestion, curation aggregation, and analysis and displays results for patient distribution in response to a crisis scenario using a 3D geospatial mapping visualization tool. The resulting capability uses an advanced, adaptive User Interface to visualize outputs from the ML algorithm, including providing explanations in a human-readable format on how the algorithm arrived at these outputs.

Conclusion: The results provide practical applications for proof of concept of AI assistance in supporting global health data processing and analysis, with applications extending to the biosurveillance, medical countermeasures, and medical logistics domains. This study has direct implications for understanding partner healthcare capabilities and integrating them into military healthcare plans to support military and medical decision-making for operational planning, crisis management and conflict healthcare delivery and the planning and execution of future health engagements.