This research develops a hybrid human-AI approach that significantly improves disease detection from electronic health records while reducing manual review requirements.

• Achieved 94-97% accuracy for detecting conditions like myocardial infarction and diabetes
• Reduced manual review workload by 76% compared to traditional methods
• Demonstrated superior performance over both pure-AI and traditional approaches
• Established a practical workflow integrating clinician expertise with LLM capabilities

This innovation addresses critical healthcare challenges by enabling more efficient disease surveillance and performance monitoring across healthcare systems while maintaining high diagnostic accuracy.

Integrating Large Language Models with Human Expertise for Disease Detection in Electronic Health Records