Value Based Monitoring of Cycles of Care

We propose to prototype a scalable, cost-effective methodology and supporting technology for rapidly developing procedures to automatically measure value of delivered healthcare for a wide range of condition-specific cycles of care, starting in Phase 1 with analysis of low back pain. To maximize scalability, our approach takes full advantage of unique asset available to the military: comprehensive longitudinal patient data available in the Military Health Services M2 Data Mart. Our approach applies a mix of fully automated and human-in-the-loop active learning methods to these data to rapidly and semi-automatically specify fully automated procedures for 1) identifying and linking all medical records related to low back pain for each patients medical history and organizing them into discrete treatment episodes, and 2) estimating six metrics for measuringvalue of healthcare delivered. We also prototype additional machine-learning-based technology enabling analysts to rapidly explore the resulting value-scored treatment episodes to identify military treatment facilities (MTFs), healthcare providers, or demographic segments where opportunities may exist to improve value of care.