SBIR Phase II: Determination of complex outcome measures using narrative clinical data to enable observational trials

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase II project is to develop a software application that identifies clinical outcomes to support high validity real-world evidence (RWE). Approaches to ensure data accuracy and protocol validity are critical to maintain safety and efficacy in healthcare. This project will analyze electronic health data to generate clinically meaningful information for personalized treatment plans for patients with multiple conditions that can confound treatment. This technology will fulfill an unmet need to improve patient outcomes, improve healthcare delivery and chronic disease care management, and reduce healthcare costs for patients. This technology can also improve regulatory and reimbursement decision-making for therapeutic approaches. This SBIR Phase II project will address the need for consideration of using additional health data to allow for individualized personalized therapeutic plans for patients with multiple co-morbidities. Subgroup analysis or individualized therapy plans for precision medicine are currently not available based upon the structure of randomized controlled trials for broad conditions like breast cancer or hypertension. This proposal seeks to identify clinical outcomes from unstructured Electronic Health Records (EHR). The proposed work is to develop analytics using natural language processing and inference to leverage the large amounts of health data from real-world evidence (RWE) and observational studies to augment data provided in randomized controlled trials (RCT). The analytic tools will allow a comparison of the effectiveness of various treatment protocols in defined cohorts of patients and develop a personalized treatment plan for an individual patient with multiple co-morbidities. The tasks include: 1) Leverage linguistic phrases extracted by natural language processing (NLP) to recognize outcome-related clinical findings to be maintained as clinical feature metadata; 2) Combine NLP and inference to accurately identify candidate clinical outcomes; and 3) Apply machine-learned and expert knowledge to accurately define complex outcome measures. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.