NESTED KNOWLEDGE INC — Department of Health and Human Services SBIR Phase I: NLM

Description

PROJECT SUMMARY A wide range of vital decisions, from reimbursement to regulatory clearance to physician and patient decision- making, depend on clinical publication. Clinical publications are the primary mechanism of evidence-based decisions, and literature reviews are thus the primary comprehensive clinical outcome comparison methodology. However, current literature review methods are outdated, unstructured, and disorganized, and potentially lifesaving data are accessible and comparable only with hundreds of hours of work in literature reviewing. Especially given the growth of technology-driven data management, the current paradigm of combining clinical outcome data fundamentally fails to communicate to general medical audiences whether any given therapy works, in transparent, comprehensive, and updatable forms, and the typical meta-analysis is unable to even demonstrate level of coverage of its search across existing indices. This problem has been recognized by many organizations, from the NIH’s Data Informatics Working Group (DIWG) to the AAAS to The National Academies of Sciences, Engineering, Medicine, which have all publicly stated that tech-driven harmonization and data visualization are necessary to effectively share research. However, searchability, visualization, and harmonization efforts have not yet permeated medical publishing. The DIWG stated that: “The colossal changes in technologies and methods for doing biomedical research have shifted the bottleneck in science productivity from data production to data management, communication, and data interpretation.” We agree that the greatest bottleneck in clinical sciences are in fact related to communication, and we believe that it is due to insufficient adoption of novel technologies in publishing, especially interactive visual methods of presenting field-wide data and automated methods of data gathering. Our vision is to create a comprehensively researched, constantly updated, easily digestible platform for dissemination of crucial data presented among scientific publications based on expert-designed, automated data extraction from existing publications. After debuting in stroke—because of our previous experience in the field—we have achieved proof-of-concept for interactive, visual meta-analyses and for partially-automated data extraction from PDFs. Now, we propose to automate the foundation of meta-analysis, the search/inclusion process, through data analytics and prediction modelling, and expand our platform to include all studies relevant to stroke research. If successful, our project would provide a field-wide research tool in stroke, as well as enabling us to create the methods that make scaling across all medical disciplines possible.PROJECT NARRATIVE Clinical publications provide the data used in decisions ranging from patient decision-making to reimbursement to regulatory approvals, but the methods used in publications to promulgate data are outdated. We have designed a software-based, visual method of communicating hundreds of studies in a single interface, and are working to automate the underlying research process. This grant would fund the expansion of our interface to cover pressing questions in stroke research, validate a search predictive analytics engine to ensure meta- analyses cover all relevant research, and lay the foundation for expansion across medical disciplines.