Precision health holds the potential to bring forth game-changing breakthroughs in medical research and practice, but it’s being held back by a dearth of AI-ready data.
As a discipline, precision health aims to improve individual and public health and wellness by precisely tailoring interventions, treatments, and care plans to the unique genetic, environmental, and lifestyle circumstances of each individual patient. Precision health is advanced through research, clinical, and public health practices that synthesize knowledge and meaning from a wide variety of high-volume data sources, including electronic health records, insurance claims, social determinants of health, wearables, environmental exposures, and omics data such as genomics, metabolomics, and epigenomics.
Due to the vast quantities of data involved, emerging technologies like artificial intelligence (AI) and machine learning (ML) have quickly become central to applying and advancing precision health, and achieving meaningful progress in the response to key public health threats and treatment of some of the world’s most burdensome chronic health conditions.
Unfortunately, a shortage of AI-ready data is hindering precision health’s development by slowing AI/ML adoption in the field and limiting the technology’s ability to integrate data points and uncover beneficial health insights. A key first step for incorporating AI-ready data practices into any discipline is developing a clear, widely accepted definition of the relatively new term. At Booz Allen, we consider AI-ready health data to be that which conforms to the FAIR Guiding Principles for scientific data management and stewardship (i.e., findable, accessible, interoperable and reusable) in addition to being equitable, protected, machine readable, and well-defined.
The factors limiting the availability of AI-ready data are many. They include lax data provenance practices; insufficient data governance, siloed data, non-standardized data (e.g., data that is not mapped to common ontologies or terminologies) and inconsistency in the collection and management of sample and patient identifiers. To harness the transformative potential of precision health, the data needs to be AI-ready.
Seeing that it will take large quantities of broadly accessible AI-ready data to realize the full transformative potential of precision health, U.S. Government agencies like the Department of Veteran’s Affairs (VA), the Department of Defense (DOD), and the National Institutes of Health (NIH) are prioritizing its generation and availability. For example, NIH is engaged in two initiatives, Bridge to Artificial Intelligence (Bridge2AI) and Artificial Intelligence/Machine Learning Consortium to Advance Health Equity and Researcher Diversity (AIM-AHEAD), that seek to advance biomedical research in part through the development of flagship AI-ready data.
It's time for all biomedical data producers and managers to recognize this imperative and follow suit.