Crucial to all of these efforts is the appropriate alignment of data sources with their intended use. Different uses have different requirements of data, and therefore different priorities in terms of the evolving clinical data utility. This challenge is magnified by the lack of lessons and best practices for how to approach data quality assurances needed to support the multiple facets of a learning health system. To address these issues and gain a better understanding of the types, sources, applications, limitations, appropriate uses, and quality improvement needs for digital health data, the IOM’s Roundtable on Value & Science-Driven Health Care convened a meeting on March 23, 2012, titled Digital Data Priorities for Continuous Learning in Health and Health Care. This meeting followed a series of related discussions summarized in the IOM publication titled Digital Infrastructure for the Learning Health System (2011), and built on a body of work done by the Roundtable on the centrality of a clinical data utility to support continuous learning and improvement in health and health care (IOM, 2010, 2011a,b,c).
Digital health data are produced in a variety of different environments, which impact greatly the characteristics of the data. Who collects the data, how it is collected, why it is collected, and what is collected are some of the ways that digital health data differ depending on their source and have implications for the use of that data. Understanding these characteristics is necessary to match data users with appropriate sources, and to understand limitations and barriers in data analysis.
The increased adoption of EHRs has given data from routine care increasing prominence as a potential component of the data utility. Data collected in the course of delivering patient care come from a variety of sources such as clinician offices, ambulatory procedure centers, hospitals, and nursing and extended care facilities. The types of data vary by care setting, but generally include both clinical and administrative elements. Clinical elements include structured fields and free text notes, laboratory results, images, and diagnostic test results. Administrative information includes process performance metrics, and details needed for billing, such as International Classification of Diseases (ICD) codes.
Also growing in importance is data originating directly from patients. These data can be captured through the use of personal health records or patient portals, in clinical records as recorded by healthcare personnel, or in records external to the health system. They can include personal reports of current health status and wellness, family history, and remote site laboratory readings, as well as health-related data such as socioeconomic, environmental, and lifestyle factors. There is increasing interest in including