Discussion

Prevailing trends in biomedical research include decreasing federal research budgets, and at the same time an increasing availability of data captured in multiple sources.  The first trend is driving innovation in the second by emphasizing the need for better integration, management, and use of available data in order to conduct research more efficiently.  By aggregating data from heterogeneous data sources and from large numbers of patients, the application of certain “big data” techniques in medicine holds the promise of facilitating new research, conducting research more efficiently (for example, the use of registries as a platform for clinical trials), answering public health questions, and serving as the foundation for a learning health care system.

The current data environment in CHD research has many assets.  These resources include numerous clinical registries, administrative and clinical research data sets, quality improvement databases, electronic health records, and others.

A number of independent efforts are working on various aspects of developing an advanced data architecture and better data integration in the field of CHD research.  These include developing data standards and definitions, and employing various methodologies for certain data linkages within and across hospital systems.

However, most CHD data sources currently remain siloed with relatively limited integration at the patient, health system, or national level.  In addition, aspects such as data governance and sharing policies, IT strategies for the development of more widespread data integration, and solutions to foster collection of longitudinal follow-up and population-based information across the lifespan have received more limited attention to date.

The Working Group discussed a number of topics related to data integration including:

• Landscape of CHD registries and other datasets;
• Advances and new developments from current CHD registries;
• Nomenclature standardization;
• Data linkages at a national level through probabilistic matching;
• Data integration strategies at the local level; and
• Models of data networks including PEDSnet, National Database of Autism Research (NDAR), research in health care delivery systems like the Cardiovascular Research Network (CVRN) and Patient Outcomes Research to Advance Learning Network (PORTAL), and distributed data networks (Mini-Sentinel, NIH Distributed Research Network, and PCORNET).

Recommendations

The Working Group made several recommendations to further the efforts of data integration in the CHD community.

• Future vision.  The Working Group acknowledged that our current conceptualization of data and data management is largely outdated.  In addition, the volume of available data will continue to increase through more wide-spread capture of genomic data and real-time physiologic data.  Our current databases and data structure will be insufficient for the task of managing data needs of the future.  The Working Group recommended that further collaboration and consultation with data scientists and experts from diverse fields and industries outside of medicine will be important in understanding and incorporating modern data storage, manipulation, and analytic techniques into short and long term data solutions for the CHD community. The importance of considering these techniques within the context of medical decision making was also discussed, as simple associations found in the data may not always be appropriate to guide the treatment of patients, and ensuring data accuracy, reliability, etc. will continue to be important concepts regardless of the data storage and integration techniques used.

• Short-term data integration. The Working Group recommended that the CHD community take steps in the near term to begin merging information from the currently available CHD data sources.  The group felt that this could facilitate further research that could not be conducted with individual datasets alone, and could also help to promote efficiency in research.  One approach to data integration could be the use of a global unique identifier (GUID, as developed by NDAR) which would be incorporated into the various registries.  Downsides to this approach include the need to collect the appropriate data elements from families, and the need to maneuver through the various data sharing policies of the different registries in order to merge data on a national level.  Another approach is to link data locally at a site within a data warehouse using existing patient identifiers, dates, and native source identifiers (as is already occurring at some sites), with subsequent sharing of linked information between hospitals or heart centers.  This approach bridges some of the hurdles identified with the GUID approach and makes linked information available for both local purposes as well as for aggregate research, but would require more investments at the local level. Either approach would also need to address the limited current data available regarding basic long term outcomes such as survival and quality of life. Funding sources would need to be identified, and data sharing and governance policies explored in either approach. The Working Group favored the second option involving local linkages and recommended further exploration and pilot testing of logistics, as well as investigating options for funding. Exploring funding and support available through NHLBI and the Pediatric Heart Network, PCORI, and other sources, was recommended

• Data standardization. The Working Group acknowledged the importance of data standardization to facilitate data pooling and analysis. The Working Group recognized the International Pediatric and Congenital Cardiac Code (IPCCC) as the standard nomenclature for this field, and recommended that IPCCC terminology and definitions should be incorporated into all relevant data sources when possible, including CHD registries, clinical data and the EHR, and research datasets, and recognized ongoing work to incorporate IPCCC terms into the ICD coding system. The Working Group recommended that sites and other stakeholders work together to develop standard elements of data collection to support longitudinal follow-up, and to identify more efficient ways to capture this information.

• Align goals with stakeholder interests. Moving toward more fully integrated data will require investment and input from multiple stakeholders, including hospital systems, researchers, national organizations and professional societies, and patients and families.  To make a case for change, the field will need to discuss and identify the value of data integration across stakeholders.  At the hospital level, one such value point is the optimization of strategic investments already made in electronic health records and clinical registries.  Data integration will provide a fuller picture of the care provided and thus may enable improved quality, reduction in errors, and increased value.  Patients and families are also critical stakeholders, and data integration can seek to align with their interests by facilitating improved care and increased participation through patient reported outcomes and a seat at the table when determining quality improvement and research priorities.

Publication Plans

The Working Group committee will develop a report of the meeting for publication in an appropriate professional journal.

Participating Division

Division of Cardiovascular Sciences

Staff Contact

Jonathan Kaltman, MD
kaltmanj@nhlbi.nih.gov
301-435-0510

Working Group Members

Chair

• Sara Pasquali, MD, University of Michigan C.S. Mott Children’s Hospital

Members

• David Bertoch, Children’s Hospital Association
• Betsy Blume, MD, Children’s Hospital Boston
• Robert Campbell, MD, Emory University
• Anthony Chang, MD, MBA, MPH, Children’s Hospital of Orange County
• Wendy Chung, MD, PhD, Columbia University
• Tiffany Colarusso, MD, MPH, Centers for Disease Control
• Lesley Curtis, PhD, Duke University
• Gregory Farber, PhD, National Institutes of Mental Health
• Sherry Farr, PhD, Centers for Disease Control
• Christopher Forrest, MD, PhD, Children’s Hospital of Philadelphia
• Michael Gaies, MD, University of Michigan
• J. William Gaynor, MD, Children’s Hospital of Philadelphia
• Alan Go, MD, Kaiser Permanente
• Paul Henchey, MS, Arbormetrix
• Jeffrey Jacobs, MD, Johns Hopkins All Childrens Heart Institute
• Gerard Martin, MD, Children’s National Medical Center
• Steven Schwartz, MD, The Hospital for Sick Children

NIH Staff

• Kristin Burns, MD
• Gail Pearson, MD, ScD                                         
• Victoria Pemberton, RN, MS, CCRC
• Mario Stylianou, PhD