The developing science of clinical informatics, driven by the growing use of electronic health records, offers an unparalleled opportunity to improve the quality of care for children with cancer. Researchers carefully analyze data gathered by both clinicians and physician scientists to improve disease detection and diagnosis, choose the most appropriate cancer treatments, minimize side effects, and provide the best available supportive care to children with cancer.

Although large and rapidly growing clinical data sets already exist in electronic form, their use is hampered by differences in data collection and storage, and a lack of secure methods to retrieve and subsequently analyze them. Charles Bailey, MD, PhD, and researchers at the CCCR have implemented several approaches to address these issues.

The first of these involves creating the components needed to support a learning health system for children, that is, a medical system and a community of clinicians and patients dedicated to continuously learning how to improve the health of children, both inside and outside the hospital. The informatics work starts with constructing a complete and accurate picture of care provided to large pediatric populations. This information, curated by an academic collaboration called PEDSnet, forms the basis of a comprehensive data trust for The Children's Hospital of Philadelphia and other pediatric hospitals, and facilitates collaborative sharing of pediatric health data and knowledge across institutions. This resource provides researchers with better ways to determine what tests and treatments are most effective, which will help clinicians to make more informed point-of-care decisions for pediatric cancer patients. By advancing a collaborative approach to learning, PEDSnet also provides ways for researchers from different institutions to work together more effectively, and for patients and families to be an integral part of the research process.

Our current work in the PEDSnet Data Coordinating Center focuses on making the core data consistent across sites, using the OMOP vocabularies and common data model as a basis for semantic interoperability. In order for the data to be useful once assembled, it is also important to understand what specific elements mean and where there may be pitfalls that could affect analyses.  Our work on assessing the operating characteristics and quality of the data helps both informatics teams to improve data collection and researchers to design studies that use the data appropriately.

In addition to work on the informatics infrastructure of collaborative networks, we are interested in exploring the scientific utility of clinical data.  We have used data collected within PEDSnet to demonstrate that primary clinical information provides a valuable complement to the kinds of billing data usually used to study healthcare delivery.  Building on this work, we have used data from electronic health records to demonstrate an association between commonly used antibiotics in young children and later obesity. We have also used information from clinical registries to show that it is possible to compare the effectiveness of different treatments in standard clinical settings, as a complement to more tightly controlled but more expensive and smaller randomized clinical trials.  Our current work in this area focuses on better defining significant patterns, called computable phenotypes, in clinical data and using these to learn about the risks and effectiveness of specific practices.

In a learning health system, it is important not just to develop new knowledge, but to be able to measure how well known goals are being met.  Our work in this area deals with creating quality measures that can be applied to large numbers of patient records automatically, so we can understand the quality of care in large populations.  Within CHOP, we are also developing ways to make better use of electronic health records locally to improve care for children with cancer. For example, an informatics-based system has improved the vaccination rate by more than 20 percent in children receiving chemotherapy. Current work deals with other aspects of supportive care to reduce errors, minimize side effects and tolerability issues, and improve patient safety.

Each of these areas of research has positive but limited impact in isolation.  However, the knowledge and experience gained from each bring us closer to meeting the critical commitment of a learning health system: That each time we care for a child, we learn how to provide better care the next time.