To change the way children's hospitals perform comparative effectiveness research and generate evidence for researchers, Child Health Corporation of America (CHCA) is in the process of collecting lab, microbiology, and radiology results for inpatient encounters to enhance its administrative and clinical database, known as the Pediatric Health Information System+ (PHIS). To tackle the daunting task of normalizing and standardizing the disparate feeds from the six participating children’s hospitals, CHCA is using a terminology management solution to assist with this interoperability initiative. Ultimately, CHCA leaders see this development work expanding beyond the realm of pure research to supporting performance improvement.
In 2011, the Overland Park, Kan.-based CHCA, the National Association of Children’s Hospitals and Related Institutions, and the National Association of Children’s Hospitals merged under a single organization, called the Children’s Hospital Association, which brings together the resources of more than 220 children’s hospitals and related institutions.
PHIS+ is a pediatric database that was created by CHCA for member hospitals. It contains clinical and financial details of more than 14 million patients. Since 1999, PHIS+ has collected data on 20.5 million patient encounters. A three-year $9 million Agency for Healthcare Research and Quality (AHRQ) grant that began in October 2010 to expand the dataset to include lab, radiology, and microbiology results for inpatient encounters to support research. Ron Keren, M.D., M.P.H, is the grant’s principal investigator, and associate professor of epidemiology, University of Pennsylvania School of Medicine, and associate professor of Pediatrics at the Children's Hospital of Philadelphia. University of Utah’s Biomedical Informatics Core lead Phase I of this project with mapping and data standardization processes required for this effort.
“The researchers told us having the lab results will help them to determine very specific diagnoses and also condition-specific outcomes,” says David Bertoch, vice president of informatics at CHCA. “The lab results will also help them with better risk adjustment techniques when developing their research models.”
Bertoch also notes that radiology results will help researchers definitively determine the diagnosis from the patient procedure, while microbiology results will help determine what type of infection, and the type of antibiotic regimen the patient was given.
Standardizing Data Feeds
When CHCA started collecting lab data from hospitals, they learned that there were hundreds of tests that each hospital performs, and it was too much to compute. So CHCA pared down the data streams, so that each hospital would submit the top 300 tests ordered. “When we were going through those tests, it was very clear that those hospitals that are not using a standard such as SNOMED, use very different, very localized ways of naming their tests, and also with some of the other things that would be considered meta data around each of the different results,” says Richard Stepanek, CHCA’s CIO.
So far, CHCA has collected five years of the laboratory data from the six participating hospitals and is in the process of classifying and connecting it back to the historical records criteria, says Stepanek. CHCA is using a terminology server (provided by the Ridgefield, Conn.-based Apelon) to normalize and standardize the disparate feeds from the children’s hospitals, and as any of those standards evolve, the database can quickly adapt to those new standards, says Stepanek.
“Some of the initial challenges were getting the extracts of the data or consistency about how to get the data out of the different systems of the different hospitals,” says Stepanek. “We went through a little trial and error around that space until we settled on a standard for collecting that data.” Also getting the data use agreements and business associate agreements from the beginning was another challenge.
After laboratory data is mapped, microbiology results will be tackled next, and then radiology. “Because most radiology results tend to have PHI embedded in the narratives, we have to go through a process of de-identifying the narratives,” says Stepanek. “We’re going to use some natural language processing to extract out some of the terminology. Then we will process the remaining discreet data through the Apelon platform and map that to some standard terminology.”
Comparative Effectiveness Research Projects
As part of the grant, four comparative effectiveness research projects will be performed as an initial pilot to demonstrate the value of the new data; however, researchers could potentially tweak these studies at any point.