As diagnosis codes change, data lost in translation – both ways
Changes in how medical diagnoses are coded under the latest international disease classification system – known as the ICD-10 codes – may complicate financial analysis, research projects and training programs that depend on look-back comparisons of health care data, report researchers at the University of Illinois at Chicago.
The report, a collaboration of researchers at UIC and at the University of Arizona, is online in the Journal of the American Medical Informatics Association (JAMIA).
Codes for diagnoses – used to justify payments, among other things – may not translate from ICD-10 back to ICD-9 in a simple way, says Andrew Boyd, assistant professor of biomedical and health information sciences at UIC and first author of the paper.
Boyd and his colleagues have been looking at issues that could come up as physicians and hospitals change from one system to the other. Previously they found that some ICD-9 codes map well to ICD-10, but many more have highly convoluted mappings, and some don’t map at all. This forward-mapping is needed for continuing payments of ongoing medical conditions.
“Now, we are taking the same methodology and looking backward,” Boyd said. Reverse-mapping from ICD-10 back to ICD-9 will be important for all sorts of retrospective analyses, he said, “because we have 30 years of data that we want. We don’t want to lose all this information.”
Clinical researchers and analysts conducting studies across datasets – and hospital administrators who manage growth and watch trends for strategic planning – will need to pull data under both the new and the old codes. Mapping back from ICD-10- to ICD-9 is just as complex as mapping from ICD-9 to ICD-10.
The researchers created a web portal tool and translation tables designed to provide guidance on ambiguous and complex translations and to reveal where analyses may be challenging or impossible. The tool lists all ICD-9-CM diagnosis codes related to the input of ICD-10-CM codes and classifies their level of complexity, which can be a one-to-one “identity,” or reciprocal (the simplest); class-to-subclass; subclass-to-class; “convoluted”; or “no mapping.”
“The health care system runs on data,” Boyd said. “We are fundamentally changing the way we record the data.” Although the new system will improve the way the data are sorted and recorded, he says, integrating it with the last 30 years of information will be difficult.
“In 20 years, it will be great, because we’ll have all this new data,” said Boyd. “But right now, the transition is going to be challenging.”
Yves A. Lussier of the University of Arizona is co-principal investigator on the study. Jianrong ‘John’ Li, Colleen Kenost and Donald Saner of the University of Arizona and Binoy Joese, Young Min Yang, Olympia A. Kalagidis, Ilir Zenku and Neil Bahroos of UIC are co-authors.
Boyd and Lussier are supported in part by Center for Clinical and Translational Sciences of the University of Illinois (NIH 1UL1RR029879-01, NIH/NCATS UL1TR000050), UIC’s Institute for Translational Health Informatics, the Office of the Vice President for Health Affairs of the University of Illinois Hospital & Health Sciences System, the Office of the Vice President for Health Sciences of the University of Arizona, the Arizona Health Sciences Center, the BIO5 Institute, the National Library of Medicine (K22 LM008308-04) and the University of Arizona Cancer Center (P30CA023074).