Informatics Evidence, Redux
William Hersh, MD
Professor and Chair
Department of Medical Informatics & Clinical Epidemiology
Oregon Health & Science University
Blog: Informatics Professor
About a year ago there was a big dustup in the informatics field concerning a study published by Romano and Stafford in Archives of Internal Medicine that purported to show that electronic health record (EHR) use was not associated with improved quality of care [1]. As honest informaticians, we need to take such research seriously, aiming to improve what we do based on the evidence. This study, however, was problematic, in that it was based on an older data set not designed for answering questions such as the one asked by Romano and Stafford. A better approach would have been to perform a prospective clinical trial that directly assessed an informatics intervention, one of which was indeed published a few months later that did show improvement in care augmented by use of an EHR [2].
Now comes a similar situation a year later with the publication of a study by McCormick et al. in Health Affairs, which uses the same data source to show that physicians who have access to computerized imaging results (not necessarily via an EHR) have a 40-70% higher likelihood of ordering imaging tests [3]. This study set off a similar conversation about whether we are jumping the gun, especially with regards to the substantial federal investment in EHR adoption through the Health Information for Clinical and Economic Health or HITECH Act. As with the Romano and Stafford study, this new study set off a lot of debate, including an exchange between the National Coordinator for Health IT and a rebuttal by the authors.
It is unfortunate to have to reiterate that we should be guided by the evidence, but given that many of us do have careers staked on the success of the HITECH Act, we must acknowledge potential biases and be as objective as possible in evaluating research results. That said, the study by McCormick truly uses a very weak methodology and certainly does not justify the sweeping conclusions by the authors in their paper or their rebuttal.