A nice writeup in the New York Times about mining EHR data to further medical knowledge.
The use of electronic records also may help scientists avoid sidestep [sic] the rising costs of medical research. “In the past, you had to set up incredibly expensive and time-consuming clinical trials to test a hypothesis,” said Nicholas Tatonetti, assistant professor of biomedical informatics at Columbia. “Now we can look at data already collected in electronic medical records and begin to tease out information.”
Recent work by Dr. Altman and Dr. Tatonetti, published in 2011, offers a compelling case study. As a graduate student at Stanford, Dr. Tatonetti devised an algorithm to look for pairs of drugs that, taken together, cause a side effect not associated with either drug alone. One pairing popped up when he used his new software to search the Food and Drug Administration’s database of adverse drug reports: Paxil, a widely used antidepressant, and Pravastatin, a cholesterol-lowering drug.
Neither was known to raise blood sugar, but Dr. Tatonetti’s results suggested they might when taken together.
But the challenges posed by this sort of research are significant. The information entered into a medical record may be wrong, and diagnostic codes are notoriously unreliable, according to Dr. Tatonetti, partly because they are also used for billing. And doctors don’t think like researchers.
“If a patient gets well after treatment, a physician may not feel the need to follow up with a lab test because it doesn’t have any clinical usefulness,” Dr. Altman said. “But that’s exactly the kind of data a researcher looks for.”
Perhaps the most pressing issue is patient privacy. Electronic health records must be “de-identified” before they can be used for research. That requires more than simply removing a name. Any information that might identify the patient must be excised. At the same time, researchers have to be able to tell when they’re looking at records from the same patient, which may be stored in several databases.
“One patient may be in as many as 20 different databases,” said Dr. William S. Dalton, founding director of the Personalized Medicine Institute at Moffitt Cancer Center, which is currently tracking more than 90,000 patients at 18 different sites around the country. Moffitt combines information from the electronic medical record with data from X-rays and other imaging studies, tumor tissue cultures and even genetic profiles.
“There’s an immense amount of information and different databases, all using different data dictionaries,” Dr. Dalton said. “And they don’t all agree.”
The results of this review suggest that journal endorsement of CONSORT may benefit the completeness of reporting of RCTs they publish. No evidence suggests that endorsement hinders the completeness of RCT reporting. However, despite relative improvements when CONSORT is endorsed by journals, the completeness of reporting of trials remains suboptimal. Journals are not sending a clear message about endorsement to authors submitting manuscripts for publication. As such, fidelity of endorsement as an ‘intervention’ has been weak to date. Journals need to take further action regarding their endorsement and implementation of CONSORT to facilitate accurate, transparent and complete reporting of trials.
Turner L, Shamseer L, Altman DG, Schulz KF, Moher D. Does use of the CONSORT Statement impact the completeness of reporting of randomised controlled trials published in medical journals? A Cochrane review. Syst Rev. 2012 Nov 29;1 (1):60. PMID: 23194585. (PDF)
Some PROSPERO data (of SRs registered in their system):
PROSPERO at one year: an evaluation of its utility. (PDF)
Systematic Reviews 2013, 2(4)
Booth A, Clarke M, Dooley G, Ghersi D, Moher D, Petticrew M,
Are not used very often. TBH, I skipped over the maths, but that's the conclusion (which is what we'd all expect, I suspect). When I demonstrate field tag use (specifically [title], [author], [journal], and sometimes [mesh] and [majr]), I usually do so within the context of efficiency, and often to address the question, "what are strategies for improving retrieval?" Worthwhile to point out to EBM practitioners, I think.
A study on pubmed search tag usage pattern: association rule mining of a full-day pubmed query log (PMID 23302604)
BMC Med Inform Decis Mak. 2013 Jan 9, 13(1):8
Mosa AS, Yoo I.