There’s an important effort underway among health care data experts to enable clinicians and medical researchers to share the same data for analytics to improve patient outcomes.
At issue is the structure of electronic health records (EHR) that were originally designed to be used in day-to-day patient care and are not set up to handle much bulkier data types such as X-ray images and genomic tests.
As a recent editorial in the Journal of the American Medical Association notes, a critical shortcoming of EHRs of today is that despite their usefulness they can’t hold and analyze much of the ancillary data that health care experts need in a timely fashion. Ancillary data could include laboratory and imaging test results. (See “Why Digital Medical Records Can’t Hold an X-Ray,” below.)
This condition persists even though available technology is already able to gather some of this information.
“EHRs were never designed to develop insights on large-scale sets of data. They help to collect information that can address inefficiencies of paper records and provide basic error-checking when you saw patients,” says Dr. Graham Hughes, chief medical officer at SAS for the SAS Center for Health Analytics and Insights. Hughes is a developmental neurobiologist and a leader in health informatics.
Addressing this problem is the focus of the Electronic Medical Records and Genomics (eMERGE) network, funded by the National Human Genome Research Institute, a division of the National Institutes of Health. It is a bioinformatics program established in 2007 with seven facilities to develop, disseminate, and apply approaches to research that result from the mapping of the human genome. The program’s coordinating center is located at Vanderbilt University.
This national consortium of scientists and organizations, using supercomputer systems, so far “has captured data sets from 56,000 individuals,” says Dr. Rongling Li, a genetic epidemiologist at the National Human Genome Research Institute, and eMERGE’s program director. Multiply 3 billion pairs of data for each of those 56,000 people and, she notes, “You can see what we mean by really big data.”
Experts say such a program is a way to move beyond the limitations of medical records.
“Even when EHRs advance, unless there are fundamental changes, they will not be able to handle large volumes of genetic data. We need to build a more fluid system,” says Dr. Justin Starren, chief of the Division of Health and Biomedical Informatics at Chicago’s Northwestern University., adding: “We could wait for the mainstream EHR vendors to solve the issue in the near term, or simply try to stuff the genomic data into the current system.”
Neither of those options seems likely, however. Hughes says he doesn’t think EHRs are the answer.
“You don’t need all [the genetic information] stored in the patient’s health record,” he says. “What you do need are new algorithms that will teach a system to say, ‘I know that I need to look at this particular gene…I know that’s a variant.’ Then signals in the EHR would provide some guidance to the doctor as to the implication of what impact these variants could have on that patient’s care.”
The Potential for Data-Driven Benefits
Discussions about data-driven health care improvements have been going on for years in political and public policy circles, not just the medical field. And they continue among experts working to come up with new data models for patient records.
Crunching vast repositories of genomic data has enormous potential for saving lives. Starren offers this example of a maternity patient:
“There was a woman who was on codeine after her delivery and, unfortunately, turned out be among the approximately 6 percent of the population that doesn’t metabolize codeine efficiently. She ended up retaining so much of it in her breast milk that her baby’s respirations were depressed and the child died.”
If there had been an easier way to analyze her gene sequence to show whether this woman was one of these “high metabolizers” during her pregnancy, and there was a process in place to flag the doctor about the variant, either the mother wouldn’t have received codeine, or wouldn’t have initially breastfed her baby.
Hughes says this kind of preventive scenario is not far-fetched. “We can already find data that allows us to suggest very specialized patterns of treatment (for example, surgery, a specific drug, exercise), determining first what’s best for a specific group overall—like 64-year-old black women—and then eventually for individuals within that group,” says Hughes. “The technology is here today, [just] not used widely.”
Once these analytics provide more easily read data, health care economics will also benefit, Li says: “When we get the right diagnosis, and provide the right dose of the right medicine, we’ll save money.”
Such data analysis might eventually help avoid malpractice suits. “It would act as smart surveillance that can troll through this information 24/7 looking for warnings, information that your care team is too harried to look for,” says Hughes.
Analytics could also lead to personalized medicine. “Think about the number of drugs people over 65 take, and how many are necessitated by a genetic influence, like cholesterol,” says Starren. “Where we’re going over the next 10 years is not just checking your blood pressure at a pharmacy. You’ll have your entire genome sequenced and your risks will be sent on to your doctor to guide your individual treatment,” Hughes says.
On the downside, algorithms allowing this kind of genetic sifting raise other issues, such as privacy and ethics. “We haven’t figured out all the unexpected consequences, in areas like insurance or employment, when each individual can be flagged as carrying ‘dangerous’ genes,” says Starren.
In the meantime, though, Starren says, “I think one of the lessons behind this is that we traditionally think of research and clinical care as two separate worlds that have nothing to do with each other.
But as medicine becomes recognized as a big data problem, the researchers and the clinical IT people will see the need to work much more closely together.”
He adds, “If you’re going to be a scientist in this century, you’ll have to follow algorithms.”
Wendy Meyeroff, of WM Medical Communications, is an experienced freelance writer based in Baltimore who specializes in health care and IT topics.
Home page illustration of chromosomes of the human genome, via National Human Genome Research Institute.