Patients and providers generate incredible amounts of data every day. Can it be put to better use in individualizing care? Endocrine News looks out how using information that already exists could dramatically improve patient care as well as scientific research.
Love ’em or hate ’em, there could be gold in those electronic health records. The University of California Health System is saving $1 million a year on a single diabetes drug — simply by monitoring the records among some 200,000 UC employees to ensure that physicians prescribe the generic form of metformin instead of the brand name.
And that’s just a taste of the promise of harnessing big data, according to pediatric endocrinologist Atul J. Butte, MD, PhD, the Priscilla Chan and Mark Zuckerberg Distinguished Professor at the University of California, San Francisco, and chief data scientist for the UC Health System.
The concerted effort to steer clinicians toward the cheaper alternative manifests in different actions. At one UC campus, administrators tweaked the EPIC records system to increase the difficulty of ordering the brand name. Another approach is that when the medical records alert pharmacists of the prescriptions, the clinicians might receive telephone calls asking for clarification of their practice patterns. “It’s known as the Hawthorne effect,” Butte tells Endocrine News. “Once people know you are looking, all of a sudden, things start to improve.”
Electronic health records make this kind of accountability possible, Butte says: “The U.S. health system is making endocrinologists document in electronic medical records everything they are doing, which some doctors love and many doctors hate. So the real question is: What can we now do with all that data? We have spent hundreds of billions of dollars putting these systems in. If we don’t use this data to improve the practice of medicine, it’s going to be a national tragedy.”
Variations in Diabetes Drug Regimens
Beyond the first diabetes drug, Butte’s team is examining prescribing patterns among the myriad diabetes drugs, combinations of drugs, and order in which clinicians choose to prescribe them. One striking preliminary finding is that among 71,000 diabetes patients in the UC Health System, there are more than 6,500 “unique medication trajectories.”
“Think about how many different ways we have to practice medicine,” Butte says. “For instance, when we start someone on diabetes medications, think about how many different ways we have of starting. Consider how many ways we have to go to the next one and the next one. And this is for a disease where we actually have consensus treatment guidelines. There is a lot of variability in practice, and one might question how much of that variability is unnecessary, or unnecessarily expensive. Maybe a few of these treatment regimens are better than the others, and it is about time we start to study that.”
“The U.S. health system is making endocrinologists document in electronic medical records everything they are doing, which some doctors love and many doctors hate. So the real question is: What can we now do with all that data? We have spent hundreds of billions of dollars putting these systems in. If we don’t use this data to improve the practice of medicine, it’s going to be a national tragedy.” – Atul J. Butte, MD, PhD, the Priscilla Chan and Mark Zuckerberg Distinguished Professor, University of California, San Francisco; chief data scientist, UC Health System
Butte likens treating a patient to playing a game of chess: the physician makes a move, the patient’s disease makes a counter-move, and it is the physician’s turn to move again. Clinicians should be better at mapping out the strategies for their moves and anticipating future countermoves. He hopes that deeper analyses of the data on the individual choices being made and how they are turning out can provide comparative effectiveness information on different approaches and regimens.
New Tests from Public Data
Butte’s current position at a large health system has him focused on electronic health records, but his previous research dealt with harnessing public molecular data available on the internet to answer new questions. For example, when he was at Stanford University, he received his first large National Institutes of Health grant to gather RNA expression data sets across many diseases, and he used the information to help create a prototype diagnostic test for preeclampsia. Although it is a large cause of maternal and fetal death, Butte says that the diagnostic testing for preeclampsia “is ancient.”
Butte’s team wrote programs to screen the publicly available data from 266 experiments for commonalities to identify potential markers that could be used in a blood test. After publishing their results, Butte and his colleagues tapped into Silicon Valley investors who supplied more than $2 million in start-up funds for a company to develop the blood test, which has advanced into clinical trials. “This is the new way science has to continue out of our labs,” he says.
New Uses for Old Drugs
Given that it costs billions of dollars to develop new drugs, Butte has also pioneered techniques of using large sets of existing molecular data to circumvent these costs by repositioning — finding new uses for — old drugs. There are famous examples of this process happening serendipitously: The antihypertensive/cardiac drugs minoxidil and sildenafil found much larger markets when they were marketed for their “side effects” of causing hair growth and combatting erectile dysfunction, respectively. “Instead of finding these new uses by accident, how about we try find them on purpose by using public data? The raw data of the best scientists in the world is sitting on the internet,” Butte says.
Butte’s teams have mined researchers’ data to match several drugs with possible new applications. For example, they showed that the anti-epileptic drug topiramate is effective against inflammatory bowel disease in a rat model. They found that the tapeworm medication niclosamide is effective treating liver cancer in a mouse model. Start-up companies are continuing the investigations of both these new uses.
“Instead of finding these new uses by accident, how about we try find them on purpose by using public data? The raw data of the best scientists in the world is sitting on the internet.” – Atul J. Butte, MD, PhD, the Priscilla Chan and Mark Zuckerberg Distinguished Professor, University of California, San Francisco; chief data scientist, UC Health System
Butte sees another huge opportunity on the horizon from the raw data from clinical trials: “Half of them fail, and when they fail, we don’t even write papers about them and release the data. That is going to change.” The European Medicines Agency is requiring researchers to publish the raw data from all clinical trials — and the availability of this data should open new avenues of research.
— Seaborg is a freelance writer based in Charlottesville, Va. He wrote about the perceived diabetes risks for patients prescribed statins in the February issue.