The Google Flu Trends project is a great example of not only the potential use for big data at a public health level, but also some of the challenges that come with it.
When armed with de-identified health data from multiple sources – rather than the smaller pool of patients an institution sees within its own walls – public health researchers can run disease-wide studies on bigger and more varied groups of patients (for a look at other examples of how different types of data analyses are being used in public health, see Stephanie Baum’s story from last month). That’s great except that, as demonstrated by Google Flu Trends, data without context doesn’t always lead to accurate assessments and predictions.
That’s just one of the challenges facing public health researchers as they leverage the growing mass of healthcare data to better understand and predict disease. In a public symposium on the use of electronic medical records for non-treatment purposes hosted by Case Western Reserve University on Friday, researchers gathered to shared thoughts on some of the most pressing issues.
When Investment Rhymes with Canada
Canada has a proud history of achievement in the areas of science and technology, and the field of biomanufacturing and life sciences is no exception.
Legal framework
Fitting with the primary topic of the event, the speakers noted wide gaps in the legal framework for data privacy, security and ownership. “The hugest opportunity is to get some of the legal paradigm and structure to support the growing data,” said Dr. David Kaelber, a primary care pediatrician and chief medical informatics officer at MetroHealth in Cleveland. “We don’t yet understand all the risk of all this data being out there.”
While aggregated, de-identified patient health information falls outside of HIPAA guidelines, questions remain in how to get data to that point. One of the big questions is, who owns the data? Physicians own the EMR that’s created for a patient, but what about the data that exists within it? Should it be owned by anyone?
The assumption of patient ownership causes some bumps in the road for public health researchers, as patients may be willing to share some parts of their data but are not willing to share other pieces of related data. “Patients do really want to be in charge of what gets shared, and that’s been a bit of a challenge,” said Anil Jain, a former staff physician at Cleveland Clinic who is now chief medical information officer at big data company Explorys.
What’s being documented
There are plenty of technical challenges, as well. “When I’m collecting this data as a physician, I’m not doing it so that someone else can do a study later; I’m doing it because I have a patient and I have 10 minutes to figure out what’s going on,” Jain said.
Thus, much of the information that researchers want is not stored or collected in electronic medical records or clinical notes for routine care, Kaelber said. That may include minor side effects, a patient’s home environment or dietary habits, for example.
On top of that is the issue of translating clinical notes into usable data. “People are working on natural language processing but, at least in my view, in most cases it certainly hasn’t reached prime time,” Kaelber said.
Standardization
Even when that kind of valuable data is collected, it isn’t always done in a way that makes it easy to aggregate. “There’s 10 different ways on one of our customers’ sites to enter the fact that a patient has quit smoking,” Jain said. “Well, if there are 10 different ways, how would anyone know whether a patient quit smoking or whether it just didn’t get that piece of data?”
While the industry has adopted data standards for certain kinds of health and demographics data, there’s still much work to be done in creating a common framework.
Personnel
EMRs and other health IT are just tools and alone are not enough to have a positive or negative impact. Their real value is determined by the people using them. “Giving a stethoscope to a clinician doesn’t make them any better at being able to detect murmurs unless you teach them what to listen for and how to distinguish between normal and abnormal. It’s the same thing with health IT,” Jain said. Clinicians need proper training on how to use new IT tools like they would anything else.
And, the industry needs more people who understanding how to analyze big data, especially when it’s been de-identified. “These data scientists are really rare to find,” Jain said.
