diagonal background pattern in shades of blue

Breakout Innovations: Mining Health Records

By Ted Alcorn

In what is commonly described as the founding act of public health science, John Snow noticed a spatial pattern to the casualties of an 1854 cholera outbreak in London and identified the source of exposure: the Broad Street pump. Today, Brian Schwartz, MD, MS, a professor in Environmental Health Sciences, is exploring the environmental influences on health at a level of complexity and sophistication that Snow never would have dreamt possible.

Schwartz’s partner in this work is the Geisinger Health System, which provides primary care services to more than 400,000 people in Pennsylvania. The electronic health records (EHR) of those individuals who do not opt out are made available for research. While public discussion has focused on how EHRs might reduce medical errors and improve diagnoses, Schwartz saw an opportunity to use the health system data to evaluate environmental exposures in daily life.

Once they are coded according to a patient’s locale, (geocoded), EHRs can reveal exposure to a spectrum of environmental variables. The features incorporated into various models range from playgrounds to abandoned coal mines to animal feeding operations. Physicians historically have done a poor job connecting their patients’ illnesses to these kinds of influences, Schwartz points out. “For a health system to actually ask questions about how the community might be contributing to health problems, I think, is unique and ahead of the curve,” he says.

Detailed data from the EHRs are a huge leap forward from self-reported health data, the previous mainstay for this kind of research. The Geisinger population is also several orders of magnitude larger than earlier sampling methods—such as random-digit dialing—would have allowed. “A lot of epidemiological modeling to date has been kind of deterministic, and one risk factor at a time,” says Schwartz.

In contrast, by analyzing a large sample with numerous variables, Schwartz and his colleagues can use complex dynamic systems modeling to explore manifold interconnections, feedbacks and unexpected properties. They are researching or initiating projects on diabetes, obesity, asthma and MRSA, among other topics.

“As more and more health care is captured by EHRs, there’s going to be an increasing ability for this kind of research,” says Schwartz. “So I actually think we’re early in this period. It’s a huge growth area, and it’s going to continue to grow for the foreseeable future.”