By Samuel Greengard
Few fields benefit more from advanced analytics than health care. Somewhere between the doctor’s office and the somewhat abstract world of epidemiological data is the real world of patients and outcomes.
“Today, in the outpatient setting, 25 percent of adverse affects are caused by poor or inadequate follow up of abnormal test results,” says Carlton Moore, associate professor of medicine at the University of North Carolina School of Medicine. In the past, “There had been no simple way to manage the process and improve outcomes.”
Moore and his colleagues at UNC Health Care, a Chapel Hill-based not-for-profit system owned by the state of North Carolina, are working to change that. Faced with medical literature that was doubling every five years and a growing pile of unstructured data—including notes, registration forms, discharge summaries, phone calls and a variety of documents— the medical center turned to an IBM Smarter Care solution. The system, similar to IBM’s Watson technology, converts data streams into actionable text that allows clinicians to access and analyze critical patient information via natural-language processing.
That capability is changing the face of medicine. With the ability to see and interpret both structured and unstructured data, doctors and staff can identify high-risk patients, understand in context what is leading to hospitalization and take preventative action.
“Using conventional methods, too many patients fall through the cracks,” Moore says. “A lot of unstructured data in medical records and other places is lost. Natural-language processing can spot things that people miss.”
About two years ago, the facility began using the technology to sort through mammography reports. “They are stored as text in electronic health records,” he explains, “and the physician has to go in and find the reference to the abnormality in order to take any action.”
After testing the Smarter Care text-extraction system with 500 files, UNC Health Care found that it was 100 percent accurate in identifying the need for follow-up visits. As a result, it is now expanding the use of the technology into a number of other key areas. They include timely follow-up for other abnormal cancer screening results, including colonoscopies; reducing costly 30-day re-admissions by identifying predictors of risks (new Medicare rules penalize facilities with high readmission rates); and engaging patients more effectively by simplifying medical data.
The health care facility also uses IBM Content Analytics to transform clinical data from electronic medical records into a simpler format that patients can better understand. This helps patients become more proactive about their own care.
Moore is planning to add other capabilities and build a more “systematic mechanism” for spotting abnormalities and generating alerts. He is also hoping to extend the analytics capabilities in the future. For instance, the technology could help physicians better understand factors that contribute to patient noncompliance with medications and recognize how different behaviors play a role in outcomes.
“We need to use every tool possible to reduce re-admissions and improve health care processes and decision making,” Moore emphasizes. “Natural-language processing is helping us transform processes and make more informed decisions.”