Using Insurance Records to Anticipate Future Medical Needs and Reduce Costs

Scientists at UCLA have determined that medical insurance company records can be used to make accurate predictions about future health conditions in their members. These predictions lead to better outcomes by enabling proactive treatments for future illness, rather than falling back on reactive treatments for each new health crisis, according to the researchers.

In recent years, “big data” analysis has been used to identify faces in photos, predict shopping tendencies, and recognize driving habits. The UCLA study published in the Journal of Biomedical Informatics illustrates that while predictive technology has made far less progress in improving medical outcomes, scientists can predict which patients are likely to be hospitalized or require specialized medicine. The highest-risk patients can then be brought in, proactively, for specialized preventative treatment.

The study used health insurance claims from thousands of insureds to predict and improve hospitalizations for patients with inflammatory bowel diseases.

“These findings are exciting because they show the potential of big data in the healthcare setting,” stated Dr. Welmoed van Deen, assistant professor of clinical medicine at USC and co-author of the study. “We [showed] it is possible to use the data to create meaningful insights.”

The researchers contend that unlike many new technologies that are expensive, this model will actually save insurance money by predicting high-cost events before they occur. Preventative treatment for at-risk patients is less expensive than preserving the status quo, which inevitably incurs expensive hospitalizations.

“This is a good example of how insurance companies, clinicians and researchers’ interests can come together to improve patient care and save money,” stated Dr. Jamie Feusner, a professor in residence in UCLA’s Department of Psychiatry.

Dr. Don Vaughn, the lead author and a computational neuroscientist at UCLA, points out that the consensus about the benefits of prevention is not new. “The problem has been in getting useful data,” he says. Medical data is fragmented among different labs, hospitals and doctors. By contrast, insurance companies—like Anthem—have a comprehensive view and keep great records, giving these companies “the potential to be heroes.”

Using claims records from more than 7,000 insureds, the research models predicted both IBD-related hospitalizations and the initiation of expensive biopharmaceuticals, with average positive predictive values of 17 percent and 11 percent, respectively — each a 200 percent improvement over chance. Further, when they used the modeling to identify four member subpopulations, the positive predictive value of predicting hospitalization increased to 20 percent.

The study’s authors maintain that their “hospitalization model, in concert with a mildly-effective interventional treatment plan” for high risk patients may “both improve patient outcomes and reduce insurance expenditures.” The say their approach provides a “roadmap for how claims data can complement traditional medical decision making with personalized, data-driven predictive medicine.”

UCLA’s Vaughn contends the results illustrate what’s possible from a cross-disciplinary team of clinical doctors, mathematicians and neuroscientists and that the program will attract interest because it “offers something for everyone: easier treatment decisions for overworked doctors, better outcomes for patients, and lower costs for insurers.”

“Predictive medicine like this has the potential to improve medical outcomes,” noted Dr. David Eagleman, a Stanford University neuroscientist and author who heads the Center for Science and Law. “Additionally, it could reduce healthcare costs and help overworked physicians. It’s a no-brainer.”