New technology could change health care system

Computer technology could potentially reduce health care costs and improve patient care in the United States, according to new research from the IU School of Informatics and Computing.

Assistant professor of computer science Kris Hauser and Ph.D. student Casey Bennett discovered that a mathematical model for predicting patient treatment options could cut health care costs in half, while doubling improved patient outcomes.

Bennett said the new data adds to previous research, which found that a computer could predict the most effective treatment options for an individual at a single point in time. A patient types in symptoms and the machine calculates a plan. This new system, he said, could look into the distant future.

“We’re thinking about the future of someone’s health status,” he said. “What we need is a system that can consider all the possibilities.”

This system, known as an artificial intelligence model, is able to search through massive amounts of complex patient data and decide a future path for treatment using complicated math, Bennett said.

“It’s real patient data that it’s working with,” he said. “It’s trying to look into the future and figure out a sequence. It can ‘think like a doctor.’”

Using data from Centerstone Research Institute, the study randomly selected 500 patient records from a pool of approximately 6,700. Bennett said he and Hauser chose to focus on patients with both chronic physical and mental illnesses.

Bennett said he selected these patients because treatment for chronic illnesses such as diabetes and heart disease makes up 90 percent of all health care costs in the U.S.

These costs are rising.

According to a 2007 report from the Congressional Budget Office, total health care costs are expected to reach about 20 percent of the gross domestic product by 2016. Federal spending on Medicare and Medicaid alone is projected to reach about 20 percent of the GDP by 2050.

“If we don’t do something, we will bankrupt this country,” Bennett said. “These kinds of technologies could be a solution.”

The research compared a “raw effect” model to the AI model. Bennett said the simpler raw effect model replicated a human doctor’s decisions, and its cost was calculated using a treatment as usual basis. In this case, Bennett said he calculated the costs of eight outpatient visits during a three to four month period.

“It’s based on what providers do with real patients when they see real doctors,” Bennett said.

Bennett and Hauser found that the usual cost of treatment was $497 compared to the AI machine’s cost of $189.

Bennett said another issue the research is attempting to improve is patient
outcomes.

On a first visit, patients receive a correct diagnosis and treatment less than 50 percent of the time, according to a 2003 New England Journal of Medicine article.

The AI model increased patient outcomes by 30 to 35 percent, Bennett said in a press release. He said with more research, the outcomes could improve by 50 percent.

He said the computer system could work for any disease, even with incomplete patient data.

“It doesn’t operate based on facts,” Bennett said. “It operates on the beliefs of what the patient’s status is. This allows us to figure out what works for each individual person.”

Pete Grogg, interim director of the IU Health Center, said the AI model would make a good supplement to their already existing health care technology.

“It could provide support back to the providers,” he said. “You’re not always able to predict what the problem would be.”

The IU Health Center currently uses electronic medical records, which generate all patient information — including symptoms, labs, diagnoses and treatment plans — into a large database, Grogg said.

He said they also use electronic systems for sending pharmacy and x-ray information across the state or country and back.

The AI, he said, could avoid unnecessary treatments or tests, which will save patients money.

“Certain symptoms have to be present before we can order a test or prescribe a medicine,” he said. “The way to control costs is to provide necessary care.”

He said the technology could also save patients money by helping them treat minor problems on their own.

“There are some things you don’t need to see a doctor for,” Grogg said.
Bennett said he is not trying to replace human doctors with machines, but wants to maximize the potential of both.

“You still need people to be comforting,” he said. “This will free doctors up to do what they do best. That will be beneficial for everyone.”

Grogg said although the concept is great, it’s too early to tell if the AI will truly function well.

“I think it’s something that could really improve the health care system,” he said. “My only concern is how well it can adapt.”