Study: Computer Model Could Curb Costs, Improve Patient Care

Study Details

For the study, researchers at Indiana University randomly selected 500 patients from the Centerstone Research Institute database. The database houses clinical data, demographics and other information on 6,700 patients, 60% to 70% of whom had both clinical depression and a concurrent physical disorder such as diabetes or hypertension.

Researchers then used the selected patients’ data to compare actual doctor performance and patient outcomes against simulated outcomes generated by an artificial intelligence framework (Mearian, Computerworld, 2/12).

Framework Details

The framework combines two mathematical modeling formulas, known as Markov Decision Processes and Dynamic Decision Networks. The framework uses the formulas and sequential decisionmaking to:

• Simulate the effects of a variety of treatment paths;
• Maintain predictions about the status of a patient’s health, even when certain data are unavailable or uncertain; and
• Refine treatment plans as new information becomes available (Medical News Today, 2/13).

The computer model is not specific to a particular disease, so it could be used to analyze any diagnosis or disorder.

Study Findings

By running the patient data through the machine-learning algorithm, researchers found that the artificial intelligence framework could have generated a 30% to 35% increase in positive patient outcomes.

Casey Bennett — a study co-author and PhD student at Indiana University — said, “And we determined that tweaking certain model parameters could enhance the outcome advantage to about 50% more improvement at about half the cost.”

For example, the cost of using the computer model to diagnose and treat a patient is about $189, compared with a cost of about $497 for usual treatment (Computerworld, 2/12).

Comments on Study

Kris Hauser — a study co-author and assistant professor of computer science at Indiana University — said, “Modeling lets us see more possibility out to a further point, which is something that is hard for a doctor to do.” Hauser added that physicians “just don’t have all of that information available to them.”

Bennett said, “[W]e believe that the most effective long-term path could be combining artificial intelligence with human clinicians. Let humans do what they do well, and let machines do what they do well. In the end, we may maximize the potential of both” (Indiana University release, 2/11).