Just what the computer ordered
Need help figuring out the right treatment plan for an injured worker? There may be an app for that.
Researchers at the University of Alberta have developed a prototype for a web-based tool that would provide health care professionals with personalized recommendations for the type of rehabilitation program a worker needs to get healthy and return to the job.
For the project, summarized in the Journal of Occupational Rehabilitation, the researchers used a form of artificial intelligence called “machine learning” to analyze injury, treatment and outcome data for 8,611 injured workers who were being assessed for their capacity to return to work. With machine learning, a computer is able to learn from data.
When put to the test, the tool’s recommendations were around 85 percent accurate, which was slightly better than actual physical and occupational therapists.
Lead author Douglas Gross, associate professor of physical therapy at the University of Alberta, used to work as a physical therapist, and he said one of the main challenges is picking the optimal rehabilitation program for injured workers.
As for why the tool beat clinicians, Gross pointed to the human element. “We’re all kind of subject to biases and errors and making mistakes,” he told me. “The computer though, it’s learning from the positive case, and once it has a model, it doesn’t make mistakes. It does the same thing every time.”
The tool is still in the testing phase, but even when ready for use, Gross said it’s supposed to supplement – not be a substitute for – the judgment of a health care professional. A clinician would perform a worker assessment, enter the information into the online program and hit a button for recommendations on the type of rehabilitation program (e.g., worksite-based program or complex pain management program). The tool offers options and evidence, but it would be up to the health care professional to interpret the results and make the final decision on the patient’s care.
“One of the things we were very careful with in developing it was to make sure this was really a resource,” Gross said. “It’s not taking away the decision-making responsibility of the health care professional.”
The opinions expressed in "On Research" do not necessarily reflect those of the National Safety Council or affiliated local Chapters.