In 1989’s film Field of Dreams, the character Shoeless Joe Jackson told Ray Kinsella, “If you build it, he will come.” That sentiment worked in the movie but doesn’t always carry over in real life, particularly in medical and health technology. We write daily about scientists, engineers, and researchers hard at work developing new technology. However, brilliant ideas and hard work do not guarantee acceptance by healthcare professionals, patients, hospitals, or insurance companies. A group of Toronto, Canada educators, researchers, and healthcare professionals recently published a study in the Journal of Medical Internet Research that surveyed the issues and obstacles in implementing machine learning applications in medicine.
We’ve written about machine learning applications that identify suicidal adolescents, uncover hidden diabetes insights, predict drug approvals and psychosis, and much more. We’ve looked at forecasts of AI application for healthcare. Previously, however, we’ve not written about implementation studies of machine learning applications in medicine.
The Toronto team acknowledged the rapid growth in AI apps but found little work that addressed implementation issues. The team studied machine learning as a general purpose technology (GPT): one that can perform diverse functions for many purposes in health care. The study breaks down implementation hurdles in two broad machine language application categories: decision-support and automation. AI makes predictions in both categories but automation applications go further and take independent action.
In the end, the researchers have a positive outlook for machine learning in healthcare. Success will depend on the support from patients, the public, and a wide range of public and private healthcare stakeholders. Tools that are more user-friendly will also help speed future adoption.
