Bridging the Implementation Gap of Artificial Intelligence in Healthcare

Author(s): Eera Bhatt

Originally published on Towards AI.

Each year, we spend so much time and money developing new machine learning models, but most of them never get used in a practical setting. Sadly, this issue is even worse in the healthcare industry.

A.I. in medicine. Because of COVID-19, a lot of us know about AI and are also familiar with its applications in medicine, but let’s summarize them for anyone who needs it. Here are just a few benefits of using AI in healthcare:

• Improves outcomes for patients & clinical teams.
• Lowers healthcare costs. (Although money does have to be spent on funding to research machine learning.)
• Helps the population’s health overall by connecting long-term patients with screening and therapy that can help them heal. Also, promotes information about disease prevention online (think COVID-19 CDC guidelines).

Implementation gap. But again, so many machine learning models are developed each year that don’t get used in a practical setting. According to Stanford, healthcare gained $6.1 billion for A.I. investments in 2022, but the industry suffered to put the A.I. into actual practice. Lots of professionals call this the “implementation gap.”

Here’s the brutally honest truth: It doesn’t matter how much clinical A.I. research is done unless we translate it into effective decisions that can help patients. This can’t be fixed until we reflect on our current strategies and collaborate across disciplines.

But why is this an issue? For any industry, there are several barriers between developing A.I. and putting it into practice, but we’ll touch on three of them.

• Uncertainty. Healthcare workers who could be using the A.I. don’t see its benefits. When there are communication gaps regarding a model, there can be uncertainty about its true accuracy and limitations.
• Time investment. It takes time and energy from IT workers to translate info from models to computers. Also, the whole process of gathering real-time patient data is very time-consuming.
• Change management issues. Oftentimes, healthcare leaders can’t properly convince users that the models work well, let alone teach them how to use the models, so their medical workers resist using the A.I. completely.

That’s cool to know, but how do we fix these issues?

Bridging the implementation gap for medicine. For the healthcare industry in particular, the first step to improvement is to create leaders who are open-minded enough to approach clinical A.I. with curiosity, not resistance.

Thankfully, since the COVID lockdown, we’ve seen lots of top institutions develop online courses to help people reach their professional goals.

I know what you’re thinking, and no, these courses are not just for computer programmers!

Courses to bridge the knowledge gap. Several professors and healthcare industry leaders have worked together to create courses about implementing clinical A.I. for patients effectively. Here are just five options to try out (in no particular order):

• Artificial Intelligence in Medicine (University of Illinois Urbana Champaign): Graduate program that gives healthcare professionals an idea of how specific machine learning models can be implemented in real medical scenarios.
• A.I. in Healthcare Specialization Coursera (Stanford University): Five courses for anyone to learn when machine learning is appropriate in the healthcare industry and how it can be implemented. Addresses patient safety as well.
• Artificial Intelligence in Healthcare: Fundamentals and Applications (Massachusetts Institute of Technology): Course for healthcare leaders, tech consultants, and entrepreneurs to navigate an era of digital medicine and to understand A.I.’s potential benefits and limitations in healthcare.
• A.I. in Healthcare: From Strategies to Implementation (Harvard University): Course that teaches students to translate machine learning into a tangible impact in healthcare by applying specific models to medical cases and understanding each type of model’s characteristics. Includes a capstone project for a stronger grasp of content.
• Data Science for A.I. in Healthcare (Johns Hopkins University): Course that trains healthcare workers and engineers to interpret machine learning research in medicine. Clarifies when and how machine learning can help patients and teaches healthcare providers how to make decisions based on this knowledge.

Again, collaboration is key here!

It’s time for a lot of us to stop thinking of medicine and computer science as two completely unrelated fields. The reality is that so much money is being funded by the government into these machine learning research projects by top universities, but a lot of these projects are yet to reach patients at all.

Collaboration and open-mindedness. So many computer science students spend years learning the technical and analytical skills required to develop helpful models. They go as far as publishing their work in top journals and sharing it at conferences to help it reach medical professionals.

But so much of this work isn’t used either because healthcare workers don’t have the resources to implement clinical A.I. in their work, or they are too skeptical to learn how to.

For computer scientists. If you are a machine learning worker or researcher who develops clinical A.I. models, be sure to share your work through publications and even web applications when appropriate! Also, take a look at the courses above if you authentically care about your clinical A.I. work making a tangible impact on patients.

Remember that the way we communicate about our work can have a huge effect on how it gets used. Articulate your model’s accuracy clearly so that healthcare workers understand how your work can help them out.

For healthcare workers. If you are a healthcare professional who is new to A.I.’s concepts, consider taking one of the courses above, or find your own practical way to learn about machine learning in your industry. Remember, it’s not like you are committing to any organizational change just by learning about it!

Further Reading:

[1] AI in health care: From strategies to implementation Corporate Learning at HMS. Available at: https://corporatelearning.hms.harvard.edu/individuals/executive-education/ai-health-care-strategies-implementation

[2] AI in Healthcare Specialization Coursera. Available at: https://shorturl.at/UwKHG

[3] AI in healthcare: The future of patient care and health management (2024) Mayo Clinic. Available at: https://mcpress.mayoclinic.org/healthy-aging/ai-in-healthcare-the-future-of-patient-care-and-health-management/

[4] Artificial Intelligence in Healthcare with MIT xPRO: Fundamentals and Applications of AI in Healthcare online course Artificial Intelligence in Healthcare with MIT xPRO | Fundamentals and Applications of AI in Healthcare Online Course. Available at: https://shorturl.at/iLz8Y

[5] Artificial Intelligence in Medicine (no date) Artificial Intelligence in Medicine. Available at: https://illinois.catalog.instructure.com/browse/grainger-college-of-engineering/ai-med/courses/ai-meds

[6] Data Science for AI in Healthcare (2020) Johns Hopkins Malone Center for Engineering in Healthcare. Available at: https://malonecenter.jhu.edu/ai-course/

[7] Johnson, M. (2024) The implementation gap: Using artificial intelligence to help patients in clinical practices, Executive and Continuing Professional Education. Available at: https://www.hsph.harvard.edu/ecpe/implementation-gap-artificial-intelligence-help-patients-clinical-practices/

[8] Overgaard, S.M. et al. (2023) ‘Implementing Quality Management Systems to Close the AI Translation Gap and Facilitate Safe, ethical, and Effective Health AI Solutions’, npj Digital Medicine, 6(1). doi:10.1038/s41746–023–00968–8.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI