While an extensive number of machine-learning models are developed, many of them are never put into use.
“All those amazing models could be helping patients,” says Santiago Romero-Brufau, M.D., Ph.D., Program Director of Implementing Health Care AI into Clinical Practice and Adjunct Assistant Professor in the Department of Biostatistics at Harvard T. H. Chan School of Public Health. “And if they’re never used, they’ll never help anyone.”
Romero-Brufau refers to the space between what is developed and what is used as an implementation gap, and notes that it’s particularly visible in the health care realm. He cites a Stanford University study that notes while health care attracted $6.1 billion in investments in artificial intelligence in 2022, the industry is at the bottom when it comes to embedding AI into practice—coming in at only 12%, compared to the average of 19%, or 31% for retail.
Navigating the distance between what is developed and what is put into practice may be challenging, but it is possible, says Romero-Brufau.
“There are a lot of things that need to happen from when you have a model that allows someone to be able to predict something about a patient to it actually being used in clinical practice, and impacting real patients,” he continued.
What are the Barriers to Bringing AI Models into Practice?
Several barriers play a role in restricting AI’s implementation in the clinical setting. However, three categories tend to be the most significant.
“One of the biggest ones is that the physicians that would use it don’t see the benefit,” says Romero-Brufau, who notes that this could be caused by issues ranging from poorly defined problems during the development of the models to the insufficient way the model’s accuracy is communicated during its implementation. As a result, the physician’s problems are not addressed.
Another barrier is difficulty with information technology. Translating the information from a model to the computer, as well as acquiring data from patients in real-time and being able to show those results to physicians, can be a challenge that takes both time and information technology expertise to work out.
The third barrier is change management issues. Often, the troubled implementation is the result of a failure to properly gain users and explain to them why and how AI models can be used. As a result, people ignore their potential. While other barriers can come into play, such as resistance from the medical establishment to embrace AI, these three commonly exacerbate the implementation gap.
Who Needs to Further Their Knowledge?
When it comes to filling leadership positions, particularly those with an AI, digital health, or technological spin, organizations are eager to recruit professionals who are skilled in applying this type of technology.
“That’s the key skill that they are looking for,” says Romero-Brufau. “How can they translate the great AI or machine-learning work into value for patients or the institution?”
Specifically designed for professionals who are responsible for their organization’s implementation of AI, the Implementing Health Care AI into Clinical Practice program pulls from diverse career paths, including:
- Clinician
- Physician Leader
- Chief Executive Officer
- Chief Information Officer
- Chief Innovation Officer
- Chief Medical Informatics Officer
- Chief Medical Officer
- Data Scientist
- Director
- Innovation Specialist
- Implementation Specialist
- Product Manager
“Anyone in any sort of leadership position at a health care institution that has to decide what vendors to use for different AI tools would need to know about this,” says Romero-Brufau, adding that those who hope to develop health care models and sell their solution also needs to understand, “how these things fit together.”
Artificial intelligence and machine learning are powerful tools with unlimited capability. However, without the education that teaches professionals to apply these technologies, clinicians and executives fall victim to the implementation gap. Being equipped with the knowledge and skills to implement AI solutions is a step toward transforming the future of patient care.
Harvard T.H. Chan School of Public Health offers Implementing Health Care AI into Clinical Practice, which focuses on providing clinicians and executives with the knowledge and skills to ensure AI solutions are successfully implemented in the clinical setting.