Speaker: Muhammad Mamdani, PhD
Reported by: Vjekoslav Hlede, PhD, CE News Editorial Team
Recent rapid advances in AI have struck the CME/CPD community with a set of challenges and opportunities. What to do? What is next? How will we handle the change? — are just some of the questions that come to our mind.
In this masterful lecture on Applied AI in Health—What the Education Field Needs to Know, Dr. Mamdani provided insight that can help us address the emerging questions. The lecture included a set of convincing examples of how AI improves healthcare delivery, reflected on what AI is and the complexity of its implementation, and discussed how the CPD community can better lead the digital transformation in the AI-enhanced world.
To listen to the entirety of this fascinating talk, click on the YouTube video on the left or go to: https://www.youtube.com/watch?v=iOelbxEXClk. The introduction starts at 6:40.
As a leader who contributed to the creation of 50+ healthcare AI-enhanced apps, a university professor, director of Temerty Centre for Artificial Intelligence Research and Education in Medicine, and Vice President for Data Science and Advanced Analytics, Unity Health Toronto, Dr. Mamdani, is one of the leading global authorities in Healthcare AI area.
Unity Health Toronto, the organization where Dr. Mamdani works, is the first healthcare system in Canada, and probably globally, to define Artificial Intelligence as one of the core strategic pillars. That official recognition of the importance of AI assures that people and resources are assigned and needed organizational and cultural changes are supported. Ultimately, that creates a context where Unity Health Toronto has a team of 30+ professionals working solely on AI solutions, where leadership is actively involved in digital transformation. Therefore, it is no surprise that Unity Health Toronto is recognized as one of the global leaders in this area.
Three messages shape this presentation:
- AI and healthcare practice: AI saves lives and reduces administrative burdens
- Nature of AI: AI is a mechanism to understand complex relationships and data. Therefore, it can help us better handle the increased complexity of healthcare practice and CPD
- AI is important for CPD professionals because AI impacts learners, teachers, and learning systems (how CPD is delivered), and we need AI-focused learning programs/courses for the healthcare workforce
AI saves lives and reduces administrative burden
Dr. Mamdani provides two convincing examples of how AI is used to help healthcare teams reduce mortality, and one example of how nursing team scheduling can be improved while reducing workload.
AI-enhanced radiology tracking board, in a matter of seconds, analyses scans for possible intracranial hemorrhages and promptly communicates findings to all involved teams. Without any delay, radiologists are informed to review scans of the affected patient, and the care team and surgical team are pinged that the patient may need urgent attention.
CHARTWatch an early-warning AI program – enables 20% reduction in unplanned mortality. The system continuously monitors available patient data and calculates the risk of dying. It informs the medical, palliative, and critical care teams as soon as it recognizes a high-risk patient (Grant, 2023). That significantly increased the ability to deliver needed care in a timely and coordinated manner, reducing unplanned mortality by 20%. Since doctors get advanced warnings about possible worsening situations, they can talk with patients before their situation worsens and explain the options and what will happen. Ultimately, patients are prepared for upcoming treatment. More
An AI-enhanced app that improved scheduling and teamwork. St. Michael emergency department has numerous rules to ensure that (1) in all areas, they have skilled nursing staff, (2) less experienced staff is paired with more experienced staff, (3) staff has learning opportunities, and so on. The emergency department is a very dynamic system, so the charge nurse would spend 90 minutes each day preparing the schedule, and the clerk would spend two hours. Despite that huge time investment, 20% of the time schedules did not follow the rules properly. The AI-enabled app almost entirely eliminated the time needed for scheduling while reducing errors from 20% to 5%. Ultimately, the nursing team is better organized and has more time to do their core task – help patients and save lives.
Ability to learn from huge amounts of data. An important mechanism that enables AI to be successful is the ability to handle and learn from a huge amount of data. During the last 15 years, Unity Health had over 24 million ambulatory visits and 2.5 million emergency department visits.
Implementation of AI in Healthcare Context
Implementation of AI is a practical, hands-on experience, Dr. Mamdani argues. While there is a wealth of opportunity for academic and theoretical work, Dr. Mamdani is primarily interested in addressing real problems in clinical practice. Therefore, Unity Health Toronto does not partner with researchers or data scientists but with clinical teams. The goal is to deliver substantial, measurable improvement and reduce mortality, readmission, length of stay, and cost.
Eliminating bias and explainability are important – but saving lives is even more important. Bias and explainability are real challenges we should try to address. However, we should also be practical. We should not aim to get unachievable goals. Dr. Mamdani illustrated that with an ethical dilemma, is it better to have an app that reduces mortality by 26% or postpone deployment of such an app until we have all bias and ethical questions addressed?
Explainability is a wealthy goal, but thrust and reliability are more important. Acetaminophen (Tylenol) is a good example. Clinicians are prescribing it widely. Yet, there is no consensus on how it works. We know it is reliable, and we thrust it.
AI quality standards hot questions. How good does the app have to be before we start using it? To answer that question, Dr. Mamdani presented a series of examples to illustrate that we tolerate quite high error rates in all aspects of healthcare delivery (diagnosis, prognosis, treatment, and communication). For example, one research found that physicians’ ability to correctly estimate that a patient will die and go to the ICU was correct less than 1/3 of the time. Another research found that more than 1 in for patients is prescribed unproven or potentially harmful treatments. Therefore, an algorithm that is just a bit better or can HELP us be a bit better than that is good enough.
Medical decisions are complex and continuous processes that involve considering hundreds of medical and socio-cultural parameters. While humans can process & 7 ± 2 parameters simultaneously (Miller, 1956). Therefore, Dr. Mamdani provides a convincing argument that to process all relevant data and make sound decisions, we need AI help.
Nature of AI
The second part of the presentation discusses the nature of AI. AI is a mechanism to understand complex relationships and data Dr. Mamdani explained. Therefore, it can help us better handle complex systems such as healthcare delivery or education of healthcare professionals.
AI and CPD
Finally, the third part of the presentation discusses the role of AI in healthcare CPD. There are two general directions: create programs to teach healthcare professionals about AI and use AI to improve CPD.
AI-enhanced medical education and CPD is a very dynamic areas. A growing number of AI Medical academic centers, courses, master programs, and centers for ethics of AI are associated with burgeoning research on AI in education. We learned about examples from the University of British Columbia, where ChatGPT is considered as one of the team members of student groups, and about learning management system providers that heavily promote their
AI capabilities and about the AMA guidance document around AI and medicine education: Advancing AI in medical education through ethics, evidence and equity, question generating bots (example), avatars that can present lectures. And much, much more. Please check the video of the presentation below.
If I need to summarize the presentation in one word, I will select ‘opportunity’. We have more than a few very convincing examples of how AI makes measurable improvements in healthcare. CPD context follows that trend. There are lots of opportunities ahead of us.
References
- Grant, K. (2023, 2023/11/20/). Canadian health care workers turn to AI for help amid a staffing crisis, Article. The Globe and Mail web edition, p. NA. Retrieved from https://link.gale.com/apps/doc/A778459189/HRCA?u=anon~e9913a23&sid=sitemap&xid=4e8dd328
- Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81.
Articles referenced in the presentation:
- Abid, A., Murugan, A., Banerjee, I., Purkayastha, S., Trivedi, H., & Gichoya, J. (2024). AI Education for Fourth-Year Medical Students: Two-Year Experience of a Web-Based, Self-Guided Curriculum and Mixed Methods Study. JMIR Medical Education, 10, e46500.
- Ayers, J. W., Poliak, A., Dredze, M., Leas, E. C., Zhu, Z., Kelley, J. B., . . . Hogarth, M. (2023). Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum. JAMA internal medicine, 183(6), 589-596.
- Charow, R., Jeyakumar, T., Younus, S., Dolatabadi, E., Salhia, M., Al-Mouaswas, D., . . . Wiljer, D. (2021). Artificial intelligence education programs for health care professionals: scoping review. JMIR Medical Education, 7(4), e31043.
- Civaner, M. M., Uncu, Y., Bulut, F., Chalil, E. G., & Tatli, A. (2022). Artificial intelligence in medical education: a cross-sectional needs assessment. BMC Medical Education, 22(1), 772. doi:10.1186/s12909-022-03852-3
- Crossnohere, N. L., Elsaid, M., Paskett, J., Bose-Brill, S., & Bridges, J. F. (2022). Guidelines for artificial intelligence in medicine: literature review and content analysis of frameworks. Journal of Medical Internet Research, 24(8), e36823.
- Gordon, M., Daniel, M., Ajiboye, A., Uraiby, H., Xu, N. Y., Bartlett, R., . . . Grafton-Clarke, C. (2024). A scoping review of artificial intelligence in medical education: BEME Guide No. 84. Medical Teacher, 1-25.
- McCoy, L. G., Nagaraj, S., Morgado, F., Harish, V., Das, S., & Celi, L. A. (2020). What do medical students actually need to know about artificial intelligence? npj Digital Medicine, 3(1), 86.
- Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81.
- Miller, K. (2021). Should AI models Be explainable? That depends. In Machine Learning (Vol. 2024): Stanford University Human Centered Artificial Intelligence. Available at ….
- Ohashi, N., & Kohno, T. (2020). Analgesic Effect of Acetaminophen: A Review of Known and Novel Mechanisms of Action. Frontiers in Pharmacology, 11, 580289. doi:https://doi.org/10.3389%2Ffphar.2020.580289
- Omiye, J. A., Lester, J. C., Spichak, S., Rotemberg, V., & Daneshjou, R. (2023). Large language models propagate race-based medicine. npj Digital Medicine, 6(1), 195. doi:10.1038/s41746-023-00939-z
- Paranjape, K., Schinkel, M., Panday, R. N., Car, J., & Nanayakkara, P. (2019). Introducing artificial intelligence training in medical education. JMIR Medical Education, 5(2), e16048.
- Pupic, N., Ghaffari-zadeh, A., Hu, R., Singla, R., Darras, K., Karwowska, A., & Forster, B. B. (2023). An evidence-based approach to artificial intelligence education for medical students: A systematic review. PLOS Digital Health, 2(11), e0000255. doi:10.1371/journal.pdig.0000255
- Tierney, A. A., Gayre, G., Hoberman, B., Mattern, B., Ballesca, M., Kipnis, P., . . . Lee, K. (2024). Ambient Artificial Intelligence Scribes to Alleviate the Burden of Clinical Documentation. NEJM Catalyst Innovations in Care Delivery, 5(3), CAT. 23.0404. doi:https://doi.org/10.1056/CAT.23.0404
- Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., . . . Polosukhin, I. (2017). Attention is all you need. Paper presented at the Advances in Neural Information Processing Systems. https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf
