From Matrix to medicine: AI applications in health care
Estimated read time: 10 minutes
For many, the term “artificial intelligence” (AI) conjures up scary images of robots and computers taking over the world, with humans left fighting for their lives under a heartless digital regime. However, whether people realize it or not, AI is already well-integrated in our daily lives. With programs like Alexa and Siri well-known around the world, it was only a matter of time before AI technology moved from guiding us through a lasagna recipe to playing a key role in healthcare settings.
AI Image generated in Canva for the prompt: "AI in health care"
But what does AI in health care actually look like today? Our imaginations can run wild with possibilities: medical droids performing surgeries or physicians using devices to diagnose conditions with a quick scan. Although not quite at that hyper-futuristic level, AI applications in health care have the potential to provide impactful benefits, and CPSA is working hard to manage regulation and innovation within this exciting new realm.
AI Medical Scribe Technology
AI applications like ChatGPT have gone viral over the last couple of years, often used by frazzled students to spark inspiration for a paper, or companies to populate their websites with dynamic content. But could this type of technology be useful for doctor’s appointments?
In Alberta, the concept of ChatGPT has been developed further into Scribeberry, an exciting AI innovation that provides a tool to improve medical practitioner burnout. Founded by Edmonton family physician Dr. Zaahir Moloo, Scribeberry provides its users with medical scribing, charting and documentation support. It is more secure than the USA-based ChatGPT since, as an Alberta-based product, it is compliant with provincial privacy legislation including the Health Insurance Portability and Accountability Act (HIPAA) and the Personal Information Protection and Electronic Documents Act (PIPEDA).
“Having a large patient panel personally, I found myself burning out - not because of the patient interaction, but because of the administrative workload, endless forms and other task-based work. This motivated me to create Scribeberry. In terms of the impact to the profession, I think AI tools are in their infancy. Tools like Scribeberry will inevitably end up augmenting physician workflows and replacing many tasks that are currently time consuming. In a broader sense, I envision AI eventually changing everything from novel drug development and drug delivery to a total overhaul in how medicine is delivered”
It is well documented that healthcare providers across Canada are experiencing burnout at an alarming level. While one in five Canadians are without a family physician, primary care providers are feeling the pressure to expand their patient panel. Scribeberry could provide some relief in these areas considering that according to a user feedback survey, 90% of Scribeberry users reported a reduction in burnout and the program saves providers an average of 1-2 hours of work per day. This reduction in administrative burden has allowed providers to accommodate an average of 30% more patients to their panel. By providing physicians with the capacity to direct more energy towards the patients they see each day, Scribeberry could result in not only improved physician wellness, but better patient experiences.
Given these positive impacts, Scribeberry has begun to develop other helpful tools that health providers can add to their repertoire. As of December 2023, they launched the early version of an add-on known as Ada, an AI clinical co-pilot. Ada can take histories, chat with patients before their visit and summarize important information to help with triage and improve efficiency during an appointment.
Despite medical scribe’s obvious benefits, it’s important for regulated members who use these types of AI programs to remember that the accountability for quality care rests with the wielder of the tool, rather than the tool itself. When using any scribe tool, regulated members have multiple responsibilities, an important one being to receive patient consent before using the scribe during an appointment. The chosen tool must adequately protect patient privacy, generated chart notes must be assessed for quality and differential diagnoses must still be checked for accuracy and bias. Before including any information in a patient’s record, regulated members must confirm accuracy of all final information. Published on Sept. 1, 2023, CPSA’s Advice to the Profession document on Artificial Intelligence in Generated Patient Record Content helps regulated members prepare before taking the step to use AI in clinical care.
CPSA’s Machine Learning Model: Opioid dispensation
In addition to providing guidance on AI usage for regulated members, CPSA has been making an impact in AI innovation. CPSA partnered with OKAKI Health (a technology company that provides professional services and custom software) and the University of Alberta to create a machine learning model to identify patients who may be at higher risk for adverse outcomes following opioid prescription. If you’ve never heard of machine learning, you’re not alone—this arm of AI is driven by data science, a field that extracts value from raw data. Machine learning is a subset of AI, referring to systems (algorithms) that learn through data and pattern recognition.
As a partner organization of Alberta’s Tracked Prescription Program (TPP), CPSA provided valuable data to develop this project. The initial study used data from 2018–2019 from more than 850,000 participants representing over 6 million opioid dispensations, and analyzed demographic characteristics like age, income and healthcare utilization to determine probability of negative outcomes within 30-days of opioid dispensation (i.e., drug-related emergency department visits, hospitalization or death). The machine learning model predicted adverse patient outcomes with 80-90% accuracy, much higher than the roughly 60% prediction accuracy exhibited by physicians using prescribing guidelines. The opioid crisis is ongoing and has led to an increasing number of deaths and harms to patients. This machine learning program can provide added information to physicians to prompt them to check in with patients and if needed, intervene for potentially life-saving results.
“CPSA is delighted to play a role in this important program. This application of machine learning technology will enhance patient safety by minimizing risks following an opioid dispensation. This work is another demonstration of the proactive and innovative approach that CPSA is taking towards medical regulation.”
CPSA is looking forward to the next steps in this program’s journey and in 2023, began assessing external validity by extending it to select opioid agonist therapy providers in the province. Feedback on the program’s usability and value in physician practice has been largely positive. Next steps include determining how physicians would like to receive opioid risk score information, whether by integration into their electronic medical records, through an additional portal or another option that would best meet their needs. Once the means of delivery has been determined, the program will further explore external validity for primary care providers then, if successful, will provide the risk scores to regulated members with the aim of improving quality of care and safety for patients receiving opioid prescriptions.
“My experience with utilizing an AI machine learning tool for predicting medication outcomes has been overwhelmingly positive. By harnessing the power of predictive analytics, we can enhance patient safety, optimize treatment strategies and inform evidence-based healthcare practices. I am grateful that the CPSA is researching AI-driven innovations that hold the potential to revolutionize healthcare delivery and improve patient outcomes across the province.”
Data Interoperability
As demonstrated by the opioid dispensation machine learning model, quality AI-generated predictions cannot exist without quality data. Usable health data is what drives analysis within AI technologies and helps create a useful end product, whether that be generating high-quality charting notes through scribe technology, accurate patient predictions through machine learning models or other applications.
CPSA has been heavily involved in the development of high-quality data interoperability through its representation on the Alberta Virtual Care Coordinating Body (AVCCB) which is chaired by CPSA Senior Medical Advisor (Health Informatics), Dr. Ewan Affleck. A high-functioning healthcare system must be able to exchange health data efficiently and safely between platforms and health providers. In Canada, there are currently no jurisdictions that have a systematic approach to health data interoperability, and data shared is often fragmented and broken up between services, custodians and health authorities. This causes a barrier for AI development, since it can be challenging to access comprehensive data and aggregate it in a meaningful way for the development and improvement of AI applications.
In response to this problem, the AVCCB proposed a set of recommendations for consideration in their 2023 report, “Interoperability Saves Lives”. This report focuses on the integration of health information and data between all medical and health providers, regardless of location or time, to provide a fully integrated and holistic view of the patient and their overall health. The application of the proposed recommendations could result in substantial benefits for patient outcomes, health and wellbeing of providers, and AI applications that can enhance the healthcare system. Since its release, the report has been reviewed by Health Canada and many other prominent agencies and has led to work with the University of Calgary on actioning the report recommendations.
AI Symposium
As detailed in our strategic directions, CPSA strives to be a leader in Alberta’s healthcare system by leveraging proactive and innovative approaches to regulation. To support this objective, CPSA’s Analytics, Innovations & Research (AIR) team worked with the Canadian Medical Protective Association (CMPA) to co-host an AI Symposium in the fall of 2023 that focused on “Innovation, safety, and AI: Developing regulatory & medico-legal approaches to AI in Canadian health care.”
Bringing together thought leaders from several industries, the AI Symposium facilitated connection and discussion on current and future AI applications, while considering its impact on patient safety and quality of care. In a judgment-free environment, attendees brainstormed new ideas and avenues for AI in health care, to be further explored and assessed. The event also identified unnecessary barriers that have not caught up to innovation and no longer serve the purpose of protection, but unexpectedly hinder safe and beneficial developments.
Attendees shared valuable information and ideas, revealing a balance between optimism and caution. In order to support the safe and appropriate use of AI, a frame shift must take place for regulators and practitioners alike. In addition to a reframe, CPSA needs to determine the regulator’s role in AI oversight and outline how members can safely adopt emerging technologies. As more technology is proposed and applied to health care, CPSA will have to adequately assess and determine when training, advice and guidance are needed, and clearly communicate that regulated members remain ultimately accountable for clinical decision-making and the care they provide. This can be completed while still investing in innovation and identifying AI’s potential applications in health care.
AI and other regulators
Learnings from the 2023 AI Symposium appear to be echoed nationally and internationally. Across the world, regulators in health care and other industries are contemplating how to ensure regulation focuses on responsible AI usage rather than AI technology itself. AI is developing quickly—there are constant announcements regarding new applications and ongoing updates on technology improvements. Regulators must keep this ever-changing nature of AI in mind as we work towards finding a balance between fulfilling legal obligations while still supporting innovation.
CPSA is a trailblazer in its approach to innovation and AI. We are the only regulator in Canada with a department focused on analytics and research, which has provided us with the capacity to take part in projects like the opioid dispensation machine learning model and AI research. CPSA has presented at conferences across the world, including the 2023 International Association of Medical Regulatory Authorities conference in Indonesia, and has spoken widely about how risk and protective factors can predict certain elements of a physician’s practice. In our unique position as a highly innovative regulator, we hope to set the stage for other regulators when moving forward, through effective AI guidelines, tools and contributions for members.
Conclusion
Overall, AI is a tool that shares commonalities with other technological innovations that have shaped medicine over the years. For example, prior to the discovery of the X-ray in 1895, a tumour or intestinal obstruction could not be identified without performing an invasive procedure. Due to its obvious benefits, X-ray technology was shared quickly and put into practice across Europe and North America within a year of its discovery. Just like the X-ray, AI applications in healthcare settings open the door to exciting benefits; however, like any technology, the potential for major improvements does not come without challenges. Due to the excitement of its discovery, the X-ray was put into practice without much research into potential risks, such as exposure to radiation. Similarly, the use of AI tools raises its own set of concerns that must not be ignored and careful thought must be made to minimize unintended consequences. Alongside effective direction from regulators, we are confident AI can be integrated to enhance and improve patient care, with limited negative impacts. CPSA is excited to continue assessing and supporting the role of AI within Alberta’s healthcare system in the years to come.
