In health care, 2017 is going out with a BANG!December 15, 2017
As always when wrapping up the year, the big question is what will the future hold for helping plan members get the health care they need, and how will it affect benefits plans? The answer for years to come will increasingly include the role of artificial intelligence (AI) in health care. As you may recall, in the October 2017 edition of The Inside Story, we discussed how AI is set to impact all areas of insurance—with health benefits no exception. In fact, AI is a rapidly emerging field in health care…
What do Rosie, Jaco, and Pumpkin have in common?
They are all “intelligent robots” currently in use in Canada’s health care system.1 Intelligent robots involve advanced electronics and computer programming plus they’re equipped with an “AI brain.” This enables intelligent robots to act like humans by, for example, interpreting complex data about a patient, determining required actions, and carrying out the actions.
- Enhanced disease diagnosis: AI is outperforming health care specialists in the diagnosis and classification of skin and breast cancers. Researchers are working toward AI technologies that have the ability to integrate nearly all types of data to accurately diagnose different diseases.
- Decreased challenges faced by imaging (x-rays, computed tomography, and magnetic resonance imaging): AI technologies built to replicate humans in making imaging diagnoses are predicted to enhance imaging productivity issues and provide second opinions to catch mistakes. Ultimately, challenges with imaging like the increasingly high volume of patient data, the shortage of skilled radiologists, and the high error rates, all have the potential to be tackled by AI technologies. In fact, in some cases, imaging diagnosis by AI technologies is already on par with, and even better than trained clinician decisions.
- More effective surgeries: Image-guided neurosurgical robotic arms guided by a surgeon from a nearby workstation are proving to be safer, less invasive, and more accurate than traditional surgery. Advantages include that robotic arms—that typically rely on CT and MR images, and GPS—are more precise than a surgeon’s hand-eye coordination alone. In addition, robot-assisted surgeries reduce strain and fatigue for the surgeon.
- Improved access to health care in rural and remote areas: Remote presence robotic systems are also known as telerobotics and telepresence robots. A physician can activate a robot remotely to check in at the nursing station and then head off to the patient’s room to interview a patient. Then, with help from a health care worker and by attaching peripheral devices like ultrasounds and electrocardiograms, the robot examines the patient. This improves access to care and reduces travel costs, while also decreasing disruption compared patients having to travel to urban areas to receive care.
- Reduced need for home health care services: Assistive robotic arms that attach to electric wheelchairs—and activated by a joystick, chin, head or eye controller, sip or puff mechanism, and even brain-to-computer interface—enable patients to perform routine tasks. This decreases reliance on caregivers while at the same time enhances independence and quality of life. And then there are robots directly caring for patients and other interactions… keep reading…
- Enhanced surgery training: Medical students training to perform surgeries can experience more authentically what surgery is like by practicing with surgical tools that use AI technology. These intelligent tools mimic real surgeries by replicating the sensation of what the actual interaction is like between surgical tools and human tissue like the pressure on the tools.
- Operational improvements: AI technologies can help improve various behind-the-scenes aspects of health care, which ultimately can lead to enhancing the patient experience. For example, AI technologies that analyze data can gain insights leading to improvements in staff workflows that translate into faster patient care.
Clearly, we’re seeing the value of AI in diagnosis, patient care, and patient outcomes, but for GSC a burning question is: How can AI make inroads regarding preventing diseases and more effectively managing them? Diseases like diabetes—which is now described as a global epidemic. In fact, the International Diabetes Federation just released new estimates on the prevalence of diabetes around the world. One in 11 adults globally are currently living with diabetes; this is 10 million more than in 2015. Canada is seeing more than 60,000 new cases of diabetes each year and type 2 diabetes is one of the fastest growing diseases in Canada.6
…And GSC plan member data mirrors the global and national statistics with a high incidence of diabetes. Fortunately, type 2 diabetes can be prevented or postponed—as well as more effectively managed—by making healthy lifestyle choices. So what role can AI play in diabetes health management?
With benefits of AI technologies like these, it’s no wonder the Senate of Canada authorized its Committee on Social Affairs, Science and Technology to develop the 2016 report called Challenge Ahead – Integrating Robotics, Artificial Intelligence and 3D Printing Technologies into Canada’s Healthcare Systems. The report concludes that that AI technologies “are going to change the way Canadians live and specifically, the way healthcare is delivered.”8
Insurers are sitting on reams of data, but how to use it for health management?
As you can imagine, a priority for GSC has been to continue to explore ways to address the issues surrounding diabetes, a prominent disease in GSC’s plan member population. Accordingly, we decided to take this opportunity to talk to the AI experts at Memotext, Amos Adler, president and founder, and Bill Simpson, director, data science. Memotext uses AI technologies to build health interventions to help people change their behaviour so that they have the best chance of reaching their health goals.
Amos explains, that the biggest void in developing health management strategies is the ability to effectively use—or put another way, to activate—large volumes of data. “Data is being generated at a hyper-exponential level and although data represents tremendous potential, we have to figure out exactly what the potential is. For example, can we activate it for health management of diabetes? Basically, it’s not enough to just have reams of data, we need to figure out how to actually apply insights to real-world scenarios.”
As Bill elaborates, “That’s where AI can help make sense of data so that the insights can lead to interventions to help prevent diabetes and improve its management for plan members. And from the plan sponsor and insurer perspective, also help decrease costs.”
Amos adds that “in terms of GSC’s plan member data, it’s important to think much broader than just aggregate claims data because AI technologies like machine learning allow us to gain insights from widely varied sources. For instance, to develop a diabetes health management intervention, GSC claims data that reflect plan members who take diabetes medications is invaluable. However, it’s important to look at any and all information that indicates how plan members behave. For instance, we’re also looking at GSC’s aggregate call centre log data, as well as GSC’s web portal use—any plan member information that is available, given confidentiality. Then we use machine learning to apply an algorithm to the data to find patterns.
Next, as Bill describes, “From the patterns, we create what are called classification models that categorize how different plan members behave—like plan members who are at risk of developing diabetes versus those who are effectively managing diabetes versus those struggling with effective management.”
And then, Amos continues, “Based on the models, additional AI technologies can make decisions regarding how to interact with plan members to deliver diabetes interventions. This is where the data insights move into action and when we get to this stage, the AI technologies will be able to deploy interventions tailored to the plan member’s specific diabetes management issues—and at a large scale. So AI allows interventions to reach all plan members who need help.”
From our discussion with Amos and Bill, it’s clear that the future seems bright for diabetes prevention and management regarding the possibilities of how AI can help. However, as more and more AI technologies are being developed, numerous issues will have to be addressed.
Bumps along the AI road
In addition to the obvious issue of privacy regarding the use of personal data, AI in health care faces a number of significant issues. For instance, a major debate is the issue of potential job loss versus job creation. Where do you draw the line between benefits of a human—for example in surgeries and patient care—versus benefits of a robotic arm or a humanoid robot?
In terms of barriers to adoption of AI in health care, experts predict a main concern will continue to be ethical issues. For instance, surrounding the delegating of medical decision-making to machines, the Chair of the Senate of Canada report explains, “Above all we will have to educate both health care workers and their patients to build trust in these new and developing technologies. For example, if artificial intelligence disagrees with the doctor’s diagnosis or method of treatment, who does the patient believe and will the doctor be prepared to accept that perhaps the robot has got it right?”
Just scratching the surface
To sum up, we turned to our very own Ned Pojskic, GSC’s pharmacy strategy leader, who feels that “it’s early days; what we need to see is AI evolve to the point where the technology can deliver very targeted interventions customized for each plan member. This is how health management interventions will become relevant to each individual and be motivating for those who, for example, are struggling with managing diabetes. As a result, these very highly sophisticated AI technologies will help GSC change the trajectory of diabetes—as well as other chronic conditions—to promote prevention and, as needed, more effective diabetes management.”
1, 8, 11“Canadian healthcare system must brace for a technological revolution,” The Senate of Canada media release, October 31, 2017. Retrieved November 2017: https://sencanada.ca/en/newsroom/soci-challenge-ahead/. “The Senate of Canada’s ground-breaking study on the role of Robotics, 3D Printing and Artificial Intelligence in the Health Care System,” The Senate of Canada infographic. Retrieved November 2017: https://sencanada.ca/en/newsroom/soci-challenge-ahead/#infographic. “Senators learn about robotics, artificial intelligence in the health-care system,” The Senate of Canada, May 29, 2017 – Parliament of Canada. Retrieved November 2017: https://sencanada.ca/en/sencaplus/news/senators-learn-about-robotics-artificial-intelligence-in-the-health-care-system. “Integrating Robotics, Artificial Intelligence and 3D Printing Technologies into Canada’s Healthcare Systems,” The Senate of Canada, Standing Senate Committee on Social Affairs, Science and Technology, October 2017. Retrieved November 2017: https://sencanada.ca/content/sen/committee/421/SOCI/reports/RoboticsAI3DFinal_Web_e.pdf .
2,3,4,5“CIHI: Humber Leads Large Community Hospitals in Patient Satisfaction,” Canadian Institute of Health Information media release, September 20, 2017. Retrieved November 2017: https://www.hrhfoundation.ca/blog/cihi-humber-leads-large-community-hospitals-in-patient-satisfaction/.
6“New IDF figures show continued increase in diabetes across the globe, reiterating the need for urgent action,” International Diabetes Federation media release, November 14, 2017. Retrieved November 2017: https://www.idf.org/news/94:new-idf-figures-show-continued-increase-in-diabetes-across-the-globe,-reiterating-the-need-for-urgent-action.html.
7Type 2 diabetes – Government of Canada, Diabetes. Retrieved November 2017: https://www.canada.ca/en/public-health/services/diseases/type-2-diabetes.html.
9“Do Mobile Phone Applications Improve Glycemic Control (HbAIc) in the Self-management of Diabetes? A Systematic Review, Meta-analysis, and GRADE of 14 Randomized Trials,” Can Hou, Ben Carter, Jonathan Hewitt, Trevor Francisa, and Sharon Mayor, Diabetes Care, Voume 39, November 2016 – US National Library of Medicine and National Institutes of Health. Retrieved November 2017: https://www.ncbi.nlm.nih.gov/pubmed/27926892.
10Artificial Intelligence for Diabetes, 1st ECAI Workshop on Artificial intelligence for Diabetes at the 22nd European Conference on Artificial Intelligence (ECAI 2016), Beatriz López, Pau Herrero, and Clare Martin, August 30, 2016. Retrieved November 2017: http://www.ecai2016.org/content/uploads/2016/08/W7-AID-2016.pdf.