LAS VEGAS—Medical education is confronted by huge opportunities with the coming of what has been termed the fourth industrial revolution—that of the intersection where “big biology and big medicine” meet “big data.”
Those were the sentiments of Michael Dake, MD, senior vice president at the University of Arizona Health Sciences in Tucson, Arizona, when he addressed delegates at the Vascular and Interventional Advances (VIVA) conference, Nov. 4–7, 2019, in Las Vegas in a keynote lecture entitled “The future of medicine and medical education: How do we prepare the next generation?”
“Big biology and big medicine is meeting big data,” said Dake, who is responsible for the integration of undergraduate and graduate education at his institution. “Some have called it the fourth industrial revolution, where we have an accelerated, exponential convergence of data science, physical science and life sciences. These are creating enormous opportunities but also challenges—for all of us. Our main task as educators is to figure out how to integrate large-scale, multidisciplinary datasets into our education.”
He spoke of a fraught but fruitful future. “We are in a period of great innovation,” Dake went on. “There are tremendous opportunities, and these will continue to grow with technological advances that promise to impact clinical practice and transform medicine and medical education. The pace of change is rapid; thinking outside the box is essential. The potential impact on medical education and medical care is enormous.”
Dake, known globally for pioneering image- guided therapies and novel approaches in interventional therapy in the fields of vascular imaging, venous thromboembolic disease, aortic aneurysms and dissection, talked about a key role for artificial intelligence (AI). “How do we drive precision medicine and data-driven healthcare into routine clinical practice?” he asked. “Where AI can really help in the future of healthcare,” he said, “is with merging these two dominant trends: precision medicine and digital medicine.”
Dake detailed how new technology platforms that use profiling, automation and computing to provide deep phenotyping and risk profiling will interact with the “expanded care space— wearables, sensors, telemedicine, social media and lifestyle metrics, data on consumer patient engagement—to enable remote monitoring of patient health status. This interaction will be mediated by machine learning algorithms.” The potential power of AI in this context, Dake said, is to provide analytics for improved decisions and clinical outcomes at lower cost.
Indeed, the cost of big data has decreased dramatically over the last decade. Showing a graph plotting the cost of data relative to speed versus data consumption, Dake demonstrated that the cost of data per second has dropped from nearly $3 per Mbps in 2004 to less than $0.1 per Mbps by 2013, while data consumption per subscriber per month rose from less than 10MB a month in 2006 to 225MB a month in 2013. “We can feel confident that at least computation comes cheaply. All this big data certainly is not as expensive to process as it was years ago,” he said. “That is good for us.”
He continued: “Everyone in any healthcare centre or academic community is aware of the buzzwords around precision medicine. There is a lot of emphasis on deep phenotyping: taking large- scale datasets, predicting complex traits and disease risk, whether it is by a variety of ‘omics’ that take large cohorts of individuals and stringently control for clinically phenotyped outcomes, or by mapping genetic overlap between different diseases involving shared pathogenic elements and comorbidity risks.”
Illustrating the idea of genetic phenotype mapping with an example, he cited a study published in Nature Genetics in 2018 by Rainer Malik of the Institute for Stroke and Dementia Research, University Hospital, LMU Munich, Munich, Germany, et al that identified a genetic overlap between stroke and related vascular traits at 32 genome loci in 521,612 people. Some 22 of these loci were new to the designation “stroke risk.” This is one example of the “tremendous opportunities” afforded to researchers with access to large datasets, Dake said.
Not without hurdles
Yet, there are also several challenges presented by the emergence of big data, Dake conceded. “The problem with real world data,” he said, “is that it is, indeed, real world,” explaining how people analytics and large-scale databanks had blurred the boundaries between medical research, clinical care and daily life, rendering every monitored event as a potential data point, every individual as a data node and research asset. Social spaces are also becoming quantifiable, and with sufficient data, investigators could reveal increasingly predictable behavior and individual risk patterns. “This blurring of private and public spaces could lead to complex ethical and legal issues,” Dake commented, noting that consent, privacy, security and surveillance were factors society was “only beginning to address,” and that they would be “increasingly important as areas of focus.”
For AI, one of the key benefits is that it can overcome the bandwidth limits of humans, Dake noted. “There is currently a data deluge,” he said. “Our cognitive bandwidth is certainly challenged and overwhelmed, but I think there is great promise in decision support, and ways that we really can impact the future of medical education for the better.”
Dake told the VIVA audience that society was on the cusp of the “era of cognitive computing and decision-support systems.” Using AI algorithms can help draw patterns and sense from such large datasets. “Clearly we have limits to individual expertise,” Dake said. “We have limits to our individual abilities to multidimensionally evaluate data, we have limits to our sensory systems, cognitive experiences and perceptions, and limits to our objective decision-making. This is where AI, deep learning and machine learning can come to help us.”
Envisioning the future of medicine as made possible by AI and big data, Dake predicted: “As we evolve from qualitative, descriptive information of variable quality and provenance to quantitative data of known provenance and validated quality, [we will] change from a complex ecosystem of largely unconnected data sources to an evolving, interconnected network of data sources for robust decisions and improved care.”
“If you are teaching today what you were five years ago, either the field is dead or you are,” the father of modern linguistics Noam Chomsky famously said. Sharing this quote, Dake added, “Noam Chomsky is 95 years of age, and he is aware of this—so should we be.”
Dake highlighted a shift in medical education, from being science-centric in the 20th century, to becoming more data-centric between 2000 and the present day. Extrapolating this trend, he said that education will next become network-centric, stressing the importance of students gaining mastery of escalating complexity and massive data.
“There is a digital Darwinism looming,” Dake warned. “Understanding data structure and application to improve decisions and outcomes will become a critical institutional competency. With major skill gaps and predicted personnel shortages, we are going to have to train a new cadre of data scientists—both medical and nonmedical— and institutions that lack adequate computation infrastructure and are not committed to this training are going to suffer ‘cognitive starvation’ and relegation to competitive irrelevance.”
So how will medical education adapt to this demand? Dake outlined his thoughts on curricular emphasis in the 21st century, believing that knowledge capture and creation, not information retention, should be the priority, and that students should be taught to distinguish between information and knowledge.
He said that he would like to see future physicians have a deep understanding of probabilistic reasoning, as well as collaboration with and management of AI applications. Lastly, he underlined the importance of cultivating empathy and compassion in the next generation of doctors, emphasizing the “very real” issue of physician burnout, and advocating for a more holistic curriculum prioritizing students’ mental health.
The University of Arizona has implemented a program to “future-proof ” its graduates, Dake recounted, taking all of these factors into account. The institution is following a health sciences strategic plan, comprised of 26 initiatives, which cover a range of topics, from reducing student debt, to increasing the number of nurses and primary care physicians, to also creating more flexible learning pathways.
While every institution will need to explore ways of tackling its own unique and specific challenges, Dake said, his keynote lecture could serve as a blueprint for building a comprehensive medical curriculum that is capable of reimagining the physicians of the future.
“Institutions should make every effort to take advantage of the strategic synergies that can arise by creating a virtuous cycle connecting education with research and patient care in a continuous feedback loop,” he concluded.
“The ultimate goal is the creation of a real-time learning health system, in which the practice and teaching environments learn from each other and are informed by research.”