The future is here, it's just not evenly distributed -William Gibson
Three weeks ago I had the opportunity to attend this year’s Exponential Medicine conference. It has taken me this long to process some of my initial thoughts, as the event was quite simply a mind-blowing four days that has altered my thinking and radically expanded my perspective.
Technological change in healthcare is happening quicker than anything that we (meaning those of us who work daily in healthcare) might think.
The reason for this acceleration, is what the noted futurist Ray Kurzweil (speaker Day 4) has called the Law of Accelerating Returns (LOAR). Kurzweil extended Moore’s Law, first observed in 1965 by Intel’s Gordon Moore, who observed that the number of transistors on an integrated circuit doubled every every two years and would do so for the foreseeable future. Kurzweil realized that Moore’s law actually preceded semiconductor technology and could be extended back through multiple technologies, transistors, vacuum tubes, electromechanical devices all the way to Charles Babbage’s Difference Engine first proposed in 1822! Kurzweil went further to postulate that it is information that powers this phenomenon and that any technology that is information enabled will double in performance every 1-2 years, and that once this starts it will not stop as underlying improvements in computation will continue to improve. This effect become even more dramatic as multiple fields become information enabled and begin to synergize and converge.
In healthcare today, we are seeing a convergence of multiple exponential technologies that are being combined to create possibilities that only years ago would be considered science fiction. The combinations of cheap and ubiquitous sensors, computational power, data aggregation platforms, big data engines, machine learning and deep learning along with ubiquitous mobile computing foretells a coming time when access to healthcare will be personal, precise, predictive, preventative and amazingly universal.
Dr. Aenor Sawyer from UCSF Digital Health Innovation Centre talked about the potential for digital health to reinvent health by connecting doctors and patients at the point of choice (rather than point of care) leveraging what she calls the panome: all the potential data now available at a patient level: genome, transcriptome, metabolome, proteome, microbiome, physiome (think sensors tracking every aspect of physiology, lifestyle and activity, antome (digital medical imaging) along with electronic medical records. Bringing this data together in the cloud opens possibilities for machine intelligence and deep learning. Dr. Jeremy Howard from Enlitic, described the capabilities of current generation deep learning systems outperforming radiologists and pathologists after relatively limited periods of training. Excitingly, this ability to use deep neural networks to detects patterns and correlations across a multitude of variables, well beyond human capability, coupled with expert clinicians opens the door to the computer augmented physician and improvements in diagnosis today. The insights derived from the aggregation and analysis of data will power advances in healthcare as well as in wellness. In fact the ability for consumers to understand their own health and the use of applied behavioral science and gamification to create compelling user experiences, based on data driven insights is likely the holy grail for chronic disease management.
Keeping all of the data secure may involve the use of the blockchain as Chelsea Barabas of MIT Media Lab proposed. Most of us, if we know anything about the blockchain, understand it in conjunction with bitcoin, or as The Economist points out: the technology that allows people who do not know or trust each other, to create a dependable ledger. What if patient data was held in a secure, dependable ‘ledger’ where patients control who has access, revealing only the data required to make a clinical decision. This intriguing potential may solve one of the biggest challenges to the era of personalized medicine: privacy. This concept has also been explored by Dr. Eric Topol and Leonard Kish who argue that patients need to own and control their own data.
Prior to ExMed, I had read many articles and books that would breathlessly describe how a smartphone app could replace your family doctor. As a GP myself, I felt that this was nonsense. I honestly could not see it. Now, I am humbled to say that the capabilities required to do this are almost here:
Ability to understand written and verbal language
Ability to process meaning from conversation
Ability to understand emotion
Ability to compare symptoms and history to a database of differential diagnosis
Ability to learn from interactions and outcomes
Ability to create compelling mobile app experiences (think games)
In medical school, I learned that 80% of a diagnosis came from a good history, imagine the capabilities of a deep learning system combined with increasing precise data powering a Bayesian probability engine. (I will leave that to another blog post…)
As you can tell by reading this far, I am very excited about the potential of exponential technologies to transform healthcare, but I could not post this without pausing to reflect some of the challenges that will occur along the way. How healthcare systems embrace the new technology without exploding costs will depend on how well leaders view and understand the full determinants and drivers of health. Dave Chase has recently written on the “Copernican” realignment required to organize health around the individual and not around providers and medical technology. Dave has written eloquently about a multitude of next-generation healthcare delivery organizations that are appearing in the US that understand this shift and are making it real. In Canada, primary care reform and medical home models are emerging but in a frustratingly slow and uneven way.
This ability for healthcare systems to resist the changes necessary to put patients first is what concerns me most deeply at this most exciting of times. It does give me pause to wonder if the new technology is most likely to flourish and develop first in areas that are not deeply rooted with entrenched self-interests, regulation and change-avoidant cultures. Two areas to watch are consumer health and wellness and emerging healthcare systems in the developing world, where in a manner similar to their adoption of cellular technology without landlines, patient centred healthcare around digital technology may be their first modern healthcare systems.
It will take me months to process what I learned and saw at Exponential Medicine, and I am sure that many blog posts will come from ideas that were planted in the many lectures by leading thinkers across genomics, genetic engineering, stem cells, AI,3D printing, robotics, big data, nanotechnology, virtual/augmented reality, wearables, sensors and many more! As important as these lectures were, the hallway, dinner and late-night conversations were even more impressive. Kudos to Dr. Daniel Kraft and his team at Singularity University, I have to say that this was the most impressive conference I have ever attended and I look forward to next year’s.
"It’s far more important to know what person the disease has, than what disease the person has.” – Hippocrates
The topic of personalized medicine is attracting a lot of attention in the scientific community and raising controversial questions about healthcare privatization, data privacy as well as its cost-benefit. Indeed, personalized medicine represents a whole new paradigm of diagnosing and treating disease and, as such, comes with many as-yet unresolved concerns.
This article takes a step back and instead considers some of the basics. It aims to serve as a primer on the promise of personalized medicine, the most prevalent ways that it is starting to be applied, what’s driving its emergence, and what this fundamentally means to the future of medicine and healthcare.
Understanding an individual’s molecules changes everything A few vignettes to help set the stage:
A 67-year- old woman battling colon cancer learns that her tumor has recurred and that conventional treatments will no longer be effective. Her condition is terminal. At the same time, the cancer research team sequences her tumor and discovers that it is over producing a specific protein that can be controlled with an existing medication for high blood pressure. The medication is prescribed and the cancer disappears.
A 40-year-old man must prepare to broach funeral arrangements for his 7-month-old son who suffers from Leigh Disease, a rare neurometabolic degenerative disorder that results in pain, neuro motor degeneration and eventually death. He receives a call from a research team indicating that his son's disease can be traced to a metabolic disruption that can be treated.
A 41-year-old woman with severe Rheumatoid Arthritis is pondering whether to start taking Azathioprine, understanding that there is a rare but severe potential adverse reaction involving suppression of her bone marrow that could be fatal. She is given the option to have a test from a local start-up to check for the gene variant that causes this complication before treatment.
These three stories share one common thread: They all illustrate how our new molecular understanding of disease at the individual level can provide cures, save lives and avoid potentially devastating adverse reactions.
These stories also highlight the first three areas of personalized medicine – oncogenomics, rare disease and pharmacogenomics – that are breaking through into conventional medicine with impressive results.
Progress in these fields is providing exciting glimpses into a new clinical world where diagnosis and treatment will come from a molecular understanding of disease and treatment at the individual level, rather than the traditional broad-based population level.
Unprecedented data volumes that hold untold opportunity This massive change is only just at its outset and is being powered by the convergence of multiple technologies. Computing power, gene sequencing, biosensors, nanotechnology, machine learning, big data and 3D printing, when applied to health, will provide the ability to understand it at a molecular level for each individual. The resulting data set available for analysis will be unprecedented in terms not only of volume (1-2 terabytes per person), but also scope, and ultimately uses.
This data will come from:
Genome – the complete set of an individual’s DNA including all of their genes – representing more than 3 billion DNA base pairs
Transcriptome – the complete set of RNA transcripts produced by the genome at any one time
Proteome – the complete set of proteins that are expressed by an individual at any one time
Metabolome – the complete set of metabolites present within an individual at any one time
Microbiome – the complete genomic information for the microbes that live inside and on the human body – approximately 100 trillion cells.
Physiome – the physiological dynamics and characteristics of the individual, as measured through sensor technology
Anatome – the individual’s unique anatomical characteristics digitally represented through advanced medical imaging
With new abilities to capture and analyze this data, healthcare will be transformed through a precise understanding of how health outcomes are influenced by genetics, environment, diet and lifestyle on an individual basis. This understanding will result in new opportunities to treat, predict and prevent disease and, at the same time, build continuous feedback between health outcomes and the molecular changes that precede them.
Changing how we understand disease This new personalized paradigm will change the way that disease is understood.
Traditionally diseases have been characterized by the physiological effects they have on a population. For many diseases, especially complex chronic diseases, there are multiple molecular causes that result in similar physiological signs and symptoms. For example, even in a rare disease like Leigh's Disease there are dozens of genetic variants causing multiple, different neurometabolic breakdowns that all clinically result in the condition that presents to a clinician as Leigh's Disease.
For a disease like Type II Diabetes there may be hundreds of different molecular causes, each with a different set of treatment options. Without the ability to differentiate at an individual level, current medical therapy applies a one-size-fits-all approach and treatment options are studied across these heterogeneous disease populations.
The impact of one-size-fits all treatments is quite varied. Some may work extremely well for one molecular variant of a disease, but may not work for others or can even be harmful. In fact, only half of the population will respond to most medications. This variability can be avoided when clinicians have an understanding of the molecular underpinnings of a disease for an individual patient. In personalized medicine only treatments that are certain to benefit the individual are prescribed.
The transformation of medicine from “one size fits all” to a personalized approach will allow doctors to be more precise, predictive and ultimately preventive in their practice.
Three gates to personalized medicine Personalized medicine is not yet an established paradigm and there are a number of gating factors that will need to be cleared before this vision becomes a reality. Interestingly, technology is not one of these gating factors. The current state and exponential progression of the underlying technologies actually bodes well for a near-term realization of personalized medicine.
Instead, the three challenges, in increasing order of difficulty will be:
Complexity – much of current health data is and incomplete. Data from sensors will come from contexts we do not yet understand, our ability share information usefully between multiple systems is inadequate. Finally, we simply do not know what to do with much of this data (yet).
Regulation – the legal frameworks for ownership of the data and algorithms derived from this data needs to be fully clarified internationally. The privacy and security mechanisms to allow robust but secure sharing and secondary use of data need to be expanded, and laws to protect individuals against discrimination must be enacted.
Culture - in a matter of years, conventional medicine paradigms will be turned over completely and patients will demand treatments based on diagnoses derived from computer algorithms – healthcare professionals will need to adapt to roles that are dramatically different to what they have historically done.
These challenges, rather than the technology, may be the biggest obstacle in the way of personalized medicine. Early recognition of the changes required and engagement of a broad set of stakeholders will be required to enable the appropriate system changes necessary for society to realize the benefits of this new and profound way of understanding health.
Originally published on TELUS Talks (www.telushealth.com)
I have just started a new job with the compelling title – Chief Innovation Officer, TELUS Health. When I tell friends and colleagues they give me a quizical look before asking some variant of “what does that mean?” While innovation is almost universally viewed as positive, there is great variation in what people mean by it; ask 25 people and you will get 25 different definitions, all representing their own outlook and biases. After a few of these conversations, I realized that I need to define innovation for myself and for my organization in a way that is clear, illuminating and actionable.
If one looks up the definition in the Oxford dictionary, innovation is defined as making “changes in something established, especially by introducing new methods, ideas or products.”
The Google ngram shows an almost exponential 400% increase in the use of the word innovation since 1940. Much of this is likely attributable to Joseph Schumpeter, who in 1939 defined invention as an act of intellectual creativity undertaken “without importance to economic analysis” while innovation is an economic decision consisting of any one of these: 1. Introduction of a new product 2. Creation of a new method of production 3. Opening of a new market 4. Conquest of a new source of materials 5. Implementation of a new form of organization Schumpeter went further in decoupling the dependance of innovation on invention when he said: “Innovation is possible without anything we should identify as invention and invention does not necessarily induce invention.”
Stepping back a bit, I think the best definition I have uncovered comes from Scott D. Anthony when he simplified this to define innovation as: “doing something different that has impact.” From this definition, we can see that like Schumpeter he sees innovation as involving significantly more than new products, but Anthony goes a step further and does not include “new” or never been done in his definition. When something is new, an act of creation, but has no economic impact, it is invention, not innovation. On the other hand, when something already discovered, i.e. “not new”, is applied in a different way and has impact, it is innovation.
This latter aspect of innovation is especially important when one understands the power of convergence, the coming together of different lines of inquiry. Convergence is truly the most fertile ground for innovation, as methods and insights gained in one area, result in application and breakthroughs in others. We live in a golden age for convergence in healthcare as we are seeing multiple exponential technologies converge: • Computing power • Networking • Gene sequencing • Biosensors • Big Data • Machine learning
Individual development in each of these areas is compounding at(or greater than Moore’s law) resulting in capabilities that were cutting edge only twenty years ago becoming affordable or even virtually free now. For example from 1992-2012, the cost of computing has decreased from $222 to $0.06 per million transistors, computing storage from $569 to $0.03 per gigabyte, bandwidth from$1245 to $23 per 1,000 Mbps. Applied to healthcare the cost of sequencing a human genome has dropped from $100M for the first genome sequenced in 2000 to $1000 today. Cameras and sensors that cost millions in the 1970’s are now 1,000x better, lighter and cheaper and found in every phone sold!
Over the next decade, using these exponential technologies, our understanding of the molecular basis of life in the individual will simply explode as we will have within our reach affordable ways to sequence not only an individual’s genome, but also their transcriptome, proteome, metabolome and microbiome. With cheap and ubiquitous sensors we will have the ability to monitor all aspects of human physiology in real-time. Through big data and machine learning advances, we will also have the power to synthesize all of this new information with existing electronic health records and realize insights and breakthroughs allowing us to predict and prevent disease..
This convergence of exponential technology will lead to a golden age of innovation in healthcare, transforming medicine from a pathology focused, one size fits all discipline to a precise, predictive and preventive model that improves vibrant life expectancy. These innovations will come from all angles: new products, diagnostics, therapies and cures, completely new capabilities, and new business models (think disruption).
The realization that we are in the midst of a golden age of healthcare innovation has led me to my own definition of innovation as I begin my new role as Chief Innovation Officer at Canada’s leading health IT company:
“Innovation is something different that has the impact to exponentially transform healthcare”.