Abstract
Tom Lawry, director of worldwide health for Microsoft, spent some time talking to editor in chief Damian Doherty about his new book Hacking Healthcare—How AI and the Intelligence Revolution will Reboot an Ailing System. Tom's book shines light on how the pandemic, despite its destructive force, galvanized a collective mindset to adapt and react in ways that we've not previously witnessed in healthcare. Part of that success was driven by the transformative insights enabled by AI. He posits that we should think about how we can sustain this force for good by addressing inequities and shortcomings in our healthcare systems, creating a more powerful and predictive framework that cares for everybody in sickness and in health.
Just as this book was coming to market the COVID-19 pandemic hit. It changed everything. It forced health systems to rethink existing processes to quickly respond to this global health crisis. For consumers, the pandemic cut to the core of how the system affects what they universally cared about, which is their health, the safety of their families, and economic security.
The good news about the pandemic is that showed that clinical and health leaders around the world are capable of agile change and transformation. Systems known for changing at glacial speed suddenly began changing at warp speed.
My new book picks up where the old one left off. It chronicles the good news of the pandemic, which includes all the things we learned from fighting a global pandemic. This includes how the pandemic accelerated the use of AI. Health leaders stepped up to fight and win the battle….The “weapons of choice” used to drive agile transformation were AI-driven solutions.
And so, after chronicling the lessons we all learned (and are continuing to learn), it looks ahead to define how we can apply our learning to solving other big challenges in healthcare, including things like chronic disease, mental health, the opioid crisis, and the other issues affecting the health and well being of citizens.
Before the pandemic, most health and clinical leaders recognized the limits and inefficiencies of the existing system and had a sincere desire to make it better. The pandemic came along and became a forcing function. It accelerated our thinking and willingness to change. It tested our ability to harness the power of the Intelligent Health Revolution to do good.
Eventually the pandemic will become an endemic issue like the chicken pox. As the pressure lessens, there will be a push among some health leaders to “return to normal.” The real question is what does the new normal for healthcare look like? Whatever your answer is, the use of AI and intelligent tools will continue to escalate. I think we'll also see growing pressure from those served by health systems to change.
Let's look at telehealth. Research and clinical literature on telehealth demonstrating its value and clinical efficacy goes back 30 years. Why did it take a pandemic for the healthcare industry to pull this forward? The majority of consumers who experienced telehealth for the first time mainly gave it high marks. Here in America this includes older populations like Medicare beneficiaries. One study shows that 93% of them were satisfied with the experience.
While there is a desire in some sectors of the healthcare industry to go back to old ways, I think the momentum for change will continue for two reasons.
First, I think the pandemic shifted consumer and employer sentiment. Most were already feeling worn down by inefficient and disjointed processes used by health systems. When COVID hit, consumers used their “lockdown” time to find in-home alternatives to manage out-of-home activities. As this happened, healthcare rose to the top of their list.
Serving the needs of people must become a generational thing rather than the historical “one size fits all” approach taken in the past. Baby boomers are more likely to accept the traditional ways healthcare works. Millennials and Gen Z'ers, on the other hand, want to do a medical consult from the same place they order dinner, which is their couch. AI and intelligent applications give us the ability to better serve the unique needs of consumers while making the system more efficient.
The second factor coming into play is how health systems will make use of AI. Traditional health systems will mainly use AI to make existing systems and processes more efficient. Intelligent Health Systems, on the other hand, will leverage AI to rethink the entire delivery model. Their focus will be on leveraging the Intelligent Health Revolution to their advantage to efficiently change how the system works across all touchpoints, experiences, and channels. This, by the way, will apply to both the consumer experience as well as rethinking the experience of clinicians and caregivers.
Tom Lawry speaks at PHM's second annual HealthFront,held in April 2022 in New York
Automating work means things done by a human today will be done by a smart machine going forward. These are jobs consisting of highly repetitive activities with little variance in the work being done.
Augmenting work means that AI will automate some aspects of work done by humans but won't replace them.
Augmenting work is where AI will shine in supporting all knowledge workers in healthcare. It's a vehicle by which we can change outdated and inefficient work processes. The important point here is that such change requires leaders to think and act differently. It requires creating a culture that embraces change. This is what I call the AI Leadership imperative.
We're turning one of the most talented, dedicated workforces on the planet into data entry clerks. A study by Stanford shows that a doctor in America spends more time doing documentation than they do with patients.
Another study by McKinsey suggests that up to one third of activities performed by clinicians today could be automated by AI. Again, these are typically the repetitive, lower value activities we force people to do that could be done by intelligent machines. So, imagine giving clinicians back a third of their time to spend with patients, do research, or get home for dinner more often.
In this regard, I often like to pose this question: Which is a better predictor of future health—your genetic code or your zip code? Research and data are increasingly pointing to how socioeconomic and environmental factors play a greater role in health status than one's DNA.
As far as the second part of your question regarding building trust, it's important to recognize that existing health systems are producing very uneven results based on a number of factors including race. The challenge for us now is to work to ensure that the biases and inequities that exist in the real or physical world of health and medicine don't cross over to the digital world through algorithms.
Another question I often pose to health providers I work with is whether their AI plans are aligned with the organization's diversity, equity, and inclusion plans. Addressing inequities in the real world without consideration of the fairness and equity of your AI or digital transformation plan is kind of like squeezing a balloon…You think you are making progress in one area but in reality, you are transferring the same issues to the digital world.
There are many articles and studies that demonstrate that bias is creeping into the digital world through algorithms. These are driving things like clinical decision support tools or pre-screening diagnostic images.
As we develop and put such algorithms into common use, we must stress-test them to ensure that their accuracy and efficacy produces benefits for all.
Health systems are increasingly using algorithms to guide clinical decision-making, but a good number of them rely on a specious assumption that race is somehow an immutable biological characteristic.
The use of race-based clinical algorithms don't just fail to “correct” for health determinants that are clinically influential, their use reinforces racial biases that exacerbate health inequity. The American Medical Association should be lauded for recently calling out race-based algorithms for what they are: sources of harm.
And so, giving visibility and voice to these issues is the starting point. We must now move towards employing a growing number of tools and processes that can be used to evaluate and “stress-test” algorithms to ensure that they perform equally well with all patients and citizens.
We also need to open up the lens on what data exists and what is used today in healthcare to assess, treat, and manage health and medical conditions. This is where social determinants of health can play a huge role in improving treatments and the health status of all.
And so, the good news is that much of this data is readily available for use by healthcare. AI is a vehicle by which we can leverage such data to better manage the health of all citizens not just some citzens.
It can come in unknowingly, through the conscious or unconscious bias of those developing and deploying it. It comes from the bias inherent in many of the data sets used to develop, train, and test an algorithm. It can also come into play based on how it is put into common use.
The key point to remember is this: AI can be deployed in keeping with all laws. It can be compliant with all regulations like HIPAA or GDPR yet still be unethical and perpetuate the biases and inequities that we're dealing with in the real world today.
As AI becomes more pervasive and impactful in the delivery of health and medical services, we must do more to champion and require that it's done in ways that are fair and responsible. I'm proud of the responsible AI efforts Microsoft brought forward long before this was a popular topic. We've had an office of responsible AI for years that has created tools, protocols, and processes which have guided our development and deployment of AI-driven solutions. We're now taking much of what we've learned and pushing resources out into the market to help others in building their own approaches to delivering AI responsibly.
Beyond AI itself, the broader discussion of health tequity must include giving voice to equal accessibility to reliable connectivity. For example, there are parts of the United States today that do not have the connectivity which is sufficient to support things like telehealth visits, another area is digital literacy. Many people have not grown up using computers. Some of this is generational. The lack of digital literacy can also be tied to social determinants.
When it comes to data in general, there were five exabytes of information created between the dawn of civilization through 2003, but that much information is now created every two days.
When it comes to health and medicine, a newly minted physician in 1950 practiced for 50 years before they saw medical knowledge double. Today this is happening roughly every 75 days.
And so we have massive amounts of medical data. This can and should be used with many other types of data, like social determinants of health, data from wearables, and environmental data. The challenge we face is that much of the data is “dirty data” that needs to be staged and cleansed to reduce the propensity for things like bias. It comes from disparate sources making it difficult to combine and use on an agile, cost-effective basis. Many systems used by healthcare today, like EMRs, were designed as closed-loop repositories to collect data on a single patient. They are not so good at ingesting or adding things like streaming wearable data or easily aggregating data to look at how to manage the health of populations such as looking for patterns among all citizens in a geography with diabetes.
When you look at how AI systems work, the number one thing to get right is the data. With this in mind, I find it interesting that I'm often asked to do keynote talks on AI at major conferences, but I've never been asked to do a keynote on creating and managing a modern data estate, which is the key to getting AI to provide value at scale.
If you want to have AI deliver measurable value across a health enterprise, the focus should be on getting your data estate in order.
While there are many working models for health systems that vary based on which country one lives in, the majority of the health delivery systems operate on a “break-fix” model. This is where citizens are guaranteed healthcare, which means that when they are sick or injured, they will be taken care of. Everyone talks about health, but if you follow the money, the majority goes to taking care of people once they have a medical problem.
Given changing demographics, growing sophistication of intelligent technology, and consumer expectations—I'm hoping there is a new model emerging. In my book Hacking Healthcare, I have a chapter titled the “Coming Care Calamity”. It highlights the collision course we are on when it comes to the unsustainability of the “break-fix” model. For example, in America a baby boomer turns 65 every 15 seconds. Those 65 and over use three times the amount of health resources as those under 65. If you look at the impact of how we poorly manage things like chronic conditions and multiply these costs by the number of those turning 65, you'll understand that we'll soon see that we see a collapse in both the economics and the outcomes of older Americans who deserve better.
Data and AI are platforms that allow us to change these dynamics. They allows us to shift the focus to improve consumer engagement and personalize care. We have the ability to do better but political and health leaders must recognize this and shift the financial incentives toward first keeping people healthy and then taking care of them when they are not.
Essentially, cloud computing and artificial intelligence are partners. One enables the other. Without cloud computing, current and future AI capabilities would be very limited.
Done right, cloud-delivered intelligence enables agile innovation at a near-limitless scale.
Creating and implementing a strategic cloud strategy is more than an efficiency play for IT. It's the engine that will drive the speed, scale, productivity, and level of innovation an organization achieves now and in the future.
The cloud offers tremendous value, but the benefits don't appear magically. Fully realizing its value requires a well-defined, value-oriented strategy and coordinated execution by IT, clinical, and business leaders.
Organizations with high cloud maturity exhibit different adoption mindsets than their peers. One study of AWS, organizations with higher cloud maturity shows that they share several traits: they were early adopters of cutting-edge technology (71 percent), they aggressively innovate (72 percent), and view technology as a competitive differentiator and critical enabler for launching and building new services and solutions (79 percent). By being the first to move, these organizations gain considerable experience, outstripping their peers in cloud outcomes.
For example, the Microsoft Cloud for Health was the first cloud-based platform designed to allow precision medicine companies a platform that accommodates rapid development of intelligent solutions while streamlining compliance with regulatory standards for health and medicine.
Looking ahead, Kai-Fu Lee is one of my favorite authors and leaders. He's a former senior executive at Microsoft, Google, and Apple. He's also the bestselling author of AI Superpowers. In his latest book AI 2041, Lee gives us a glimpse of what an AI-driven world will look like two decades from now. Here's what he has to say about the Intelligent Health Revolution: “When we look back in (the year) 2041, we will likely see healthcare as the industry most transformed by AI.”
I'd love to have a few pints with Sir Isaac Newton, who created the field of predictive science when, in 1687, he penned the Mathematical Principles of Natural Philosophy. This treatise established concepts of modeling and put forward the notion that mathematical equations could be used to predict how systems evolve.
The second would probably be Ada Lovelace. Ada was the daughter of the poet Lord Byron and one of the first recognized female mathematicians. She was the first to postulate that machines had applications beyond pure calculation and published the first known algorithm, in the early 1800's, intended to be carried out by a machine. As a result, she is often regarded as the first computer programmer. She also exemplifies the early days of AI and computer programming when women drove most of the progress that we build on today. I remain puzzled by the changes where today women are in the minority of clinical and data-science leaders driving us forward.