Last month, we kicked off a new research study on the future of innovation. In the weeks that followed, we’ve had many fascinating conversations with industry leaders on the topic.
Note: Union Healthcare Insight members will get access to our full report released in the fall, as well as a facilitated discussion on the topic at our inaugural in-person event early next year. We are also hosting a public webinar to preview our research in November; free registration is available for all of our webinar events here.
But in the meantime, I wanted to offer some valuable lessons we have been learning, or re-learning, about the challenge of navigating a subject as broad an buzz-worthy as innovation.
To keep from getting lost in the sheer volume of information and opinions out there, here are three guiding principles we are currently using to ground our research in pragmatic, actionable insight. In an effort to move beyond the merely conceptional, I’ll be pairing each principle with an example from the specific segment of healthcare innovation that we are finding to be most susceptible to handwaving and hype at present: artificial intelligence (AI). Fear not, this blog post won’t be another breathless account of how AI is going to revolutionize healthcare as we know it; nor am I throwing cold water on an exciting frontier (one that happens to be supporting the valuation of many worthy digital health startups). My aim is to help you derive the most insight possible from what you're reading and what you're hearing about innovation.
Principle #1: Start with a problems-first approach
The phrase “don’t bring problems, bring solutions” is everywhere in the corporate world. But while it may be good advice for early-career professionals trying to get ahead, we threw it overboard right away in designing our innovation research. Much of the existing innovation content chronicles new products and services. It's a natural consequence of there being so many players eager to market their latest and greatest offering, but when the information market on innovation tilts so hard toward listicles of XYZ new great technologies or clinical therapies, that makes it hard to discern which innovations are true solutions, and which are merely hammers in search of a nail. As one digital tech expert said to us: “every press release for a new product will tell you all about how this tool is going to solve your problems, but it’s usually long on platitudes and short on details.”
Yes. It’s true that many innovations can generate their own demand—no one was clamoring for a smartphone until it was released, and now we can’t live without them. So keeping track of promising new tech with unclear market application/uptake has value. But not only can we already count on the general information market to profile cool new tech (remember: listicles aplenty), but that demand is much easier to muster when resources are abundant. Unfortunately, resource abundance is not a reality for many healthcare strategic buyers at the moment.
In 2020 and 2021, obvious problems aside, the innovation market was in many ways booming thanks to a huge spike in the amount of money floating around the healthcare system. Volumes were low, letting insurers hold onto a lot of premium dollars; investment in every facet of healthcare tech and innovation skyrocketed; and even provider organizations, grappling with Covid, got an influx of CARES Act relief dollars. Two years later, those revenue streams are drying up, and the buying power of purchasers and investors is somewhat down. Margins are tight, and most strategic buyers aren’t interested in keeping up with the Joneses—they face large, specific challenges that require thoughtful, tailored solutions. So yes, by all means let's keep an eye on unconventional inventions for their own sake, but also, let's prioritize identifying the innovations that directly address the top issues faced by industry leaders today. They may not feel 'new'; top issues include workforce, affordability, and risk. However, they have real urgency, making for a hot market for emerging innovations.
Centering on the problem to be solved is especially important when discussing AI. Versatile and transformative, AI can appear to be a panacea for all of our healthcare woes. But with a technology as flexible as AI (really more of a concept than a technology, but we’ll get into that later), we’re seeing a lot of developers creating AI products, then trying to make a case for its ability to solve whatever the problem of the day is, even if that’s not what it was designed or most suitable for. Stakeholders are left to sift through various offerings trying to figure out which, if any, will most effectively advance their strategic goals without a clear understanding of how the product was engineered or for what purpose. Given the challenges we outlined above, some of the most promising tools are those that can extend FTEs, streamline back-office functions, and support population health efforts.
Of course, even the above are still broad, directional categories. One way to help inform the tough choices strategic buyers have to make, even when comparing tools that address a specific enterprise problem, is by using more precise language about how the innovations operate. Which brings us to our next principle…
Principle #2: Be specific
When we talk about healthcare innovations, it’s easy to fall into the trap of making blanket statements about entire fields or subsectors. We’re used to having conversation about technologies that are functional, like medical devices or new pharmaceuticals; more and more though innovation conversations are about changes that are foundational, like virtual care and AI. These ‘foundational’ capabilities are chassis upon which countless other derivative innovations can be built, making it hard to have nuanced conversations, or even ones in which everyone is talking about the same thing at the same time.
This specificity feels particularly absent in conversations around AI. The use of AI itself is not new in healthcare. In fact, researchers have been applying AI to medicine since at least the 1970s. It also doesn’t stand alone as an “innovation” the way, say, a wearable smartwatch or CDM software program would. Rather, AI is a field of study and methodology for building ‘intelligent’ outputs in various forms. That means there are many flavors of AI; in fact, the product of any AI model can be completely altered at the input level (i.e. what data it’s given), the analysis level (i.e. what it’s searching for and how it’s weighing the findings), and the output level (i.e. what it’s programmed to produce). So when we’re talking about AI, we’re actually talking about an almost limitless constellation of variables and corresponding results.
One basic distinction that we have been finding valuable to conversations around AI in healthcare is to make explicit the difference between ‘generative’ AI and ‘predictive’ AI. In the simplest terms, generative AI uses massive datasets to make an output that is completely new. Programs like ChatGPT or Midjourney create text or images based on given data but that do not already exist in the world. Predictive AI, as the name suggests, also intakes data but instead uses it to make prediction about future or uncertain outcomes. For example, programmers could build an algorithm to predict if it will rain on a given day, or if a banking transaction was made fraudulently, or (bringing it back to healthcare) which subsets of patients are most likely to be avoidably hospitalized. While generative AI tends to get most of the attention these days, it may actually be predictive AI that will have a more significant impact on healthcare in the near term. Which brings us to our final principle…
Principle #3: Pick a relevant timeframe
In the forward-looking world of innovation, nailing down a timeframe is key, and different timeframes can uncover different insights. If you’re in the business of day-to-day operations, having a pulse on the innovation landscape of the current moment has a lot of value. If you’re a casual reader just trying to imagine the future, or if you hold a position that necessitates very long-term visioning (perhaps in venture capital, R&D, academia, or brand strategy), thinking about what innovation will look like in 10, 20, or even 50 years’ time might be the relevant positioning for you. However, for many healthcare executives overseeing innovation or strategy, we believe the most useful time horizon is generally two to five years: close enough that we can make reasonable predictions about the availability of technology and the conditions in which it will need to operate, but far enough away to allow for proactive rather than reactive planning. Not coincidentally, that two-to-five-year timeframe is the one we’ve chosen for our study.
With our research we hope to uncover not only which innovations are going to move the industry forward, but also which are getting more hype than they deserve. As we've started to tease these ideas out with our innovation and strategy experts, we've noticed (perhaps unsurprisingly) that AI was ending up on both sides of the conversation. Some wrote it off as vaporware, while others touted it as the most transformative thing to hit healthcare in decades. And yet, when we overlayed the context of time, these positions were often less contradictory than they initially appeared. Proponents were thinking further in the future about the benefits that could be unlocked if AI lived up to its promised potential. Detractors were primarily considering the countless regulatory, security, and resource barriers AI would have to overcome in the short-term, not to mention the long list of challenges taking up all bandwidth healthcare leaders have at the moment. It’s not so much that the perspectives were at odds, but rather than the technology was being considered at different timescales and maturity levels.
Let’s apply this lens to the conversation we were having about generative vs. predictive AI. Generative AI has received significant attention this year not only because tools like ChatGPT have taken it from the realms of science fiction to reality for the average user, but also because its ability to generate completely novel outputs feels fundamentally new to healthcare. But that novelty makes implementation exceedingly difficult in the near-term as developers struggle with uncovering appropriate use cases for such tools—and that’s before tackling the barriers mentioned above. Hence the reasonable skepticism of our experts. Predictive AI, on the other hand, is the logical extension of clinical decision support, population tracking, and workflow tools that are already in use across the healthcare system. While it might not appear as ground-breaking as its generative cousin, predictive AI can be a practical tool for healthcare organizations in the short-term, as well as a steppingstone into the future imagined by AI promoters.
That's all I've got for today, but our innovation research is still in full swing. If you’re a healthcare leader with a perspective on the future of healthcare innovation, reach out to me at email@example.com—I’d love to chat. And if you’re interested in more innovation content, register for our Future of Innovation webinar on November 16th or contact us to learn more about membership options.