Better data should lead to more informed decisions. But that data is locked away in papers. We are freeing it
As a part of my series about “The Future of Healthcare” I had the pleasure of interviewing Matthew Michelson, the founder and CEO of Evid Science. Evid Science works to democratize access to patient-level therapy evidence by unlocking the results published in the medical literature. Given the scope of this task (more than 7,000 medical papers are published daily!), Evid Science leverages Artificial Intelligence to “read and understand” the literature, taming this mountain of hugely useful information. A scientist and technologist by training, Dr. Michelson has spent most of his career at the intersection of science and product development, bringing new data-driven products to market based on scientific achievements.
Thank you so much for joining us! Can you tell us a story about what brought you to this specific career path?
I’ve studied a lot of topics and done many things. I actually started as a creative writing major as an undergraduate, but then switched to computer science, largely driven by my love of video games. I’ve always loved science and language, so when I learned I could analyze language with computers, I gravitated toward that. I’ve also worked in a variety of industries — from defense contracting to finance to marketing automation and now healthcare — but my emphasis has always been on creating brand new products based on brand new technologies. So, now I am an entrepreneur, blending my passion for language and computers, with a pressing need to unlock medical evidence.
Can you share the most interesting story that happened to you since you began your career?
It’s hard to pick just one, but a funny story that sticks out had to do with a company I worked on that was commercializing technology related to identifying synonyms from large data sets. The company leveraged the technology to match resumes to job descriptions, automatically. One of the things it “learned” (since it was based on machine learning) was that certain words went together and could indicate someone’s skills for a particular job. It was neat because it could go so much deeper than key word matching.
So, we are humming along with the product and then someone was hiring for a System Administrator, which is someone who manages your IT, computers, internet, etc. And the system kept suggesting bartenders for the job. When we dug into it, it turns out the system thought that people who had “servers” or “server” in their skill set would be good matches for the job. The people with “servers” were correct, since servers in this context means computers. But the people with “server” weren’t. They were the bartenders who had experience serving food! But the machine thought server and servers were the same, and it is a rare enough skill that the system thought was very important for the job! So that was funny, and pretty enlightening as to how the machines can do their task pretty well and still fail in funny ways that would be obvious to people. (No idea if any bartenders ever ended up getting hired. Who knows? At Evid, our Principal Scientist is a sculptor too.)
Can you tell us about your “Idea That Might Change The World”?
At Evid Science we are unlocking the results in the medical literature. Many people are surprised to find out that healthcare decisions aren’t always made using the latest results and data. This isn’t because anyone is malicious or incompetent, it’s simply because it’s overwhelming to do so. One study estimates it would take almost 90% of a person’s waking day to read all of the relevant papers in one field. So, that only leaves 10% of the time to eat, sleep and work! And that study was pretty old! Now, more than 7,000 medical papers are published daily, so it’s literally impossible to stay current on all of the results. The goal of Evid Science is to use machines to read the papers, like a person, and pull out the main results. Then people can ask questions such as how effective therapies are relative to other potential medications, how safe they are, etc. And since the machine actually can read that many papers, the results become timely and broad. Essentially, we are using machines to overcome a herculean task that can have huge importance.
How do you think this will change the world?
In healthcare, we see two trends converging. First, there is the notion of evidence-based decision making. Essentially, the idea is that decisions in healthcare should be made based on data and results, rather than what the answer was historically. In other words, if there is a newer and better treatment, we should use that (if the evidence supports it). Unfortunately, this isn’t always the case because it’s so hard to stay on top of the evidence to make those determinations.
Second, there is the concept of personalized healthcare — treat me as an individual, rather than the “average” patient. If the guidelines say to treat patients with X, that’s the average case, and, for instance, may not be appropriate for an older patient with multiple diseases at once. In that case, maybe we treat this patient differently because she is co-morbid and elderly, so the current therapy should be adjusted.
Now, imagine a future where we can combine these two approaches! We can tailor the evidence on a patient-by-patient basis, because now we have all of the evidence at our fingertips. So, in the last example, we find evidence for treatments that are specific for people with the same co-morbidities who are also elderly.
In a nutshell, better data should lead to more informed decisions. But that data is locked away in papers. We are freeing it.
I suppose one issue is that technology such as ours can unleash huge volumes of information that was previously hard to access. And like anything, too much information can be challenging to deal with. It might cause information overload, or perhaps people can mis-analyze the volumes of data, for instance.
Was there a “tipping point” that led you to this idea? Can you tell us that story?
I am very fortunate because I have access to amazing healthcare, personally. I have terrific doctors and many physician family members. But even with the best doctors and access, I was shocked by how little the evidence plays a role in healthcare decision making. As a scientist (and data scientist, especially!) it was crazy to me that such important decisions didn’t have sufficient data to back them up. And I was further shocked that it was mostly due to the challenge of getting that data in the first place. So, we wrote a proposal to the National Institutes of Health to fund research to see if it would even be possible to automate access to the evidence. And clearly it was. That happened a few years ago and formed the genesis of Evid Science.
What do you need to lead this idea to widespread adoption?
Our challenges are human, not technical. We’re confident we can overcome any technical challenges in teaching the machines to read. The deeper challenge is that people are, and rightfully so, skeptical of machines, especially as that relates to healthcare. We need to gain their trust, and then adoption will happen. This isn’t unique to Evid Science, by the way. I think this is true for AI in general. We need more transparency and trust.
What are your “5 Things I Wish Someone Told Me Before I Started” and why. (Please share a story or example for each.)
1. If you build it, they probably won’t come…This is becoming common knowledge, but a big lesson I’ve learned from the various start-ups I’ve worked on is that it’s not the case that, “if you build it, they will come.” This only works for ghost baseball-players. More likely, if you just go ahead and build it, you’ve built the wrong thing or no one will come anyway. You need to go out and talk to people you think will use your idea, get their feedback first, etc. With Evid Science, we initially thought it would be a point-of-care decision-support system for doctors. But the last thing doctors want is another technology tool that pulls them out of their already busy workflow. So instead, we focus more on the research side of the world to add value. We came to this conclusion after having more than 30 conversations with doctors at all levels (new residents to Chiefs), life sciences researchers, insurance companies, etc. You need guidance to build useful things.
2. It’s not very likely that someone will steal your idea… Going hand-in-hand with above is that I’ve learned to share all of my ideas freely. First, I realized that if my idea is simple enough for someone to steal, then it probably wasn’t a protectable business anyway. And second, it’s much more about the execution than the idea. Computer science is a great example of this! Look at Kaggle (the website for data science competitions), and you will see, everyone is attacking the same idea (e.g., given this data, solve this problem) but they are all doing it differently. And some win while others don’t. Share your idea and get the feedback. Then do it better than anyone else.
3. If you just build technology that’s interesting, that’s research.It’s only a product once it solves a real problem. In other words, if your goal is to build something cool, that’s great, but if your goal is create a product or a company, start with the problem first, and not the technology. It’s much harder to massage your technology to fit a problem, then to start with a problem and create technology to solve it. If you do this in combination with #1 above, you are much more likely to create a product people want and care about. At one point, we built some really amazing software to predict interests of people based upon what they said about themselves on social media. We wrote some academic papers about this (essentially computational social science), and then started hunting around for ways to apply it practically. We tried to start a business that commercialized the technology, but we didn’t really have a business problem to solve. So we quickly shuttered that company. In that case, we literally started with research, buts it’s a good example of how hard it is to fit whiz-bang technology into business. Do it the other way around.
4. Communicate to win!Communication is crucial. It’s a huge advantage to write well, present well, etc. You will gain wider audiences, more customers, and happier employees if you are clear and your content is interested. Yet, so many people assume these skills are innate. They aren’t! Just like anything, they take work and practice. So, if you weak at communicating, find a way to practice. When I was a graduate student, my Ph.D. advisor gave me tons and tons of feedback on my writing and my presentations. Obviously, we talked a ton about the research, but he also cared deeply about honing our communication skills. And this was unusual, I think, because we were in a purely technical field. But it paid off tremendously! I didn’t realize it at the time, but it’s been a huge advantage for me.
5. Frankly, you’ve just got to do it all. When you first start a company, there is a feeling that there are so many functions — sales, marketing, product development, engineering, HR — that your instincts might be to hire out as much functionality as you can. I’ve found the most value, however, in trying and failing to do many of the jobs myself. You will get better at them (so you can do these tasks, if needed!) and you will learn what you are looking for when hiring. I am not a great salesperson, but I practice and get better. And, most importantly, I learned what we need in a great business development lead. We used that then, to make our amazing hire in that area.
The future of work is a common theme. What can one do to “future proof” their career?
The most important thing is to have a growth mindset, so you can be flexible to change and willing to learn. Those two components are key. The future isn’t knowable, so you need to be able to adapt to the new situations and learn how to be productive in them. I’ve probably learned a dozen programming languages in my life, depending on where I was working and what I was working on. I’m not an expert in most, but I was proficient enough to use them, in a team, and be productive. Who knows, maybe aliens will invade and we’ll all need to learn Klingon to keep our jobs!
Based on the future trends in your industry, if you had a million dollars, what would you invest in?
I would invest in Evid Science! Cheesy answer, I realize — but I truly believe in our mission to unlock the evidence in the literature. It’s such an important, but underutilized, resource, and the technical challenges are totally solvable. So why shouldn’t that be a focus?
Which principles or philosophies have guided your life? Your career?
Relativism — nothing should be thought of as absolute. I am always willing to change my mind if something can convince me. Similarly, most people make decisions that are “locally optimal.” They make the best decision they can, at that time (including myself!) so I don’t begrudge someone if that choice is a mistake later, because who could have known at the time? Concretely, in software development sometimes you look back on some code someone has written, and you feel very judgey about it (what an idiot to have done that!). But that’s super unfair — that was probably the best way to do something at that time, given all of the constraints around time limitations, etc. I’ve found that being flexible in my thinking has benefitted my career and my life.
Can you share with our readers what you think are the most important “success habits” or “success mindsets”?
I’ve repeated it a lot, but flexibility and willingness to learn new things. I’m a writer turned scientist turned entrepreneur, but the common theme is a love of learning new things and a willingness to adapt to what suited and interested me. Whether I am successful is debatable, but I am certainly happy with my work and the experiences I’ve had in the past. So, simply be open to change and to learning for that change, good things will come!
Some very well known VCs read this column. If you had 60 seconds to make a pitch to a VC, what would you say?
The future of healthcare is data, and some of the most important data is evidence. We will have the largest dataset of healthcare evidence in the world, and which is only growing (If you read that slowly, takes about 60 seconds!)
How can our readers follow you on social media?
We are on Twitter, but only lightly. It’s better to follow our company and employees on LinkedIn, and read our amazing blog! https://www.evidscience.com/category/blog/
Thank you so much for joining us. This was very inspirational.
— — — —
About the Author:
Christina D. Warner is a healthcare marketer at Walgreens Boots Alliance. She is a Duke Business School alumnus, and has innovated commercially for Northwestern Feinberg School of Medicine, Veniti (now Boston Scientific) and Goldman Sachs. Christina is a regular columnist for Authority Magazine and Thrive Global and and has been quoted in many national publications. You can download her free ‘How To Get Into the C-Suite and More: top secrets from CEO’s, political figures, and best-selling authors. Connect with Christina atLinkedInor Twitter