With Clement Goehrs, MD
CEO and Co-founder of Synapse Medicine
In this Digital Health Wire Q&A, we sat down with Synapse Medicine CEO and Co-Founder Clement Goehrs, MD to discuss the challenge of accessing up-to-date drug information and how that data can be used to improve care delivery.
With more medications hitting the market every week, providers face the impossible task of tracking countless new interactions, and software developers don’t have it any easier as they look to equip providers with the right tools for writing safe prescriptions. Dr. Goehrs co-founded Synapse in 2017 to help them do just that, with easy-to-implement UI components making real-time drug data and decision support more accessible than ever.
Can you give the audience a quick introduction to Synapse Medicine and the story arch that brought us to your current solution set?
To put it simply, our mission is to make it as easy as possible to access medication information – wherever it is – and to help providers make the best clinical decisions using that data.
During my time as a physician, I saw first hand how difficult it was to find information for optimizing a prescription, and that we’re also facing a huge public health problem due to people dying from avoidable medication errors. Those are the problems we’re aiming to solve.
And when I say that we’re trying to make drug information easy to find, I mean for our end users (providers, pharmacists, nurse physicians), as well as our clients (usually EHRs, ePrecribers, or telemedicine companies – any type of software company building for clinicians that wants to add a drug information component).
Can you give us a deeper dive into Synapse Medicine’s clinical decision support solution, and walk us through what that user experience looks like?
One of the things that we understood very quickly was that providers don’t want to have multiple tools. That means that if you’re providing some kind of clinical decision support, and you want your end user to have the best experience, you also need to have a seamless integration inside the EHR.
You also have to be able to provide a wide range of tools without making the solution overly complex, and we do that with what we call components. Our components leverage our APIs with a UI layered on top for specific use cases, like drug interactions or side effects. That makes them easy to integrate and customize, but the user doesn’t even notice it’s a third party feature.
To give you an example, if a physician starts a prescription within the EHR, as they begin entering drugs they’ll start to see safety notifications on the side effects, potential interactions, and dosages. All of that is communicated in a way that’s easy to understand.
No physician wants to see a pile of alerts, but they do want to have info on how to help their patients. So instead of just saying “there’s a severe interaction” like most tools do, we take it a step further by saying “you might want to divide that dose in half to avoid these side effects, and here are the publications supporting that decision.” That explanation is super important.
What are some of the big trends that you’re keeping your eye on, and how do you see them continuing to unfold going forward?
One of the main trends that I’m seeing is that there are a lot of legacy players, very big EHRs and software companies, that aren’t able to innovate at the pace they would like to because they have so much on their plate.
These companies face so many regulatory hurdles and have so many things they need to build, so with all the innovations and AI progress happening every day, more of them have been saying “we can’t keep building every single use case, and we are going to start working with companies that are focused on this specific problem.”
As more EHRs and telehealth companies start to integrate with other players to get the best of both worlds – breadth and depth of features – there’s a big opportunity for the startups working on those use cases.
Has there been anything that surprised you about the recent AI momentum, or as you implemented any new technology into your own platform?
AI is a fundamental part of a lot of what we do, and of course we continue to learn more about it every day. One project that we did with France’s equivalent of the FDA, and probably one of the first healthcare AI projects deployed at that scale, was to help monitor vaccine side effects and other drug interactions on a national level.
As we developed that algorithm, every time we made a major jump in performance, it was never because of the mathematical model. It was always because we ended up back at the data.
We understood the data very well from a medical point of view, and also had physicians on our team that could point out things like why we shouldn’t train the model on certain data. That’s what really enabled most of the fine tuning.
So for the short answer to your question: I’ve been surprised by how much truly understanding your field, and truly understanding your specific use case, is actually what makes AI smarter. It’s not just more data and a bigger model. I think the real progress wiIl come from the AI teams that have physicians and data scientists working together on performance.
What advice would you give to providers or startups that are thinking about improving their own prescribing strategy?
Number one, I would say to think deeply about the end user, about the physicians and the prescribers. That may seem trivial, but that experience is so important, and it’s surprising how often it gets overlooked.
Number two, think about scalability, particularly around adding secure data. What worked for 10 physicians in the beginning might not work as you scale. You’ll probably need better data, and you’ll probably want to do something with that data. If that data isn’t secure from the beginning, you’re going to lose a lot of time if you have to rebuild everything.
That brings me to number three: I wouldn’t recommend doing that by yourself. There are new drugs every week. There’s drug information all over the place, a lot of terminology, and few standards tying it all together. It’s a complex mess, and it would be a mistake to only consider the resources you would need to build it, because it’s maintaining it over time that gets really painful. You can probably avoid a lot of pain by having someone else do it.