Recorded at the Health 2.0 Conference in Las Vegas, this HealthTech with Purpose episode features Cos Fantis, CEO of UK startup Encode Health, in conversation with Mindbowser CEO Ayush Jain. Cos, a medically trained doctor with an interest in neurosurgery and a former lead surgical officer at Touch Surgery, explains the problem Encode set out to solve: hospitals ordering surgical and medical inventory against fixed stock levels, often tracked in spreadsheets, which leads to waste and products that expire before use.
Encode connects an electronic health record system with an inventory management system to match clinical demand with medical supply. Cos walks through how the platform reads GS1 barcodes and data matrix labels, uses optical character recognition to onboard new products, gives hospitals and suppliers the same live view of stock, and forecasts demand per surgeon and per procedure, a lesson carried over from Touch Surgery. He also covers the harder parts: producing evidence in an evidence-based field, reassuring clinicians about security, handling different compliance rules in the UK, Europe and the US, watching for bias in training data, and early work on generative AI for clinical handovers and post-operative letters.
Cos describes hospital ordering as surprisingly basic. Stock levels are set, a supply chain manager checks them and orders, and the result is waste, or devices that expire on the shelf before anyone uses them. Encode’s approach is to connect an electronic health record system with an inventory management system, so ordering follows what clinicians are actually scheduled to do: “we’re matching clinical demand with medical supply.” On top of that link sit forecasting, expiry alerts and inventory optimized by specialty.
The hard part of that design is the EHR connection itself, which has to be rebuilt for every system a product meets. Teams facing it can have Mindbowser build the EHR integration for their product, or use ConnectHealth, which connects a product to 20+ EHRs through one integration model and runs inside the customer’s own cloud.
By Cos’s account, citing a UK audit he asks listeners not to hold him to precisely, around 90% of UK hospitals have no electronic inventory management system. Products are scanned with a handheld scanner into Excel, and supply chain managers and suppliers reconcile the numbers back and forth. Encode gives both the hospital and the supplier a dashboard over the same live inventory, “so there’s no dispute about what’s happening.”
The data comes from labels already on the products. Most carry GS1 barcodes, QR codes or data matrix codes holding the product number, expiry date and GTIN, and Encode matches that data to the product itself. For products the repository hasn’t seen, it uses optical character recognition to read the label and onboard them.
The lesson Cos brought from Touch Surgery, where he trained surgeons through an app, is that every surgeon performs the same procedure a little differently. Encode maps that: in a department with ten orthopedic surgeons, it doesn’t just map the procedure, it maps how each surgeon performs it, and forecasts inventory on a contextual basis, per person and per time of year.
Selling to an NHS trust means meeting specific regulations, and Cos’s approach is to comply and then try to exceed them. Because compliance needs differ between Europe, the UK and the US, he argues one team can’t realistically master all of them, so Encode works with external groups that have done it before. Miguel Amador of Complear makes the same case for regulatory compliance support across Europe, the US and the UK in another episode.
Cos is candid that Encode is still learning how to apply AI. The team works to keep each dataset specific to how a model is trained and to watch for bias. Beyond OCR, it is developing a generative AI product to help with clinical handovers and to write post-operative letters, with datasets that can be specific to a surgeon or a department. He uses his own field as the example of why that is hard: brain tumor classification changed in 2021 to incorporate molecular diagnostics, and sourcing those small, specific datasets is difficult. The same careful, workflow-specific approach applies to any team building AI for clinical documentation.
[The first 30 seconds of the recording are not captioned.]
It’s an honor, because of the great things that Encode Health is doing. At the HealthTech with Purpose podcast, we’re always looking for interesting stories, especially people who are disrupting health tech, and you’re coming all the way from the UK, [inaudible] healthcare both in Europe and the US. So I was really interested and keen for this talk today. Without any further words, let’s start. Tell us more about yourself and Encode Health.
Sure. Well, I’m medically trained, but before that, I think your listeners would like to know, I went into the army. I was a [inaudible] driver in a conscription model. I was lucky enough to get into medicine, and even luckier to get into St Andrews. That’s the university Prince William went to. I was the first in my family to go on to further education, so it’s been a very interesting ride for both myself and my family [inaudible].
At St Andrews, I was always interested in neurosurgery, and I set up my portfolio to go into neurosurgery. I’m currently doing research with the University of Buckingham on the applications of artificial intelligence in neurosurgery. Before that, I was working with a company called Touch Surgery, where I was the lead surgical officer. I engaged with some of the best surgeons in the world and with medical device corporates, and we trained surgeons how to do surgery through an app. We were the first group to create that narrative around how surgeries would be performed, because surgeons perform surgery very differently, and those small, minute differences add up [inaudible]. So we created that baseline at Touch Surgery, and then I founded Encode Health.
So what are the main goals of Encode Health that you started?
At Encode Health, we believe in personalizing healthcare and providing access to all. We’re still developing; we’re still at the very early stages. I’d like to think that what we’ve done is connect an electronic health record system with an inventory management system. We provide forecasting capabilities there, and we allow hospital supply chains to optimize their inventory based on clinical needs. The way we depict it is that we’re matching clinical demand with medical supply.
Demand and supply, basic economics, but it’s not very well done, and you’d be surprised how basic the way of ordering actually is in hospitals. They set certain levels, and the supply chain manager or representative will go and look at the levels and then just order. This can create waste. It can create situations where certain types of devices or inventory expire before they’re used. So we can create alerts around that, and we can help with an intelligent way of optimizing inventory based on which particular specialty requires certain types of devices and inventory. We contextualize every bit of the field. I said that in a very long-winded way.
And I’d like to say again that we’re at the very beginning stages of development. We’ve been very lucky to have won one of the largest grants that the UK offers. It’s called the Smart award, and that’s helped us develop the platform using different types of artificial intelligence. We’re looking at the regulatory landscape as well, to make sure we’re complying with the direction the regulations are going.
I think you’ve mentioned a lot of different threads that I’d like to talk more about. Maybe let’s start by digging deeper into the problem. It’s a very particular problem that you’ve picked. Many times companies are solving from different viewpoints, especially around patient care, doctors, payers and so on. But this one piece you’ve picked up, connecting the supply chains and optimizing them, is something seldom talked about. So tell us more about this problem. What did you see when you went deeper into it, what was the turning point, and how did you start?
I could talk about the technology and understanding how mobile phones have helped with them, and augmenting that with various types of labeling, for example the QR code and the data matrix that most products should have. So that has helped.
So I imagine most pharmacies would have software that basically manages the orders or the supply chain. What I think is magical about Encode is that it connects that with what’s going on on the clinician side. Is that true? How do you do that?
Yeah. Pharmacies have a manageable supply chain; they’re very good at that. But when it comes to the actual inventory used in the hospital, in the surgical department or the medical department, if you’re an orthopedic surgeon in the UK, for example, [inaudible] having a particular product for a particular patient at a particular time. That’s part of the supply chain issue.
I mentioned that there’s a technology aspect, but there’s also the actual handling of the inventory itself. I’ll talk about that a little more. In the UK there was an audit, and it’s something incredible, and please don’t hold me to it exactly, but around 90% of hospitals in the UK don’t have an electronic inventory management system in place. What they have is Excel spreadsheets. They might have a handheld scanner, scanning the products, and it goes into an Excel spreadsheet or something similar. Then that is used with the information from suppliers, and maybe there’s a dialogue between the supply chain manager at the hospital and the suppliers themselves to match up what’s happening. What we provide is a dashboard for the supply chain representative and also a dashboard for the supplier, so that they have live inventory at both ends, and there’s no dispute about what’s happening.
The way we do that is by leveraging the technology, leveraging the labels that are already on the products. Most of these products use GS1, which means Global Standard One. It’s the way the information is organized within those labels: within the QR code, the barcode, the data matrix. We gather the information in that code and match it with the actual labeling of the product itself. The QR code holds a certain amount of information: things like the product number, the expiry date, the Global Trade Item Number (GTIN). Then we match it with what the product actually is, and we use optical character recognition (OCR) to onboard newer products as they come along. So we’ve created a repository that can be accessed by [inaudible], by all that need to access that data. I think I’ve given you a lot of information.
Yeah, thank you. What I imagine is the beauty of the whole model: once you have more and more data, which can be used as an input to building the system better, you can create more predictions and create more value overall.
Exactly. It comes back to the lessons we learned: certain surgeons do the surgery a little differently compared to others. So every time a particular surgeon does a particular type of procedure, we can map that. Even though there might be ten orthopedic surgeons in one department, for example, we don’t just map the procedure itself. We map how the procedures are performed, and provide the forecasting on a very contextualized basis, per person and per time of the year.
Okay, great. So Cos, one of the things I noted down to discuss with you: once you got to the thesis of the solution, that this is how it would all work, what challenges did you face as an entrepreneur building this solution, and how did you find ways to overcome them?
Well, healthcare is very evidence-based, so one challenge is being able to create the evidence. For example, one of the claims we make is that we will reduce time per patient by at least three hours, because of the way we manage the data that’s shared between the patient, the clinician and the supplier. Obviously all of this information is governed by [inaudible] data protection. What I would say is one of the difficulties is human factors: engaging with the surgeons, making sure we reassure the clinicians, whoever we’re engaging with, that this is safe, and that, as best we can, we’re providing a very secure system, comparable to other secure systems.
So how did you prove that? Because eventually skeptics will always be there, and they would ask how you are more secure, or how you really take care of privacy.
There are certain regulations that need to be complied with to engage with a trust or an NHS hospital, and as best we can, we comply with those regulations and then try to exceed them, with applications of AI in certain areas. That’s the way we go about it. We’ve engaged with various groups that help us comply with the regulations as well. So instead of our team onerously and constantly trying to cover everything alone, we engage with external groups too, to make sure we’re complying with those regulations and haven’t missed anything.
The compliance needs in Europe are different from the compliance needs in the UK, and from how they’re required in the US. So it’s difficult for one team, one group, to understand and apply all of those. That’s why, in general, we engage with other groups that have done this before [inaudible].
I see that AI is a very fundamental aspect of your whole solution. I’ve also seen many companies struggle to get the real power of AI, especially when you don’t have that much data or that much knowledge to apply. So how has your journey been, applying the AI aspect?
We’re still learning. We’re still looking at ways of applying AI. There are certain standards that we meet to ensure safety, but it’s a very wide open question. We could talk about the safety of consenting for data, incorporating that into how the algorithms learn, creating the right kind of statistics that we need, and also going back to our users and understanding what their needs are. So there are quite a few aspects that become part of daily life, to ensure that we’re providing [inaudible] and meeting the right regulations. Have I answered your question? I may have [inaudible].
What I can summarize from what you said is that you use AI in different aspects, and the learning is still going on. Some of the aspects seem solved, some still need work, and you are still figuring out your AI.
That is exactly right.
Because I think you have to go delicately, because of the security and the ethics.
Exactly, exactly. You hit the nail on the head, and we do have to go delicately. We do have to make sure that each dataset is specific to the way the AI is trained. We try as best we can to watch out for any kind of biases that may occur, and we try to investigate any issues around them. For example, as I mentioned previously, we’ve got optical character recognition, which is an AI tool that we use to onboard products. But we’re also developing an AI product that can help give handovers to clinicians, and also write letters, post-operative letters, for example.
So, like generative AI?
Yes, exactly. The dataset that’s used can be specific per surgeon, or it can be specific to the department. For example, in neurosurgery, which is my area of interest, the way primary brain tumors are classified changed in 2021. It now incorporates molecular diagnostics, rather than just looking under the microscope and seeing whether it’s a certain type of tumor. Incorporating these very small but specific amounts of data can be very difficult: sourcing the data, developing the handovers that are important, and making sure patients go through an appropriate pathway. We’re still developing that. I have to clarify that we are still developing, and we’re designing certain pathways, working with leading clinicians both in the UK and [inaudible].
That’s fine. Healthcare definitely takes time. We were discussing during the panels today that healthcare requires time and attention to solve problems. You can’t just go fast.
I think once we’ve solved certain types of problems, especially with datasets, we can actually make it incredibly efficient. But that’s just my opinion. That’s my hope, and that’s the direction we’re going with Encode.
So how is your team set up? Tell us more about your team.
You may have seen my business development partner, Carlon, [inaudible] earlier. He’s an orthopedic surgeon. He and I share the vision for Encode Health because of our own personal histories and experiences, engaging with large corporates and clinically at the same time. So we’ve both got that understanding of what is good, what is not good and what will work, and we engage with groups that would potentially want to work with us. That’s the core leadership team.
Then we’ve got the development team, working incredibly hard to make sure the back-end architecture is correct, and ensuring that the data is placed in the appropriate ways, which means securely. And we have a design team as well, to make sure the human factors, when it comes to the user interface, are appropriate and not over-engineered in any way, so somebody can come to the platform, intuitively scan the product, look at the forecasting capabilities and have a pleasant time doing it.
User experience is super important as well.
Exactly, exactly. Because I really do want people to be able to come onto the platform and just have a pleasant time doing their work, something akin to, I’m not going to say names, [inaudible] or so.
Like social media?
Yeah, yeah, things that they really use. Exactly. I think there’s a standard people are so used to. You can’t just throw any kind of user interface at people and say, look, a wonderful thing, you should go here and here as well, and expect to get what you’re asking for.
I think we’ve been seeing more and more companies start creating value just based on user experience. The way they disrupt is simply by having a better user experience than what existed from the status quo companies. Great. So this is my last question. You are in an industry that is constantly evolving, especially post-COVID; new things are happening every day. What are your habits to keep yourself abreast and up to date with what’s happening: the learnings, the possibilities, the hype and so on?
Well, that’s an interesting point. Like I said, I’m a forever student in many ways. I’m doing research through the University of Buckingham, and even though I’m looking at neurosurgery specifically, I can see other specialties developing in certain ways as well. I mentioned molecular diagnostics. It’s not just neurosurgery doing that; other specialties are applying and understanding the genome and the details that can be gleaned from it, and applying that to AI. That, for me, is the way I try to keep ahead of things: trying to understand how genomics is developing with reference to research, and how artificial intelligence is developing with reference to research.
What are those sources? Can you name any sources you usually follow, that you find pretty legit and would recommend to other health founders?
I do some very basic things, like going to PubMed. I look at the other large healthcare datasets as well, and look at the summaries of those. I restrict it to the last year or two, because the amount of information that’s advancing is just faster. In the last five years of my research, I can see the difference between 2020 and 2023. Some of the research done before 2020 was still based on just looking through the microscope, and now the research being done is very specific, using artificial intelligence algorithms to make sure the biopsies from particular patients are exactly what they say they are, and not only that, to research even further to understand new types of genetic expression per tumor. So I’m very interested in keeping up to date with genetics, with how artificial intelligence is interpreting the results, and with how that will help provide more precise and personalized healthcare treatment.
All right, makes sense. Thanks so much for your answers. I wish you all the best in building Encode and doing great things.
Thank you so much. Thank you.
At the time of recording (August 2023), Dr. Costas “Cos” Fantis was founder and CEO of Encode Health, a UK startup founded in 2019 that connects electronic health record data with hospital inventory management to match clinical demand with medical supply and track medical devices. Medically trained at the University of St Andrews, with a specialist interest in neurosurgery, he was doing research at the University of Buckingham on applications of artificial intelligence in neurosurgery. He was previously lead surgical officer at Touch Surgery in London, which trained surgeons through an app, and completed a fellowship in clinical education at St Thomas’ Hospital. He co-founded Encode Health with orthopaedic surgeon Dr. Carlon Fitzpatrick and Chief Technical Officer Marwan Kathayer.
Building a product that has to pull data out of the EHR and act on it somewhere else? Tell us which systems you’re connecting and we’ll walk you through how we’d approach it.
Partner with us to design, build, and scale digital solutions that drive better outcomes.
Global Tech Teams LLC, 525 Washington Blvd, Industrious at Newport Tower, Jersey City, NJ 07310, United States.
Let’s discuss your goals, workflows, and next steps in a focused consultation call.