Dan Reynolds, the editor in chief of Risk & Insurance, recently sat down with Nan-Wei Gong, PhD, the founder and CEO of FIGUR8. What follows is a transcript of that discussion, edited for length and clarity.

Risk & Insurance:  Nice to see you again Nan-Wei. If you don’t mind, tell us about the origins of FIGUR8.

Dr. Nan Wei Gong: The origin story of FIGUR8, is a result of decades of research with a foundation in collaboration between MIT, Mass General Hospital and the Boston Red Sox. The question was, “How can we use decades of research in biomechanics, the results of which were usually only available to elite athletes, and make that available to the general population?”

R&I:  How do you view the role of AI and data analytics in the workers’ comp sphere and within your approach to the business?

NWG: So, there are so many AI companies out there. You can’t go to San Francisco, where I just was, without seeing a billboard on the highway that is saying, in effect, “We have an AI solution.”

The biggest problem with an AI solution is that it’s fundamentally a large language model where you need very good data input to train the model to have meaningful data output. So what we have at FIGUR8 is a data set that is foundational and can feed into the field of medicine that we are in, to create optimal prediction, analysis, and accurate outcome measurements. We treat AI as a very different tool as opposed to, just replacing, speech, or human decisions. For us, AI is a tool that can produce something that is much more meaningful and insightful in this area of medicine, which is joint and muscle function, or musculoskeletal, health.

R&I: How confident should we be in general that enough companies have good data to start with?

NWG: We don’t have a lot of data to start with in musculoskeletal care. And that may be true in multiple areas of medicine. That’s because if you are forced to rely on human bias and subjective feedback about the patient’s self-reported outcomes, you really don’t know how to accurately assess the progression of injury recovery. There has not been a dataset that is science driven, or an objective tool from which every dataset would build on from the same baseline. It has not existed. So to use AI relying on today’s typical dataset outside of FIGUR8; in computer language, you would call it garbage in, garbage out. It takes the type of effort that we have at FIGUR8 to bring together a data set and a measurement tool that is reproducible, scalable, and really actionable.

R&I: It’s a heavy lift to get to the right dataset. Right? I’m just imagining that some companies can’t put the right amount of resources into that.

NWG: Absolutely. Nobody wants to make the lift. It took us 7 years to get to the point where we not only have the technology, we have user interfaces that are consumer friendly. We care a lot about workflow optimization. We also care a lot about user interaction. And, we care a lot about the accuracy of the data. Tens of millions of dollars of investment went into bringing the platform together. And now we are scaling up commercially with a lot of great partners in workers’ compensation as well as many provider partners. The heavy lift has been a multiyear effort with tens of millions in investment, and not every company can afford that.

R&I: We hear a lot about evidence-based medicine and workers’ comp. Can you talk about musculoskeletal injury and how you monitor that using evidence-based medicine?

NWG: Imagine an EKG or ECG in the clinic, or even a digital thermometer. Before having those tools, you would have used your hand, your eyes, and you depended on a lot of observation to monitor progress.

Practicing precision medicine means having personalized treatment, because everyone has a different baseline. Each of us has multiple variables in recovery when injured. Without a tool that can give you objective, and preferably digitized, information, it’s impossible to have precision medicine because everyone will tell you when they’re in a maximum amount of pain.

With a tool like FIGUR8 and the data that we generate, you can have personalized treatment. Here’s an example. Our solution can measure the week over week progress of a musculoskeletal injury. Normally, a PT will tell you, come back in 2 weeks, 4 weeks, or 6 weeks, and then see how you do, based on how you feel. With FIGUR8, you can detect a variance in a matter of days instead of weeks or months, and you can course-correct the treatment very, very quickly. That’s opposed to the rigidity of cookie cutter programs. That is the basis of precision medicine, having a more precise understanding of a disease or the progression of a recovery. So, data first, and then the treatment programs follow.

R&I: You’ve talked before about the importance of equity in health care. What does that concept mean to you?

NWG: Thanks for asking.  It’s a very good question. If you think about a very rural area where there is no specialist and you have to drive 5 hours to see someone with an MD, that’s a dynamic that exists in many places in America. Health equity for us really is asking how can we help solve the access problem that I mentioned earlier. We want to provide access to technology and access to our tool so that the data can be transferred to a specialist that could be in Boston or New York City, while you are in rural Oklahoma. You can then get the experts to give you feedback or second opinions on whether the treatment is effective or not. That is the digital transformation that happened through the Internet decades ago, but not just in health care. We talk about the internet of things; let’s also create an internet of medical devices.

R&I: Do you feel on the topic of equity that you are seeing improved access from the patients that you’re working with?

NWG: Absolutely. A lot of our provider partners are not in the major cities. We are very well received in cities that have fewer medical resources. They’re excited that there is a company out there that is serving these underserved areas.

R&I: You’re capturing people’s degree of mobility, right? How do you achieve that?

NWG: If you think about what we describe in the field of medicine as muscular and skeletal health: There are two components to it. We measure biomechanics. That’s how your joints move, and not just one joint.

When we talk about joint motion, we are looking at multiple joints and how they interact with each other. For example, if you have a lower extremity screening completed, we look at the major joints in your entire lower body. When you have a knee injury, you likely compensate, and your hip will in turn move differently, and then you have a pain in one hip, but it’s actually because the original knee injury makes you compensate.

We measure joint motion and how all of the joints move dynamically, which is different from someone talking about measuring them with a goniometer, which is a static measurement. We are recording a series of dynamic, multiple joint movements.

We also measure muscle function because musculoskeletal health has a soft tissue component to it. We measure how your muscle contracts during those dynamic movements. So if you have an injury and you have muscle atrophy or you have a weakness in a certain area, we can see the difference. The treatment can then focus on strengthening the weaknesses or increasing the flexibility of the limited joint so that you reach movement functionality.  So it’s joint motion and muscle function, during dynamic movement. That’s how we measure it.

R&I:  And the data that you’re collecting — you mentioned large language models — my understanding is that they work with unstructured data. With the measurements that you’re producing, does that fall under unstructured data, or are you getting something that’s a little bit easier to work with?

NWG: You put that unstructured data in the AI model, and it will show you the data sets that are the most sensitive to this recovery trajectory.

That’s really how we could leverage data in a way that has never been done before because the wealth of information that we are collecting tells you so much more. There is a lot more to discover than we have discovered so far. I would say it’s just the beginning of our data collection and data analysis journey at this point in time. &