What the Tech? The Future of Medical Device Predictive Analytics & Big Data

This What the Tech? event focused on big data in the MedTech industry and featured Join Analytics CEO Bruno Horta, Ph.D.; Nested Knowledge President Keith Kallmes; NXT Biomedical R&D Engineer Joe Passman; and Womble Bond Dickinson IP Transactions Partner Karthika Perumal, Ph.D. Womble Bond Dickinson IP attorney Tyler Train moderated the discussion, which took place live in Womble Bond Dickinson’s Irvine, Calif. Office as well as virtually.

Big Data has arrived in the MedTech industry, expanding opportunities for start-ups and well-established medical device companies alike. With much more data—and far more specific data—at their fingertips, researchers and MedTech entrepreneurs can make far better decisions about developing and marketing new products.

But when does information overload become a problem? More data isn’t necessarily helpful if its owners don’t know how to harness it properly. So how can medical device companies best use data to identify the best ideas, research and develop high-quality new products, and successfully bring those products to market? And how can MedTech companies best protect their increasingly valuable data assets?

“I’ve Got an Idea! Now What?”

Every innovation starts with an idea. But what is the next step to bringing an idea to fruition as a fully realized, marketable medical device? And at this initial stage, is data necessarily the goal, or just a byproduct?

Kallmes said the answer to these questions may vary, but in general, medical device companies seek data and concepts flow from that. 

“You could be a massive device company or an entrepreneur, but you need to find an unmet clinical need to move forward. You start with the data and then you move to the product, not the other way around. Otherwise you're going to be solving a problem that possibly doesn't exist,” he said. Also, investors will want to see data up front before committing capital.

The type of data product developers may seek can vary, but Kallmes said that in the medical device industry, the first step usually is consulting published literature on therapies that may outperform products that are being used in the marketplace. The second step is a bit trickier and more subjective, but Kallmes said companies need to talk to doctors and find out why the therapy in question isn’t being used.

“That’s customer discovery data—that is harder to get than the literature,” Kallmes said. 

Passman said, “The Stanford Biodesign Process was kind of indoctrinated in me when I was at Edwards Life Sciences, and that process centers on finding the unmet clinical need.” In his work in heart failure, that core focus involves developing ways to solve that one central problem. 

Determining if a Market Exists for a New Medical Device

Obviously, the end goal in product development is to take that product to market. So how should medical device companies determine if such a market exists for a proposed new device?

Market research is important, Passman said, and tools such as Google Surveys have made it much easier and less expensive for start-ups to solicit this type of important feedback. But there also is a more nuanced aspect of determining if there is a market for an idea. He regularly talks to doctors and clinicians to get that first-hand perspective on whether an idea might become a marketable product. 

From an IP portfolio management perspective, Perumal recommends developing an initial IP strategy early in the process, once an unmet clinical need has been identified. She encourages revisiting and refining this strategy on a regular basis as the company moves through its product development.

“We don't wait for that great, innovative step and then start developing strategy. We usually work at an early stage with our clients,” she said.

Securing Investment

Good ideas, quality data and sophisticated research all are vital in developing new medical devices. But successful product development requires one more key ingredient—the investment capital to get the project off the ground. But where should this money come from?

“Smart money is always the best,” Kallmes said. “If you can get money from people who eventually are going to use your product, that’s probably the best. If you can get smart money, those people will pay you to help you.”

The next most desirable type of funding, he said, is non-dilutive financing, where the company does not give up any equity in exchange for funding. For example, government grants are a good source of non-dilutive funding, he said.

“The big downside to government grants is timelines,” Kallmes said. The typical funding cycle takes two to three years from initial application to receiving money. So most companies need to find venture capital or angel investors during this period.

Passman said, “It's best to get as scrappy as possible before you ask (for funding). There's always ways to get further down the road with minimal spend.” For example, finding collaborators working on similar types of research can move projects forward without a large initial outlay of money.

Decisions in Data Collection

How do you decide what type of data you want to collect?

Passman said when he is working on a new medical device, he considers two basic types of data: well-established, consistent data types (ex. heart rate, blood oxygenation) and novel diagnostics/biomarkers.

“When you look at a problem that's worth solving, you try to talk to your clinicians and say, ‘Do you think that this type of information would help me solve this problem?’” he said. “You can't just throw data at a problem—it actually has to have some physiologic link. And if the data through predictive analytics is potentially synergistic, then you maybe have a winner.”

For example, in Passman’s work on finding ways to reduce heart failure hospitalizations, he is exploring the body’s signals that may give doctors and patients an early warning that heart failure is imminent. The goal is to reduce hospitalizations, improve patient outcomes, reduce healthcare costs, and save lives through better diagnostics.

Data scientists such as Horta also have to decide how they will handle data from different types of sources.

Horta said that on one hand, some data comes from well-established data sources—for example, data collected by a mobile phone’s accelerometer or microphone. “You have several sensors you know in my mobile phone and there are many people trying to develop apps to connect to physiological data using predictive analytics,” he said.

“In the case of novel signals, it's a bit more complicated because one has to validate not only the connection between the signal and the physiology, but also the precision and accuracy of the sensors that are being used to collect those signals,” he said. Having both a new signal and a new sensor produces accuracy challenges, so researchers need to make sure results can be reproduced and that sensors deliver accurate information. 

Separating Good Data from Bad

Train noted the long-held data science saying that “Garbage in equals garbage out.” But in complex medical device research and development, how do engineers, data scientists and product developers ensure they are getting the best possible information to drive decision-making?

Passman said good data starts with good tools. Before passing along data to data scientists like Horta, he makes sure all the sensors are performing to spec in a well-known, controlled setting before using them to collect important data.

“We serve our clients in terms of helping them interpreting the data,” Horta said. “So I like to have the opportunity to talk to the engineers that are acquiring this data, because sometimes we can give hints in terms of what could improve the data collection so as to avoid the problem of having useless data. So if this conversation happens, then we move much faster in terms of designing experimental setups that will collect the data in a proper way.”

Perumal said that from a legal standpoint, to get clean data, make sure that you develop the appropriate contracts to acquire all the rights required to access the data and control how the data is collected at the start of the project.

So what type of datasets should medical device companies choose—commercially available datasets? Private datasets? Even novel datasets?

“I think it depends on exactly what point you are at in the device development process,” Kallmes said. “So in terms of being able to predict whether or not you'll pass or fail those tests that you're going to do at a third party for you for regulatory reasons, it will save you a lot of time and money if you know whether or not you'll pass those tests.”

Then, once the product is on the market, companies need a combination of economic and health data to convince payers to buy or reimburse for the medical device.

Data Interpretation

Once an engineer like Passman develops a dataset, he passes it along to a data scientist like Horta to interpret it. The two have worked together on developing and analyzing large datasets related to heart failure. 

“You have all this engineering knowledge, but then you have these guys that basically study lines on a computer all day and they say, ‘Can we make sense out of those lines?’ I think that's where really cool algorithms can come out,” Passman said.

“It’s almost like seeing the Matrix,” Train said. “It’s like something that only data scientists can see into.”

Horta said his role depends on having an open line of communication with the scientists behind the idea, as well as the engineers collecting the data. This type of teamwork is vital in producing accurate, useful analyses.

Kallmes added that medical device companies need to carefully define data elements and present the data scientists with the cleanest possible datasets.

“People underestimate the value of clean datasets,” Perumal said. “The value of having a clean, reliable data set makes a big difference in valuation.”

Protecting Data Assets

Researching and developing this type of information is a highly expensive, difficult process. So medical device companies need to make sure they are taking the proper legal steps to protect their key research data. Doing so may require a combination of IP strategies, Train said—patent, copyright, trade secrets, etc.

“I usually ask clients to look to take a very holistic view of what do they want to protect,” Perumal said. “What aspects have the most value? What is the timeline of protection? Where are your customers located? And how do they want to get to market?”

An IP strategy needs to fit into the company’s overall business approach, she said. For example, if a company is sharing information with prospective business partners, then a strategy built on trade secrets may not be the best approach. But patents also have limitations, so crafting an IP strategy is a customized process that involves examining a medical device company’s specific needs.

“For several clients right now, their IP strategy is a combination of copyrights to the code and their database, patents to some of their algorithms and how it's being applied to solve a clinical need, and then trade secrets for how the whole system actually works,” Perumal said.

She also said companies should manage their trade secrets disclosures—not every business partner needs to know every detail of the company’s operations. Well-crafted non-disclosure agreements can be useful in these situations, she said, as well as among co-founders and employees. She noted that the biggest source of trade secret litigation comes from former employees. 

In addition, Perumal said one common mistake she sees inventors make is waiting too long to seek patent protections. “I strongly recommend placing your stake in the ground by filing that provisional application while you’re looking to see what is out there (in terms of investments).”

Train said, “Sometimes (companies) think they need a full prototype before they file a patent application. But they don’t. Provisional patent applications can be filed very quickly and relatively inexpensively.”

Next Steps for Data Evolution in MedTech

Kallmes said that building on existing work often is the most efficient use of resources. Expanding indication allows companies to skip having to develop a new device and years of clinical studies just by working with existing data. 

“Pharma has been doing this for years. Viagra started as a vasodilator and they found a better use for it to make money. I think devices are going to do the same,” he said. Passman said that on the cardiovascular front, companies are working to create less invasive devices.

Big Data will continue to grow in importance in the medical device sector. The winners will be the companies that figure how to best use this deluge of information to make better scientific and business decisions.

This article is based on a Feb. 24 Womble Bond Dickinson thought leadership series event. 

For more insights, opinions and events on how innovation will drive industry in the near future, click below to visit our What the Tech? hub.