Copy to clipboard 
Despite its technical nature, healthcare moves at a rather stately pace in adopting new care models and technologies. Innovative materials, robotic surgery tools and a fulsome ASC market have long loomed on the orthopedic horizon. In speaking to industry insiders, we believe many of these trends are reaching critical mass and are about to exert real force on the market.
Artificial intelligence (AI) is one of the most rapidly-developing technologies in orthopedics. It is ubiquitous in our daily lives, from autonomous vacuum cleaners to the panoply of “personal assistants” on smart devices. For some of us, however, AI still conjures ideas more suited to science fiction.
To get a sense of what AI is doing in orthopedics today, we spoke with two executives immersed in the field. Rob Kraal is the Vice President and General Manager of Zimmer Biomet’s Connected Health Group. Peter Verrillo is the founder and CEO of Enhatch, a software company with an intelligent surgery ecosystem. The two represent the spectrum of companies introducing AI in orthopedics, from the traditional implant giants to the startup software companies.
So, how would they describe artificial intelligence and machine learning to a layperson?
Python vs. PowerPoint
Peter Verrillo offered a simple definition of artificial intelligence. “Here’s the loose definition of how I describe AI: it’s anything that a human would normally have had to do but is now a software solution that a computer interprets,” he said. “We have optical character recognition on our phones. You take a picture, and it will tell you the text. It doesn’t feel like artificial intelligence, but it still is, fundamentally.”
As consumers of technology, we tend to move the goal posts on AI as it creeps into our lives and elevates our expectations. Verrillo pointed out how normal it is for our phones to remind us to eat better or “know” to suggest the fastest route to your child’s school at a particular time.
Verrillo’s working definition isn’t far from Microsoft’s description of the technology: a computer system that mimics human cognitive functions like learning and problem solving through math and logic. But where does machine learning come into play?
“Machine learning is really the substance,” said Rob Kraal. He continued, “It is a mathematical model designed to run on near real-time data to, in our case, provide clinical insight.” He relayed a tongue-in-cheek description of the difference between machine learning and AI: Machine learning is written in Python, and artificial intelligence is written in PowerPoint.
Machine learning models require robust and varied data sets. Google, a company awash in data, recently suspended an engineer who claimed that the company’s LaMDA chatbot had become sentient. Verrillo said he does not doubt that Google has enough data to train a genuinely astonishing AI. “They have your Gmail, every search you do, Google Messaging and Android Messaging. That’s all training the system how to respond.”
Responding to Unmet Needs
According to Kraal, AI in orthopedics is currently very good at prediction. These predictive abilities have surgeon-focused and patient-focused components. Models built from wearable devices and consumer electronics can predict things like gait speed at 90 days post-surgery, a surrogate outcome measure.
“On the surgeon side, we can alert them with WalkAI when patients are directionally off course, and they can intervene in a timely manner,” Kraal said. “On the patient side, they just don’t know if they’re on or off track. From our surveys, that is the number one unmet patient need. We will give them a scorecard based on machine learning to show how they’re doing compared to their cohort.”
More companies are leveraging artificial intelligence to bring value to customers in new ways. Internally, we saw a fivefold increase in AI-related orthopedic product news in 2018. Those announcements and developments have kept a brisk pace since then.
Recent announcements have come from companies across the revenue spectrum. Zimmer Biomet’s Omni Suite is an AI-enabled intelligent operating room designed to optimize surgical workflow and procedural efficiency by automating manual tasks and streamlining unnecessary technology and redundant hardware. Corin commenced a significant upgrade to the CorinRegistry data warehouse of radiographic hip and knee images. These data and images will drive the machine learning to identify at-risk patients. THINK Surgical announced a collaboration with Concordia University to use machine learning for advanced image registration. The results may ultimately inform aspects of THINK Surgical’s next robotic solution.
But can clinically-focused enabling technologies improve patient outcomes? It is still a matter of debate in orthopedics but shouldn’t be for much longer, according to Kraal.
He said, “There’s a building body of evidence to support that robotics, for instance, do improve outcomes. Right now, we’re at the beginning of the adoption curve. It will be a couple more years of building an ecosystem and body of clinical evidence to change everyone’s opinion.
With a connected ecosystem following the entire patient journey, the evidence just becomes overwhelming. It’s not going to take 10 years. It’ll just be a couple of years of evidence. I’m convinced of it.”
Ethical Considerations for AI
In orthopedics, AI’s voracious appetite for data favors the largest global players like Zimmer Biomet. Kraal said the company’s scale is also an essential factor in guarding against bias in its models.
“Building responsible AI representative of populations is something we think about a lot. You can easily imagine, specifically in America, that you would get a cohort of patients with access to different care levels and the best tech. We must be conscious of that. Our data science team constantly talks about avoiding bias in their models,” Kraal said.
Appropriate consent and use of data are crucial considerations for these technologies in healthcare. Companies leveraging AI must be fully transparent with the way data is used. Consent must come from both the provider and patient. Kraal adamantly opposes breaching that consent and trust for the sake of marketing.
He said, “Patients must be able to trust that we’re not going to sell them things. I think that is the worst possible thing that could happen in healthcare. Companies with access to patients can’t use that as a marketing tool. That’s a horrible, horrible road to go down.”
Zimmer Biomet has set up a data governance committee led by a data privacy officer. Checks and balances like this will become increasingly important in the digital era of healthcare. As it has in our everyday lives, artificial intelligence will rapidly become commonplace in the orthopedic industry.
Kraal said he sees connected medicine pushing informed consumerism into healthcare. He recalled his personal experience as a pediatric ICU nurse and the significant gap in outcomes between hospitals in two different areas of the United States.
He said, “I often thought if these parents knew there was another facility with better outcomes … They just didn’t know. It was just proximity to care that led to different outcomes. In orthopedics, consumers are not informed. That’s the most exciting thing about what we’re doing. Eventually, consumers will have more understanding of disease progression. They will be more informed about their choices for care.”
