AI in the OR

Artificial intelligence can be a tool to save lives and increase efficiencies in hospital operating rooms that today often rely on outdated processes, according to issue experts in Silicon Valley and beyond.

While some operating rooms do have advanced robotics and diagnostic equipment, most, arguably, have less advanced assistive technology than a typical American living room equipped with smart TVs, connected thermostats, robot vacuums, and software-controlled lights.

Most processes used in ORs are completely manual, leveraging expensive personnel, like surgical techs and nurses, to perform the full range of tasks. For example: The “surgical count,” which keeps track of surgical instruments, is often maintained on a dry-erase board with a marker. Another board is also used for the “safety checklist”—which aims to ensure the procedure is done at the correct site on the patient, among other critical aspects of an operation.

“Ask any orthopedic surgeon, ‘How efficient is your operating room?’ and you will likely be met with a smirk, indicating some combination of disgust, disappointment, and disillusionment,” Drs. Charles DeCook and Jeremy Statton wrote in a 2022 paper on OR efficiency.

The U.S. Centers for Medicare and Medicaid Services (CMS) estimates that by 2028 the nation’s healthcare system will pay between $505 billion and $696 billion annually for surgical reoperations, readmissions, waste, and preventable safety events. Each year, 6.6 million surgical site infections occur globally; 200,000 result in fatalities. A total of 1,411 surgical “sentinel events” in the United States—such as a procedure being performed on the wrong side of the body—were reported in 2023 to the Joint Commission. That was a record high; 88% of the incidents happened in hospitals and 18% were connected to the death of the patient, the organization said in its 2023 review of sentinel event data.

Insurers, of course, want to reduce their risks by preventing costly medical errors. In the United States, the average annual premiums for medical malpractice insurance for small hospitals and clinics can run up to $15,000 per physician, while a policy for a large hospital may exceed $1 million.

“Less efficient surgeries mean that surgeons get tired more quickly, which can mean they are more likely to make surgical mistakes. This can also lead to decreased hospital profitability,” says Beracah Stortvedt, digital health and emerging med-tech practice co-founder at Marsh McLennan–West. “Extra time on each surgery means not being able to bill for as many surgeries—and that’s a hit to the bottom line. From a malpractice standpoint and a risk standpoint, and from a general cost perspective, those inefficiencies create big problems.”

Artificial intelligence (AI) and other digital tools can significantly reduce those surgical inefficiencies that are driving up risk and cost at hospitals and health systems, Stortvedt adds.

“I think 99 out of 100 physicians, you give them better information to help them make a medical decision, you’re going to get a better medical decision,” says Pete Reilly, North America healthcare practice leader and chief sales officer for Hub International. “And that’s incredibly encouraging. And at least long term, that’s going to be the most positive effect for healthcare, both patients and clinicians, and then as a consequence, all of us in the insurance industry, for better outcomes clinically, financially, risk management-wise.”

Replacing the Subjective with the Objective

A growing network of entrepreneurs and data scientists aims to find ways to protect healthcare professionals from preventable risks and liabilities using advanced technology, including AI.

Last year, one of the most powerful venture capital firms in Silicon Valley, Andreesen Horowitz, cited AI as the key fix for such healthcare industry problems.

“Hundreds of thousands of people die each year because of avoidable human errors. Modern technology has little penetration in healthcare. The industry still largely relies on paper and fax machines for communication,” according to the firm’s report, “Hey Tech, It’s Time to Build. In Healthcare.” “In healthcare, AI stands to dramatically improve the previously unpleasant economics, making building in the industry far more compelling. Furthermore, human-driven services tend to scale linearly with each incremental human added, but AI-driven services can scale exponentially.”

Within the extended list of benefits the Andreesen Horowitz report says AI could bring to healthcare is resource optimization at hospitals. AI tools can predict patient needs, optimize hospital operations, and improve resource allocation, leading to more efficient care delivery, according to authors Daisy Wolf and Vijay Pande, partners in the firm’s Bio+Health team.

Surgical care could be transformed by better data with the help of technology, according to a growing chorus of experts. For example, computer vision systems that can generate a comprehensive video record of a surgical procedure.

The market for digitizing surgery and operating rooms is rapidly expanding and projected to surpass $90 billion globally by 2026.

Among the companies working in this space: Proximie provides a platform that can collect and analyze data from any system in an operating room, creating what it calls a “single source of truth”; Apella promises computer vision technology that tracks the progress of a surgery, collecting data to improve efficiency in the OR; and nSight Surgical’s objective data platform focuses on containing costs and improving quality in surgeries.

Today, even in the most advanced hospitals in the United States, key details of a procedure are entered manually into electronic medical records (EMRs). Many smaller ambulatory surgery centers, unable to afford EMR solutions, remain largely paper-driven.

These widely adopted methods require a circulating nurse to document information, when they have time to do so amid a stream of surgical tasks. There’s no place in these existing systems to record a great deal of important information needed to create an objective record of the surgery. That means EMR data from operating rooms is often inaccurate, says nSight CEO Nathaniel Smith.

From its base in San Francisco, venture capital-funded startup nSight Surgical uses video-based AI technology to build what it calls “the first truly objective healthcare record.”

nSight uses AI to watch streaming videos in real time from cameras in the OR and turn that information into passively documented objective records that track costs, surgical quality, and efficiency metrics. That data encompasses documentation of the inventory of tools used during surgery and breakdowns of procedures into discrete phases, such as records of when patients were brought into and out of ORs, among other information.

The system feeds all the collected data, including video, to the server with AI-based software capable of processing it in real time. It can then inform better decision-making for subsequent surgeries.

For example: Identifying “idle time” and “late starts” during surgeries helps to optimize procedure flow and schedules. And detecting infection-related events and behaviors enables hospitals to reduce reinfection rates and thus downstream readmission rates.

nSight digitizes the surgical count and safety checklist process—away from dry-erase boards—via its digital whiteboard. It provides a retained record showing that staff conducted the safety checklist. A transcript, which anonymizes participant voices, is also maintained as objective evidence of the process.

By the end of 2025 nSight expects to have placed its AI technology within 100 ORs, spread across top-tier health systems.

According to Smith, a year from now the system will be able to generate accurate insights into what worked and what didn’t during surgeries based on the information collected from 100 health system ORs.

“Ultimately, it comes down to OR preparedness, which is paramount to efficiency; one simply doesn’t exist without the other,” Smith says.

The Insurance POV

It is too early to say how this technology will affect medical professional liability, according to Hub International’s Reilly, but the signs are promising.

“The amount of [meaningful] data that AI will pull from diagnosis through procedure to post-op analysis, there’s a great deal of encouragement that will allow for better and more consistent underwriting,” he says.

One question is how far to extend the boundaries of what is considered meaningful data that AI can find. Does it go beyond clinical data? While the insurance industry can be slow to change, it will need to understand the data and what it actually shows, Reilly says.

To illustrate the point, he cites this theoretical example: “If AI can go out and search social media, not only take all the clinical data that it gets from the provider institution, but if you see that the surgeon has just bought themselves two Maseratis and had been vacationing in some glamorous destination, is that indicative of a provider who’s on the downside of their career or is less focused on clinical expertise? Who knows? AI can certainly collect across those platforms a whole lot of information that historically actuaries and underwriters and brokers have not paid attention to.”

AI-supplied data could also help underwriters identify risks in covering a medical institution or provider that could lead to a nuclear verdict, Reilly adds.

Alongside the optimism, he offers a note of caution—AI systems are plugged into networks, with the vulnerabilities that creates when it comes to protected healthcare information. The February 2024 data breach of Change Healthcare and a small number of other incidents “already scared the daylights out of all of us,” Reilly says. “As AI grows, we’re just going to become a richer and richer target for that stuff. So, we can’t forget about that downside of this incredibly powerful tool,” he notes.