WellSpan Health has launched an AI agent named Ana, who's connecting with at-risk patients for screenings that might otherwise be forgotten or ignored.
Healthcare organizations are now using AI to have conversations with patients that doctors and nurses might not have time for—and closing critical population health care gaps that could save lives.
WellSpan Health launched an AI platform roughly one month ago that calls selected at-risk patients to schedule colorectal cancer screenings. The AI agent, called Ana and developed by digital health startup Hippocratic AI, asks patients if they would agree to take the test and, if they agree, arranges to mail a testing kit to their homes.
The so-called AI “agent” is designed to replace either a mailing, a cold-call program often launched out of a health system’s call center or—if the health system doesn’t have those resources—the conversation that a doctor would have with a patient during an annual checkup.
None of those tactics has a high level of success. And as a result, at-risk patients often don’t have those screenings when they should, if at all, increasing the chances of a serious health issue down the road.
WellSpan Health executives decided to use the program to reach out to patients at risk of developing colorectal cancer, identified by their birthdate and family history. Of particular concern were Spanish-speaking patients, who might miss the mailer or the conversation with a doctor because of language issues.
“It’s an opportunity for us to reach out to people in their homes, in their own time, and have a conversation with them that we maybe couldn't staff with a human,” says R. Hal Baker, WellSpan’s SVP and chief digital and innovation officer. “But [with] the right AI, that conversation could be had.”
R. Hal Baker, SVP and chief digital and innovation officer at WellSpan Health. Photo courtesy WellSpan Health.
Baker, a primary care physician, says the program takes pressure off of doctors to fit that task into an already busy care visit and gives them more time to make that visit meaningful.
“Most of my visits are 20 minutes,” he notes. “And while it may take only 30 seconds to have a conversation about colorectal cancer screening, … if I can take any of these things like mammograms and colorectal cancer screening and COVID shots and flu shots out of the visit, and they can happen reliably outside of coming to the office, that gives me more time to discuss what can only be discussed in the exam room.”
Kasey Paulus, MBA, RN, CENP, WellSpan’s SVP and chief nursing executive, says Ana works because she’s designed to be empathetic and engaging. And she can speak Spanish, in which not many doctors or nurses are fluent.
“A sensitive AI that speaks your language [is better than] a well-meaning human who can’t speak Spanish,” Baker says.
And it works. Of the first 455 Spanish-speaking patients contacted by Ana, 15% agreed to screening, Baker says, and the net promoter score was higher than that of English-speaking patients (of which 6% agreed to a screening).
“Not bad for an automated reach-out cold call to people who weren’t expecting it,” he says.
Kasey Paulus, MBA, RN, CENP, SVP and chief nursing executive at WellSpan Health. Photo courtesy WellSpan Health.
Eventually, the health system will see results in clinical outcomes. More screenings will lead to more cancers detected and treated early, improving the quality of life for patients and reducing deaths. Financially, more screenings might boost initial costs but lead to less expensive medical treatments and long-term care later on.
Baker and Paulus say WellSpan worked closely with Hippocratic AI to develop Ana, going through every scenario that the AI agent would face and every question that a patient might ask. Baker says he even tried to confuse Ana during a test by announcing that his birthday was on New Year’s Day.
“We have a very creative and collaborative multidisciplinary team, and those individuals do their best to try to break it before we launch it,” he says. “So we're looking for ways [in which] somebody might trip it up that we didn't even think of, because inevitably that will happen.”
For example, Baker says, “We had to very quickly realize we had needed to add the ability for somebody to say, ‘Please don't call me again.’ “
In launching Ana, Baker says the health system was very careful to make sure that patients know they’re talking to an AI agent. For the first 100 phone calls, a nurse was also on the line to make sure things ran smoothly.
“What we found out was that [the nurse] didn’t have to” be in on the call, he says. “What is novel here is that we have now moved the human in the loop to the human on the tail for our next thousand calls,” meaning a nurse will review the call within a few hours.
With the platform now up and running, Baker and Paulus say they’re giving Ana another task: connecting with patients who are coming into WellSpan for a colonoscopy. Ana will call them ahead of that appointment and go through everything the patient needs to do before the procedure. Again, that task would have been handled by a nurse or call center, if at all.
“The last 48 hours of coaching are really critical,” Paulus says. “And AI can help us where we can't always have somebody on call 24/7 365.”
She says the health system will see the benefits in reduced cancellations and procedures that are started and cut short or unsuccessful.
Baker says the platform may scale up in time as WellSpan explores how Ana can effectively interact with patients and support their healthcare journey. They may use the platform for more population health outreach, or to help patients prepare for other procedures, or even to check up on them and coordinate care after a procedure.
“We recognize that we're boldly going into some uncharted territory and that our AI is not replacing our clinical staff but augmenting the work that they just don't have the capacity to do,” he says.
There has been a generational shift in what to expect from nursing as a profession, according to this CNE.
On this episode of HL Shorts, we hear from Melanie Heuston, chief nurse executive at WVU Medicine, and HealthLeaders Exchange member, about what new graduate nurses are expecting from the job in 2024. Tune in to hear her insights.
The HealthLeaders Exchange is an exclusive, executive community for sharing ideas, solutions, and insights.
Hear more about how nurses and nurse leaders can help patients engage with their healthcare, in and out of the hospital.
Nurses wear many hats in the industry, but their primary focus is on patient interaction and care delivery.
Both nurses and nurse leaders play pivotal roles in making sure that patients engage with preventative care measures and patient education programs. It's also the CNO's job to ensure the effectiveness of remote patient monitoring programs and to find creative solutions for dealing with social determinants of health.
According to Cassie Lewis, chief nursing officer at Bon Secours’ Richmond market, part of Bon Secours Mercy Health, and HealthLeaders Exchange member, improving patient engagement begins with the day-to-day interactions between the patient and the bedside nurses who care for them.
"Those subtle interactions that they can do with our patients everyday can truly make such a difference in how patients view our healthcare system and how they view their care," Lewis said. "Trying to emphasize this for our teams and making sure…they understand the importance and what they can do for our patients is the first step in helping them engage in a more meaningful way."
Building trust and communication between nurses and patients is also critical to patient engagement, according to Lewis.
"One solution that we have really tried to emphasize here is the importance of bedside shift reports and good handoffs," Lewis said. "It's really important to start developing trust with the patients and including them in their care, versus keeping them on the outside looking in."
Lewis recommended that CNOs leverage technology to build robust, individualized patient education and remote patient monitoring programs, while considering social determinants of health. Clinical nurse specialists, nurse practitioners, and nurse educators all play a key role in these processes.
"We know that it's not a one-stop shop for every single patient that comes in our door," Lewis said, "and while we have great things we can augment with technology […], [we need] to take that one step further."
Listen to this week's episode of the HealthLeaders Podcast to hear more about how nurses and nurse leaders can help patients engage with their healthcare, in and out of the hospital.
A new survey finds that poor or ineffective technology is costing the healthcare industry $8 billion a year. Here are the eight biggest culprits.
Healthcare organizations are losing $8 billion a year to ineffective and outdated IT, and few have the money to improve that technology.
That’s the key takeaway from a new survey of more than 900 healthcare professionals by Black Book Research. The study, the third in Black Book’s “What’s Hot and What’s Not in Healthcare IT Investments” series, finds that bad IT investments have jumped significantly since 2017, when those costs were estimated at $1.7 billion, and budget limitations are keeping healthcare leaders from correcting those problems.
"Three-quarters of IT leaders surveyed indicated that they have no plans to allocate funds for replacing these flawed systems in 2025, reflecting a broader trend of financial constraints across the sector," Black Book President Doug Brown said in a press release.
"CIOs are understandably cautious about replacing underperforming systems when the ROI is uncertain, given the track record of many healthcare IT vendors failing to meet expectations. Without clear evidence that a new investment will deliver tangible financial or operational improvements, justifying the expense becomes challenging."
According to the survey, tech limitations are tied to one of more of five key reasons: poor user experience (almost have of those surveyed cited this), lack of interoperability (24%), cost (20%), lack of flexibility (6%) and alert fatigue (2%).
The findings will disappoint healthcare leaders who are counting on their IT platforms—especially their EHRs—to support innovations like AI and virtual care. An ineffective tech platform not only cuts into the ROI of a new program, but adds to the stress and frustration that haunt nearly every health system and hospital and cause burnout and workforce shortages.
Black Book’s study lists eight IT adventures that have plagued healthcare leaders the most:
Overly complex or unintuitive EHRs. Either a scapegoat or a savior for healthcare organizations collecting and managing their data, EHRs haven’t yet fulfilled their promise. According to Black Book, more than three-quarters of those surveyed are still having issues, often with navigation or workflow design, leading to “click fatigue.” In addition, at a time when many smaller health systems and hospitals are being acquired by larger, more stable networks, 91% of small medical practices in the survey say the haven’t been able to transition smoothly to the larger system’s EHR.
Bad telehealth. Virtual care saw a surge in popularity during the pandemic, but in many cases those platforms were adopted quickly and without due diligence. As a result, more than 80% of survey respondents said those telehealth tools are not synching with their EHRs, creating dreaded data silos and duplicate information, and impeding workflows.
Clunky RCM systems. Healthcare organizations have for years tried to automate their revenue cycle management operations to improve efficiency, capture lost reimbursements and reduce manual administrative tasks. Unfortunately, the technology has met expectations. Some 70% of executives surveyed said their RCM tech is either outdated or unable to integrate new tools like AI, leading to longer claims processing times and higher denial rates. Also, just more than 60% said poor claims scrubbing and denial management capabilities are resulting in lost revenue.
Uncooperative HIEs. Health information exchanges offer the potential to connect health systems and enable data sharing. But 28% of medical practices said their EHRs aren’t synching well with the HIE, and 23% cite a lack of data standardization and integration.
Poorly integrated CDS tools. Providers often rely on clinical decision support tech to improve their decision-making and boost clinical outcomes. But according to the survey, 80% say their CDS tools don’t integrate with the EHR, and first-generation tools often generate excessive or unnecessary alerts, leading to “alarm fatigue.”
Lack of patient engagement support. Patient engagement technology, including portals and messaging platforms, are designed to improve the patient-provider relationship. But 77% of hospital executives surveyed said their portals aren’t meeting the needs of their patients, resulting in ineffective communication and engagement. And 88% of those surveyed said smaller, niche products don’t have the integration or mobile-friendly capabilities they need.
Hyped-up AI. AI might be able to address many of healthcare’s biggest pain points, but the technology isn’t there yet. A whopping 96% of executives surveyed said they are facing challenges with ROI, and 92% said they can’t yet rely on the accuracy of the tools and find actionable results. Some 85%, meanwhile, said the tools they’re using to automate diagnostics or treatment planning aren’t yet capable of handling complex or real-world clinical environments.
Interoperability issues. Finally, 31% of the executives surveyed said they’re not happy with their data interoperability vendors, the chief complaints being slow updates and poor API support. This despite federal efforts to create a nationwide interoperability grid, through TEFCA. Many are struggling to adopt (FHIR) Fast Healthcare Interoperability Resources standards, and 8% say they’re stuck with technology that isn’t, well, interoperable.
The W-2 on-demand staffing model is a win-win for health systems, nurses, and patients, according to this CNO.
A lot has changed in nursing in the past few years, specifically in staffing.
Healthcare has become more virtual, and flexible scheduling models have replaced more traditional ones, and it's up to CNOs to consider innovative staffing solutions to address these challenges.
"COVID taught us that a lot of things can be done remotely," Garnica said. "We have a lot of telehealth, we have a lot of things out there outside of just acute care in the hospital, and that's attractive to a lot of nurses."
Garnica also explained how the competitiveness of the market has increased since the pandemic, which does not help with the nursing shortage.
"Everybody needs nurses, and we're not only competing with hospitals," Garnica said, "we're competing with all of those other venues as well."
Nurses are now looking for flexible schedule options beyond just full and part time, and they are looking for strong benefits.
"That's a challenge for hospitals who traditionally had believed that your core staffing should be about 85%," Garnica said, "[with] very little contingent staffing."
Due to these changes, SSM Health is trying something different.
The W-2 on-demand model
During the pandemic, Garnica explained that SSM Health had to quickly shift to other staffing models out of necessity. The health system began using external or third-party staffing agencies, with contracted nurses who were often coming from faraway places across the country.
"They're doing their contract, [then] they're exiting, and […] although it's a short-term solution, you have a staff and a nurse at the bedside," Garnica said. "It creates some long-term challenges not only with cost, but just around longevity, loyalty, [and] sustainability for outcomes."
Enter the on-demand model.
"An on-demand model gives us the ability to adjust our staffing needs very quickly, use nurses when we need them, flex them, [and] maybe offer [them] to other areas when we don't need them," Garnica said. "It also created a lot of loyalty and longevity with local nurses."
Garnica said the health system now gets local nurses from St. Louis through their partnership with ShiftMed, and since they all work in the area and in other local markets, the nurses can build relationships with one another.
"They're helping to sort of recruit and sustain one another," Garnica said. "So, you're getting this wonderful on demand support that understands the market, they're there to support your ministry, [and] they have relationships with your own staff."
Beginning the process
Ultimately, this new model led SSM Health to convert more than 100 on-demand nurses into full-time staff, and according to Garnica, the first step in making this change was helping the nurses to understand why it was happening.
"Helping our staff understand why we were making that shift and then engaging them in the process was probably one of our biggest successes," Garnica said, "because again, they are our best recruiters."
Garnica said that leadership had many discussions with core staff, and that they provided tools to help welcome and onboard the on-demand nurses. The goal was to have the core staff build relationships and express why SSM Health St. Mary's is a good place to work, Garnica said, and to make the on-demand nurses consider what it would be like to be a permanent employee.
"We offered some unique PRN options and things that we thought would be a nice complement for those folks," Garnica said, "and I think our conversion of those nurses to onboard to our own staff, really I would credit our nursing staff. They really were our best recruiters."
Garnica attributes the decisions of many on-demand nurses to stay on full time to the core staff who made it happen, and also to the idea that the on-demand nurses get to trial the environment beforehand.
"I think there's a lot to be said for when you're being recruited from a peer," Garnica said. "It's not scary to make a leap when you get to try it out."
The on-demand model also addresses the idea of nursing being the same in every health system.
"When you have folks that get to come be a part of your culture, be a part of your hospital, and be a part of the things that you're doing in an ongoing basis […]," Garnica said, "I think it makes it a lot easier because that jump isn't scary."
Outcomes
According to Garnica, the core staff reaction to this strategy has been positive.
"They are really proud of themselves, we update them on folks that they were able to help get moved over," Garnica said. "I think there's a lot of pride and excitement in that."
The nurses also appreciate the relationships that they get to build with more local nurses, and they feel as if they are part of a staffing solution, rather than feeling like a victim of staffing challenges.
"When you're short staffed, it's really easy to fall into a victim mentality [that] feels hopeless sometimes," Garnica said, "and I think this enables them to feel like they were part of the solution."
From a financial standpoint, there are also several benefits, according to Garnica.
"I think looking at areas of turnover and retention, obviously that's a really high-cost business when you're bringing folks in and you're turning them over," Garnica said. "Anytime we can reduce that turnover rate, we're obviously going to save money."
This model also cuts down on onboarding costs, Garnica explained.
"When you have an on-demand partner whose nurse is onboarding, they've probably already been working at our hospital for months, maybe a year," Garnica said, "so that onboarding is much quicker because they're already pretty acclimated."
The health system has also saved money on external agency contracts, and in general by stabilizing their workforce.
"We've had significant improvements in HAIs, patient experience, and those areas, because we don't have such an unstable workforce anymore," Garnica said. "We treat these folks as our own, they are part of our staff, we embed them in our culture."
Garnica said there have also been overall improvements in productivity and morale among the staff.
"When you stabilize the workforce, when nurses come to work, knowing that we're doing everything we can to get support into the building for them and they've built those relationships," Garnica said, "they know that the on-demand nurses are their partners, [and] there's a lot of positive energy around that."
For CNOs who might want to consider using this strategy, Garnica recommends being open minded, and to shift mindsets to understand that the market has changed. CNOs should focus on building those relationships and partnerships, and engaging staff in decision making to help bring about change.
"You have to learn to live with staffing [looking] different than it used to," Garnica said. "There are creative options out there […]. Once you get it going, you can see a lot of really positive outcomes."
Virtual care programs for veterans are seeing mixed results. That could help healthcare organizations better understand what works and what doesn’t.
Recent efforts to improve healthcare access for veterans offers insight into which virtual care strategies are working and which ones aren’t.
The success (or lack of) of programs launched by the Department of Veterans Affairs (VA) and the Veterans Health Administration (VHA) could help health systems and hospitals better understand the direct-to-consumer telehealth market.
For instance, the VA recently announced that its tele-emergency care (tele-EC) platform will now be available to veterans across the country, after the success of pilot programs in selected regions. The program, part of VA Health Connect, enables veterans to connect with care providers on-demand through a smartphone and associated app.
“Veterans can now be evaluated for possible emergencies from the comfort of their home,” VA Under Secretary for Health Shereef Elnahal, MD, said in a press release. “Sometimes, you’re not sure whether what you’re experiencing is a minor emergency or not — and tele-emergency care can help you resolve those questions. Veterans can get immediate, virtual triage with a VA medical provider who has direct access to their medical records. This avoids having to potentially drive to the nearest emergency department and wait to be evaluated, if appropriate.”
"I think it's a noble idea,” GAO Healthcare Director Alyssa Hundrup told a Virginia TV station in a recent interview.. “They've put in an effort but, unfortunately, it has yet to be used. VA really needs to be looking at the effectiveness of these sites, where they are, how they're using them, are they getting the word out to communicate with the veterans the availability of these? Otherwise, these sites are sitting there being unused and it's a real missed opportunity.”
So why is the tele-ICU program working but ATLAS is struggling? The issue may be similar to why so-called disruptors like Walmart, Walgreens and CVS Health are struggling to find a healthcare niche with retail clinics.
Tele-ICU is working because it gives veterans access to needed healthcare services from wherever they are, including and especially their homes. ATLAS, meanwhile, still requires veterans to travel to a specific site for healthcare.
That strategy works well for veterans in remote locations where broadband availability and even phone service is weak, and that does address a key barrier to care. According to the GAO report, those 10 ATLAS sites where veterans did access care were successful in helping those veterans and eliminating the need for long drives to the nearest VA center and long waits.
The VHA has responded to the GAO report by saying it will transition from a pilot to a grant program, adding financial sustainability to the equation, but the GAO is also asking the agency to develop benchmarks to measure the success of the ATLAS program on an ongoing basis, much like it does for other telehealth programs. Those benchmarks could help the VHA understand why veterans aren’t going to certain ATLAS sites and enable the agency to create sustainable virtual care programs that will attract veterans.
The examples set by the VA and VHA could also help healthcare leaders to understand how and where consumers want to access care. Consumers, like veterans, prefer on-demand virtual platforms for urgent care needs, while the success of retail clinics and stand-alone services is more nuanced, driven by factors that aren’t yet fully understood.
It’s clear that virtual care can address access challenges—the VA has conducted more than 9 million telehealth visits in each of the past two years—but simply putting together a virtual care platform doesn’t guarantee success. Providers need to understand how, when, where and why patients want to access care and create programs that address the needs and eliminate the barriers.
The expectations of nursing have changed drastically in recent years, and CNOs need to pivot their strategies to sustain their workforce.
HealthLeaders spoke to Melanie Heuston, chief nurse executive at WVU Medicine and HealthLeaders Exchange member, about how CNOs can keep up with changing workforce expectations and support their nurses. Tune in to hear her insights.
The HealthLeaders Exchange is an exclusive, executive community for sharing ideas, solutions, and insights.
TGH and CEO John Couris successfully navigated Hurricane Helene using lessons learned from Hurricane Ian.
Editor’s note: On October 10, Tampa General Hospital was hit by Hurricane Milton. The system remained open as emergency response teams were activated to support communities impacted by the storm. “Preparing for Hurricane Milton was an incredible effort by the entire team and a true test of our resources, but it ensured we could continue to provide exceptional care for our patients in a high-quality, safe and uninterrupted environment before, during and after the storm,” John Couris, president and CEO of Tampa General Hospital said on its website. “Tampa General is open to support communities impacted and particularly our first responders. Working together, we will come back from Milton stronger than ever.”
When it comes to preparing for hurricanes and other natural disasters, hospitals must be ready to protect patients, staff, and facilities.
Tampa General Hospital (TGH) successfully navigated Hurricane Ian in 2022, avoiding damage despite being in the storm's path. Unfortunately, history repeated itself this year.
On September 26, Hurricane Helene made landfall in Perry, Florida, and caused flash flooding in the western part of the state, but luckily, the hospital's vulnerable Level 1 trauma center remained dry and unscathed.
“We are pleased to share that our Tampa General Hospital health system stood strong against the storm. In the wake of the hurricane, we remain focused on providing world-class care to our patients and supporting emergency response efforts across our community,” the health system’s website says.
How did they do it? Well, CEO John Couris was prepared.
Here’s how to prepare your facility for an emergency storm, based on TGH’s experiences with Hurrican Ian:
1. Start Early: Preparation Begins Well Before the Storm
TGH started preparing for Hurricane Ian a week before it made landfall. This included stockpiling critical supplies like food, water, and medical resources.
According to Couris, “[We] began bringing in our emergency supplies, med-surge supplies, and emergency water about a week out. We also set up our AquaFence, which can push back about 15 feet of storm surge.”
Actionable Steps:
Review emergency supply stockpiles well ahead of time.
Secure critical infrastructure, such as flood barriers or AquaFences.
Test backup systems like emergency generators to ensure functionality.
Three days before Ian, TGH activated its emergency command center, part of a larger 8,000-square-foot facility that uses AI and predictive analytics. “We have a dedicated emergency management team that takes over, with leadership staying on-site to make real-time decisions,” Couris explained.
Actionable Steps:
Establish a centralized command center to coordinate response efforts.
Ensure your leadership team stays on-site during the storm for fast decision-making.
Create a two-tier staffing model (Team A for the storm, Team B for post-storm relief).
3. Practice Disaster Drills Year-Round
A key to TGH’s success was regular disaster training. Couris emphasized, “All critical elements worked because we practice this stuff all year long. We conduct mass casualty drills and review standard procedures constantly.”
Actionable Steps:
Conduct regular disaster preparedness drills, including mass casualty simulations.
Train staff on emergency protocols and refine procedures through tabletop exercises.
Make emergency readiness part of routine operations.
4. Plan for Water Shortages
A major lesson learned from Ian was the critical importance of water. At the time, several hospitals had to evacuate due to a lack of water pressure, even though their facilities withstood the storm.
Couris pointed out that during Ian, “Water was the primary reason why patients had to be transferred. We are developing plans to ensure TGH has its own emergency water supply.”
Actionable Steps:
Evaluate your facility’s water usage and plan for emergency water storage.
Consider tanker trucks or alternative water sources in the event of supply interruptions.
Develop contingency plans to maintain fire suppression and sanitation systems if municipal water is compromised.
5. Make Sleeping Arrangements for Staff a Priority
TGH had 2,000 staff members caring for nearly 900 patients during Hurricane Ian. Couris noted that while their emergency plan worked, sleeping arrangements for staff always pose a challenge. “Hospitals aren’t designed to house thousands of staff for days, so we’re always learning how to do that better.”
Actionable Steps:
Designate sleeping areas and provide adequate bedding for on-site staff.
Anticipate long stays and ensure facilities are equipped with basic amenities like showers and meals.
Continually refine staff accommodation plans based on feedback from previous events.
6. Collaborate with Other Hospitals for Patient Transfers
After the Ian, TGH played a crucial role in assisting hospitals to the south, transferring over 50 patients from affected areas. Couris explained, “We had five helicopters and ambulances running patients from Fort Myers up to Tampa.”
Actionable Steps:
Establish relationships with nearby hospitals for mutual aid agreements.
Plan for patient transfers, including air and ground transportation logistics.
Ensure your hospital is prepared to accept patients from neighboring facilities in case of evacuations.
7. Constantly Review and Improve Your Disaster Plan
Couris’s final piece of advice is to make disaster preparedness an ongoing effort. “My advice to people is practice, practice, practice. Review your processes and systems consistently. The only way to stay ready is to make this part of your routine operations.”
Actionable Steps:
Regularly review and update emergency plans based on new insights and lessons learned.
Involve all levels of staff in disaster preparedness to build a culture of readiness.
Keep refining your plans to ensure your hospital is ready for the next storm.
By following these strategies, hospitals can better prepare for the impact of natural disasters, ensuring the safety of patients, staff, and facilities.
Healthcare leaders are moving quickly to keep AI growth under control, but are they handling the governance question effectively?
As healthcare organizations move swiftly to embrace AI, leadership is struggling to understand how to make sure governance isn’t pushed aside.
But what does governance really mean in a hospital or health system? And who gets to decide how and where AI is used?
At the recent HIMSS AI in Healthcare Forum in Boston, issues of compliance and liability were front and center for health system executives looking to chart a clear and effective AI strategy. Sunil Dadlani, chief information and digital officer for the Atlantic Health System, said AI regulation must be handled carefully, so that it doesn’t curb innovation.
The challenge lies in deciding where innovation has to take a step back so that compliance and liability can be addressed.
As Albert Marinez, chief analytics officer at the Cleveland Clinic, said, AI introduces “the art of the possible” to healthcare. “We know that there are problems that we can solve with generative AI that we could never solve before,” he said at the HIMSS event.
“Healthcare should be proactive in the establishment and enforcement of AI governance and guidelines,” Jim Barr, MD, Atlantic Health’s vice president of physician value-based programs and CMO of ACOs, said in an e-mail to HealthLeaders. “Governmental oversight will occur, but those in healthcare should display our ability to fully understand the issues and regulate ourselves.”
“Your reason to use AI tools can’t be just the need to say we’re on the cutting edge,” he added. “With ACOs the challenge is designing and managing successful implementation while continually measuring impact and ROI. You need to take into consideration the existing pain points for clinicians, practices and patients, their willingness to change, deploy a transparent QA/validation process to build trust, and a clear customer value proposition.”
Developing a Governance Strategy
So where does governance fit into a health system’s strategy?
Ravi Parikh, MD, MPP, an assistant professor of medicine and health policy at the University of Pennsylvania, assistant professor of medical ethics and health policy at the Perelman School of Medicine and director of the Human-Algorithm Collaboration Lab, says federal efforts to establish a governance framework have resulted in vague guidelines that are a good starting point, but not enough.
“They're sort of general guidelines on monitoring for bias and monitoring for performance drift,” he says. “But how that gets operationalized is actually really variable.”
The first step for many healthcare organizations is the creation of a governance committee, charged with managing how the health system negotiates vendor contracts as well as how AI is developed, tested, used and—most importantly—monitored.
At the HIMSS summit, Shahidul Mannan, chief data, analytics and AI officer at Orlando Health, said many health systems are using AI in small programs across the enterprise, but leadership will have to create an engine to pull everything together on the same track. The trick is deciding who sits in the engine.
Parikh says current committees are “very ad-hoc,” with a mixture of executives from areas such as clinical care, IT, legal, and finance. Few are including the patient voice, which could be a critical oversight as Ai products flood the consumer marketplace and patients ask for AI capabilities to plan and manage their healthcare.
Patrick Thomas, director of digital innovation in pediatric surgery at the University of Nebraska Medical Center, wondered at the HIMSS event whether healthcare leadership is even ready to govern AI for its patients. Patients and providers are doing their own research, he noted, forcing decision-makers to try to keep up.
Understanding the Value of Data
Beyond the makeup of a governance committee, a key function of that committee is to understand data and data analytics, especially when outsourcing AI technology.
In dealing with vendors, health systems need to understand what datasets are used and how that data can affect outcomes. For instance, a company that relies on data from a decidedly white population might not help a hospital or health system whose patient population is ethnically diverse.
And when errors, such as hallucinations, occur, it may be hard to get a vendor to correct them.
“it’s actually really difficult to respond to these hallucinations by modifying the algorithm,” Parikh says. “You might be able to fine-tune and sort of say ‘Hey, we want to avoid this type of output’ and there's certain reward-based mechanisms to do that, but usually that's not in the health system’s control. Usually it's a developer who's having to respond to feedback that they're getting from the health system and then doing some things behind the hood that we don't honestly know about.”
A governance committee also has to be perpetual, and that will cost time and money that smaller organizations don’t have. Many standards now being considered are for basic AI functions, rather than generative or predictive AI, which hasn’t matured enough to be used in healthcare. But those tools will come along soon, and the rules for governing them will have to evolve.
Parikh isn’t convinced that health systems or the federal government will be able to draft standards for an ever-evolving AI landscape. Instead, he expects organizations like the Coalition for Health AI (CHAI), the Trustworthy & Responsible AI Network (TRAIN), or the Digital Medicine Society (DiME) to create standards and adjust them as the technology evolves.
He also says the federal government could, in time, require healthcare organizations to become accredited to use different types of AI, possibly as part of a quality improvement program or even payment policy.
“We [could] have these accreditation systems that signal to developers which institutions are robust for both validating and deploying [AI] technology and which of those might not be certified for large language model generation … but might be more certified for other types of predictive AI solutions,” he says. “I suspect that people are going to realize that some health systems just have more capacity for governance and more data availability to be deploying these tools. And that's a good thing for patients because we don't want to be rolling these things out for patients where errors might be promulgated.”
Cassie Lewis, chief nursing officer at Bon Secours Mercy Health Richmond market, chats with nursing editor G Hatfield about how nurses and nurse leaders can help patients engage with their healthcare, in and out of the hospital.
