Pittsburgh’s two hospital giants, Allegheny Health Network (AHN) and the University of Pittsburgh Medical Center (UPMC) are rolling out artificial intelligence tools across their facilities — in radiology, in patient safety, and in the mountains of administrative paperwork.
Backers say it could be a lifeline for an industry still shorthanded after COVID. Critics, including a CMU researcher and a health care union, want to know who’s really being served.
Health care is the largest employment sector in Pittsburgh. As capital-intensive data center projects break ground across the state, Pittsburgh’s robust research and development fields are tapping the amassing computing power for new AI technologies that promise to reshape health care.
Lance Lindauer, executive director of the Partnership to Advance Responsible Technology, said hospital systems are strategizing these innovations around three main goals:
Imaging and diagnosis
Pathologists, people who study and identify diseases, can now use AI as “another arrow in the quiver” to assist with making diagnoses, said Lindauer.
“Being able to look at data quicker and hopefully being able to help with more accurate disease discovery or cancer screening” can lead to improved health outcomes, he said. “Maybe human eyes might miss something, but with AI you could say, you know what, maybe that does look a bit abnormal.”
“Maybe human eyes might miss something, but with AI you could say, you know what, maybe that does look a bit abnormal.”Lance Lindauer
One such tool, developed by the University of Pittsburgh’s Computational Pathology and AI Center of Excellence, is called WSI Genie. WSI, or whole slide imaging, is a technique in which a thin slice of bodily tissue is scanned and turned into a high-resolution digital image. This new tool is a “machine learning framework” that can offer insight into these images before being assessed by a human.
Another tool in this bucket comes from Ibex, a UPMC-partnered company which builds AI for cancer detection. Ibex has said in releases and on its website that prostate biopsies are “digitized and analyzed” by the tool prior to review by a pathologist, noting that the process is still in the research phase and hasn’t yet cleared Food and Drug Administration approvals.
Patient safety
AI is also being used in the hospital to build systems that enhance patient safety. One such tool is known as ambient video — “just a fancy term for saying, ‘There are things that are listening to us to take in the ambient surroundings and provide an action,’” said Dr. Richard Medford, AHN’s chief digital information officer.
AHN has partnered with a company called care.ai to install video cameras in hospital rooms throughout the system. About two-thirds have been fitted and AHN expects to finish the work by the end of the year.
The care.ai video system has two modes of operation: active and passive.
“The active form is very much like how you would communicate with Microsoft Teams or Zoom,” Medford said. “There’s a camera and you can communicate virtually with someone. … They can beam in from any one location to provide that similar care.”
Medford said AHN hasn’t yet adopted this technology but intends to use it to allow Pittsburgh-based specialists to “beam into” more remote hospitals in the network.
In passive mode, the patient’s condition is constantly monitored through the care.ai camera. “What this camera and this AI platform can do is monitor this patient and say, ‘Hey, this patient is at risk of falling.’ And the idea is, 45 seconds before it detects that fall, it alerts a clinician or a nurse,” said Medford.
The care.ai cameras, already active at AHN, allow the system to expand the reach of “telesitter” nurses. “Now that we’re able to monitor 20 patients at a time … That nurse extends their reach and capability so they can monitor multiple people instead of actually sitting in the room,” Medford said.
Note-taking apps and administrative burden
The final strand of AI technologies transforming health care are those aimed at reducing the administrative burden of running a hospital.
A generative AI tool developed by Pittsburgh-based startup Abridge is producing written medical charts from recordings of doctor-patient conversations.
“Abridge … was astoundingly accurate for me,” said Dr. Laxmi Shah, a neurologist currently on sabbatical from her practice. “Having to see a large number of patients, you want to complete your notes in a timely manner. … I think this allowed a better work-life balance as a working mom.”
“Being able to spend more time talking to the patient rather than having to type my notes, it allowed me to … focus more on the patient interaction itself.”
Laxmi Shah
Shah began using Abridge while practicing at AHN in 2024 after seeing several of her colleagues use it. “I was able to be done with my notes before I left for the day. … Being able to spend more time talking to the patient rather than having to type my notes, it allowed me to … focus more on the patient interaction itself.”
“I’ve never seen something take off like ambient listening in my career,” said Medford. The technology allows doctors to avoid “pajama time … doing my notes all up through the wee hours of the night,” he said, allowing better records, more family time for physicians and less risk of burnout.
Medford said Abridge reduces administrative burden by “bridging the divide” between the health care providers at AHN and the insurance providers at parent company Highmark by addressing the criteria the insurance arm needs.
“So there’s not any need to go back and forth with the insurance company to make sure that it’s approved.”
Patient triage and follow-up care are another two administrative tasks soon to be taken up by AI. “We can create our own call centers, and as patients are calling in, have them interact with a voice bot, to say, have we triaged you appropriately? Do you have the right level of care? Can we schedule you for an appointment? Without ever involving a human in all that stuff,” said Medford.
Burnout and wait times
Much of the innovation in AI medical technology seeks to address the same core issue: Hospital workers are overburdened, leading to burnout and staff shortages. Hospitals across Pennsylvania are experiencing high vacancy rates in several key job categories, and the Pennsylvania Department of Labor estimates that hospitals across the state will need nearly 9,000 more registered nurses and more than 11,000 new nursing assistants each year through 2032 “to replace those leaving the workforce and meet increasing demand for care.”
Lindauer points out AI’s potential to alleviate the workload placed on providers. “The general kind of comment you traditionally hear in health care settings right now is: ‘It’s really hard to deal with some staffing shortages,’” he said. “If we can’t hire more doctors, AI is seen as one way to … give them time back in their day.”
Before the COVID-19 pandemic, over 30% of nurses who left their job cited burnout as a reason for leaving. Since then the problem has only worsened, leading to endemic staff shortages throughout U.S. hospitals.
Understaffing can have a negative impact on the quality of care, particularly on patient wait times. In 2024, 70% of Pennsylvania hospitals reported that wait times for emergency room visits had increased since the previous year, and 68% reported that wait times for appointments had increased similarly.
“Patients are frustrated when they want to see a health care provider and are told they have to wait three months for a visit,” said Melissa DeiCas, a certified nurse-midwife at UPMC Magee-Womens Hospital.
To cut down patient wait times, DeiCas said, UPMC has incentivized providers to see more patients including supplemental pay for meeting targets for new patient visits within a certain time period.
DeiCas has not personally adopted any AI tools into her practice. “While there could be some really good things about [AI], I don’t feel comfortable enough with it yet to be using it.” She said she hasn’t felt pressured by management to adopt new tools. “But I will say, that is something I worry about … The pressure to do shorter visits, to see more people, that kind of thing I worry about.”
DeiCas’ concern is echoed by Service Employees International Union (SEIU) Healthcare PA, the largest health care workers’ union in the commonwealth which represents over 25,000 hospital employees.
“In recent decades, we have seen a relentless drive to decrease the time we can spend with our patients in order to cut labor costs and increase profits for health systems and insurance companies,” SEIU’s statement to Public Source said. Workers recognize potential benefits in AI technology, according to the statement, but they fear it could reduce their contact with patients while increasing, rather than reducing, workloads.
“We fear that AI systems could be rolled out in a way that creates more tasks to be piled on top of fewer health care workers, causing further exhaustion and depletion of health care workers.”
Public Source contacted UPMC for comment on the role of AI in alleviating the hospital staffing crisis, but the health care system’s spokesperson did not reply.
John Zimmerman, a professor at Carnegie Mellon’s Human-Computer Interaction Institute, raised concerns about AI’s role in shaping the health care industry.
“I would say most of the innovation happening in health care is exploitive,” he said.
That might be more to do with the for-profit economic model underlying health care in the U.S. than the technology itself, he said. “You have clinicians who are generally interested in patients and health outcomes, but the enterprise isn’t. … We see a lot of optimization in ways that have nothing to do with health care outcomes.”
Zimmerman teaches a course at CMU called “AI Innovation in the Healthcare Sector,” in which students contemplate new technology applications that could improve patient outcomes while being financially viable.
“I do maybe worry that we’re going to create a multi-tiered system where people on the lower socioeconomic rungs are never allowed to get to a human service provider. I think we’re gonna make health care worse and cheaper.”
‘Putting the humanity back in health care’
As AI reshapes the economics of hospital labor, unionized workers at UPMC Magee-Women’s Hospital are advocating for better staffing standards to address the ongoing worker shortage. Their solution: more humans, not automation.
The union, which won an organizing vote in August, is negotiating for their first contract with UPMC.
“The nurses’ goal in contract negotiations is the same as what drove them to form their union last year — to put the humanity back in health care,” the union said in a January press release.
Rather than AI, the union seeks the establishment of staffing ratios set by national professional organizations like Association of Women’s Health, Obstetric and Neonatal Nurses (AWHONN). For example, the union wants one nurse to one woman who is laboring or being induced.
“It’s up to your leaders, including myself, to advocate for our providers to say: That time back doesn’t necessarily mean more patients. That time back means less burnout, less moral injury.”
Richard medford
DeiCas is hopeful that these new standards can pave the way toward a stronger workforce. Midwifery is a “high-touch, low-tech” discipline, said DeiCas, one in which a strong patient-provider relationship is paramount. And building that relationship, she says, requires time and attention that cannot be rushed. “You can’t build trust with somebody … without time.”
Medford, of AHN, said it’s up to hospital leadership to ensure that AI is implemented responsibly into the health care environment. “[AI] may give you some time back, but then it’s up to your leaders, including myself, to advocate for our providers to say: That time back doesn’t necessarily mean more patients. That time back means less burnout, less moral injury, less whatever it is to the provider so that they can continue to provide for their patients.”
Will Curvan is a Pittsburgh-based freelance journalist and can be reached at willcurvan@gmail.com.
This story was fact-checked by Joya Nasr.
MOST READ STORIES
Republish our articles for free, online or in print, under a Creative Commons license.
