Across the world, artificial intelligence (AI) technologies are becoming interwoven with the fabric of daily life. Health care is no exception: Professionals are beginning to use AI tools like large language models (LLMs) and predictive analytics to simplify administrative tasks, transcribe medical information, parse datasets for large-scale research, and even assist with diagnoses.
Technological advances may not prioritize marginalized populations and, so far, AI hasn’t been extensively integrated into health care in correctional facilities such as prisons, jails, and detention centers. Correctional health care is a complex landscape, partly because incarcerated people have a disproportionately high rate of health problems, but they also have reduced access and fewer options in health care compared to non-incarcerated people.
In other settings, AI has shown some promise in helping more people access health care. Experts say this technology also could inadvertently worsen systemic inequity and mistrust through algorithmic bias. But if these concerns can be addressed, and AI can be implemented with care, could it potentially improve health care for people in the justice system?
A University of Colorado Anschutz Department of Medicine physician thinks so — and he’s actively advocating for researchers to explore responsibly integrating AI into correctional health care, which he hopes might ease some of the challenges that incarcerated patients face without reinforcing structural barriers.
Over the last two years, Lawrence Haber, MD, professor of clinical practice at CU Anschutz and a hospitalist at Denver Health, has published nearly a dozen papers relating to health care, human rights, safety, equity, and inclusion for incarcerated people. Most recently, he was the lead author of a viewpoint piece in JAMA Internal Medicine where he argued for the potential value of AI in carceral health care.
“I’m in favor of making health care for incarcerated people as similar as possible to health care for others in the community,” says Haber. “People who are incarcerated have a constitutional right to health care, and they deserve to have their serious medical needs addressed.”
In his new paper, Haber says that the U.S. incarcerates a larger share of its population than any other nation, with about 2 million people detained in prisons and roughly 7 million people cycling through jails each year.
These millions of incarcerated people often have higher-than-average health care needs. They tend to suffer from mental health conditions, substance use disorders, and various types of chronic medical diseases at a much higher rate than the general population.
And at the same time, incarcerated patients have restricted choices regarding their health care. Haber says they have fewer clinicians and specialists to choose from due to workforce shortages in correctional systems, and they may also face general digital access restrictions. The result is more fragmented, lower-quality care than what people in the community would typically receive.
During his time as a hospital medicine physician at the University of California San Francisco and the San Francisco General Hospital, Haber spent over 10 years personally caring for people who were incarcerated or hospitalized. San Francisco General was the county’s referral site for jail medical services, where patients received medical evaluations en route to or from correctional facilities. This is where he first saw that these patients were systematically treated differently from other patients.
Haber says he was encountering situations and decisions that he had not come across before with other patients, and he found there was a lack of guidance and literature about health care in these settings. So, he set out to change that by furthering research and advocacy in this area. His time at San Francisco General fueled a passion for carceral health care that continues to this day. At Denver Health, he and his colleagues care for hospitalized patients coming from local jails and state prisons.
Haber writes in his new paper that certain AI tools could improve carceral care across a myriad of dimensions.
First, there are often shortages of clinicians and other resources, especially in mental health and specialty areas of medicine, in correctional facilities. Haber suggests that predictive analytics could help identify patients at risk for mental health crises or suicide in settings where there are too few clinicians and it might be easy to miss vulnerable patients. There’s some evidence that machine learning models could detect suicidal behavior in prison detainees, for example.
Haber also argues that AI chatbots and other LLM-based tools could help educate patients, encourage them to stick with treatment plans, and empower them to advocate for their health. And he notes that clinicians could use AI decision-support tools to provide evidence-based care tailored to each patient’s needs.
While patients in the justice system do sometimes have access to telemedicine, Haber believes AI could improve and expand this access by helping with tasks like multilingual translation, creating documents, and facilitating consultations. In addition to all of these possibilities, wearable medical devices with adequate privacy protections could provide valuable information about disease outbreaks or other issues within incarcerated communities.
There are many possibilities for integrating AI into correctional health care settings, but there are barriers to implementing them. Most people use AI through computers and mobile phone applications, but people who are incarcerated often don’t have access to these. They may be able to use personal tablets while incarcerated, but they’re only allowed to use them in specific, limited ways. Most of the time, they can’t access their electronic health information.
Haber suggests incarcerated patients should be allowed greater access to secure platforms where they can view their health records and get information about their health conditions. He also emphasizes that in these settings, it will be critical to keep clinical AI technologies separate from any kind of correctional surveillance infrastructure so it’s easier for patients to trust the AI rather than seeing it as a tool of the carceral system.
Another place where AI could make a difference is in the transitional period when people are re-entering society after being released from a correctional facility. During this time, patients are especially vulnerable and have a heightened risk of being hospitalized or dying. They frequently experience unemployment, poverty, and a lack of stable housing, and they may struggle with substance use or other health conditions, especially if they’ve lost contact with their carceral care providers and don’t have primary care referrals or medications to provide ongoing support.
In these situations, Haber suggests AI tools could identify people at higher risk for these problems, connect them with needed resources, remind them to take their medications, and help them schedule appointments. AI-integrated virtual reality could also help recently released individuals with other practical skills like prepping for job interviews or managing daily life.
When working with incarcerated or formerly incarcerated populations, AI tools need to be carefully designed so they are free of bias, says Haber. He suggests that any tools used in correctional settings should have shared oversight by independent bodies, such as public health agencies, in addition to correctional authorities. These tools should also be transparent in how they use their data and be continually inspected and audited for bias. Haber stresses that there need to be adequate privacy protections in place for incarcerated patients.
As long as these concerns can be adequately addressed, Haber says, AI could reshape the way health care is conducted in correctional facilities and potentially improve care for patients. And he sees these AI tools as part of a broader progression toward better health care across the entire medical system.
“I don't think that we inherently should shy away from trying to use technological advances to improve care for marginalized populations, as long as we do it with our eyes open to the potential risks this kind of technology can pose,” he says.