Literature Review

What does it mean to be human in the age of AI? The Hastings Center for Bioethics - The Big Question; host Vardit Ravitsky with Insoo Hyun; 9/3/25 What are the moral and philosophical challenges we face as AI advances and humans navigate aging, end-of-life care, and cognitive decline? “We’re at a turning point because for the first time in the history of either our planet or our species we are seeing the emergence of a nonhuman intelligence,” says Hastings Center President Vardit Ravitsky in conversation with Hastings Center Fellow Insoo Hyun of the Museum of Science in Boston in the latest episode of The Big Question. 

Irish man uses ChatGPT for sore throat advice, later diagnosed with stage 4 cancer TimesNow; by Ashima Sharda Mahindra; 9/1/25 A 37-year-old Irish man from Kerry sought advice on ChatGPT on chronic sore throat and swallowing difficulties – and was ultimately diagnosed with stage 4 esophageal adenocarcinoma, a rare and aggressive throat cancer. Warren Tierney said ChatGPT had reassured him that cancer was “highly unlikely,” which led him to a delay in seeking professional medical care. Months later, Warren was diagnosed with cancer, which has a five-year survival rate between five and ten per cent globally.

Can AI simplify surgical instructions as effectively as humans? Enhanced surgical instructions using large language modelsJAMA Network / JAMA Surgery / Research Letter; by Maxwell Sahhar, Joseph E. Nassar, Anne-Emilie Rouffiac, Kaitlyn Crow, Manjot Singh, Michael J. Farias, Bassel G. Diebo, Alan H. Daniels; 8/25Human authors and a chatbot improved documents from a 9th-grade to 6th-grade reading level on most measures. Preoperative instructions showed less improvement by a chatbot. Inconsistencies occurred at rates of 1.1 and 0.6 per document for human authors and a chatbot, respectively. At least 1 inconsistency was found in 20 of 43 human-simplified (47%) and 11 of 43 chatbot-simplified (26%) instructions, with the higher rate of inconsistencies from humans associated with nonclinically significant omissions. Similar rates of critical inconsistencies were observed between human authors and a chatbot.Publisher's note: AI was effective in this study. If AI can simpligy surgical instructions, how would it do with hospice admission and other instructions?

Top ten tips palliative care clinicians should know about the future of generative intelligenceJournal of Palliative Medicine; by Pragat Patel, Raj Patel, Gilad Kuperman, Monica Consolandi, David Collett, Shunichi Nakagawa, Nicolas Burry, Karl B Bezak, Jake Strand, Mihir Kamdar, Judith E Nelson, William E Rosa, Craig D Blinderman; 8/25Artificial intelligence (AI) is transforming health care and has implications for palliative care (PC) and serious illness communication (SIC). This article integrates interdisciplinary and interprofessional expertise, providing key tips for PC clinicians to best leverage AI-driven tools to enhance PC practices, care efficiency, and patient- and family-centered outcomes. AI-driven tools range from predictive analytics to improve intervention delivery to AI-assisted communication and documentation of goals of care. Beyond these potential benefits are important ethical, logistical, and equity considerations that must be carefully addressed.

When patients arrive with answersJAMA Network; by Kumara Raja Sundar; 7/24/25Patients arriving with researched information is not new. They have long brought newspaper clippings, internet search results, or notes from conversations with family. Potential solutions passed along in WhatsApp threads have at times been an integral part of my clinical conversations. Information seeking outside the health care setting has always been part of the landscape of care. But something about this moment feels different. Generative artificial intelligence (AI), with tools like ChatGPT, offers information in ways that feel uniquely conversational and tailored. Their tone invites dialogue. Their confidence implies competence. Increasingly, patients are bringing AI-generated insights into my clinic and are sometimes confident enough to challenge my assessment and plan. I heard these tools were helpful, but I understood their appeal only after using them myself. Recent studies have bolstered this claim: large language models (LLMs) show surprising strength in reasoning and relational tone. After seeing it firsthand, my reaction was simple: “Man, I get why my patients like it.”

Ethical obligations to inform patients about use of AI toolsJAMA; by Michelle M. Mello, Danton Char, Sonnet H. Xu2; 7/25When a health care organization decides to deploy an AI tool, it should decide whether notification or consent is ethically required. To decide that disclosure is not required at the point of care is not to conclude that organizations have no transparency obligations concerning their use of AI. Two complementary steps can help discharge these obligations. First, health care organizations’ websites and written materials distributed to patients should explain how the organization uses AI to deliver better, safer, more efficient care and protect staff well-being. This kind of transparency honors patients’ desire to know how their care may be affected by AI and may help some make decisions about where to seek care. 

Opportunities and barriers to artificial intelligence adoption in palliative/hospice care for underrepresented groups-A technology acceptance model–based reviewJournal of Hospice and Palliative Nursing; by Xu, Tuzhen; Rose, Gloria M.; 8/25Underrepresented groups (URGs) in the United States, including African Americans, Latino/Hispanic Americans, Asian Pacific Islanders, and Native Americans, face significant barriers to accessing hospice and palliative care. Factors such as language barriers, cultural perceptions, and mistrust in healthcare systems contribute to the underutilization of these services. Recent advancements in artificial intelligence (AI) offer potential solutions to these challenges by enhancing cultural sensitivity, improving communication, and personalizing care. However, barriers such as limited generalizability, biases in data, and challenges in infrastructure were noted, hindering the full adoption of AI in hospice settings.

Creating user personas to represent the needs of dementia caregivers who support medication management at home: Persona development and qualitative studyJMIR Aging; by Anna Jolliff, Priya Loganathar, Richard J Holden, Anna Linden, Himalaya Patel, Jessica R Lee, Aaron Ganci, Noll Campbell, Malaz Boustani, Nicole E Werner; 7/25Caregiver-assisted medication management plays a critical role in promoting medication adherence and quality of life for people living with Alzheimer disease or related dementias (ADRD). Caregivers in this study demonstrated a range of characteristics and values that informed their approach to medication management. They used a combination of technology-based strategies and strategies situated in their physical environments to manage medications. The personas created can be used to inform interventions, such as digital tools, that address caregivers' unmet needs.

Why AI isn't your advantage — your people are Entrepreneur; by Bidhan Baruah; 8/6/25 ... AI can indeed be beneficial, but only if you have the right people who can guide it thoughtfully and strategically. Key Takeaways:

Laying the foundation for health IT innovation: 5 essential cornerstones Becker's Health IT; by Pally Parrent; 8/1/25 ... A recent survey by Modern Healthcare and Nordic reveals both industry progress and opportunity across critical IT foundations like governance, infrastructure, data, and training. Investing in these foundations is important but success also requires five reinforcing cornerstones that support foundational efforts for enterprise-wide impact: strategy, cybersecurity, integration and interoperability, cultural commitment, and change management. ...

Pitfalls of large language models in medical ethics reasoningnpj Digital Medicine; by Shelly Soffer, Vera Sorin, Girish N. Nadkarni, Eyal Klang; 7/25 Large language models (LLMs), such as ChatGPT-o1, display subtle blind spots in complex reasoning tasks. We illustrate these pitfalls with lateral thinking puzzles and medical ethics scenarios. Our observations indicate that patterns in training data may contribute to cognitive biases, limiting the models’ ability to navigate nuanced ethical situations. Recognizing these tendencies is crucial for responsible AI deployment in clinical contexts.

Evaluating a large language model in translating patient instructions to Spanish using a standardized frameworkJAMA Pediatrics; Mondira Ray, MD, MBI; Daniel J. Kats, MD, MBI; Joss Moorkens, PhD; Dinesh Rai, MD; Nate Shaar, BA; Diane Quinones, MS, RN, CPNP; Alejandro Vermeulen, BFA, CMI; Camila M. Mateo, MD, MPH; Ryan C. L. Brewster, MD; Alisa Khan, MD, MPH; Benjamin Rader, PhD; John S. Brownstein, PhD; Jonathan D. Hron, MD; 7/25Patients and caregivers who use languages other than English in the US encounter barriers to accessing language-concordant written instructions after clinical visits. Large language models (LLMs), such as OpenAI’s GPT-4o ... can translate Spanish translations of real-world personalized written patient instructions that are comparable in quality to those generated by professional human translators. Independent professional medical translators preferred the GPT-4o–generated translations over the human translations, and error analysis revealed a higher rate of mistranslation errors in the human translations. These findings demonstrate GPT-4o’s ability to produce quality translations in Spanish ...

CIOs’ tech wishlist: What IT leaders would buy with a blank check Becker's Health IT; by Naomi Diaz; 7/14/25 If hospital and health system CIOs had unlimited resources, where would they place their biggest technology bets? Becker’s asked several IT leaders: If you had a blank check to invest in one technology tomorrow, what would it be and why? [Only a few responses are listed here. Go to the source article for the IT leaders' name, organization, role, and answer.]