Continuity of Care and AI at the Hand-Off

Continuity of care is easy to endorse and hard to deliver. A well-established framework divides it into three parts, the evidence links it to lower mortality and cost, and AI now produces many of the documents that carry it from one setting to the next.

In Brief

The Problem. Care in the United States is delivered across many providers, settings, and payers, and continuity breaks most often at the hand-off, where information, plans, and responsibility must move with the patient.

Why It Persists. Fragmented delivery, uneven data exchange, workforce shortages, and payment that still rewards individual services all work against continuity, and AI-drafted notes and summaries now add a new point where errors can travel with the record.

What Boards Should Do. Treat continuity as an outcome the board monitors, including as an outcome of AI deployment, and ask management to report how transitions of care are measured, which documents AI drafts, and who reviews them.

What Continuity of Care Means

In its ideal form, care is continuous: a patient moves from prevention to treatment and recovery without gaps, and every provider works from the same understanding of the patient. Continuity has long been a cornerstone of effective care and an objective of health systems worldwide. It means more than eliminating gaps. It means care that is coherent, connected, and centered on the patient, delivered over time and across settings by professionals who work together.

The clearest framework comes from a 2003 review in the BMJ by Jeannie Haggerty and colleagues. They found that the term meant different things to different disciplines, identified two core elements common to all of them (care of an individual patient, and care over time), and described three types of continuity (Haggerty et al., 2003):

• Informational continuity: the use of information about past events and personal circumstances to make current care appropriate for the individual. It is not only access to records. It is an understanding of the patient's history in context: background, preferences, and health trajectory.
• Management continuity: a consistent and coherent approach to managing a health condition that responds to the patient's changing needs. Every provider involved works toward shared, agreed goals under a common plan.
• Relational continuity: an ongoing therapeutic relationship between a patient and one or more providers, built on trust, mutual understanding, and a shared history.

The framework remains useful because each type fails in a different way. Informational continuity fails when records are incomplete or do not travel. Management continuity fails when providers work from different plans. Relational continuity fails when the patient no longer has a regular clinician.

The idea is not new, but it matters more as chronic disease, an aging population, and medical specialization multiply the number of providers each patient sees. My argument in this article has changed since 2022. Continuity is no longer only a matter of clinical practice and payment design. Artificial intelligence (AI) now drafts many of the documents through which information moves between providers. That makes continuity an outcome of AI deployment, and therefore a matter for board oversight.

Why Continuity Matters

Continuity affects both the patient's experience and the performance of the system. Patients who build long-term relationships with their clinicians tend to trust them more, follow treatment plans, and take up preventive care. Patients with chronic conditions, who see multiple providers across settings, are the most exposed when continuity fails.

The evidence is strongest where it matters most. A 2018 systematic review found that 18 of 22 eligible studies reported a statistically significant reduction in mortality with greater continuity of care, with both generalist and specialist doctors; the studies were observational (Pereira Gray et al., 2018). In the United States, an analysis of Medicare claims for about 1.4 million beneficiaries found that patients in the highest quintile of primary care physician continuity had 14.1 percent lower adjusted expenditures and 16.1 percent lower odds of hospitalization than those in the lowest quintile (Bazemore et al., 2018).

Continuity also improves the quality of decisions. A clinician who understands a patient's history and circumstances is better placed to make an accurate diagnosis and an effective plan. These are the reasons continuity belongs among the outcomes a board watches, not only among the values it endorses.

How to Measure It

Measuring continuity is harder than endorsing it, and no single measure captures it. The three types suggest different indicators, and measures of outcomes and use complete the picture. The right column shows where AI now affects each.

Type	Indicators	Where AI now enters
Informational	Completeness, accuracy, and accessibility of records across settings; the extent of electronic health record (EHR) interoperability; the frequency of information exchange between providers	AI-drafted notes, discharge summaries, referral letters, and patient instructions
Management	Adherence to evidence-based care pathways; use of coordinated care plans; use of transitional care services between settings	Models that recommend pathways or flag patients for follow-up
Relational	Strength and duration of the patient-provider relationship; patient-reported trust, satisfaction, and quality of communication	Whether time saved by documentation tools returns to patients
Outcomes and use	Hospital readmission rates; emergency department visits; control of chronic conditions	Changes in these measures after AI deployment

Current Challenges in the United States

Achieving continuity in the United States is difficult, and the causes are largely structural. Six challenges stand out.

Fragmentation of care. Patients receive care from many providers in many settings, and important information is lost between them. In a study of Medicare claims for 1.79 million fee-for-service beneficiaries, the typical beneficiary saw a median of two primary care physicians and five specialists working in four different practices each year, and only a median of 35 percent of a beneficiary's visits were with the physician most responsible for his or her care (Pham et al., 2007). Fragmentation is most harmful for patients with chronic disease who need consistent management.

Interoperability and information exchange. EHR adoption made patient information more available, but systems still do not always exchange it well. In 2023, 70 percent of non-federal acute care hospitals reported engaging, at least sometimes, in all four domains of interoperability (finding, sending, receiving, and integrating information), up from 46 percent in 2018; larger and system-affiliated hospitals engaged more than small, rural, and independent ones (Office of the National Coordinator for Health Information Technology, 2024). Gaps in exchange weaken informational continuity and lead to repeated tests and treatment.

Workforce shortages and distribution. The Association of American Medical Colleges projects a shortage of 13,500 to 86,000 physicians by 2036, including 20,200 to 40,400 in primary care (Southwick, 2024). Many patients have no regular source of care, which disrupts relational continuity, and the uneven distribution of clinicians widens disparities in access.

Transitions of care. The move from hospital to home or to a long-term care facility is the most vulnerable point. Among Medicare fee-for-service beneficiaries discharged from a hospital, 19.6 percent were rehospitalized within 30 days; half of those rehospitalized after a medical discharge had no physician office visit billed in between, and unplanned rehospitalizations cost Medicare an estimated USD 17.4 billion in 2004 (Jencks et al., 2009). Poor communication at these transitions leads to medication errors, avoidable readmissions, and worse outcomes.

Healthcare financing. Fee-for-service payment rewards the quantity of services rather than coordination over time. It encourages overuse of some services and underuse of others, including preventive care and chronic disease management. It still dominates: in 2023, 54.8 percent of United States health care payments were fee-for-service or fee-for-service linked to quality and value (Health Care Payment Learning and Action Network, 2025).

Social determinants of health. Income, education, housing, and neighborhood affect a person's ability to reach and stay engaged with care. Healthy People 2030 defines these determinants as the conditions in the environments where people are born, live, learn, work, play, worship, and age, and organizes them into five domains: economic stability, education access and quality, health care access and quality, neighborhood and built environment, and social and community context (Office of Disease Prevention and Health Promotion, n.d.). Addressing them is difficult, and it is essential to equitable, continuous care.

Commercial Practices, Models, and Initiatives

Providers, insurers, and other private organizations have developed several models to improve continuity. The evidence for each is real but more mixed than the original version of this article suggested.

Patient-centered medical homes (PCMHs). In this model, a primary care provider supported by a team serves as the patient's main point of contact and coordinates care across providers and settings, and many insurers pay incentives to practices that meet the criteria. A synthesis of 11 PCMH initiatives found no change in most outcomes studied, including emergency department and inpatient use, but a 4.2 percent reduction in total spending (excluding pharmacy) among higher-morbidity patients, with wide variation across initiatives (Sinaiko et al., 2017). The model works best where the patients are sickest.

Accountable care organizations (ACOs). ACOs are groups of providers that take responsibility for the quality and cost of care for a defined population, which rewards value over volume. Commercial insurers offer shared savings contracts, and Medicare's program provides the clearest public results. In performance year 2024, 476 Medicare Shared Savings Program ACOs, serving 10.3 million assigned beneficiaries, generated USD 2.5 billion in net savings to Medicare, and 75 percent of them earned shared savings (Centers for Medicare and Medicaid Services, 2025).

Digital health and telehealth. Telehealth extends the reach of providers to patients in rural and underserved areas, and digital platforms support communication and information exchange between providers. Its future in Medicare is settled for now: Congress extended the Medicare telehealth flexibilities through December 31, 2027 (American Academy of Sleep Medicine, 2026).

Care transition programs. Because transitions are the weakest point, many organizations assign care coordinators who work with the patient, the family, and the providers on a transition plan. The approach has trial evidence. In a randomized controlled trial of the Care Transitions Intervention with 750 patients, in which a coach supported patients after discharge, nonelective rehospitalization fell from 11.9 percent to 8.3 percent at 30 days and from 22.5 percent to 16.7 percent at 90 days (Coleman et al., 2006).

Value-based payment models. Pay-for-performance, bundled payments, and capitation reward providers for quality and cost-effective care, which aligns incentives with continuity. In 2023, 28.5 percent of United States health care payments flowed through models with downside risk or population-based payment, led by Medicare Advantage at 43 percent and traditional Medicare at 33.7 percent; commercial plans stood at 21.6 percent (Health Care Payment Learning and Action Network, 2025).

Future initiatives. Organizations are using data and predictive analytics to identify patients at risk of fragmented care and to target interventions, and insurers are addressing social needs such as food insecurity and housing instability. Both rely increasingly on AI, which is where continuity now meets governance.

AI Now Sits at the Hand-Off

AI has moved into exactly the points where continuity is most fragile. Ambient clinical documentation drafts the clinical note from the conversation between patient and clinician, and it accounted for about USD 600 million of healthcare AI spending in 2025 (Yap et al., 2025). AI tools also draft discharge summaries, referral letters, and patient instructions. Each of these documents is how information travels to the next provider.

These tools can strengthen informational continuity. A complete, timely summary is better than a late or missing one. They can also weaken it. An error in an AI-drafted summary does not stay in one encounter. It moves with the patient to every provider who relies on the record. The evidence so far supports review rather than avoidance. In a blinded comparison of 100 hospital medicine encounters at the University of California, San Francisco, 22 attending physicians found that AI-drafted discharge narratives contained more omissions (1.75 versus 0.86 per summary) and more inaccuracies (0.93 versus 0.65) than physician-written ones, although overall harm scores were low and overall quality was similar. The authors concluded that AI drafts reviewed by physicians may be a viable option (Williams et al., 2025). The question for the board is not whether to use these tools. It is who reviews what they produce, and how errors are found and reported.

AI touches the other two types as well. A model that recommends a care pathway can support management continuity if every provider follows the same plan, and undermine it if different settings use different tools that reach different conclusions. Relational continuity is the least visible effect. Documentation tools are often justified by the time they return to clinicians. In the early rollout of ambient scribes at The Permanente Medical Group, physicians reported more personal and effective patient interactions and less after-hours clerical work (Tierney et al., 2024). Whether that time goes to patients across the organization is a measure worth tracking.

Regulators and accreditors have begun to set expectations. California's AB 1979 takes effect on January 1, 2027. It requires facilities, clinics, and physician offices to take reasonable steps so that a licensed provider exercises independent judgment whenever AI output informs care (Montano & Tran, 2026). The Joint Commission launched its voluntary Responsible Use of AI in Healthcare certification on June 1, 2026, and its standards include monitoring and validation of safety performance (Guerra, 2026).

Education, Preventive Health, and Wellness

Education, prevention, and wellness are part of continuity, not separate from it. Each depends on, and strengthens, the relationship between patient and provider.

Patient education. A patient who understands the condition, the value of preventive measures, and the reasons for the treatment plan manages chronic disease better. Providers teach through counseling on disease management, medication, and lifestyle; organizations add written and online resources and workshops; insurers explain coverage and how to reach care. The need is large. In the most recent national assessment, 36 percent of United States adults had basic or below basic health literacy (Kutner et al., 2006).

Preventive health. Vaccinations, screenings, and regular check-ups prevent disease or find it early, and a consistent relationship with a provider makes it more likely that a patient receives them. Federal law supports coverage. On June 27, 2025, the Supreme Court upheld, six to three, the Affordable Care Act requirement that most private insurers cover preventive services recommended by the U.S. Preventive Services Task Force without cost sharing, a benefit that about 100 million privately insured people use each year (Sobel et al., 2025). Some insurers add incentives, such as premium reductions, for members who use preventive care.

Wellness. Wellness encompasses physical, mental, and social well-being: healthy eating, physical activity, stress management, and social connection. A long-term relationship with a provider allows ongoing, personal discussion of these behaviors and referral to resources. Insurers can support wellness through coverage for services such as nutrition counseling and mental health support, and through incentives for healthy behavior.

Economic Models and Incentives

Payment shapes continuity more than any other single factor, and the models now in use differ sharply in what they reward.

Model	How it pays	Effect on continuity
Fee-for-service	A payment for each service rendered	Rewards volume; does not reward prevention, chronic disease management, or coordination
Pay-for-performance	Bonuses for meeting quality measures	Rewards what is measured; continuity improves only if measured
Bundled payment	One payment for all services in an episode of care	Rewards coordination within the episode, including the transition home
Capitation	A set amount per patient regardless of services	Rewards prevention and management over time
Shared savings and risk-based contracts	Providers share savings below a cost target, or bear losses above it, while meeting quality benchmarks	Rewards managing the whole population efficiently
Patient incentives	Reduced premiums or cost sharing for engaging with a primary care provider or a disease management program	Rewards the patient's side of the relationship

The shift toward these models is real, as the ACO and payment data above show, but it brings its own risks: the need for reliable performance measurement, the risk that providers avoid high-risk patients, and the infrastructure and resources the models require. AI adds a fourth. A model that predicts cost or risk can, if poorly governed, steer attention away from the patients whom continuity most benefits.

Bridging Commercial and Public Insurers

The United States system combines public and private coverage, and continuity across the whole system depends on how well the two work together. In 2024, 66.1 percent of the population had private coverage and 35.5 percent had public coverage at some point in the year, including 19.1 percent with Medicare and 17.6 percent with Medicaid (Bunch & Ketema, 2025). Many patients move between the two as their age, employment, and income change, and continuity is at risk at each move.

Understanding the differences. Commercial insurers can be more flexible and quicker to change how they pay and deliver care, and many have invested in value-based models, digital health, and social needs. Public insurers such as Medicare and Medicaid serve older adults, people with low incomes, and people with disabilities, focus strongly on access and equity, and are accountable to the public and to elected officials.

Promoting collaboration and learning. Commercial insurers can learn from public programs' experience with vulnerable populations and access. Public programs can learn from commercial innovations in delivery and financing.

Harmonizing policy. Aligning quality measures and reporting across payers reduces the burden on providers and keeps attention on the same goals. The Core Quality Measures Collaborative, convened by America's Health Insurance Plans with the Centers for Medicare and Medicaid Services and others, released seven aligned core measure sets in 2016 for this purpose (Centers for Medicare and Medicaid Services, 2016).

Public-private partnerships. Partnerships can combine the strengths of both sectors in shared digital platforms, joint work on social needs, and investment in community health resources. National data exchange is one working example. The Trusted Exchange Framework and Common Agreement (TEFCA), a federal framework operated through private networks, went live on December 12, 2023; by the end of 2025, about 464 million documents had been exchanged through 11 designated networks (Assistant Secretary for Technology Policy, 2026).

Policy and regulation. Legislators and regulators can encourage alignment by rewarding collaboration, removing barriers, and enabling data sharing that protects privacy. Bridging the sectors is difficult because they serve different populations, are funded differently, and answer to different authorities. Shared goals, aligned measures, and common data infrastructure make it possible.

Future Directions

Ten developments will shape continuity over the coming decade. I listed most of them in 2022; the evidence since then has sharpened several.

• Patient-centered care. Payment and delivery continue to shift from the volume of services toward the patient's experience and outcomes, which aligns directly with continuity.
• Technology. EHRs, telehealth, and digital tools improve communication and coordination. AI is now the most consequential of these tools, and the least governed at the point of hand-off.
• Social determinants of health. Continuity initiatives will need to address housing, education, and food alongside medical care.
• Health literacy. As patients take a more active role, they need the skills to understand information and make decisions, and AI-drafted patient instructions must be written for the reader who will use them.
• Policy and regulation. Rules that support value-based care, collaboration across payers, and the safe use of technology will shape what providers can do.
• Team-based care. Physicians, specialists, nurses, pharmacists, and social workers will increasingly share responsibility for a patient's medical, social, economic, and behavioral needs.
• A broader role for primary care. Primary care providers are often the first point of contact and are well placed to coordinate care, manage chronic conditions, and provide prevention, though the projected shortage will test that role.
• Innovative payment. Risk-sharing arrangements, outcome-based payments, and incentives for addressing social needs will continue to develop.
• Interoperability and data sharing. Technical exchange and the policies that govern it, while protecting privacy, will remain central.
• Equity and access. Continuity must reach every patient regardless of income, race, ethnicity, or location, through coverage, attention to disparities, and investment in underserved areas.

The Strongest Objection

The strongest objection is that continuity of care is a clinical and operational matter that belongs to management and the medical staff, not to the board. Directors cannot judge a discharge summary, and they should not try. I agree that they should not try. Nose in, fingers out. But the board does decide which AI systems the organization deploys, approves the budgets that pay for them, and is responsible for overseeing the risks that are central to the organization's mission. Delaware's Supreme Court has held that directors must make a good-faith effort to put in place a board-level system for monitoring compliance risks critical to a company's operations (Jones Day, 2019). A system that drafts the documents on which every subsequent provider relies is central to the mission of a health system. Once California requires that a licensed provider exercise independent judgment where AI informs care, and an accreditor assesses how AI safety is monitored, the board will be asked what it knew. Asking management for a report on continuity at transitions is a modest way to know it.

What the Board Should Expect

Directors do not need to evaluate documentation tools. They should expect management to report how AI affects transitions of care, and they should see that report on a schedule. The report should show which documents are AI-drafted and who reviews them before they leave the organization. It should show how errors are detected and corrected across settings. And it should show whether readmissions, follow-up completion, and patient-reported continuity have changed since deployment. I would encourage boards to treat continuity as an outcome of AI deployment, not only as a quality topic. The same tool that saves an hour of documentation can, if unreviewed, carry an error across three settings.

For the Board

• Which notes, discharge summaries, referral letters, and instructions are drafted by AI, and who reviews each before it is used?
• How are errors in AI-drafted documents detected, reported, and corrected across settings?
• How will we show that a licensed provider exercised independent judgment where AI output informed care?
• What has happened to readmissions, follow-up completion, and patient-reported continuity since these tools were deployed?
• Do any AI models that predict cost or risk affect which patients receive care coordination, and how is that reviewed?
• Which of our value-based contracts reward continuity, and how is performance under them reported to the board?

Find board questions on clinical AI and care transitions in the Avancer AI Governance Navigator.

References

American Academy of Sleep Medicine. (2026, February 6). Congress ends partial government shutdown and extends telehealth flexibilities and the work GPCI floor through 2027. https://aasm.org/congress-ends-partial-government-shutdown-and-extends-telehealth-flexibilities-and-the-work-gpci-floor-through-2027/

Assistant Secretary for Technology Policy. (2026, February 11). Data liquidity, affordability, and access: The history and growth of TEFCA. U.S. Department of Health and Human Services. Retrieved October 21, 2026, from https://healthit.gov/resources/data-liquidity-affordability-and-access-the-history-growth-of-tefca/

Bazemore, A., Petterson, S., Peterson, L. E., Bruno, R., Chung, Y., & Phillips, R. L., Jr. (2018). Higher primary care physician continuity is associated with lower costs and hospitalizations. Annals of Family Medicine, 16(6), 492–497. https://doi.org/10.1370/afm.2308

Bunch, L. N., & Ketema, H. (2025, September 9). Health insurance coverage in the United States: 2024 (Report P60-288). U.S. Census Bureau. https://www.census.gov/library/publications/2025/demo/p60-288.html

Centers for Medicare and Medicaid Services. (2016, February 16). Core Quality Measures Collaborative release [Fact sheet]. https://www.cms.gov/newsroom/fact-sheets/core-quality-measures-collaborative-release

Centers for Medicare and Medicaid Services. (2025, September 29). Medicare Shared Savings Program accountable care organizations: Updated performance year 2024 financial and quality results [Fact sheet]. https://www.cms.gov/files/document/fact-sheet-ssp-py24-financial-and-quality-results.pdf

Coleman, E. A., Parry, C., Chalmers, S., & Min, S. (2006). The care transitions intervention: Results of a randomized controlled trial. Archives of Internal Medicine, 166(17), 1822–1828. https://caretransitions.org/files/wp-content/uploads/2015/06/39_the-care-transitions-intervention-results-of-a-randomized-controlled-trial.pdf

Guerra, A. (2026, June 1). Joint Commission launches AI certification as adoption outpaces oversight. Health System CIO. https://healthsystemcio.com/2026/06/01/joint-commission-ai-certification/

Haggerty, J. L., Reid, R. J., Freeman, G. K., Starfield, B. H., Adair, C. E., & McKendry, R. (2003). Continuity of care: A multidisciplinary review. BMJ, 327(7425), 1219–1221. https://doi.org/10.1136/bmj.327.7425.1219

Health Care Payment Learning and Action Network. (2025). 2024 measurement effort [Infographic]. https://hcp-lan.org/wp-content/uploads/2025/10/2024-APM-Infographic.pdf

Jencks, S. F., Williams, M. V., & Coleman, E. A. (2009). Rehospitalizations among patients in the Medicare fee-for-service program. New England Journal of Medicine, 360(14), 1418–1428. https://doi.org/10.1056/NEJMsa0803563

Jones Day. (2019, August). Delaware Supreme Court reinforces directors' oversight obligations on mission-critical subjects. https://www.jonesday.com/en/insights/2019/08/delaware-court-reinforces-directors-oversight

Kutner, M., Greenberg, E., Jin, Y., & Paulsen, C. (2006). The health literacy of America's adults: Results from the 2003 National Assessment of Adult Literacy (NCES 2006-483). National Center for Education Statistics. https://nces.ed.gov/use-work/resource-library/report/statistical-analysis-report/health-literacy-americas-adults-results-2003-national-assessment-adult-literacy

Montano, M., & Tran, S. (2026, October 8). Newsom approves 3 health care AI bills, vetoes 2. California Health Care Foundation. https://chcf.org/blog/2026/10/08/newsom-approves-3-health-care-ai-bills-vetoes-2

Office of Disease Prevention and Health Promotion. (n.d.). Social determinants of health. Healthy People 2030. U.S. Department of Health and Human Services. Retrieved October 21, 2026, from https://odphp.health.gov/healthypeople/priority-areas/social-determinants-health

Office of the National Coordinator for Health Information Technology. (2024, May 24). 70% of hospitals participate in healthcare interoperability. U.S. Department of Health and Human Services. https://healthit.gov/news/70-hospitals-participate-healthcare-interoperability/

Pereira Gray, D. J., Sidaway-Lee, K., White, E., Thorne, A., & Evans, P. H. (2018). Continuity of care with doctors: A matter of life and death? A systematic review of continuity of care and mortality. BMJ Open, 8(6), Article e021161. https://bmjopen.bmj.com/content/8/6/e021161

Pham, H. H., Schrag, D., O'Malley, A. S., Wu, B., & Bach, P. B. (2007). Care patterns in Medicare and their implications for pay for performance. New England Journal of Medicine, 356(11), 1130–1139. https://doi.org/10.1056/NEJMsa063979

Sinaiko, A. D., Landrum, M. B., Meyers, D. J., Alidina, S., Maeng, D. D., Friedberg, M. W., Kern, L. M., Edwards, A. M., Flieger, S. P., Houck, P. R., Peele, P., Reid, R. J., McGraves-Lloyd, K., Finison, K., & Rosenthal, M. B. (2017). Synthesis of research on patient-centered medical homes brings systematic differences into relief. Health Affairs, 36(3), 500–508. https://doi.org/10.1377/hlthaff.2016.1235

Sobel, L., Dawson, L., & Salganicoff, A. (2025). Kennedy v. Braidwood: The Supreme Court upheld ACA preventive services but that's not the end of the story. KFF. https://www.kff.org/policy-watch/kennedy-v-braidwood-the-supreme-court-upheld-aca-preventive-services-but-thats-not-the-end-of-the-story/

Southwick, R. (2024, March 22). Physician shortage could reach 86,000, AAMC projects. Chief Healthcare Executive. https://chiefhealthcareexecutive.com/view/physician-shortage-could-reach-86-000-aamc-projects

Tierney, A. A., Gayre, G., Hoberman, B., Mattern, B., Ballesca, M., Kipnis, P., Liu, V., & Lee, K. (2024). Ambient artificial intelligence scribes to alleviate the burden of clinical documentation. NEJM Catalyst Innovations in Care Delivery, 5(3). https://doi.org/10.1056/CAT.23.0404

Williams, C. Y. K., Raghu Subramanian, C., Ali, S. S., Apolinario, M., Askin, E., Barish, P., Cheng, M., Deardorff, W. J., Donthi, N., Ganeshan, S., Huang, O., Kantor, M. A., Lai, A. R., Manchanda, A., Moore, K. A., Muniyappa, A. N., Nair, G., Patel, P. P., Santhosh, L., . . . Rosner, B. I. (2025). Physician- and large language model-generated hospital discharge summaries. JAMA Internal Medicine, 185(7), 818–825. https://doi.org/10.1001/jamainternmed.2025.0821

Yap, G., Xiao, D., Hu, J., Sanday, J., & Beatty, C. (2025, October 21). 2025: The state of AI in healthcare. Menlo Ventures. https://menlovc.com/perspective/2025-the-state-of-ai-in-healthcare/