This article first appeared in NEJM Catalyst. It is co-authored with  Brian Zhao and Erin Duffy, Ph.D., M.P.H.

 Vol. 7 No. 8 | August 2026 

DOI: 10.1056/CAT.25.0509 

This article critically examines the potential impacts of artificial intelligence (AI) on prices, total spending, and the rate of spending growth in the areas of prescription drug innovation and expanded patient access to care — including developments in remote patient monitoring, chronic care management, direct- to- consumer health care, clinical decision support, and nonclinical administrative labor. Based on observed industry practices, economics, and public policy, the article presents a thought exercise on the most likely effects for patients, providers, payers, and the broader health care system. The authors argue that under the still- dominant fee- for- service payment model, as well as the highly consolidated hospital and insurance markets, AI is more likely to increase total costs and spending growth in the short to medium term rather than slow them, even as it delivers substantial access and clinical quality improvements for patients. The cost- bending potential of AI varies distinctly by fee- for- service versus value- based payment. Regulators need to introduce policy and reimbursement levers for AI to slow cost growth. The authors conclude that, without significant changes in the health care system’s financial incentives and market structures, AI will not slow cost growth. 

---

 The Promises and Limitations of Artificial Intelligence for Health Care Costs 

There has been a surge of optimism in recent years around the potential for artificial intelligence (AI) to reduce U.S. health care spending and slow its rate of growth. A widely cited 2023 analysis estimated that wider adoption of AI across the U.S. health care industry could drive 5%–10% savings, or US$200–US$300 billion annually in total spending reductions, through reductions in administrative costs and gains in clinical productivity.1 

Policy makers, payers, and patients feel an urgency for AI- enabled productivity and cost control. U.S. national health expenditures increased from US$3 trillion in 2014 to nearly US$4.9 trillion in 2023 — an increase of more than 60% — averaging roughly 5% annually, well above the rate of general inflation over the same period.2 After several years of lower growth rates following the Patient Protection and Affordable Care Act’s enactment in 2010, per capita spending climbed from approximately US$9500 to more than US$14,000 over the next 10 years.3 Health care now consumes nearly 18% of U.S. gross domestic product, by far the highest share among comparable high- income nations, with the gap driven primarily by higher prices for hospital services and prescription drugs rather than by greater utilization of services.4 

Industry analysts and consulting firms have largely expressed optimism over AI’s cost- control potential, identifying generative AI — which uses machine learning models trained on vast datasets to create new original content — as a potentially transformative tool for increasing clinical productivity and reducing back- office costs. Those costs account for approximately 25% of spending.5 The federal government is eager to deploy AI as a key lever in slowing the growth rate of health care costs, as shown by the U.S. Department of Health and Human Services’ 2025 request for information.6 

AI is being adopted by all types of health care organizations, including providers, insurers, and drugmakers, for a wide range of tasks. It seems logical in theory that AI, by enabling tasks to be performed faster, more accurately, and with less human labor, should translate to lower costs. But savings from productivity gains in U.S. health care have a long history of being absorbed by the health care industry rather than being passed through to patients or payers.7 

We are optimistic that AI developments will yield clinical value and some administrative efficiency. However, the impact of AI on total health care spending, spending growth, and out- of-pocket affordability for patients is less certain. 

We are optimistic that artificial intelligence developments will yield clinical value and some administrative efficiency. However, the impact of artificial intelligence on total health care spending, spending growth, and out- of- pocket affordability for patients is less certain.” 

Whether AI serves as a deflationary or inflationary force in health care will depend on the payment model. Under fee- for- service payment, which reimburses providers per unit of service rendered, including visits, tests, procedures, and prescriptions, and which remains the predominant payment system, the financial incentive is to deliver more billable services. This increased spending is largely captured by provider organizations. In contrast, under value- based payment models, such as capitation, shared savings, bundled payments, and full- risk contracts, the financial incentives are inverted. Still, questions remain about where savings will accrue.

In this article, we present a conceptual analysis of how AI is likely to affect U.S. health care prices, total spending, and the rate of spending growth across three areas — prescription drug innovation, expanded access to care, and nonclinical administrative labor. We primarily focus on AI’s potential to slow the rate of spending growth — sometimes called bending the cost curve — rather than on producing absolute spending reductions, which are unlikely. We argue that this is highly contingent on payment models, market structures, and policy levers. 

The Impact of Accelerated Prescription Drug Development on Pharmaceutical Spending 

First, we consider the impact on pharmaceutical spending, which is already one of the fastest-growing areas of health care spending, growing by 13% to US$915 billion between 2024 and 2025. Prices paid in the United States are the highest in the world.8 

AI is widely seen as having the potential to accelerate prescription drug discovery and development by enabling better and faster target selection and reducing the labor and timelines required to bring new treatments to market.9 

If this plays out as anticipated, the likely near- term scenario is the coming to market of many new prescription drugs and a larger population of patients eligible for high- cost specialty therapies. New drugs tend to be more expensive when they enter the market, and they lead to increased overall pharmaceutical spending. Although prescription drugs generally deliver meaningful clinical benefits to patients, many do not generate enough offsetting savings from improvements in patients’ health to lower the total cost of care.10 

In addition, it can take many years of drug treatment to prevent or mitigate costly downstream complications of chronic conditions. Given the frequent movement of patients between health plans, the cost benefit of an effective treatment may not accrue to the payer that pays for the treatment. That complicates the incentives and the impact on the total cost of care. 

With these factors in mind, the most likely outcome of many new AI- developed prescription drugs is increased total spending, even as patients benefit clinically. 

On the other hand, there is a potential for cost savings. AI may reduce the costs of drug discovery and development by lessening the time and effort involved in matching drugs to appropriate patients for clinical trials.11 If that happens, some offsetting long- term savings are possible, particularly for chronic disease drugs, the value of which in preventing disease complications accrues over years, even if shorter- term spending rises. 

Artificial intelligence may reduce the costs of drug development by lessening the time and effort involved in matching drugs to appropriate patients for clinical trials. If that happens, some offsetting long- term savings are possible, particularly for chronic disease drugs.” 

The Impact of Increased Patient Access on Costs 

Historically, there have been persistent barriers for U.S. patients in accessing care, including long wait times for primary care appointments, even longer delays for specialty referrals, geographic disparities in clinician availability, and significant out- of- pocket costs that deter visits even among insured patients. 

AI is now being deployed across a wide range of applications that promise to expand access. These include: 

• AI scribes such as Abridge, Suki, and Ambience that increase clinician capacity; 
• AI- enabled remote patient monitoring (RPM) and chronic care management (CCM) that extend clinical oversight beyond the four walls of the health system; 
• A growing set of direct- to- consumer (DTC) “AI doctor” services, such as Doctronic and Counsel Health, which offer chat- based medical interactions on demand; and 
• AI clinical decision support (CDS) tools that improve the speed and accuracy of clinician actions. 

The companies offering these tools and services range from legacy information technology companies to venture- backed start- ups. These services are being deployed by hospitals and health systems, integrated payers, and newer entrants that sometimes operate outside of the traditional insurance reimbursement channels. 

While AI is likely to expand access to care in ways that generate benefits to patients and reduce disparities in health equity, we believe many of these tools are likely to increase total spending on clinical care. That inflationary impact is largely related to the continued dominance of fee-for- service payment arrangements. But the cost impact also depends on which types of care are expanded and for which patients. 

AI scribe services illustrate the divergent ways AI could affect costs. The economic value proposition of AI scribes for physician groups and health systems is to free physicians from hands-on medical charting, enabling them to see more patients in a day.12 Under a fee- for- service model, more visits mean more professional fees, diagnostic tests, referrals, and prescriptions. Greater access is valuable for patients, and primary and preventive care can reduce costs over time.13 However, many physician visits do not result in lower subsequent costs, especially when those visits are with lower- risk patients.14 

As a result, the net effect of AI scribes enabling more physician visits is an increase in total spending and potentially in the rate of cost growth, particularly in fee- for- service payment models where providers have no financial incentive to lower costs. 

In contrast, under value- based payment models, using AI scribes to free up more physician time can enable physicians to deliver higher- value services. Such services include longer visits with complex patients and targeted follow- ups with patients between visits and for those who are overdue for preventive care. In other words, if the freed- up capacity is directed toward actions that can bend the cost curve rather than drive up encounter volumes and billing, longer- term savings are possible through the use of AI tools.  

If physician capacity freed up by artificial intelligence scribe services is directed toward actions that can bend the cost curve rather than drive up encounter volumes and billing, longer- term savings are possible through the use of artificial intelligence tools.” 

The Potential for Artificial Intelligence–Enabled Remote Patient Monitoring and Chronic Care Management to Cut Costs 

Prior to AI, it was logistically challenging and time- consuming for clinicians to deliver RPM and CCM, requiring them to observe and respond to patient data feeds. Thus, offering these services was not very profitable or attractive to providers, even though it could result in better access to and quality of care for patients. 

AI has dramatically reduced the cost and effort needed to deliver RPM and CCM by sharply reducing the amount of clinician time required to analyze and monitor data and patient messages. Standing orders can enable AI assistants to automate responses and interventions when biometrics change. 

However, RPM and CCM only lower costs when offered to select high- risk patients. Thus, broad deployment of AI- assisted RPM for more patients is likely to result in higher expenditures related to the service itself and increased ordering of potentially unnecessary interventions, particularly under fee- for- service payment models.15 

This type of cost concern may have been a reason for UnitedHealthcare’s now- paused decision, originally set to take effect on January 1, 2026, to narrow reimbursement for many RPM services in 2026. In its original decision, the insurer was going to cover RPM only for hypertensive disorders of pregnancy and chronic heart failure, citing a lack of evidence for other conditions.16 

With proper risk stratification and patient selection and engagement, AI- enabled RPM and CCM have the potential to prevent acute medical events and reduce costs long term. These tools need to be applied properly and under the right payment models. 

Will Direct- to- Consumer Artificial Intelligence Care Replace or Escalate Encounters with Human Clinicians? 

DTC health care is based on the idea of getting patients more involved in initiating their own care, often in ways that bypass the traditional clinician- mediated care model. Early examples have largely centered around discretionary lab testing, such as 23andMe for genetic testing and Function Health for comprehensive biomarker screening. 

More recently, there have been major investments in DTC AI applications that expand access to a broader range of clinical services. These include: 

• AI doctor platforms offering on- demand chat- based diagnosis and treatment recommendations 
• Consumer- facing medical data consolidators integrating personal health information across various sources, such as ChatGPT Health 
• Targeted solutions, such as AI- enabled weight loss and glucagon- like peptide 1 prescribing solutions, mental and behavioral health counseling, and dermatology diagnosis and triage tools 

These services are often cash pay for consumers, though several companies are now seeking integration with health plan reimbursement. 

Currently, most of the AI doctor tools that offer consumers medical advice and care fall back on a human clinician. In other words, patients communicate with the AI tool, often delivered as an online chat- based experience, about their medical issues, while escalations are handled by a human clinician. 

However, many of these companies plan to deliver more AI doctor care completely autonomously, moving from generic medical advice and triage to drug prescribing, laboratory test ordering, provision of more complex diagnoses, and care planning. Many of these new companies and platforms are advertising significantly lower prices for AI doctor services than the prices for virtual or in- person care with a human clinician. 

Demand for these tools is bolstered by the long wait times and other access barriers in the traditional clinician- directed care model. While the AI interactions are generally cash pay and are not reimbursable by health plans currently, new payment models are being contemplated. Unlike human clinicians, who are limited by appointment schedules and work hours, the capacity of AI doctors is limited only by computing capacity. 

The concept of creating nearly unlimited access at lower prices per encounter presents an interesting thought exercise in terms of the impact on total cost of care and spending growth. AI DTC care that replaces utilization of more expensive care by human clinicians could have a deflationary spending impact. However, if it leads to substantially more escalations to human specialists, it would produce higher total spending. 

Artificial intelligence direct- to- consumer care that replaces utilization of more expensive care by human clinicians could have a deflationary spending impact. However, if it leads to substantially more escalations to human specialists, it would produce higher total spending.” 

The evidence suggests that incidental clinical findings, which surface in consumers’ interactions with DTC AI tools, frequently trigger cascades of additional testing and care, which can be costly and often wasteful.17 With DTC AI tools, the cascades initially will be triggered by the consumer themselves, such as ordering DTC testing or going to their primary care physician with concerns. However, if and when more advanced AI doctors become able to prescribe and order tests, the inflationary feedback loop would become stronger. 

Meanwhile, oversight mechanisms for AI- provided care remain nascent. Insurers have begun scrutinizing their coverage for AI- delivered services, creating overall uncertainty about where this market will go and its ultimate cost impact. 

Who Will Pocket the Savings from Clinical Decision Support Tools? 

Clinical decision support tools powered by AI are rapidly expanding across clinical applications. These include: 

• Diagnostic chatbots integrated into clinician workflows 
• Prescribing support tools 
• Risk- stratification models for targeting preventive care 
• Imaging interpretation for radiology workflows and cancer screening 
• Other medical prediction models 

These tools are being deployed primarily by hospitals and health systems and insurers, with the goal of improving the quality, accuracy, consistency, and timeliness of clinical decisions. The economic impact of these tools depends on the behavior that is being changed and in which direction, and the payment model. 

On one hand, there could be a cost- reduction impact. If AI CDS tools reduce diagnostic errors and discourage care that is not evidence- based, that could decrease adverse events that drive up spending. AI- assisted early detection of cancers and other serious conditions could also reduce downstream treatment costs by identifying disease at a more manageable and less costly stage. 

Alternatively, AI CDS tools could encourage clinicians to do more comprehensive workups, surface incidental findings that trigger additional testing, and prescribe more expensive prescription drugs that have only marginal benefit. That would drive up clinical care costs without clearly improving outcomes. 

AI diagnostic and imaging interpretation is one of the more mature CDS applications, with strong early adoption in radiology and pathology. This AI application highlights the potential differences in cost outcomes. These tools have sensitivity and accuracy that now equals or exceeds the abilities of human specialists. Thus, there is a strong case for using these tools to extend specialist- level diagnostic capability into primary care settings, rural hospitals, and underserved populations. 

However, the payment environment in which these tools are applied, fee- for- service or value-based payment, will substantially determine their cost impacts, because financial incentives shape the recommendations that clinicians and their organizations act on. 

Under fee- for- service models, the financial logic for providers favors action on any new findings, such as follow- up imaging and testing, specialist referrals, and procedures, which are all billable services. Faster and cheaper interpretation or diagnosis lowers the threshold for ordering imaging and testing. 

Where the AI tools can fully replace specialist labor, the savings potential is substantial. But in concentrated, noncompetitive health care markets, the likely scenario is that cost savings from productivity gains will be absorbed by the imaging facilities and health systems as margin rather than translating to lower prices for payers and patients. 

In value- based payment models, AI diagnostic and imaging interpretation tools may lead to entirely different results. Without the financial incentive to deliver downstream services beyond what the clinical guidelines require, the same diagnostic and imaging findings are more likely to be monitored, potentially supported by new low- cost AI RPM tools. This could produce savings that are consistent with clinical guidelines. The cost savings through productivity gains could then be allocated to managing existing patients more effectively. 

Artificial Intelligence–Generated Administrative Savings May Not Go to Consumers 

Administrative costs account for approximately 15%–25% of total U.S. health care spending,18 a far larger share than in other wealthy nations. High administrative costs have long been identified as a target for potential savings, but little or no progress has been made in reducing them. 

AI applications are being adopted across a wide range of administrative tasks to reduce labor and other costs. These tasks include: 

• Patient scheduling, intake, and communications 
• Revenue cycle tasks, including coding, documentation review, billing, claims, and prior authorization submission 
• Payer- side tasks, including claims adjudication, payer prior authorization, and utilization management 

Almost all major payers and health systems have put administrative automation and AI among their top- priority investment areas. For example, UnitedHealth Group has announced US$1.5 billion in investments in AI initiatives across the company.19 

If these AI automation efforts are successful, they will result in lower administrative labor costs and higher health care industry profit margins. In an industry generally characterized by low margins and low labor productivity, this could generate substantial enterprise value. 

But it is uncertain whether or not market forces will result in any of these administrative cost savings being passed on to consumers in the form of lower prices. Productivity gains theoretically could moderate prices and spending growth over time, though competitive pressures to share savings with consumers are limited in concentrated, noncompetitive hospital markets20 and in the equally, if not more, concentrated insurance markets.21 

Competitive pressures to share savings with consumers are limited in concentrated, noncompetitive hospital markets and in the equally, if not more, concentrated insurance markets.” 

Despite recent transparency rules, pricing remains opaque across most segments of U.S. health care, and providers have even used the transparency rules to negotiate higher rates.22 Outside DTC cash- pay markets and generic drugs, where consumers shop on price, or meaningful competition has emerged, it is rare for prices to fall in the U.S. health care system. 

Without different market structures and economic incentives to compel health care organizations to pass savings to consumers, it is unlikely that lower administration costs produced by AI tools will lead to lower prices, lower spending, or reduced cost growth. 

Artificial Intelligence–Driven Prior Authorization Could Reduce Costs if Used Responsibly 

Among AI administrative applications, prior authorization is unique in its potential for cost savings and has become a focal point for AI innovation in health care. 

Prior authorization is a source of significant administration friction between payers and providers. It consumes major time and effort on both sides, with providers submitting and appealing claims and payers adjudicating the prior authorization requests. 

AI is now being deployed on both sides to automate the prior authorization process and, in theory, reduce processing times, improve decision consistency, and reduce the labor costs involved. 

Beyond the administrative savings, improving payers’ ability to deny coverage for low- value services could reduce total spending. But this cost- saving potential must be balanced against concerns around responsible use. AI tools that systematically deny appropriate care to lower insurers’ payments for medical services will face regulatory and litigation pressure, as seen in recent cases involving AI- driven coverage denials by Medicare Advantage plans.23 

Again, the impact of AI administration tools on overall costs and spending growth will depend on market structures and the way these solutions are deployed. 

Artificial Intelligence Will Not Produce Cost Savings without Payment and Market Reforms 

Despite the prevailing optimism about AI reducing U.S. health care spending, we believe AI’s effect on prices, total costs, and the rate of spending growth ultimately will depend less on the technology itself than on the payment models and financial incentives of the organizations adopting these AI tools. Another key factor is the competitiveness of the markets in which they operate. 

In the fee- for- service examples described above, AI is likely to create more — and more profitable — encounters. It may also increase referrals for diagnostic testing and procedures. In contrast, in a value- based payment setting, AI may be deployed to lower the number of encounters, select lower-cost drugs, and implement more cost- effective diagnostic approaches. 

AI is also useful for performing analytics, such as improving Medicare Advantage star rating metrics and coding intensity, as a result of both improved accuracy and more exhaustive capture of complexity and service levels. But, as coding intensity increases, so do federal payments to Medicare Advantage plans.24 

AI will certainly improve the profitability of the health care organizations that use AI, as long as they apply it competently. As most health care payments in the United States are fee- for- service, we think that it is more likely that AI will lead to more services, higher profit margins, higher total costs, and higher spending growth. 

For AI to produce cost savings, the most direct policy and reimbursement lever is the continued expansion of value- based payment models. Every patient served under capitation or full- risk contracting is a patient for whom AI- driven productivity gains are more likely to translate to reduced total spending. 

Policy makers are also exploring additional levers, such as specific reimbursement for AI applications and AI- delivered services, as in the Advancing Chronic Care with Effective, Scalable Solutions (ACCESS) model developed by the U.S. Centers for Medicare and Medicaid Services.25 The ACCESS model reimburses disease management applications for Medicare beneficiaries with or without human supervision as long as patients achieve clinical outcome goals. 

Although there are many potential AI benefits to patients in the form of more and better prescription drugs, greater access to care, and improved health outcomes, we have little reason to believe that AI will result in lower prices to consumers, reduced total spending, or lower medical inflation. 

To bend the cost curve, no single lever is sufficient. Financial incentives and market structure need to change, and that is one thing AI cannot do at this time. 

Brian Zhao. Investor, Venrock, Palo Alto, CA, USA 

Erin Duffy, Ph.D., M.P.H. Managing Director, Leonard D. Schaeffer Institute for Public Policy & Government Service, University of Southern California, Los Angeles, Los Angeles, CA, USA 

Bob Kocher can be contacted at bkocher@venrock.com. 

Disclosures: Bob Kocher is a partner at the venture capital firm Venrock. He invests in companies that use artificial intelligence and is a board observer at Suki, which sells Ambient note- writing software to health systems. Author disclosures are available at catalyst .nejm .org. 

References 

1. Sahni N, Stein G, Zemmel R, Cutler DM. The potential impact of artificial intelligence on healthcare spending. NBER Working Paper 30857. Cambridge, MA: National Bureau of Economic Research, 2023. DOI: 10 .3386 /w30857. 
2. U.S. Centers for Medicare & Medicaid Services. National health expenditure data: historical. Washington, D.C.: U.S. Department of Health and Human Services, 2024 (https://www .cms .gov /research -statistics -data -and -systems /statistics -trends -and -reports /nationalhealthexpenddata / nationalhealthexpendhistorical). 
3. Buntin MB, Graves J. How the ACA dented the cost curve. Health Aff 2020;39:403- 412. DOI: 10 .1377 / hlthaff .2019 .01478. 
4. Cutler DM. What is the US health spending problem? Health Aff (Millwood) 2018;37:493- 497. DOI: 10 .1377 /hlthaff .2017 .1626. 
5. Lamb J, Dahab A, Jones J. Generative AI in healthcare: adoption matures as agentic AI emerges. McKinsey & Company. April 16, 2026 (https://www .mckinsey .com /industries /healthcare /our -insights / generative -ai -in -healthcare -current -trends -and -future -outlook). 
6. U.S. Department of Health and Human Services. HHS announces request for information to harness artificial intelligence to deflate health care costs and make America healthy again. Press release. December 19, 2025 (https://www .hhs .gov /press -room /hhs -ai -rfi .html). 
7. Pearson E, Frakt A. Administrative costs and health information technology. JAMA 2018;320:537- 538. DOI: 10 .1001 /jama .2018 .10819. 
8. Organisation for Economic Co- operation and Development. Health at a glance 2025: OECD indicators. Paris: OECD Publishing, 2025. DOI: 10 .1787 /8f9e3f98 -en. 
9. Jayatunga MK, Xie W, Ruder L, Schulze U, Meier C. AI in small- molecule drug discovery: a coming wave? Nat Rev Drug Discov 2022;21:175- 176. DOI: 10 .1038 /d41573 -022 -00025 -1. 
10. Dunn AC, Fernando L, Liebman E. How much are medical innovations worth? A detailed analysis using cost- effectiveness studies. Suitland, MD: Bureau of Economic Analysis, 2024 (https://www .bea .gov / research /papers /2024 /how -much -are -medical -innovations -worth -detailed -analysis -using -cost). 
11. Hutson M. How AI is being used to accelerate clinical trials. Nature 2024;627:S2- S5. DOI: 10 .1038 / d41586 -024 -00753 -x. 
12. Olson KD, Meeker D, Troup M, et al. Use of ambient AI scribes to reduce administrative burden and professional burnout. JAMA Netw Open 2025;8:e2534976. DOI: 10 .1001 /jamanetworkopen .2025 .34976. 
13. Friedberg MW, Hussey PS, Schneider EC. Primary care: a critical review of the evidence on quality and costs of health care. Health Aff (Millwood) 2010;29:766- 772. DOI: 10 .1377 /hlthaff .2010 .0025. 
14. Mafi JN, Russell K, Bortz BA, Dachary M, Hazel WA, Fendrick AM. Low- cost, high- volume health services contribute the most to unnecessary health spending. Health Aff (Millwood) 2017;36:1701- 1704. DOI: 10 .1377 /hlthaff .2017 .0385. 
15. Mecklai K, Smith N, Stern AD, Kramer DB. Remote patient monitoring — overdue or overused? N Engl J Med 2021;384:1384- 1386. DOI: 10 .1056 /NEJMp2033275. 
16. Beavins E. UnitedHealthcare pauses effort to cut RPM coverage after stating the tech has “no evidence.” Fierce Healthcare. December 18, 2025 (www .fiercehealthcare .com /health -tech / unitedhealthcare -pauses -effort -cut -rpm -coverage -after -stating -tech -has -no -evidence). 
17. Ganguli I, Simpkin AL, Lupo C, et al. Cascades of care after incidental findings in a US national survey of physicians. JAMA Netw Open 2019;2:e1913325. DOI: 10 .1001 /jamanetworkopen .2019 .13325. 
18. Chernew M, Mintz H. Administrative expenses in the US health care system: why so high? JAMA 2021;326:1679- 1680. DOI: 10 .1001 /jama .2021 .17318. 
19. Minemyer P. UnitedHealth Group spotlights AI investments as part of operational turnaround. Fierce Healthcare. April 21, 2026 (https://www .fiercehealthcare .com /payers /unitedhealth -group -boosts -2026 -outlook -it -posts -q1 -earnings -revenue -beat). 
20. Godwin J, Levinson Z, Neuman T. One or two health systems controlled the entire market for inpatient hospital care in nearly half of metropolitan areas in 2024. KFF. March 27, 2026 (https://www .kff .org / health -costs /one -or -two -health -systems -controlled -the -entire -market -for -inpatient -hospital -care -in -nearly -half -of -metropolitan -areas /). 
21. U.S. Government Accountability Office. Private health insurance: market concentration generally increased from 2011 through 2022. November 14, 2024 (https://www .gao .gov /products /gao -25 -107194). 
22. Barnes K, Glied SA, Handel BR, Kim G. The impact of price transparency in outpatient provider markets. NBER Working Paper 32580. Cambridge, MA: National Bureau of Economic Research, 2024. DOI: 10 .3386 /w32580. 
23. Mello MM, Trotsyuk AA, Mahamadou AJD, Char D. The AI arms race in health insurance utilization review: promises of efficiency and risks of supercharged flaws. Health Aff (Millwood) 2026;45:6- 13. DOI: 10 .1377 /hlthaff .2025 .00897. 
24. Medicare Payment Advisory Commission. Estimating Medicare Advantage coding intensity and favorable selection. In: Report to the Congress: Medicare and the Health Care Delivery System. Washington, DC: MedPAC, March 2024: 415- 463. 
25. U.S. Centers for Medicare & Medicaid Services. ACCESS (Advancing Chronic Care with Effective, Scalable Solutions) model. April 28, 2026 (https://www .cms .gov /priorities /innovation /innovation -models /access).