The Fund supports networks of state health policy decision makers to help identify, inspire, and inform policy leaders.
The Milbank Memorial Fund supports two state leadership programs for legislative and executive branch state government officials committed to improving population health.
The Fund identifies and shares policy ideas and analysis to advance state health leadership, strong primary care, and sustainable health care costs.
Keep up with news and updates from the Milbank Memorial Fund. And read the latest blogs from our thought leaders, including Fund President Christopher F. Koller.
The Fund publishes The Milbank Quarterly, as well as reports, issues briefs, and case studies on topics important to health policy leaders.
The Milbank Memorial Fund is is a foundation that works to improve population health and health equity.
February 4, 2025
Quarterly Opinion
Patricia Hong
Ezekiel J. Emanuel
Jan 26, 2023
Back to The Milbank Quarterly Opinion
The United States is experiencing a mental health crisis. Over the past decade, there has been a troubling increase in depression, overdoses, and suicides. In 2023, the percent of adults ever diagnosed with depression reached 29% (59 million adults), a 10% rise since 2015. Even with an over 14% decrease in overdose fatalities in the past year, overdose deaths have nearly doubled since 2015. Between 2011 and 2022, the suicide rate increased by 16%. An estimated 50 million Americans are struggling with mental health issues, including over 15 million with serious mental illness, such as schizophrenia and bipolar disorder.
This crisis is exacerbated by a mental health workforce shortage. Even with the most optimistic estimates of 56,500 psychiatrists, 181,000 psychologists, and 115,000 social workers specializing in mental health, there are still only 356,500 mental health clinicians. For a population of over 330 million, that is roughly 1 per 1,000 people, or 1 per 140 individuals with mental health issues. Geographic disparities worsen the situation. Over half of the US population lives in a mental health workforce shortage area.1 This disparity is even more stark in rural areas, where more than 50% of counties lack a single psychiatrist. In 2021, 69% of rural counties lacked a psychiatric mental health nurse practitioner as compared to 31% of urban counties.
Consequently, nearly half of all adults with mental illness do not receive care. Those who do receive care wait an average of 48 days. This gap in mental health care contributed to a $478 billion cost to the economy in 2024, and is projected to reach $1.3 trillion by 2040 due to lost productivity and premature death.
Current solutions to address the workforce gap are insufficient. The US Department of Health and Human Services (DHHS) is increasing the number of psychiatry residency positions by a mere 100 in 2026. However, it takes eight post-college years to train a psychiatrist. By 2036, the shortage nationwide is projected to grow to over 51,000 psychiatrists and hundreds of thousands of other mental health professionals. While a step toward progress, the increase in residency positions won’t meet the demand soon enough. Similarly, addressing low reimbursement rates for mental health services may attract more individuals to the field, but increasing payment parity may not be enough to overcome the workforce challenges.
Other potential solutions focus on addressing the maldistribution of professionals. Standardizing and making permanent physician licensure across state lines would facilitate telemedicine, which is often limited by state licensure requirements and only permitted through a temporary federal exemption. Integrating behavioral health into primary care—whether through routine mental health services or geographic co-location—can serve as a force multiplier, enhancing care delivery.
While these strategies should become the norm to better reach underserved populations, they alone are not sufficient to combat the workforce crisis.
Fortunately, artificial intelligence (AI) technology has opened a new avenue of opportunity to augment the mental health workforce and increase access to care. AI-enabled tools can ease the administrative burden on clinicians, allowing them to spend more time seeing patients, and reducing burnout in the existing workforce. Moreover, AI chatbots have the potential to deliver scalable mental health support. Now is the time to adopt technology and get creative in how we approach the mental health workforce shortage.
Perhaps most promising at this time, AI can help alleviate the claims burden on overworked clinicians. Claims submission hassles are widely recognized as a top reason for why psychiatrists have the lowest rates of insurance acceptance among medical specialties. Psychiatrists spend an average of 16 hours per week on administrative tasks, including insurance claims and EHR documentation. Even minor mistakes in the billing process can lead to claim rejections, requiring additional time to correct and resubmit paperwork. These time-consuming responsibilities ultimately divert attention from patient care.
AI tools can automate many of these tasks, including filling out forms and flagging inconsistencies or errors. This can reduce claim rejections, eliminate hours of back-and-forth between clinicians and insurance companies, and speed up reimbursement.
By streamlining medical documentation, AI can also help ease the burden on mental health clinicians. The demands of electronic health records (EHR) often contribute to burnout, as clinicians are forced to juggle patient care with documentation-related tasks. Burnout is a serious issue contributing to both medical errors and a concerning rate of professionals leaving the field altogether. The American Psychiatric Association reports that 2 out of 5 psychiatrists experience burnout. In a study of over 450 physicians, those meeting criteria for burnout were more than twice as likely to leave their institution within two years compared to those who did not. From 2020 to 2022, psychiatrists saw an 860% increase in monthly electronic health record (EHR) message volume compared to 2018-2020, a surge driven by the Covid-19 pandemic. When patients don’t receive timely responses, dissatisfaction grows, further fueling clinician burnout. In fact, many clinicians now spend as much or more time managing messages and documenting records as they do with patients.
AI-powered tools can automate significant portions of progress notes, treatment plans, and message responses, reducing the time clinicians spend on EHR tasks. These systems can also flag inconsistencies in lab results or gaps in care, prompting proactive attention to potential issues. In addition, AI-driven scheduling tools can match available appointment slots with patients’ preferences and availability, reducing the need for constant coordination between patients and clinicians. This savings in time helps prevent burnout and improves efficiency by allowing clinicians to focus more on their core mission—providing high-quality care to patients and addressing the growing demand for care.
Reducing administrative work and documentation will give mental health clinicians more time with patients and increase satisfaction with their day-to-day work. But implementing these changes still does not solve the wider problems of access and affordability. AI chatbots have the potential to help fill these gaps, providing scalable, affordable care to those in need.
In 1966, computer scientist Joseph Weizenbaum developed ELIZA, a program that simulated Carl Rogers’ concept of reflective listening in Rogerian therapy. ELIZA used simple pattern-matching to mirror users’ emotions with open-ended, non-directive responses. To his surprise, users found the chatbot convincing and even meaningful. Ironically, psychiatrists wrote in an academic journal that year that a chatbot could help scale the therapy workforce. ELIZA displayed the potential of AI to replicate human-like interactions and the therapeutic relationships users could form with the program.
Artificial intelligence has come a long way since 1966. We have moved from pattern-matching and rules-based methods to highly fluent chatbots trained through large language models. These chatbots have shown conversational capability across an enormous range of human-like expression. With these developments, rigorously tested AI chatbots are an underappreciated solution to the shortage of therapists and may help address key challenges in mental health care: access, affordability, and the delivery of consistent, empathetic care. We are quickly iterating toward AI that can effectively improve mental health care access and outcomes. These AI-driven tools are unlikely to replace humans but can significantly complement the existing mental health infrastructure. AI chatbots can augment human therapists by providing patients with some therapy before an in-person session, supporting patients between sessions, acting as delegated agents for therapists, providing recommendations, or helping review transcripts.
A major benefit of AI chatbots is access. They can dramatically increase the availability of therapy support, particularly when patients need urgent help or are unable to access traditional services due to schedules, waitlists, or geographical barriers. Without chatbots, the alternative for many isn’t simply a delayed appointment with a highly-qualified psychiatrist or psychologist—it’s no care at all. Chatbots are available 24/7, offering immediate support in times of emotional vulnerability or crisis, ensuring that help is a few clicks away.
Another advantage of AI chatbots is their affordability. At an average cost of $100-200 per session, traditional therapy can be prohibitively expensive and out of reach for many individuals. Even with insurance, the cost of therapy sessions quickly adds up. Nearly two-thirds of psychologists report that they rarely serve individuals with low incomes. AI chatbots can provide support at a fraction of the cost, at around $20 monthly or even at no cost, allowing individuals to seek care more easily.
Finally, chatbots offer the potential for delivering standardized, high-quality care. Current Centers for Medicare and Medicaid Services quality measures for psychotherapy focus primarily on visit numbers rather than assessing outcomes such as symptom severity or relapse, fidelity to evidence-based methods, or the appropriateness of care. Chatbots can be trained to follow evidence-based methodologies and be measured against more meaningful measures of quality. Unlike with human therapists, chatbots can also be monitored continuously for quality through transcript analysis and review.
With guardrails in place, chatbots can provide individuals with mental health support and talk therapy as a first line of care that supplements traditional therapy and the wider mental health care system. While more rigorous evidence is needed regarding the safety and efficacy of chatbots, studies are demonstrating that chatbots can improve depression symptoms, deliver cognitive behavioral therapy, and build therapeutic relationships. AI-driven tools can guide individuals through difficult emotions, offer coping strategies, assist with reframing thoughts, and provide resources for further support. In doing so, they bridge the gap in care and ease the burden on the mental health workforce, allowing human therapists to focus on more complex cases requiring greater professional intervention. Additionally, users increasingly report positive views of the chatbot support provided. Some individuals may also find it easier to disclose more fully to chatbots compared to a medical professional, with whom they may worry about confidentiality or stigma. At a time when demand for mental health services is higher than ever, AI chatbots offer a lifeline by helping to reduce strain on the system and potentially improving overall access to care for millions of people.
Critics of AI have raised concerns about the use of chatbots and other AI tools, citing issues such as increased prior authorization denials, biased decision-making, and harmful content. While these concerns are valid and oversight, along with continuous refinement, is essential for ensuring ethical use, AI chatbots are constantly and quickly improving. In the current mental health crisis, there is no other solution that can meet the growing demand. For many people in the US, the alternatives are delayed in-person care or no care at all. Lack of care is far from benign, and AI chatbots are likely to be an improvement.
AI has the potential to play a role in addressing some of the challenges in mental health care, particularly in areas where there are gaps in access and quality. A shortage of skilled clinicians and heavy burdens on existing clinicians have created barriers to care and access disparities. Policy changes and workforce expansion are worthy long-term aspirations, but people need care now. AI could help by easing administrative burdens and improving access to quality care, providing relief to a system under intense strain.
Patricia Hong holds an ScM from Brown University and a BA from Wake Forest University.
Ezekiel J. Emanuel is vice provost for global initiatives and chair of the Department of Medical Ethics and Health Policy at the University of Pennsylvania. From January 2009 to January 2011, he served as special advisor for health policy to the director of the White House Office of Management and Budget. Since 1997 he was chair of the Department of Bioethics at The Clinical Center of the National Institutes of Health and a breast oncologist. Emanuel received his MD from Harvard Medical School and his PhD in political philosophy from Harvard University. After completing his internship and residency in internal medicine at Boston’s Beth Israel Hospital and his oncology fellowship at the Dana-Farber Cancer Institute, he joined the faculty at the Dana-Farber Cancer Institute. He has since been a visiting professor at the University of Pittsburgh School of Medicine, UCLA, and New York University Law School, the Brin Professor at Johns Hopkins Medical School, and the Kovitz Professor at Stanford Medical School. Emanuel has written and edited 9 books and over 200 scientific articles. He is currently a columnist for the New York Times.
