top of page

The Collapse of Quality of Care in the Age of Innovation

  • Jun 2
  • 15 min read

Updated: 6 days ago

By Raji Mohanam, Editor, Women in the AI Vanguard Magazine

June 3, 2026



We are living through the most advanced moment in the history of American medicine, yet still cannot deliver the high quality of care patients expect and deserve. The United States spends more on healthcare than any other country in the world, yet the Commonwealth Fund consistently ranks our health outcomes last among comparable wealthy nations. The numbers behind this paradox are staggering: an estimated 795,000 Americans die or are permanently disabled each year from diagnostic errors alone, and women and patients of color are 20 to 30 percent more likely to be misdiagnosed than white men.


These numbers translate into the small and sometimes large failures patients and their familes witness unfolding in front of them when they are at their most vulnerable. The failures show up worst and first in our Emergency Room (ER) experiences. Patients wait an average of two hours and forty-four minutes just to be seen, and that is only the national mean; the wait in DC stretches past five hours.


Once finally seen, patients move through encounters that feel transactional at best and contemptuous at worst. Rushed, incomplete interactions with physicians who have not fully read patient charts are commonplace. Nurses, when they have time at all, often do not have answers to the questions a frightened patient needs to ask. Pain is dismissed. The elderly are talked past. Immigrant patients are talked over. Family members standing at the bedside are absorbed into the same indifference. Discharge instructions are handed across the bed in a language the patient may not read, with no one verifying that anything has been understood. Layer the growing healthcare staffing shortage on top of all this, and the steady decline of our current care system comes into even sharper focus.


Ironically, we are living through the most significant burst of medical innovation in a generation.

Large language models are entering clinical workflows. Predictive algorithms are reshaping triage. Federated data networks are making population-level learning newly possible. The technology available today is precisely the technology that could rebuild the quality of care that has been slowly breaking down, but only if the leadership of American medicine choose to point it in that direction. So far, that leadership seems to have chosen otherwise.

The quality of care in America is being deprioritized, in real time, across our healthcare system at the exact moment in history when the most advanced technological and clinical tools to revive it are at our fingertips.

Quality of Care, Sidelined

American healthcare has always known what good quality of care is. We have defined it, studied it, and taught it to others. Yet we have chosen to stop prioritizing it and measuring it, and, as we know, what you don’t measure doesn’t exist.


In 2001, the Institute of Medicine published “Crossing the Quality Chasm,” a landmark report that defined healthcare quality across six domains the Agency for Healthcare Research and Quality (AHRQ) has used as its operating framework ever since. The domains are simple to state and surprisingly hard to deliver. Care should be safe, meaning it avoids preventable harm to patients. It should be effective, meaning it matches services to scientific evidence and avoids both underuse and overuse. It should be patient-centered, meaning it honors the values, preferences, and individual needs of each person. It should be timely, meaning it reduces waits and harmful delays. It should be efficient, meaning it avoids waste. And it should be equitable, meaning it does not vary in quality on the basis of gender, ethnicity, geographic location, or socioeconomic status. This is the American framework. It has been on every accreditation form, in every nursing curriculum, and on every hospital quality dashboard for nearly twenty-five years.


The evidence base for what works inside these domains is enormous and well predates the framework itself. When my colleague Saumya RamaRao at the Population Council and I co-authored a 2003 review of a parallel quality-of-care framework, looking across implementation programs in five countries, the conclusion that emerged was simple. The interventions that most reliably improved patient outcomes were the ones that improved the client-provider interaction. Good quality of care produced more satisfied patients, greater knowledge of their own bodies and choices, and longer, more effective adherence to treatment. The same finding has been replicated in dozens of studies inside the United States since. The American failure isn’t a knowledge problem. It’s a will problem.


Set the AHRQ six domains against an American emergency room in 2026 and the gap is staggering. Safety is compromised every day by chronic understaffing, by the diagnostic errors that injure 795,000 Americans a year, and by the fatigue that fourteen-hour shifts produce. Effectiveness varies wildly by institution and by shift. Patient-centeredness has become a luxury that depends on which nurses are on duty that particular day and whether the unit has the staffing to allow time at the bedside. Timeliness has collapsed under ED boarding and the travel-nurse contracting model, in which hospitals fill chronic vacancies with short-term contract nurses on rotating eight- to thirteen-week assignments, an arrangement adopted to survive the post-pandemic workforce collapse. Efficiency has been redefined as throughput, the speed at which patients move through the system, which compresses the very interactions that prevent readmissions and downstream harm. And equity, the last domain on the AHRQ list, is the one we have abandoned most thoroughly: women wait longer for pain medication, the elderly are talked past, immigrant patients are talked over, and patients of color are more likely to be misdiagnosed.


The combined effect shows up in concrete moments: a discharged patient sent home with three new medications, a wound-care need, and no central nurse to call when the situation deteriorates. Care that should have been safe, effective, patient-centered, timely, efficient, and equitable was none of those. We had the definitions for good quality care and knew how to plan on providing it, but we chose to stop building toward it. We measured length of stay, readmission rate, infection rate, and patient satisfaction surveys that asked more questions about the food than about whether the patient experienced quality interactions with the providers or understood her care plan. We stopped asking the harder questions because the answers would have demanded that we focus on structural change.


The History Of The Breakdown

The systematic dismantling of quality of care didn’t happen by accident. American healthcare reorganized itself, beginning in the 1990s and accelerated after the 2008 financial crisis, around a logic that treats care delivery as a margin-management problem. Hospital systems consolidated. The American Hospital Association’s own data shows that more than 70 percent of U.S. hospital beds now sit inside consolidated systems. Private equity firms began acquiring physician practices, emergency-department staffing companies, dermatology and ophthalmology chains, even hospice agencies. The staff nurse model gave way to the travel nurse model, which solved a short-term problem by destroying the continuity of care that good practice requires. Productivity metrics rewarded throughput, the speed at which patients moved through the system, and the time a nurse was permitted to spend at any single bedside compressed.


Researchers like Elisabeth Rosenthal documented the transformation in An American Sickness. Anne Case and Angus Deaton, in Deaths of Despair and the Future of Capitalism, argued that American healthcare had become a tax on working-class life, extracting more and delivering less.

The political economy of disinvestment in care wasn’t a hidden process. It was an explicit one. We chose it, repeatedly, through legislation and acquisition and reimbursement design. And the dimensions that were hardest to count were the dimensions that disappeared.

The conversation in 2026 keeps treating the current crisis as if it were a recent failure rather than the predictable outcome of thirty years of deliberate policy and design choices. The ongoing federal cuts now arriving at NIH and NCI are accelerating a dismantling that was already underway. The Brookings Institution has documented how the current administration’s NIH and FDA reductions will lengthen the queue of care delays already visible in oncology and emergency medicine. Science magazine’s reporting on NIH grant rates, now at one in twenty-five from one in ten a few years ago, tells the rest of the story about how much research and innovation in caregiving the country is choosing not to fund. And the proposed restructuring of HHS, with thousands of terminations and the consolidation of 27 institutes into 8, is the kind of decision whose effects show up in patient corridors three years later, in ways that look like indifference but are actually the predictable downstream effect of the policy choice.


When Budget Cuts Reach the Bedside

What does this look like at the ground level? It looks like a much longer wait in the ER. A nurse who has been on shift for fourteen hours because the unit lost two travel contracts last month. A discharge instruction sheet that’s been translated badly from English because the language-access budget was the first line item cut. A delayed clinical trial enrollment because the institute that funded the trial site lost its contract administrator. A canceled surgery because the staff shortage that day pushed the case to next month. A misread imaging study because the radiologist who would have caught a malignancy is now reading for three different hospitals on a contract that pays him to be fast rather than careful.


The policy decisions made in Washington eventually show up in our lives without announcing themselves as policy. And too often, they reveal themselves as tragic consequences, especially for women or others who may be overlooked by the health system.

The Uneven Burden

The dimensions of quality care are not equally distributed in their absence. The patients who require the most relational investment lose the most when the relational layer is the first thing cut.

The research is consistent on this. Women wait an average of thirty minutes longer than men for pain medication in American emergency rooms.

Nurses are 10 percent less likely to record women’s pain scores at all. When clinicians are handed identical case charts with only the patient’s gender changed, they consistently rate the woman’s pain as less severe. The Washington Post’s 2022 investigation into how doctors dismiss women’s pain pulled the data together for a general audience, but the pattern has been visible in the research literature for decades, dating back at least to Hoffmann and Tarzian’s 2001 paper “The Girl Who Cried Pain.”


Elderly patients carry an additional layer of dismissal. The medical literature has documented a persistent ageism in ER triage and pain management. Linguistic minorities, particularly patients who don’t process English as a first language, face yet another. A patient who can articulate her symptoms confidently in fluent English may receive a different response than the same patient whispering in Hindi or Tagalog or Mandarin with a daughter translating beside her. The cumulative effect on an elderly immigrant woman with cancer is a triple discount on the credibility her symptoms are accorded by the system charged with treating them.


This is the population for whom quality of care matters most. The dimensions of good care become most visible, and most consequential, when the patient is least likely to be seen by the system meant to treat her. The American failure to extend equal care to its own elderly, female, and immigrant patients is an old pattern but has just come into sharper focus in 2026.


Moral Injury

The cost on the clinician side is significant. The conversation in American healthcare keeps describing nurse and physician departure as “burnout,” and the word is wrong. Wendy Dean and Simon Talbot, who introduced the term moral injury into the medical workforce conversation, argued that what clinicians are experiencing isn’t the exhaustion of working hard. It’s the wound of knowing the right thing to do and being prevented by structural forces from doing it.


A nurse who walked into nursing school precisely because she wanted to be the person at the bedside in the hardest moment of someone’s life, and who is then assigned a workload that makes that kind of care impossible, doesn’t burn out in the casual sense. She is being asked, every shift, to choose between the standard she trained for, and the productivity demanded by the system that employs her. The Annals of Emergency Medicine commentary on what’s driving the nursing shortage elucidates this directly. The system is asking clinicians to absorb the gap that the political economy created. They’re leaving the profession because the moral cost of staying has become unsustainable.


Lessons From Abroad

While American healthcare has been dismantling the relational architecture of care, other countries have been steadily building it. Their advantage is not that they have more advanced AI than we do. Their advantage is direction. They are deploying AI inside care models they have already articulated, rather than letting AI define the model by default. The Netherlands, the Nordic countries, Japan, Singapore, and Rwanda each illustrate a different version of that approach.


The Dutch national healthcare system, for example, has been investing in a care model that places continuity, patient experience, and equity at the center, and is now deploying AI in service of that model rather than as a substitute for it. Recent reporting on the Netherlands describes hospitals collaborating on AI-supported triage between facilities, with patient-data protections and explainability built into the procurement standards. The Netherlands is not deploying AI in the hope that it solves their care crisis. They are deploying AI as one tool within a care model they have already articulated.


The Nordic countries have launched what they call the FederatedHealth network, a federated health-data infrastructure that allows AI tools to learn from population-level data without compromising patient privacy or sovereignty. Nordic Innovation has explicitly framed the network as an ethical alternative to the American model, which has tended to either over-protect data to the point that it can’t be used for learning, or under-protect it to the point that it becomes commercial leverage.

The Nordic countries are choosing a third way, and the AI tools they’re building are designed to read patient records in ways that respect linguistic and cultural context. The contrast with the American emergency room that can’t deliver a discharge instruction in a patient’s first language is glaring.

Japan, facing the most accelerated aging demographic in the world, has pioneered hybrid human-and-technology elder care models that augment nursing relationships rather than replace them. Singapore’s HealthHub platform integrates patient-facing AI with the country’s continuity-of-care architecture. Rwanda has rebuilt its community-health-worker network from nothing after the 1994 genocide, and global health scholars now study Rwanda as a model of how to embed care continuity at population scale on a small budget. Each of these examples represents a different point in the design space, but they share a common starting question. They ask what the care relationship should look like, and then they build the technology and the workforce model in service of the answer. The American approach has been to deploy the technology first, optimize for cost, and hope that quality of care emerges as a byproduct. It does not. Care is a relationship. A relationship requires deliberate design.


What AI Could Actually Fix

The most important thing AI could do for American healthcare in the next decade is rebuild the AHRQ quality domains that have been hollowed out since their adoption. The technology is real, the use cases are real, and the gains are achievable. The question is whether the will exists to point the tools at the right targets.


Patient-centered care, the domain most directly compromised by understaffing and the throughput model, is where AI could close the largest immediate gap. Real-time translation tools embedded in the discharge process could deliver complex care instructions in any of forty languages, at the appropriate health-literacy level for the patient receiving them. Symptom-explanation tools could replace the discharge paperwork no one reads. Patient-facing question-answering systems like AI chatbots, properly governed, could give a discharged cancer patient a place to go at 2 a.m. when she’s not sure whether her symptom is normal or an emergency.


The relational layer of patient-centered care, the part sacrificed first to the throughput model, could be partially revived by AI-assisted documentation.

The single largest reason clinicians spend less time at the bedside is the documentation burden.

Ambient AI scribes, already in pilot at several major systems, can return that time to the bedside if hospital leadership chooses to reinvest the freed minutes in relational care rather than reassigning them to higher patient loads. That is a leadership choice, not a technology choice. And it’s the highest-leverage choice available. The review RamaRao and I wrote concluded, after looking across five countries and a decade of intervention data, that the interventions which improved client-provider interactions showed the greatest promise of any quality lever tested. AI tools that return time to the bedside are aimed at exactly that lever. The evidence for what they could accomplish is more than twenty years old. What we’ve been missing is the institutional decision to act on it.


Effective and timely care, the AHRQ domains stripped by the travel-nurse contracting model and chronic ED boarding, could be supported by AI-driven care coordination platforms that maintain continuity across staff handoffs, identify patients who haven’t followed up, and trigger interventions before deterioration. Predictive deterioration models, several of which have shown real promise in academic medical center pilots, could catch the patient whose pain is being dismissed before the dismissal becomes a code.


Each of these is achievable with technology that exists. The Coalition for Health AI has published consensus standards. The WHO has issued ethical guidance. There are good bias-audit methodologies the field can use. The Coalition’s framework explicitly demands explainability, human-in-the-loop oversight, and equity weighting at deployment. None of this is theoretical. The American hospital systems that have chosen not to deploy these tools at scale, or that have deployed them in service of throughput rather than relationship, have made a leadership choice. That choice is now visible in our hospital corridors.


Where Are the Innovators?

As the Editor of Women in the AI Vanguard Magazine, I’m always on the lookout for the maverick innovators who can pierce the darkness with the light of their vision.

Where are the healthcare vanguards at the helm of the ship who can course-correct to avoid inevitable calamity?

Sadly, the institutions best positioned to lead American care delivery in the age of innovation are the ones whose research budgets are now under threat. NCI Comprehensive Cancer Centers, of which there are only 57, receive substantial federal investment, generate enormous data, and host some of the most credentialed leadership in American medicine. Several of these institutions are world leaders in research innovation. Almost none of them are recognized as leaders in care delivery innovation.

The gap is concrete. The dimensions of quality care are not being piloted at scale in our hospital systems the way they should be. Atul Gawande has been calling for system-level redesign for years. A few institutions, like Cleveland Clinic with its empathy training initiative and Mayo with parts of a care-team model, have piloted small efforts. But sustained investment to meaningfully scale any of this has been absent. The leadership cadre that could announce a public commitment to rebuilding quality of care, fund the AI tools that would revive the dimensions, and model what 2030s American care delivery should look like has not, as of this writing, stepped forward.


This is not a research question. The research is done. Doyle, Lennox and Bell published the BMJ Open systematic reviewthirteen years ago that established the link between patient experience and clinical outcomes. The Compassionomicsevidence base is over 250 studies deep. The Schwartz Center for Compassionate Healthcare, the Beryl Institute, the Institute for Healthcare Improvement, and the Lown Institute all have curricula and frameworks ready to deploy. What’s missing is the institutional will to make the framework the operating principle rather than a footnote on the About Us page. The cancer centers have the prestige, the data, and the convening power. The federal cuts have given them a reason to demonstrate that prestige is more than a brand. The moment for that demonstration is now.


The Way Back

Care, properly understood, is a relationship. A relationship requires deliberate design. We knew this once.

We used to help the rest of the world build programs around this insight. The 2026 budget cuts are accelerating a collective shift away from this core understanding and have removed us from global leadership on what quality healthcare should look like. Other countries are, right now, building the architecture we walked away from.


We have the evidence. We have the technology, AI and digital and otherwise. We have decades of writing, on the ethics of care, from Arthur Kleinman on care as moral experience, from Paul Farmer on structural violence in medicine, that all converges on the same insight. The design, in 2026, has to be carried by leadership willing to make care-delivery innovation the equal of research innovation. The architecture of good care is still available to us. The will is the open question.



Raji Mohanam is the editor of Women in the AI Vanguard Magazine, where she highlights the women and systems shaping ethical, responsible innovation that can meet human needs. This essay draws on her longtime interest in quality-of-care research and programming, and on the conviction that the ‘care’ in healthcare is something we design, not something we hope for. She writes here as both an editor and a patient advocate for her mother, a cancer survivor.



Sources and further reading

Agency for Healthcare Research and Quality (AHRQ), “The Six Domains of Health Care Quality,” drawn from the Institute of Medicine, Crossing the Quality Chasm: A New Health System for the 21st Century, 2001.



Saumya RamaRao and Raji Mohanam, “The Quality of Family Planning Programs: Concepts, Measurements, Interventions, and Effects,” Studies in Family Planning, 34(4): 227-248, 2003.




Helen Riess, “The Science of Empathy,” Journal of Patient Experience, 2017.


Stephen Trzeciak and Anthony Mazzarelli, Compassionomics, 2019.



Arthur Kleinman, The Soul of Care, 2019.


Anne Case and Angus Deaton, Deaths of Despair and the Future of Capitalism, 2020.


Elisabeth Rosenthal, An American Sickness, 2017.


Wendy Dean and Simon Talbot, “Reframing Clinician Distress: Moral Injury Not Burnout,” Federal Practitioner, 2019.






American Medical Women’s Association, “Gender Bias in Emergency Care.”


Diane E. Hoffmann and Anita J. Tarzian, “The Girl Who Cried Pain: A Bias Against Women in the Treatment of Pain,”Journal of Law, Medicine and Ethics, 2001.


David E. Newman-Toker et al., “Burden of Serious Harms From Diagnostic Error in the USA,” BMJ Quality and Safety, 2024. Johns Hopkins press summary here.


Coalition for Health AI (CHAI), consensus standards on health AI deployment.

World Health Organization, “Ethics and Governance of Artificial Intelligence for Health,” 2021, updated 2024.




Annals of Emergency Medicine,“What Is Driving the Nursing Shortage?”

bottom of page