In December 2009, in the midst of the fight over adoption of the Affordable Care Act, President Obama laid out a central goal for health policy for the next decade. The country, he said, had to “bend the health care cost curve.” The president argued that containing the seemingly inexorable rise of medical spending was critical to the success of the ACA – not to mention the well-being of tens of millions of families, businesses and governments.

What’s remarkable, and little noticed outside of a small circle of health economists, is that we largely succeeded in that endeavor. To be sure, the cost of health care is higher than it was in 2009. But it costs far less than any forecast of spending at the time suggested was possible. Indeed, the difference between actual spending and the sums projected is in the trillions of dollars.

The Trillion That Didn’t Get Spent

In 2010, medical outlays commanded 17.2 percent of GDP. Using historical data as a guide and taking account of the recently passed Affordable Care Act, the actuaries at the Centers for Medicare and Medicaid Services projected that medical spending would exceed 21 percent of GDP by 2024. At the time, most experts found that estimate to be reasonable.

Yet in 2024 spending as a share of GDP reached only 18 percent, one-sixth below forecast. This works out to nearly $1 trillion less spent on medical care in 2024 alone, and $6.7 trillion less cumulatively since 2011.

Why was the money not spent? Was it an accident of circumstance, never likely to be repeated? Here, we summarize our analysis to be found in the Brookings Papers on Economic Activity.

Why Economists Were Pessimistic

The slowdown in medical spending growth is particularly unexpected because several plausible theories all predicted medical spending would rise more rapidly than GDP. The first is Baumol’s cost disease, named after the late Will Baumol of Princeton and New York University. Since productivity almost inevitably grows more slowly in services than in goods industries, wages and prices in the service sector must rise over time to retain workers. Much of health care operates as a service industry – it takes a professional an hour to have a psychotherapy visit, just as it always has. Thus, this theory predicts relative price increases over time.

A second relevant theory asserts that health care is a “luxury good,” so spending grows disproportionately to income as countries grow richer. And yet a third is built around the idea that technological change steadily expands the pathologies that medicine can treat, and thus expands the demand for care. This tendency is amplified by a payment system in which patients pay little for each extra visit, thus making them virtually indifferent to rising costs at the time they seek out care, while providers earn more for doing more, incentivizing them to offer treatments that may not be worth it.

Moreover, reality meshed nicely with these theories for many decades. Health care spending grew over two percentage points faster annually than GDP from 1960 (when good recordkeeping began) through 2010.

There was one precedent for a sustained slowdown. In the 1990s, after the failure of the Clinton administration’s health plan, businesses pushed private insurers to restrict access to some provider networks, thereby allowing the insurers to pit providers – hospitals, surgeons, rehab facilities – against each other to bargain down prices. This worked, at least for a while: from 1993 through 2000, medical spending did not rise relative to GDP.

However, the suppression did not last. Patients pushed back against restrictions on their access to desired providers, and insurers responded by loosening them, leading to a rebound in spending. Catch-up price increases kicked in after 2000, and by 2004 the gap was no more. There was no evidence that spending growth had diverged from the pre-1990s trend.

This Time Is Different

The current slowdown is more likely sustainable. No 14-year stretch since reliable recordkeeping has been in place has seen medical spending grow this slowly relative to the economy. And for the first time since international data have been systematically compiled by the OECD, health outlays in the U.S. as a portion of GDP rose less than in the average rich country.

The slowdown, by the way, is not an artifact of the Great Recession or demographic trends. Nor is it confined to particular insurance providers. Spending per enrollee was below forecast for Medicare, Medicaid and private insurance alike.

Digging just a bit, it’s worth noting that the slowdown has been most pronounced for hospital care and prescription drugs, and less so for physician services and long-term care for people with disabilities and the elderly. So what is going on?

We identify four primary drivers. First, medical technology has matured, perceptibly shifting from advances that add to cost to innovations that reduce total spending. Second, patients are using less care, presumably because they are pushed to economize by higher out-of-pocket costs and tighter restrictions on access imposed by insurers. Third, adjusted for demographic factors, Americans are healthier on average and thus need less care. Fourth, medical price increases have slowed as markets for some medical goods and services have become more competitive.

Medical Technology Has Matured

Medical technology has been a game changer, extending both the length and quality of life well beyond what our grandparents thought possible. Take one of the most important areas of advance: treatment of acute heart disease.

A half-century ago, a heart attack was something one either survived or not, and medical care didn’t change the odds much. Mortality rates were fierce and the quality of life for those who did survive was diminished.

When President Eisenhower suffered a heart attack in 1955, he received care that was standard for the day: a long hospital stay (seven weeks) followed by months at home in bed. Bed rest is not very expensive. What costs a lot is the surgery, imaging, monitoring and medications that might accompany the bed rest – and there was little of that to be had in 1955.

The first medical interventions for severe heart disease were dramatic and expensive. In the 1960s, surgeons learned to bypass blocked arteries. This involves opening the chest through the ribs and creating a new blood flow bypassing the occluded artery. The procedure is costly, requires weeks of recovery in hospitals and nursing facilities, and poses a substantial risk of death or disability. Still, because the alternative was so much worse, open-heart surgery was used frequently. Hospitals dedicated wings to it, and spending on heart disease rose immensely.

But the treatments that followed this initial wave were better and cheaper. Balloons and stents delivered through a small opening in the groin allowed surgeons to unclog arteries without open-heart surgery. Drugs dissolved clots that once required hours in an operating room to bypass. Each advance reduced the need for extended intensive care and the costs that come with it.

Some advances raised total medical spending, as the technology was sufficiently safe to be employed in many more patients. In other cases, the reductions in traumatic surgery and intensive care were enough to offset these costs. But even when net spending rose, it increased less rapidly than it had during the period in which open-heart surgery dominated treatment.

Then came improvements in prevention. By the 1990s, a daily pill could lower the cholesterol deposits in arteries that caused heart attacks in the first place. The adoption of statins and other cardiovascular drugs alone explains nearly all of the enormous decline in hospitalizations for heart attacks since 2000. Statins were relatively expensive at first: Lipitor became the best-selling drug in the world. But in 2011 the patent expired, generic versions poured in, and Pfizer’s revenue from Lipitor fell over 80 percent.

This is the predictable life cycle of medical technology. When a condition first becomes treatable, the new care is costly and carries substantial side effects. Moreover, when new treatment means going from nothing to something, spending obviously rises. But for the follow-ups to breakthrough treatments to compete, they must improve efficacy, lower side effects or reduce costs. Meanwhile, doctors and nurses master techniques, ancillary machinery improves and drugs lose patent protection. Ultimately, this process bends toward cheaper care, prevention and perhaps a cure.

For most of the postwar era, American medicine was early in this technology diffusion cycle. First treatments were arriving in waves – for heart disease, stroke, cancer, premature birth – each of them expensive. This history dominates the period from which forecasters predicted costs, and it explains why history-grounded analyses still forecast that the climb will continue. What is missed is that many treatment protocols have crossed into the later stages where cheaper strategies emerge.

Prevention and Cure

Heart disease tells the story of prevention. Cure, a related development, can reduce spending even more. The polio vaccine is the classic example of cure; a more recent one is treatment for hepatitis C, a chronic infection that ultimately leads to liver failure. In 2012, estimated costs to manage the disease ranged from $10,000 to over $42,000 for patients with end-stage liver disease. Then, in December 2013, Sovaldi – the first of a breakthrough class of direct-acting antivirals – was released by Gilead Sciences and shown to cure hepatitis C with minimal side effects.

Priced at over $80,000 for a course of therapy, Sovaldi was expensive. But even at its peak price, Sovaldi was cheaper than the cost of managing the untreated disease year after year. Thus, insurers reluctantly approved the costly treatment – first for those near liver failure, where rapid treatment is essential. But with time, competition from other antivirals drove down the price. Hepatitis C treatments now sell for less than a third of what Sovaldi cost when it was first on the market. The result is the potential to eradicate hepatitis C entirely. And a disease that’s cured costs nothing to treat.

Note, too, that prevention need not be medical, it can be societal. The classic example is smoking cessation. In 1965, 42 percent of American adults smoked and the typical smoker went through two packs of unfiltered cigarettes daily. Today, only 11 percent smoke. And the behavioral change is yielding ever-rising benefits. The decline in smoking in the 1970s, 1980s and 1990s is only now showing up as fewer respiratory and cancer admissions.

However, not all behavioral changes will save money. The large increase in obesity in the past few decades has been a huge driver of spending on cardiovascular, respiratory and musculoskeletal ailments. Reversing the trend toward increasing weight – a likely outcome from the rapid diffusion of GLP-1 agonist drugs like Ozempic – will surely improve length and quality of life enormously, and might save money as well.

Less Expensive Therapies

Like the replacement of open-heart surgery with stenting, innovations that reduce the intensity of treatment can also lead to substantial cost reductions even if they do not prevent disease. Stenting as an alternative to bypass surgery was already mentioned. An equally classic example: minimally invasive surgery across a wide variety of pathologies.

Consider total hip and knee replacements. There are just under 1 million such surgeries performed annually in the Medicare population. In the 2000s, hip and knee replacements were “open” procedures typically requiring a 10-inch incision and five days in the hospital, along with additional recovery in a nursing facility for about a third of patients. The procedure was expensive, and quality of life after successful surgery was mixed.

Innovation in joint replacements since then have dramatically reduced cost. Minimally invasive surgery allows surgeons to operate through multiple small incisions as opposed to one large entry, yielding less trauma. The new approach was supported by a confluence of treatment changes ranging from miniaturization of the instruments to more targeted anesthesia to aggressive physical therapy earlier in the recovery.

Savings beget savings. Less tissue damage from minimally invasive surgery reduces infection risk and speeds recovery. Patients are often discharged to their homes rather than to rehab facilities. These innovations lead to faster recovery, often making it possible to eliminate overnight hospital stays entirely. Hip and knee replacements moved first to hospital outpatient departments and then to ambulatory surgical centers. By 2024, more than three-quarters of total hip and knee replacements were done on an outpatient basis, reducing the cost by 40 percent compared to the equivalent open procedure.

Organizational Improvements

Innovations in the organization of care can save as much as innovation in treatment regimes. Patients often arrive at hospital emergency departments with vague symptoms – dizziness, nausea, nonspecific chest pain. Such patients used to be routinely admitted for a day or two of observation, since discharging them without a clear diagnosis could prove a terrible mistake.

Increasingly, many such patients are now monitored in hospital observation areas near emergency departments. The price paid for observation care is far below that for hospitalization. Some of that gap reflects true savings – workup in a hospital is more extensive. Some follows from the reality that many patients can be discharged sooner. What’s clear is that in increasingly burdened emergency departments, observation facilities can save a bundle without degrading patient outcomes.

Technology is also making it easier for care to be delivered by workers who lack training as physicians. In particular, nurse practitioners and physician assistants can now see more patients and handle a large number of routine situations with only a physician backup and perhaps the help of AI. For example, routine medication refills can be handled by people with more modest training, as can visits for monitoring after an acute event. Awakening to growing proof of efficacy and cost savings, states have steadily widened what these clinicians may do on their own.

Utilization Reductions

A second reason for lower-than-expected spending is reduced use of care, even when the care itself has not changed in efficacy or cost. This shows up in many medical settings, notably in imaging. Advanced imaging – such as MRI and CT scans whose use ballooned in the 1990s and early 2000s – grew much less rapidly starting in the late 2000s. In some clinical settings, imaging has even declined. The change was enormous. Had U.S. imaging kept pace with growth in other rich countries after 2008, there would be about 70 percent more CT scans today. Far more costly MRI and PET scans followed similar paths.

By the same token, the use of on-patent branded drugs slowed, even as new branded drugs continued to be developed. Take the case of a new class of cholesterol treatments called PCSK9 inhibitors, which are more effective than statins for many patients but far more expensive. In the past, drugs that worked somewhat better were often widely adopted, even if the cost was much higher. But insurers put PCSK9 inhibitors on restrictive prior-authorization lists, subject to evidence that statins were tried first, and when PCSK9 inhibitors were authorized they carried a higher burden of cost-sharing. The net effect was much less use than extrapolation from past experience would have implied.

Utilization management is particularly important in medical care because most of the incentives are to do more when in doubt. Insured patients want to get healthier, and copayments are far below full cost. Thus, they often demand more care than justified by the benefits. And since doctors were paid on a fee-for-service basis, they earned more by doing more, too.

On top of that, physicians were afraid that if they did not provide the care requested and patients thought they were receiving inferior service, they might be sued for malpractice. Not surprisingly, use exceeded what studies suggested was appropriate.

In the private sector, purchasers (mostly employers offering medical coverage) have been after insurers for decades to reduce their premiums. In the public sector, agencies have a parallel interest in avoiding tax and fee increases. In the past two decades, insurers have responded.

First, many private insurers sharply increased out-of-pocket costs, forcing patients to decide if the additional care was worth more of their own money. Enrollment in high-deductible plans – where the patient pays the first few thousand dollars before the insurer pays anything – rose from 13 percent of the privately insured to one-third by 2025. Those patients spend less, but not because they are healthier.

Note that this is not a free lunch for anybody. Moving into a high-deductible plan leads patients to forgo valuable and wasteful care alike, and rarely leads them to shop for a cheaper provider. In addition, some insurers imposed selectively high cost-sharing on specific therapies. For example, the most basic therapies for rheumatoid arthritis might be covered at low cost-sharing. But more effective medications may come at a much higher price. These “specialty tiers” may force payments of tens of thousands of dollars out-of-pocket before insurance pays anything.

In addition, insurers increasingly required doctors to obtain approval prior to delivery of care. Such prior authorization requirements can effectively stifle individuals and their clinicians who would prefer treatment that is more expensive. For example, by 2021 more than half of diagnostic-imaging spending in many Medicare Advantage plans needed approval before the scan could proceed. For branded prescription drugs, nearly half of all spending now requires prior authorization, roughly double the share in 2011.

In some cases, these restrictions are justified: before using an expensive medication, one should be sure that a cheaper medication does not work just as well. In other cases, the decisions seem capricious – cases where an insurer with no detailed knowledge of the patient’s circumstances overrules a physician who knows a great deal. And in all cases, adjudicating these prior authorization requests imposes substantial time and hassle costs on the system. Estimates suggest that administration of prior authorization rules costs the medical system upwards of $35 billion annually.

Changing the structure of physician payments has also contributed to the trend to do less. The shift from volume-based to value-based compensation gave doctors a new reason to evaluate patterns of care. Such arrangements are typically called “accountable care” contracts, as the group of providers is accountable for the costs of care along with the treatment of the patients.

In a typical policy, a group of physicians is given a target for average spending for their patients as a whole. They are allowed to keep some or all of the savings if average spending is below the target, provided the practice meets quality standards. More than half of physicians now work under some such arrangement for at least some of their patients. And naturally, physicians lean toward less expensive care when they are subject to such reimbursement arrangements.

A note here: higher cost-sharing, prior-authorization and accountable-care arrangements explain part of the spending slowdown, but not all of it. For example, the rate of growth of imaging fell even in traditional fee-for-service Medicare, which has no deductibles and no prior authorization.

What explains the rest of the decline of utilization is not certain. There are a number of theories. One is that there has been a shift in norms that some have termed the “value zeitgeist,” perhaps spurred by President Obama’s call to bend the cost curve. The idea is that physicians have absorbed a new concept of medical practice, one that incorporates cost as well as benefit to the patient. For example, oncologists now talk openly about the “financial toxicity” in addition to the “clinical toxicity” of the drugs they prescribe.

It is possible that utilization restrictions are approaching their limits, as both high cost-sharing and ubiquitous prior authorization face popular backlash. Unlike the 1990s experience, though, even a complete unwinding of these restrictions would not return spending to the peaks. And in any case, prior authorization is continuing to expand, and enrollment in high cost-sharing health plans shows no signs of tailing off. Even traditional Medicare, which has never imposed prior authorization, has proposed implementing restrictions.

More likely, federal and state authorities will intervene to make the prior authorization process function more smoothly. For example, many states have set deadlines by which insurers must decide on prior authorization requests – the answer does not have to be “yes,” but it must be timely. Similarly, improving information-sharing could allow insurers to automate prior authorization requests in the same way that credit card authorization at retail checkout takes so little time that customers barely notice it.

High cost-sharing is more likely to come under scrutiny than prior authorization. There have been proposals to limit it, for example to reduce high deductibles or at least to spread out the payment obligation over longer periods. A key issue here will be how to address the additional pressure on insurance premiums that would result from reductions in other revenues.

Slower Price Growth

Total outlays for medical care include the patient payment and the insurer payment. We term that total the full “price.” Because there are so many health insurance policies – multiple insurers, each with multiple policies, and many different provider groups – and the contracts are trade secrets, we do not know the precise rate of price increases in medical care. The pace does appear to have slackened, though, in recent years.

Data from the Bureau of Labor Statistics indicate that price growth on that combined bill slowed from about 1-2 percent above the general rate of inflation to roughly par with general inflation. Over time, this adds up. A percentage point a year, accumulated across 15 years, translates into hundreds of billions of dollars – about a quarter of the total slowdown in health care spending.

Fighting back, some providers have been able to impose hidden price increases to recapture some of the surplus generated by slower price growth. For example, hospitals have bought physician practices and relabeled them “hospital outpatient departments,” for which insurer payments are higher than for services delivered in physicians’ offices. Similarly, providers have learned to “upcode” patients’ records, making them appear more ill than they really are. But there’s good reason to believe that these subterfuges have failed to offset slower price growth entirely.

Consider, too, that the slowdown was not uniform across payers. Medicare and Medicaid set prices administratively, and the government simply decided to impose smaller increases. For example, the Affordable Care Act built in a productivity-related slowing of the growth in Medicare hospital payments, and later budget rules went even further. Legally lacking the discretion to avoid caring for such patients, providers accepted what was offered.

For private insurers, negotiating leverage was key to slowing medical inflation. Here, the same forces noted earlier strengthened insurers’ hands. Fewer hospital stays meant hospitals would often give up more in the bargaining to remain in the insurance companies’ good graces. Patients with more of their own money at risk led pharmaceutical companies to agree to larger “rebates” if given a more favorable position on the insurance companies’ formularies. And the growing power to deny claims gave insurers more leverage over negotiated payments. Slower commercial insurance price growth thus in part complements the demand slowdown and the increasing price sensitivity of consumers, payers and providers.

Bent – But Not Enough

The health care cost curve has indeed bent enough to make a difference. But even with all that has been done, it has not bent enough. The U.S. still spends more on medical care per person than any other country, with no payoff in terms of better outcomes. And because the population is aging, the federal government will continue to spend more, even if spending per beneficiary rises no faster than per capita income. Employers will face increasing premiums, which is especially problematic for workers with low- and middle-income wages since medical coverage represents a higher share of total compensation. Those workers will be at risk of wage reductions as employers cope with rising total compensation, or of losing their health insurance entirely and looking to government for help with medical bills. Slowing cost increases to date is thus a good first step, but more needs to be done.

The Future of Cost Savings

The most important item on the agenda for cost savings is to continue to reorient medical innovation from a narrow focus on improving what can be treated to one focused on value-based health improvement. Thanks to the accomplishments to date in this regard, we believe that the U.S. has entered a period of fundamentally lower trend growth in cost.

There will still be periods of significant spending growth in aggregate, particularly when the system integrates blockbuster technologies. A recent example: GLP-1 receptor agonists – think Ozempic – that treat obesity, a condition so widespread and so difficult to manage otherwise without invasive bariatric surgery that spending on GLP-1 drugs is bound to rise. The hope is that GLP-1s will save money downstream by preventing or delaying a host of pathologies the way that statins did. In the short run, however, they drive up spending.

For every GLP-1 example, there are other conditions that may become much cheaper to treat. Certain cancers can be prevented and others reduced through preventive care. Substance abuse can be reduced (perhaps with GLP-1s), leading to great potential savings in future treatment. And cognitive impairment that is now very costly to manage may respond to new drugs.

Artificial intelligence also offers the opportunity for technology-induced cost savings. The first target for AI in health care might be reduced administrative costs – billing, coding and hours spent on the phone seeking authorization is the kind of work that AI appears well-positioned to do. AI may also become an important input to identifying cost-saving drugs, in predicting who needs which type of therapy, and in remote monitoring of patients who can recover more cheaply and comfortably at home rather than in a hospital or nursing facility.

* * *

The Affordable Care Act, President Obama’s signature legislation to reform health care, helped to bend the health care cost curve. So did the actions of other presidents and countless insurers, businesses and individuals. While the impact took a long time to show up in the data, their efforts have been far more successful than almost anyone expected.

We are not passive observers. Whether the cost curve keeps bending depends on which policies we adopt and how the medical system responds. Medical care is an economic system, and like all economic systems, we can have a good deal of sway over how it performs.