Health Care Policy


Following Trauma

First-in-kind study offers unique opportunity to advance care for veterans and survivors of trauma

More than 2.6 million servicemen and women have deployed to Iraq or Afghanistan since September 11, 2001. Many veterans return home from their service with symptoms of posttraumatic stress, depression, chronic pain and traumatic brain injury. These symptoms are also common among civilian trauma survivors.

Get more HMS news here

The National Institutes of Health, in response to an executive order from President Obama to initiate major research efforts to better understand and treat these disorders, has just provided $21 million for the most comprehensive longitudinal study of trauma survivors ever performed. The five-year-long project, which will be known as the Aurora study, is funded by the National Institute of Mental Health.

The research will utilize the efforts of 19 institutions and more than 40 scientists. Three of the four Aurora principal investigators are Harvard faculty: Ronald Kessler, the McNeil Family Professor of Health Care Policy at Harvard Medical School, who will be responsible for the social epidemiological aspects of the study; Karestan Koenan, professor of psychiatric epidemiology at the Harvard T.H. Chan School of Public Health who will be responsible for the genetic epidemiological aspects of the study; and Kerry Ressler, professor of psychiatry at HMS and the James and Patricia Poitras Chair in Psychiatry and chief scientific officer at McLean Hospital, who will be responsible for the neurophysiology and neuroimaging aspects of the study.

Trauma survivors will be enrolled in the study in the immediate aftermath of trauma and followed longitudinally for one year using sophisticated adaptive sampling methods to perform a comprehensive, state-of-the-art assessment of genomic, neuroimaging, physiologic, neurocognitive, psychophysical, behavioral and self-report markers.

In addition to its unparalleled scope, the study differs from previous studies in that it will assess neuropsychiatric effects of trauma broadly rather than focus on only one or a few diseases.


“We want to be patient-centered and not diagnosis-centered,” said Samuel McLean, lead principal investigator of the study and an emergency medical physician at the University of North Carolina at Chapel Hill.

“We also recognize that a trauma survivor with depression, who also has a lot of posttraumatic stress symptoms and traumatic brain injury symptoms, is likely to be a lot different in terms of underlying biology than a trauma survivor with depression alone. We believe that only by looking at an individual globally can we achieve great advances in treatment,” McLean said.

The study will use a wealth of biologic data collected during the critical time after trauma exposure to chart onset and progress of disorders and to search for treatment intervention targets. Psychosocial specifiers will also be examined in an effort to determine ways in which behavioral or environmental interventions might be useful in prevention and treatment.

Another major goal of the study is to develop predictive analytic tools that clinicians in emergency departments, military hospitals and military aid stations can use at the bedside to identify trauma survivors at high risk of persistent sufferings.

“We need these tools urgently so that trauma survivors at high risk can be identified for early preventive treatments,” Kessler said.

Study investigators are currently pursuing additional foundation and philanthropic support to expand the study in ways that would allow targeted models to be developed for important trauma subtypes, including sexual assault, physical assault and motor vehicle collisions involving the death of a loved one.  

This story is adapted from a news release from the University of North Carolina.



Doc Versus Machine

Head-to-head comparison reveals human physicians vastly outperform virtual ones

Increasingly powerful computers using ever-more sophisticated programs are challenging human supremacy in areas as diverse as playing chess and making emotionally compelling music. But can digital diagnosticians match, or even outperform, human physicians? 

The answer, according to a new study led by researchers at Harvard Medical School, is “not quite.” 

The findings, published Oct. 10 in JAMA Internal Medicine, show that physicians’ performance is vastly superior and that doctors make a correct diagnosis more than twice as often as 23 commonly used symptom-checker apps. The analysis is believed to provide the first direct comparison between human-made and computer-based diagnoses.

Diagnostic errors stem from failure to recognize a disease or to do so in a timely manner. Physicians make such errors roughly 10 to 15 percent of the time, researchers say. 

Over the last two decades, computer-based checklists and other fail-safe digital apps have been increasingly used to reduce medication errors or streamline infection-prevention protocols. Lately, experts have wondered whether computers might also help improve clinical diagnoses and reduce diagnostic errors. Each year, hundreds of millions of people use Internet programs or apps to check their symptoms or to self-diagnose. Yet how these computerized symptom-checkers fare against physicians has not been well studied.

In the study, 234 internal medicine physicians were asked to evaluate 45 clinical cases, involving both common and uncommon conditions with varying degrees of severity. For each scenario, physicians had to identify the most likely diagnosis along with two additional possible diagnoses. Each clinical vignette was solved by at least 20 physicians.

The physicians outperformed the symptom-checker apps, listing the correct diagnosis first 72 percent of the time, compared with 34 percent of the time for the digital platforms. Eighty-four percent of clinicians listed the correct diagnosis in the top three possibilities, compared with 51 percent for the digital symptom-checkers.

The difference between physician and computer performance was most dramatic in more severe and less common conditions. It was smaller for less acute and more common illnesses.

“While the computer programs were clearly inferior to physicians in terms of diagnostic accuracy, it will be critical to study future generations of computer programs that may be more accurate,” said senior investigator Ateev Mehrotra, an associate professor of health care policy at HMS.

Despite outperforming the machines, physicians still made errors in about 15 percent of cases. Researchers say developing computer-based algorithms to be used in conjunction with human decision-making may help further reduce diagnostic errors.

“Clinical diagnosis is currently as much art as it is science, but there is great promise for technology to help augment clinical diagnoses,” Mehrotra said. “That is the true value proposition of these tools.”



Allaying Fears

Hospitals' transition to electronic health records appears safe for patients 

As waves of hospitals move from older methods of record keeping to new digital electronic health record (EHR) systems, many medical professionals express fears that implementing an EHR system in their hospital will have dire results, including more errors and higher patient mortality.

But these fears are largely unfounded, researchers from Harvard Medical School and the Harvard T.H. Chan School of Public Health have found after studying a diverse group of U.S. hospitals that implemented new EHR systems during 2011 and 2012.

Get more HMS news here

Better technology and monetary incentives from the federal government promoting more advanced record keeping have meant that EHR adoption has accelerated, both in the U.S. and abroad.

However, implementing a new EHR system can be one of the most disruptive events a hospital can experience. Workflow disruptions can create numerous frustrations and distractions as providers learn new systems, potentially leading to errors. Many providers fear that patient care could suffer as a result.

One prior study on a hospital’s EHR implementation even reported increases in mortality in the months following activation of the new system.

In the current study, led by Michael Barnett, assistant professor at the Harvard Chan School, researchers sought to address this concern by studying the short-term impact of implementing EHR systems nationally. They identified 17 hospitals that implemented a new EHR system in a single day, known as “going live.” Then, using Medicare data from 2010 to 2012, they compared patient outcomes before and after EHR implementation to the same trends in other nearby hospitals.

Because these hospitals implemented their systems in a single day, this created a “natural experiment” to study the effect of EHR implementation in similar groups of patients admitted before and after activation of a new system.

The researchers found that there was no difference in the rates of inpatient mortality, adverse safety events and readmissions in hospitals implementing EHR systems before and after going live, compared to the control group.

There was also no change when examining groups of patients and hospitals that might have been at higher risk for problems, such as sicker patients or hospitals who transitioned from paper to electronic charts, versus those just switching from one electronic system to another.

“Physicians’ tremendous frustration in switching to new electronic health records can spill over into concerns that patient care is actually worse because of these systems,” said Barnett, who is also a primary care physician at Brigham and Women’s hospital. “Happily, our results suggest that switching to a new system is a challenge that hospitals are prepared to handle safely.”

This should be encouraging to doctors, practices and hospitals planning their own implementations, the researchers said.

“Having recently witnessed firsthand how disruptive an EHR implementation can be, it is reassuring to know that hospital safeguards prevent patients from being harmed,” said Anupam Jena, senior author of the study and the Ruth L. Newhouse Associate Professor of Health Care Policy at HMS.

This research was supported by grants from the Office of the Director at the National Institutes of Health (NIH early independence award 1DP5OD017897-01 and Health Resources and Services Administration T32-HP10251).


Worth Their Salt

Publicly available databases reveal gender-based salary disparities in medicine

Women physicians earn an average of $20,000 per year less than men even after adjusting for factors likely to influence income, according to what is probably the largest study of salary differences between male and female medical school faculty members.

The study, which analyzed data for physicians employed at 24 public medical schools, appears online in JAMA Internal Medicine.

Get more HMS News here

“More than raising attention to salary sex differences in medicine, our findings highlight the fact that these differences persist even when we account for detailed factors that influence income and reflect academic productivity,” said Anupam Jena, the HMS Ruth L. Newhouse Associate Professor of Health Care Policy, and an internist at Massachusetts General Hospital.

“The fact that we observed these income differences among physicians who are public employees raises issues that may have state regulatory implications,” he said.

While several previous studies have documented salary differences between male and female physicians working in academic medicine, those investigations have been relatively small, restricted to specific specialties or depended on participants’ responses to survey questions.

The current study started with employee information—including names, titles and salaries—from public medical schools in 12 states that make such information available online. The researchers merged that data on physician faculty salaries with information from the Doximity database of more than 700,000 U.S. physicians.

That database includes demographic and professional factors including age, gender, faculty rank, university affiliation, specialty, year of residency completion, type of clinical practice (based on receiving Medicare payments) and several indicators of research activity.

Among their final sample of almost 10,250 physician faculty members—35 percent of whom were women—the unadjusted average annual salaries of women were almost 20 percent lower ($206,641 versus $257,947) than those of male physicians.

After adjusting for the factors mentioned above, female physician faculty members still received salaries 8 percent lower than those of comparable male physicians ($227,783 versus $247,661).

Adjusted salary disparities were greatest for orthopedic surgery, obstetrics/gynecology (one of the specialties female physicians were most likely to enter), other surgical subspecialties and cardiology, and lowest for family medicine and emergency medicine.

Adjusted average salaries for women in radiology were slightly higher than for men.

Disparities also varied among medical schools, with adjusted average salaries for male physicians being significantly higher at nine schools—the greatest disparities occurring at schools in the western U.S.

 “Our findings also highlight how non-traditional data sets like Doximity and public employee salary information can be used to investigate questions that historically have been difficult to evaluate due to lack of access to large-scale data,” said co-author Daniel Blumenthal, HMS research fellow in medicine and clinical fellow in the Mass General Division of Cardiology.

The study was supported by National Institutes of Health Early Independence Award 1DP P50D017897-01.

Adapted from a Mass General news release.


Unfulfilled Promise

Early claims of success of “pay-for-performance” incentives have not paid off

Early studies have suggested that paying doctors bonuses for meeting targets for certain health care measures would lead to improved health outcomes for patients, but the findings were not repeated in later, more rigorous studies, according to new research. 

In a paper published June 23 in Preventing Chronic Disease, Stephen Soumerai, Harvard Medical School professor of population medicine at Harvard Pilgrim Health Care Institute, and Huseyin Naci at the London School of Economics analyzed the latest results.

This study is the latest in a series of papers from Soumerai and colleagues exploring how uncorrected bias and unreliable study design may contribute to flip-flops in health care recommendations and policies, and offering recommendations on what researchers, educators and journalists might do to address these issues.

Get more HMS news here

Soumerai discussed the paper and the challenges of research design with Harvard Medicine News.

HMN: Why are the promising conclusions from early studies of pay-for-performance so different from the findings of later studies?

When the initial national programs in pay-for-performance were rolled out, many early studies suggested that the policies were contributing to improved health outcomes, including findings which suggested that giving physicians cash incentives to take patient blood pressure contributed to improved patient mortality rates.

The problem was that these studies didn’t correct for other changes that were taking place in medicine at the same time. Failing to account for these ongoing baseline changes is called history bias. Uncorrected, these biases led to false, optimistic conclusions because the policies were taking credit for improvements in medical care that were already occurring.

Subsequent, stronger research—including quasi-experimental studies, randomized control trials and systematic reviews that can control for these biases—consistently overturned the early, optimistic findings.

One international systematic review found that not only was there little evidence to support pay-for-performance’s effects on quality of medical care but some studies found that it sometimes had the unintended consequence of discouraging doctors from treating the sickest patients.

HMN: What happened next?

Unfortunately, that initial, weak research on the supposed success of pay-for-performance in health care was published in top medical journals and hastened the adoption of these policies around the world.

Despite its unfulfilled promise and discouraging evidence, this costly and ineffective approach to improving health care is a widespread component of current national and international health care policies. It is entrenched in many policies created by the Affordable Care Act. This has led to billions of dollars in wasted resources around the world.

Remarkably, most of these initiatives were started after the evidence of failure or very weak effects were published.

A strong interrupted time series study of the percentage of study patients who began antihypertensive drug treatment from January 2001 through July 2006 (originally published in doi: 10.1136/bmj.d108.) Dashed line indicates when the United Kingdom’s pay-for-performance (PfP) policy was implemented (April 2004). Controlling for the improvement in treatment that was happening anyway over many years, there was no additional benefit of the $2 billion PfP policy. A single average point before and after the program (week-post design) would have contributed false hopes of an improvement in quality of care.

HMN: How common is this?

This isn’t an isolated incident; it’s a pattern that repeats itself throughout health care effectiveness research. Weak research designs are the most common reason for biased health care effectiveness research.

This paper continues a line of research we’ve been pursuing that shows how stronger research designs can avoid these costly mistakes.

HMN: What’s the solution?

Science can be a useful tool for designing and implementing effective, efficient health care, but only if we use reliable research methods such as randomized, controlled trials; interrupted time-series designs; and systematic reviews—and if we conduct rigorous pilot tests of expensive policies.

Investments of private and taxpayer funds should be based on solid evidence of safety and efficacy. The alternative—the present system—relies on weak and uncontrolled research designs, misleads policy makers and the public, and will ultimately lead to unsustainable costs, unhappy clinicians and policies that may damage—rather than improve—the quality of medical care.

Everyone involved in the process that brings this information to light—the researchers who design the studies, the journal editors who accept and reject papers and the mainstream journalists who cover health care and policy—all have a responsibility to get this work right. 


Opioid Unknowns

U.S. hospitals vary nearly twofold in opioid prescribing rates

Nearly 15 percent of opioid-naïve patients hospitalized under Medicare are discharged with a new prescription for opioids, according to a study published today in JAMA Internal Medicine

Among those patients who received a prescription, 40 percent were still taking opioids 90 days after discharge. The rate of prescription varied almost twofold between hospitals, with some hospitals discharging as many as 20 percent of patients with a prescription for opioids.

Despite growing concern about the public health costs of long-term opioid use, little has been known about how often the painkillers are prescribed. Even appropriate short-term prescriptions can lead to long-term use and, potentially, abuse.

“Every day, 44 people in the United States die from prescription drug overdoses, especially opioid overdose,” said Anupam Jena, lead author of the study, the Ruth L. Newhouse Associate Professor of Health Care Policy at Harvard Medical School, and a physician at Massachusetts General Hospital. “It’s critical that we understand hospital prescribing patterns so that we can make sure we are prescribing these medications safely and effectively without fueling this deadly crisis.”

Jena said that the researchers were especially interested to see whether there was a relationship between Medicare incentives to encourage hospital pain management and the rate of opioid prescriptions by hospitals.

Currently, hospitals that score well on patient–reported pain metrics receive cash bonuses. Could this policy encourage hospitals to overprescribe opioids?

While the study, which analyzed hospitalizations under Medicare in 2011, did find that the hospitals that scored highest on pain control measures also had the highest rates of opioid prescribing, the link was quite modest and did not explain most of the variation between hospitals.

While the study did adjust for differences in patient mix, diagnosis and other clinical factors, Jena said, it was possible that some hospitals had a high percentage of patients with more severe pain. However, he said it was more likely that the observed differences in prescribing rates were related to differences in how individual doctors prescribe or to differences in the prescribing culture of different hospitals or geographic regions.

Short-term prescriptions for opioids, while well-intentioned, can lead to long-term use.  “Opioids have an important role to play in specific forms of acute pain, but the high risks of long-term use mean that adherence to proper prescribing guidelines is critical,” said study co-author Pinar Karaca-Mandic, associate professor at the University of Minnesota’s School of Public Health Division of Health Policy and Management.

Dana Goldman, professor and the Leonard D. Schaeffer Chair and Director of the Schaeffer Center for Health Policy and Economics at the University of Southern California also contributed to the study.


Testing Transparency

Giving patients a tool to search for health care prices did not result in lower spending

If patients knew the price of different health care options, they could make better decisions about their medical care and help cut health care costs by shopping for lower-priced care—at least that’s the hope.

With that in mind, more than half of U.S. states have passed laws establishing price transparency websites, and many employers have offered price transparency tools to their employees.

But does it work?

Limited research has been conducted on whether price transparency tools actually help control spending.  In one of the first studies of its kind, appearing today in JAMA, a team of researchers from Harvard Medical School found that access to a price transparency tool was not associated with lower health care spending.

Get more HMS news here

The researchers studied the Truven Health Analytics Treatment Cost Calculator, an online price transparency tool, which tells users how much they would pay out of pocket for services such as X-rays, lab tests, outpatient surgeries or physician office visits. The out-of-pocket cost estimates are based on the user’s health plan benefits and on how much they have already spent on health care during the year.

Two large national companies offered this tool to their employees in 2011 and 2012. The researchers compared the care of the almost 149,000 employees who were offered the price transparency tool to 296,000 employees from other companies who were not offered the tool.

Overall, having access to the tool was not associated with reduction in outpatient spending, and patients did not switch from more expensive outpatient hospital-based care to lower-cost settings. When the researchers looked at only those with higher deductibles—who would be expected to have greater price shopping incentives—they also found no evidence of reduction in spending.

“Despite large variation in health care prices, prevalence of high-deductible health plans and widespread interest in price transparency, we did not find evidence that offering price transparency to employees generated savings,” said Sunita Desai, a research fellow in health care policy at Harvard Medical School and lead author on the study.

There were several potential reasons.  Despite aggressive promotion, only 10 percent of the employees who were offered the tool used it. Further, when they did they use the tool, more than half the searches were for relatively expensive services of over $1,000.

“For expensive care that exceeds their deductible, patients may not see any reason to switch. They do not save by choosing a lower-cost provider, even if the health plan does,” said Ateev Mehrotra, associate professor of health care policy at Harvard Medical School and senior author on the study.

The tools can still provide patients with valuable information, including their expected out-of-pocket costs, their deductible, and their health plan’s provider network, the researchers noted.

“People might use the tools more—and focus more on choosing lower-priced care options— if they are combined with additional health plan benefit features that give greater incentive to price shop,” said Desai.

This work was supported by a grant from the Laura and John Arnold Foundation and the Marshall J. Seidman Program for Studies in Health Economics and Health Care Policy at Harvard Medical School.


Sparing Livers

Curing hepatitis C will create transplant opportunities for patients with other illnesses

Recently developed treatments that cure hepatitis C virus (HCV) will create new opportunities for people with other liver diseases to receive transplanted livers.

Only one-third of Americans who need liver transplants receive them and shortages are expected to rise as the transplant waiting list continues to grow even as the supply of organs remains flat.

New research suggests that the benefits of new HCV treatments could spill over to many other diseases that cause end-stage liver failure.

Get more HMS news here

The findings, from a study led by researchers at Harvard Medical School, are published May 3 in the American Journal of Managed Care.

In the United States, the most common reason for needing a liver transplant is cirrhosis caused by HCV, followed by cirrhosis from long-term alcohol use.

Other reasons include nonalcoholic fatty liver disease, other forms of chronic hepatitis, genetic forms of liver disease and acute liver failure from drugs such as acetaminophen (Tylenol).

“The inadequate supply of liver donors in the United States is a real problem,” said Anupam B. Jena, associate professor of health care policy at Harvard Medical School and an internist at Massachusetts General Hospital, who led the research.

“People die every day of liver disease because a suitable organ never materializes. By curing patients of HCV before they become sick enough to need a new liver, new HCV drugs shorten the waiting lists and make more livers available to patients with other illnesses,” he said. 

Get more HMS news here

Jena and his colleagues at the Leonard D. Schaeffer Center for Health Policy and Economics at the University of Southern California, Precision Health Economics and the University of Chicago developed an epidemiologic-economic model that combined data on trends in chronic liver disease with liver transplant allocation models to estimate the potential effects of systematic HCV screening and treatment on the demand for liver transplants in the United States.

The researchers found that systematic HCV screening and treatment would not only reduce rates of end-stage liver disease due to HCV infection but would also spare approximately 10,500 livers from being transplanted into HCV-infected patients over a twenty-year period (2015-2035).

An estimated 7,300 of these livers would be transplanted into liver disease patients without HCV. The remaining 3,200 livers would be transplanted into people who were not screened for HCV or who have HCV but did not respond to existing therapies.

The study has implications for a broad range of diseases aside from liver disease, said study co-author Darius Lakdawalla, Quintiles Chair in Pharmaceutical Development and Regulatory Innovation at the University of Southern California. 

“For any disease in which organ transplants are in short supply, our study suggests a novel pathway by which treatment of a single disease may save lives of those with other diseases by sparing organs for transplant,” Lakdawalla said.

For example, he said, improvements in coronary artery disease management will mean that people with heart failure arising from this condition will require fewer heart transplants in the future; those hearts will be spared and can be transplanted into people with other forms of heart failure. 

But not all spillovers are benefits.

The study also highlighted how increasing rates of diseases such as diabetes and hypertension may have unintended consequences for patients with other diseases. By raising the demand for kidney transplants due to end-stage kidney disease, for example, both diabetes and hypertension crowd out transplant opportunities for patients with other forms of end-stage kidney disease.

Study co-authors included Tomas J. Philipson from the University of Chicago; Warren Stevens from Precision Health Economics, a healthcare consultancy to life science firms  (Jena and Philipson are consultants for PHE); Yuri and Gonzalez, Steven E. Mark and Timothy Juday from AbbVie, Inc., which develops and markets HCV treatments. 

Funding for this study came from AbbVie, Inc.


Quality Concerns

A new study shows corporate chains tend to buy and sell lower-quality nursing homes.  Does care improve? 

Over the last two decades, thousands of nursing homes have been bought and sold by corporate chains. A new study has found that these transactions are related to lower quality of care.

The results of this study suggest that chains bought and sold nursing homes that were already having quality problems and these quality issues persisted after the transaction.

Get more HMS news here

In nursing homes that underwent a transaction, quality problems were present both before and after the transaction occurred.

The research, led by David Grabowski, professor of health care policy at Harvard Medical School, measures the implications of corporate ownership changes, which policymakers have long worried negatively impacts nursing home quality of care. The findings are published in the May issue of the journal Health Affairs.

“A large number of mergers, sales and acquisitions have occurred over the past two decades among nursing home chains, and we wanted to see how residents living in these nursing homes were affected by these transactions,” Grabowski said.

According to the study, around 1,200 to 2,000 nursing homes in the U.S. (7 to 13 percent) reported a transaction annually. The most commonly reported were mergers across chains. For the 10 largest chains, which contain 12 percent of all nursing homes, tremendous “churn” occurred in the study period; only two of the 10 chains did not experience any change in corporate ownership.

In spite of all these transactions, the researchers found that the organizational structure of the nursing home industry has remained relatively constant. For example, the proportion of nursing homes owned by a chain is the same today as it was in the late 1990s.

Grabowski and colleagues at the University of Michigan, the University of Rochester and Vanderbilt University studied the quality implications of these transactions.

“By comparing the quality of care in facilities that undergo a transaction with those that do not, we were able to calculate the quality implications of these transactions,” said senior author Jane Banaszak-Holl, associate professor of health management and policy at the University of Michigan.

The authors found that nursing homes that underwent a chain-related transaction had more government deficiency citations than nursing homes that did not experience a transaction. Given the presence of low quality prior to these transactions, the authors could not conclude that the chain transactions led to declines in quality. Rather, the results suggested that these transactions are indicators of low quality nursing homes.

This result raises important issues for nursing home policy regarding ownership accountability, oversight and transparency. Going forward, the number of chain-related transactions could be a useful indicator of potential quality to consumers and their advocates, and best practices should include requiring nursing homes to make the notification of an impending sale publicly available.

The authors also suggest that more detailed data on chain ownership and quality is needed for both consumers and regulators. Policymakers should consider legislation that requires more detailed and comprehensive reporting of ownership for nursing home chains.

This study was funded by the National Institute on Aging (R01 AG042418).


ACA Payment Reform Achieves Early Gains

Early Medicare Shared Savings Program results are promising

Accountable care organizations that joined the Medicare Shared Savings Program (MSSP) when it launched in 2012 achieved modest savings while maintaining or improving performance on measures of quality of patient care in 2013, the first full year of the program, researchers at Harvard Medical School found in the first rigorous examination of this key health care payment reform program.

These early adopters lowered spending by 1.4 percent in 2013 relative to a control group of non-ACO providers in the same areas, which represents a $238 million reduction in spending.

Get more HMS news here

These savings provide more evidence of early promising results from accountable care organization initiatives in Medicare, of which the MSSP is the largest, but the results also tell a more complex story about the pattern of savings across different types and cohorts of ACOs.

The findings were published in the New England Journal of Medicine

ACOs are groups of health care providers who agree to provide care to a population of patients under a global budget known as a benchmark. ACOs that hold spending below the benchmark and perform well on measures of quality of care share in the savings.

In contrast with other ACO programs such as the Pioneer model, MSSP participants are not required to reimburse Medicare if spending exceeds the benchmark.

The first two cohorts of provider groups (220 in total) entered the MSSP in mid-2012 or at the beginning of 2013. Since then, the program has expanded and currently includes over 430 participants.

While the ACOs that joined in 2012 cut spending by $238 million, the next cohort of ACOs that joined in 2013 achieved no savings in their first full year in the program, suggesting that the early success of the first participants may not be replicated by the subsequent waves of ACOs that have joined the MSSP.

In addition, because Medicare paid out $244 million in shared-savings bonuses to ACOs in the 2012 cohorts, the lower spending in that cohort did not constitute net savings to Medicare.

 “These results suggest that ACOs with no downside risk can achieve savings, but that savings to Medicare and society may be slow to develop,” said J. Michael McWilliams, the Warren Alpert Associate Professor of Health Care Policy at HMS and lead author of the study.

“But the incentives for ACOs to lower spending are currently very weak, so savings may accelerate if the incentives are strengthened.”

In particular, the current method for setting an ACO’s benchmark diminishes its incentives to save. Specifically, if an ACO lowers spending now, it is penalized with a lower benchmark later.

According to the authors, severing the link between an ACO’s benchmark and its previous savings could go a long way toward rewarding ACOs adequately for curbing wasteful practices and allowing the returns necessary for ACOs to invest in more efficient systems of care.

The investigators also found that independent primary care groups participating in the MSSP achieved significantly greater savings than hospital-integrated groups.

“Some have presumed that forming a large hospital system that owns a lot of outpatient practices is a prerequisite for ACO success,” McWilliams said. “We do not find this to be the case.”

One reason for this finding, the authors note, is that independent physician groups have stronger incentives to prevent hospitalizations than hospital-owned groups, since their shared-savings bonuses from doing so are not offset by foregone organizational profits from the reduction in hospital care.

Finally, the authors found that ACOs in the MSSP with high spending for their region achieved greater savings than ACOs with spending below the regional average. This suggests that ACOs with more opportunities to cut spending had an easier time doing so.

Recently, the Centers for Medicare and Medicaid Services proposed transitioning to a system in which an ACO’s benchmark would be primarily based on average spending in its region.

Because the participation of high-spending ACOs in the voluntary MSSP is particularly valuable for lowering overall spending in Medicare, the authors caution against moving to such regional benchmarks too quickly. Doing so could prompt ACOs with high spending for their region to leave the program, thereby diminishing program-wide savings.

“These early results are encouraging overall,” said McWilliams, who is also a practicing general internist and HMS associate professor of medicine at Brigham and Women’s Hospital. “But building on the initial success of ACO models in Medicare will require stronger incentives and rigorous evaluations to identify groups of systematically successful ACOs whose organizational models and strategies can be disseminated.”

This research was supported by grants from the National Institute on Aging of the National Institutes of Health (P01 AG032952 and F30 AG044106-01A1) and from the Laura and John Arnold Foundation.