In Conversation With… Paul McGann, MD
Editor's note: Dr. McGann is the Chief Medical Officer for Quality Improvement at the Centers for Medicare & Medicaid Services (CMS). He also serves as the Co-Director of CMS Partnership for Patients and the Co-Director of the Transforming Clinical Practices Initiative. We spoke with him about his work at CMS and with the Partnership for Patients initiative.
Dr. Robert M. Wachter: How has your background, first as a geriatrician and internist and second coming from Canada, influenced your thinking about the programs that you help run?
Dr. Paul McGann: All of the various places I've been and worked certainly have influenced my thinking about how we can go into the future with a better health care system and a safer health care system. My previous experiences, I would say especially at the Sticht Center on Aging at Wake Forest University, really showed me about health care transformation and started to give me hints about the safety issues in our country. The experience at the frontline, especially in different cultures, helped point the way to what's important and what to work and focus on to generate value and quality. When you work in different cultures, it enables you to see that processes are brought to bear—and customs, habits, and cultures—don't always contribute to value and quality. In the late 1990s, I started to become aware that better quality and coordinated care is actually less expensive than its opposite, which was kind of counterintuitive, at least at that time. Better quality care is more efficient, higher quality care. We were trying to change systems back in the 1990s without labeling it as such. The maturity in the field over the last 10 years is that these are now explicit goals, and it's widely accepted that better quality and better coordinated care is actually less expensive than its opposite.
RW: Tell us about the Partnership for Patients. What's the theory behind it and how is it structured?
PM: Let me talk about the terminology. HEN stands for Hospital Engagement Network, and that was the contracting vehicle by which we executed the main engine or lever of the Partnership for Patients. But thought of broadly, the Partnership for Patients was much more than HENs. It was a nationwide public–private collaboration. It was launched in April 2011 at the US Department of Health and Human Services. The aim was hospital safety. Writ large, we were answering the question: What could we do at national scale to make hospitals safer quickly? We came upon two goals to define what we meant by safety: reducing all-cause patient harm (all payer in all hospitals) and reducing all-cause 30-day readmissions.
We spent quite a long time calculating what we felt was best-in-class performance at the time from the medical literature. That's where we came to the best goal statement that I've ever worked on in quality improvement, which is 40/20 by 14. So we committed, compared to a 2010 baseline, to work to reduce all-cause preventable hospital-acquired conditions by 40% and all-cause 30-day readmissions by 20% by the end of calendar 2014. I realize now, even more than I did then, how powerful it is to make a commitment like that at a national level.
There are three principal engines by which we can achieve those goals. The first is to form powerful partnerships at the federal level. These involve reaching out to the various levers and arms of influence in the federal government, particularly the Department of Health and Human Services but not necessarily limited to it, and asking our colleagues: what can we do in our daily work that would turn energy and commitment and resources to achieving the 40/20 aim?
Then a broader set of partnerships outside the government—this is where the public–private partnership comes in, where we would reach out to other organizations that could influence patient safety, some of them quite surprising, that would join with us to achieve these aims. In the end, we got more than 5000 partners to sign up. One of the most surprising measures was to reduce the rate of early elective delivery in this country. The two principal outside partners for that one were the March of Dimes and the American College of Obstetricians and Gynecologists. Some tremendous results happened. But there are many other partners from the US Department of Veterans Affairs, to APIC [Association for Professionals in Infection Control and Epidemiology] and SHEA [The Society for Healthcare Epidemiology of America] in the hospital-acquired infections category, to the AFL-CIO. The American Nurses Association joined with us. The patient safety movement, Blue Cross Blue Shield? the list goes on and on.
Finally, as I mentioned at the beginning, we were also able to provide direct funding to entities that felt that these goals were worth the intense effort required to achieve them. We funded 26 hospital engagement networks directly; their goals were to work as hard as they could with their resources to make those goals happen in the vast majority of hospitals in the United States. So those were the three principal engines.
RW: That goal statement and the numbers were crucial in many ways. I'm trying to imagine myself as a fly on the wall in those discussions. It was an era when some of the early data about patient safety was somewhat disappointing. The needle hadn't moved very much. I imagine you wanted ambitious and audacious goals, but I also imagine you didn't want to fail. What were those discussions like?
PM: That's a really perceptive question. It took us more than 6 months, a group of us that were doing this while we were trying to do our main jobs. I was Deputy Chief Medical Officer of the Agency at the time. We got people together from the CDC, AHRQ, the Office of the Assistant Secretary of Health all in HHS, and several components in CMS. People who knew about safety, knew how to measure it, and knew about the payment incentives that were coming from the Affordable Care Act. We met in very intensive ways. Many of us were familiar with the literature, and we started pulling out the studies individually to look at this, in many cases harm by harm. We realized that the best way to encapsulate this was to try to express it as a single harm rate per thousand discharges. We felt we had the ability to do that with some of the datasets in the federal government, even though it hadn't really been looked at with these datasets in that way before.
One of the things I know you're interested in is what can the federal government in particular bring to this field? The first way is that we have access to datasets that no other entity has. In this case, we decided to turn some of those resources that many of us were already working on for other reasons—for example Hospital Compare and hospital public reporting at CMS—into a deliberate and systematic attempt to calculate the number of all-cause harms per thousand discharges. We also knew that there were varying estimates of preventability of those individual harms. That's why we put the very important modifier in the goal that we were going to achieve a 40% reduction in preventable hospital-acquired conditions. That concept of preventability changes over time, so we knew that was a moving target. But we felt it was important to acknowledge what providers tell us, and that is that there are some harms that appear to not be preventable, and we wanted to make it clear that we were aiming to reduce preventable harm.
The 30-day readmissions? Well, in my field geriatric medicine, multiple 30-day readmissions is a harm. We're in a unique position at Medicare to measure those readmissions. I also had the advantage of being very much involved in three QIO [Quality Improvement Organizations] statements of work that created this as a priority. We also were aided by the Affordable Care Act, which instituted the readmissions penalty. So readmissions became a very special part of our discussions. We had been involved in several pilot studies and knew people like Eric Coleman and Mary Naylor who had been already working and publishing in this field. So we thought we had a pretty good handle on what was possible with concerted effort.
Now the big proviso is that I've had experience in large-scale quality improvement collaborations in the nursing home field. That's where I got my start at CMS, back in the early 2000s. There are 16,000 nursing homes in the United States, and I did several projects that had 7,000 to 10,000 nursing homes in them. I had learned from that experience that just because there had been a randomized controlled trial involving a few hundred patients didn't mean you could necessarily reproduce that study, multiplied by 1,000 or 10,000. This issue of scaling is a really important concept that people who are interested in national improvement need to think about because it isn't as simple as you just multiply it by some factor number. Scaling comes with a lot of serious perturbations in how the interventions can work. You have to attend to those to be successful.
RW: The more general experience is that things often don't scale. They worked in a few places because of charismatic leaders and the Hawthorne effect; then you try to scale them to 10,000 nursing homes and it doesn't work. What were the steps that you took to make sure that this would scale?
PM: I'm very glad you asked this. I actually was able to distill what we believe the elements of that were into seven words and one action. Let me give you the seven words: goals, partnerships, alignment, culture, leadership, flexibility, bidirectionality—that last one is particularly important at CMS. And then my action, which kind of hits the essence of the answer to your question, is it's important to do many, many things all at the same time. That theory of action I just described in less than a minute is a lot easier for me to say now than it would have been for me to say in 2011. We weren't exactly sure when we started this journey how it would work. But we certainly learned a lot as the project unfolded.
RW: What does bidirectionality mean?
PM: Bidirectionality, from the perspective of the federal government and agencies I worked at, is the concept that although CMS policies, including payment policies, are very important to outcomes, structure, and how a system works, it's equally important that program policies and even data come in the other direction.
When I was in practice in the 1990s, it always seemed that information was going in one direction—that is from the policies and payment regulations of the Agency—down into the field. As you engage in large-scale national quality improvement collaboratives, like the Partnership for Patients and the Transforming Clinical Practice Initiative (TCPI), I believe that bidirectionality is crucial. So this idea that we will come to the table in good faith; we will put down the data that we have access to. We will work very hard to reduce the lag time and the turnaround time of that data. But it's equally—and I'm starting to wonder if it might be more—important that the providers that we're working with in quality improvement bring their data to the table and teach us what they know and what they see from the data systems that they're increasingly involved with as meaningful use catches on and EHRs and health IT spread around the health care system.
So that's what we mean by bidirectionality: that the old tradition of the Medicare demo, where we measure it using our claims database and the results are the results, was met in the Partnership for Patients—I would say halfway—by a bidirectional flow of data, information, and learning. Together, if both of those directions can be considered simultaneously, the ability to move things more quickly is enhanced.
RW: That's really impressive. I mean it takes a tremendous amount of humility and openness to what the lessons are from the field, and I'm not sure that was always the core structure of the agency. Having talked to Mary Dixon-Woods and Peter Pronovost about how Michigan worked, it strikes me that you've scaled up those lessons as well. How do you take advantage of things like checklists and structures but also change the culture at scale?
PM: I want to couple that with this notion, and again this is a unique characteristic of the federal government, CMS in particular, of the alignment of multiple payment incentives, both for additional funding and also payment penalties, various survey, certification, and accreditation guidelines. I would extend that now to subspecialty accreditation, maintenance of certification, and alignment across many different components of the federal government—all in service of simple-to-understand aims, like 40/20 by 14. That multiple-lever approach and multiple areas of alignment all in the same 3 years, it's incredibly powerful. Now it's very hard, as we're learning now in the evaluation phase, to evaluate that, to assign attribution to that, and I would call that actually the biggest challenge of my professional life. I understand it, but to measure it, to prove it, and especially to calculate it statistically is a huge, huge challenge. I now understand why randomized controlled trials are designed the way they are—because they're much, much easier to evaluate.
RW: Sure. But they're also designed to evaluate easier problems at nowhere near the scale. Let's talk about the results first and then we'll get into measurement. So you had these ambitious goals, and now we've passed 2014. How well did you do?
PM: The evaluation is still in progress. The AHRQ National Scorecard was the bedrock of this. We realized that the data from the field was going to be hard to analyze. We realized, as Peter Pronovost has pointed out, that it wasn't standardized and that it really wouldn't by itself withstand the scrutiny that this project would have. Many months were spent before this was launched developing the AHRQ National Scorecard and the ability to express harm as number of harms per thousand discharges. I think we did a pretty good job on that. It's not perfect, but we've been measuring it.
The baseline year was 2010 when none of the Partnership for Patients interventions were in the field, and the initiative had not yet been announced. We also have national harm rate estimates now for 2011, 2012, 2013, and preliminary from 2014. We have analyzed with the 2010 baseline and through the end of calendar 2014, and this is a nationally representative sample of hospitals, harm rates done by chart review. Harm rates were calculated by clinical data abstraction from approximately 30,000 charts reviewed each year. It's a mammoth effort to estimate patient safety with a nationally representative sample of hospitals. The results showed a decline in the harm rate per thousand discharges from the baseline of 145 harms per thousand, to 121 harms per thousand by the end of calendar 2014. The results are available now on the AHRQ website. And, harm by harm, we've calculated what the impact of that would be if that nationally representative sample was true nationally in approximately 5000 hospitals. The result would be a correlation with about $19.8 billion of savings, compared to the expenditures predicted to have occurred if the 2010 harm rate had continued unchanged over those years (2011–2014). And an estimated 87,000 lives saved, with a total of 2.1 million adverse events avoided. So that is the initial estimate of national impact, without a statement of specific attribution. The next phase of the question is—if this is the estimate of the nationally observed improvement in patient safety in short-stay acute-care hospitals, what caused the improvement? That's the research question that people are working on now.
RW: When you read the Pronovost and Jha critique, did any parts of that ring true and make you wish you had done something differently with the program?
PM: That's a fair question. After we get the final results of the evaluation, I'll be in a better position to answer. We think the strategy that we had is robust, given that we stated goals that we wanted to be large scale and national. We wanted to go fast. We wanted to involve as many hospitals as possible. It's obvious from the enrollment numbers that we achieved large enrollment. We got 3700 hospitals to report regularly on multiple causes of harm, up to 11 in a hospital depending on how many applied to them, every single month for 3 years. I don't think that's ever been done before, and we made a stated goal when we started this project to achieve that result of enrollment and engagement. So that was successful.
I believe a critical part of that success was the flexible approach to measurement that we had, as opposed to a less flexible or even 100% standardized approach that Peter was advocating. So we could have had 100% standardized approach; many people have done that before. But we wouldn't have had 3700 hospitals reporting data on harms every single month for 3 years if we had done that. So it's a tradeoff. We were going for large-scale and quick results, and we allowed flexibility. I think that strategy worked. That having been said, it is true in terms of calculating and collating the results that a more standardized measure structure would have made that easier.
Now I do want to point out something at this point, and that is some people have come away from that New England Journal of Medicine opinion piece with the idea that there was no standardization of any kind. That is an inaccurate view. For example, for the hospital-acquired infections, when possible, NHSN [National Healthcare Safety Network] data were used. The NHSN website and data collection system and its measures are as standardized as they come. Thousands of hospitals, in fact in some states it's mandated, use that for things like central line–associated bloodstream infections, catheter–associated urinary tract infections, and surgical site infections. So to say that there was no standardization in the Partnership for Patients is to not understand the measurements. That having been said, we have more than 500 different types of individual measures across the 11 types of harm. So there is a lot of measure diversity in this project, which does make counting all the results from all 3700 hospitals difficult.
The Partnership for Patients 2.0 statement of work does have a suggested standardized list of measures. Another very interesting thing happened. As the community of practice worked together in a learning network, we observed trends toward measure standardization over time. We intend to measure these trends and to report on them and ask the question: Are there things we could do to accelerate spontaneous or voluntary measure standardization over time? In the United States with 5000 hospitals, if you look at how people do quality improvement and work on patient safety, the fact is it is not standardized, and people do use a lot of different measures. We're starting to run on the idea that large national collaboratives like this can actually serve as a motivator and an accelerant to measure standardization, which would make future evaluations of future patient safety efforts, not just by us but by others, easier as well. This is another example of that concept of bidirectionality that I was talking about.
RW: As you think about the new version of this over the next decade, large-scale efforts to improve care in safety or in other domains, the landscape is very different, the payment mechanisms are moving swiftly toward value. We're thinking about populations, there's more system-ness in the health care delivery system. And now nearly everybody has electronic health records, which was not true at the start of this program. As you look at these trends, how does that change your thinking and the Agency's thinking about the way to structure these programs going into the future?
PM: That's an excellent question. What I'm seeing and HHS actually went to this space and used one of the words in my list of seven words, and that is establishing goals. Not in a direct quality sense but more in how the payment system should evolve. Secretary Burwell's January 2015 NEJM perspective discussed the new goals to move Medicare payments into alternative payment models and away from fee-for-service such that we get 30% of payments in alternative payment models by the end of 2016—a big goal in a short period of time—and 50% of payments into alternative payment models by the end of 2018. On March 3, 2016, CMS estimated that the first ambitious target (30% of Medicare payments tied to alternative payment models) was actually hit in January 2016—11 months ahead of schedule. Also released the same day was an overview of the 10 alternative payment models that contributed to progress of the 30% goal. I believe that quality programs and other reduction of harm programs, like the Partnership for Patients, more and more are going to start to align themselves with these system and payment goals so that they aren't as independent of payment systems as they've been over the last many years. And this will mean greater financial rewards for those providers who excel at quality and safety in the future.
RW: In a world of Kaiser Permanentes, Cleveland Clinics, and Geisingers, does the role of the federal government to establish the collaboratives and learning networks go down because these large systems have internal competencies to do that? One of the points you've made that's really quite impressive is that part of the goal of federal involvement is that bidirectionality. The fact that you were involved in making this work happen and facilitated it, that it provides an information stream back to you that informs what you do in payment policies and other structures that you are uniquely set up to organize.
PM: Exactly, and of course it's a big country out there and there are Geisingers and Kaiser Permanentes, but there are also thousands of rural providers and providers in underserved areas. I know there is a tendency for them to come together into learning networks now and that's good. We applaud that. But for the foreseeable future, there are enough small rural primary care providers and providers in underserved areas that there will be a lot of opportunity for the federal government to help them with large learning collaboratives such as we're doing now. And in fact, with TCPI [Transforming Clinical Practice Initiative] we made that a priority. We wanted each application to come in with at least 20% of the clinicians in their proposal to be from sources like those that would have a hard time getting that help in any other way. And don't forget—CMS is now a major national funder of innovative new payment models to improve quality and reduce cost, through the CMS Innovation Center, established by the Affordable Care Act in 2010. So, in much the same way that NIH leads the way in basic science research for the country, CMS has now established itself as a major force in developing newer and better systems of reimbursement, and in evaluating the impact and efficacy of those systems. Commercial insurers are now starting to adopt some of the new models that CMS has developed, and that will remain an important role for CMS for the foreseeable future.
RW: Any other things that you want to talk about?
PM: If there was one thought I could leave with your readers or listeners, it is that this concept of alignment is more powerful than I realized when we started it. When you have 10 or 20 or 30 government agencies and private organizations all working toward shared aims and aligning programs that they do anyway, without necessarily changing the funding, it is very powerful. For example, the American College of Obstetricians and Gynecologists had no funding of any kind involved in the effort to reduce early elective deliveries, they just aligned what they were doing and saying with their members to what we wanted to achieve and put that energy to reducing early elective deliveries. It's amazing what happens when you can align 10 or 20 levers like that over the same period of time in terms of going toward a common aim. The single thought I would leave with your readers is that alignment is a good thing, and synergy is an even better thing. To the extent that your readers can get involved in these projects where there is alignment and synergy, I think they'll see a multiplicative effect on the results and the impact they can have.
Previous interviews can be heard by subscribing to the Podcast