My path from there led very clearly to family medicine but also a public health degree; working in three different community health centers, and having the pleasure of getting to help start one in Baltimore; and seeing all along the way the failings of our current payment models to deliver the kind of care that populations and patients needed in primary care. Then I had the opportunity to work with Bob while setting up that community health center in Baltimore to come to the Graham Center as a scholar, learn a lot from him and ultimately be a deputy director there, then the director; and now a co-director for our Center of Professionalism and Value in Washington D.C. continuing to work on these same issues.

 

JL: As I understand it, you spent the last few years at least, if not beyond, working on advancing a discussion on how to use payment and policy options for social risk adjustment. Can you take us back to the beginning of this work and share what was the rationale for doing it?

 

BP: This started for me back in the early 2000s and trips to the United Kingdom and New Zealand where I found that they were adjusting payments based on the social risk of their populations and using geographic measures of deprivation to do that. In both countries, they not only enhance payments to healthcare but to social services as well. It really winds up funneling more resources to the practices who are caring for those populations, so it makes it viable for them to serve those populations, but also there’s just intrinsic value in being able to identify and meet peoples’ social needs.

 

We came back and have written a series of papers over the last 15 years, and when Andrew and I were at the Robert Graham Center, actually developed a related deprivation index for that use in the United States. In January of 2021, we wound up having a workshop and publishing it as a Health Affairs blog about the real opportunities with demonstration projects through the Innovation Center at CMS [Centers for Medicare & Medicaid Services] to do this.

 

It led to the workshop series that we held last year really focused on Medicare and Medicaid policy but trying to influence all payers and health systems and thinking about how to do it. It’s been so gratifying to see CMMI [The Center for Medicare & Medicaid Innovation] and CMS come out now with three different demonstration models that use the deprivation index to do this very thing. We’re hoping that it can lead to this more broadly.

 

JL: Andrew, anything you want to add to that?

​

AB: It was global experience, similar to Bob, more in lower- and middle-income countries like South Africa, Honduras, and Bolivia where, despite far lower resources spent per capita on health care, you were supporting community-based organization primary care partnerships and had funding models, absent pure fee-for-service, that propped up interventions that would address social needs. Then with Bob working on a number of different indices, watching as we work through health professional shortage area process redesign, the ability for the policymakers federally to direct resources to get workforce teams into places where they were needed was not matched by the service side payments. In a time before, we were really talking about the social determinants of health or labelling them as such, recognizing that we needed simpler ways for payers to understand how to direct those needs. Then Bob has filled in the rest beautifully.

 

JL: Based on those discussions and convenings as you mentioned with policy and practice leaders focused on Medicare and Medicaid policy, can you summarize for our listeners: what are some key strategies of focus areas for implementing social risk for payment adjustment in Medicare and Medicaid? And to the extent there were strategies mentioned that would be more fruitful or less fruitful, could you share a few of each?

​

AB: I’d say after three convenings now, there’s never prefect consensus. But these were opportunities for leaders of major commercial plans as well as federal staff, folks from academia and beyond to get together and try to build consensus.

 

I think the general agreement that came out of some of the workshops included the notion that there’s a real gap in knowledge around and use of these small area deprivation indices, and the agreement that they’re ideal when you’re trying to adjust payments for social risk because they decrease this burden in data collection. It is so hard, speaking as a clinician, to reliably, consistently gather information from individuals about their social risk. As important as that is, it’s difficult. And it’s equally difficult to sustain it so that it can prop up a payment model, and especially for the most disadvantaged settings.

 

The small area deprivation indices: the group came together and agreed with big concerns about gaming, and [the need for] any kind of new payment adjustment to reduce that potential, and [small area-level indices] offered transparency for payers and providers because they were built on publicly available common data that they could all see. In presentations, there was an agreement that two of these small area indices in particular, Area Deprivation Index [ADI] and Social Deprivation Index [SDI], were felt to be particularly reliable for the most at-risk patients, and that’s been shown in a number of different studies. Again, you still had challenges and the tradeoff – particularly for disadvantaged persons living in lower risk areas that were identified by these ADI or SVI indices – that you might actually miss those living in those spaces, but that was a tradeoff that would have to be wrestled with in other ways.

 

Specifically, the group addressed for Medicare that we would have to maintain budget neutrality, but that funding adjustments would make necessary the need to look at how you could build out infrastructure and benefit dollars. If you looked at models, there was a particular fondness for the CMMI HEART payment model’s inclusion of neighborhood social risk and comorbidities. There was a lot of conversation about the benefits of the current Maryland all-payer HEART payments; and that even as we were working on the payment side, that clinicians and teams in the health systems should already be trying to adapt their practices to address social needs directly and in partnering with community benefit organizations. Bob, I’m going to let you jump in; there was a lot more said in those early conversations.

 

BP: I think one of the things that the group acknowledged fairly quickly was that the people most in need of this support are least trusting of healthcare systems and payers, and are not as willing to share their social needs. And the people who are also at most risk are those least likely to come in for health care and be screened for social needs. So we wanted a process that would be highly reliable and that would apply to the entire patient population that a practice was taking care of, with the risk of some false negatives, people who live in more affluent areas but have social needs. We’re a little bit comforted by the fact that those folks tend to have better life outcomes regardless just because of their neighborhoods. The other [concern] is that people with high social needs, their lives are in such flux that if they have housing this month, they might not next month and relying on the healthcare system as the process for screening them and assigning them services is just not very reliable either.

 

As Andrew alluded to, the model that CMMI rolled out in Maryland, the HEART payment model, which is focused on health equity, combined the Area Deprivation Index with medical complexity scores. A study led by Kate Sapra and Steve Cha (HHS) that showed that that combination of people who were in the top quintile of both ADI and medical complexity had 24% higher health care costs down the road. That gave them the authority to build into this model.

 

The group also pointed to the Massachusetts model which has been operating now for five years, not so much for how the payments are adjusted but for how the funds flow. There’s a lot of concerns about just putting more money into the healthcare system, so the thing that Massachusetts did – that the convenings enjoyed – was the fact that they created a half billion-dollar pot of money. It’s outside of the health care system, but that physicians can basically write prescriptions against. So if a patient has needs, they can write a prescription that gets them into a community-based organization and brings resources to that community-based organization to meet their needs. Those two models were the ones that people would like to see expanded.

 

The last thing I’d add is that, with the growing interest in adjusting payments for social risk, there is an inclination to create proprietary measures. So we could see an arms race coming where every health system developed their own, every managed care organization developed their own, every Medicaid program vendor developed their own and sold that back to the health care system, and no one was willing to work off of the same indices. In primary care, when we have an average of 9 to 13 different payers, we just couldn’t afford to have 9 to 13 different ways that we were being paid to adjust risk. So we really just wanted to blow that out of the water and come out with a declared gold standard in an effort to develop even better so that there was a level playing field.

 

AB: Bob, I think we’d agree that since we broke the two convenings and conversation into a Medicare- and Medicaid-focused conversation, you better believe that came up in the Medicaid conversation. We really can’t afford as much as 50 states innovating drive some good in the health system, in this instance we can’t afford to have 50 different approaches, particularly with modern movement across state lines, to social risk payment adjustment.

 

To double down on what Bob said, it came up on the Medicare side, how Medicare Advantage had a particular opportunity to fund both medical and non-medical providers alike through their pathways – Massachusetts really highlighting the way. And this relationship – there should be a curvilinear relationship with how you do adjustments, not just thresholds. You really see with increasing need remarkably different resource requirements to achieve health.

 

BP: To that last point, one of the questions that was a little bit unresolved, other than to say it should be curvilinear, is how much do you pay? How much do you adjust payments? A lot of the models so far have been adjusted based on downstream costs, rather than on what does it actually cost to put someone in a house? What does it cost to feed someone?

 

We had some preliminary information for the convenings, but we have another study that puts a price point on it: what does it cost for the average patient to meet their housing needs, transportation needs, food needs? How much comes from federal payers and how much needs to actually help support the practices in managing that effort? We’re glad that those things are coming together so if we make decisions with other payers that the price point can be right too. 

 

JL: I think what I’m hearing in some ways is if we accept the idea that we should adjust payment somehow for social drivers, there are different ways to do it; the individuals and populations who perhaps most need these data to be captured are least trusting; those groups and organizations that take care, the so called safety-net, often are not resourced, so if you put the burden on clinicians and organizations in some implicit way you may be biasing away from the groups that really need to capture that information. All that said, [some] people would say, don’t you think that in the ideal world we’d have individual level data on everybody?

 

Do you see a world in which we use these small area indices alongside individually captured? How would you guide someone who says [area-level indices are] great because it’s pragmatic, it’s universal, we can use it now, but don’t we want to go to something better in the future?

 

BP: It’s a fair question, Josh, and it’s one that came up in the convenings. Our point is, you need a process for identifying people at high-risk and for allocating payments sufficient to meet those populations’ needs. You need individual data to figure out what does this person actually need? What is the thing that needs solving for them and what do we allocate resources to?

 

The small area indices we think are helpful for giving a practice enough resource, and then they screen the patients individually, and they can then use that resource to meet specific needs and to partner with community-based organizations to achieve that. So, it’s a “yes – and.” What we came out pretty clear with was a consensus that we shouldn’t use those individual level need measures as the mechanism for adjusting payments, but the mechanism for making the right prescription.

 

AB: I would just really emphasize the “and”. It is too easy to get into debates of individual versus area. They both offer so much, to your point, to a future ideal state of really understanding need. As clinicians, which the three of us are, we have to have individual data to really tailor any sort of resources or funding that come through an ecological adjustment, really down to the patient.

 

Also, there’s precedent for “and”. When you look at the ways health professional shortage areas are designated – there are geographic designations, there are population designations. One makes the case for the latter, or even point designations, based on the particularly small information either about individual groups or populations. This has been done, and in the age of big data, we’re only getting better at capturing multiple points for the individual. I am optimistic that we’ll have better capture going forward. But I still think “and” will exist for the reliability, transparency, and reduction in game-ability that the small area indices still offer.

 

JL: Let’s go one level up. More recently, I’ve heard this line of thinking that if we try to adjust but we fix it on the downstream costs, you end up codifying things like historical disparities and care, barriers to access. You may in the pursuit of precision in your adjustment and predicting costs actually end up predicting things that don’t actually service equity. How are you thinking about this issue?

 

BP: The first is if we adjust payments based on downstream costs, then you’re risking giving inadequate resources to the clinician who is screening those patients but then cannot actually meet their need. That’s the real concern is that you’re going to take [clinician] burnout and turn it into something even worse because now I have an obligation, now I am accountable for screening for these needs, but I still don’t have enough resource to meet them.

 

The second issue is, and this is what the managed care organizations in the communities really taught us, is that right now underserved areas suffer from the fact that pricing of services in those neighborhoods is based on historical expenditures. Historical expenditures are based on inadequate access, so you wind up in this vicious cycle of underpricing underserved areas and managed care organizations don’t want to go there. Practices can’t afford to locate there. So, one of the other benefits of doing this adjustment based on small area deprivation indices is that you actually create new markets. You create incentives for people to care for those populations because they’re going to get more resource. So I think that actually, to the most pressing issue that you raised of not solving the equity issue, actually may turn out to be one of the most potent products of this.

 

AB: The only thing I would add, Josh, is, the problem you mentioned, it’s not a uniquely U.S. problem. But the fragmentation of our payers and the absence of patience that comes with it only heightens the emphasis on cost and historical-to-current cost comparisons. It does take patience to imagine investing with an accountability to the fidelity of fund flow, to the community-based organizations, to the clinicians, and an acceptance that you’re playing a long game where you’re going to see real changes. And frankly, since you’re using small area indices, you’ll see small area outcomes hopefully start to rise with patience. But it’s understandable that payers who regularly see their clients, their beneficiaries, move from one to another payer, struggle to move beyond the cost-based evaluations only.

 

JL: Whether it’s through your work and international context or in these convenings, are there any things that people who are interested in this can look ahead to; things where you can anticipate a little bit of turbulence as we implement these things?

 

AB: Where whole health systems are working currently, it is those with more of a national approach. And even a state like Vermont, and its blueprint for health. It took a long game approach to addressing social risk and supplying more resources to connect particularly foundational primary care health care side elements with community-based organizations that are meeting social needs; and bring them together and provide the superstructure and support that started to see differences, but differences measured over time. Frankly, I think the most overt comparison I’d make now – at best a low- and middle-income nation – Costa Rica, which now has greater longevity than the United States despite a fraction of the investment in health care because it has really played the long game in addressing local and social risk and building up primary care infrastructure. But not just primary care in the medical sense, primary health care infrastructure: combining public health, primary care, community-based organizations working together to enhance that longevity. Singapore also has passed us in longevity. I won’t say that it has achieved primary health care at its highest, but one of the things that they have long played is that long game. We’re working for decade to decade changes in population health not just what we can achieve in cost savings over the next year, if that’s our measurement of success.

 

BP: I think there have been some lessons from the [CMMI] demonstrations. The HEART payments that started a year ago in Maryland. One practice got a half a million-dollar check in the first quarter of the year and their question was, “what am I allowed to do with this?”

 

So there’s a need to help practices, even once you hand them a check, to understand: this [payment] is so you can hire community health workers, this is so that you can hire behavioralists, this is so you can work with that foodbank in your community to be able to refer patients.

 

There’s a need to help the community, too. A lot of these community-based organizations are starved of resources, so having them inundated with new patients is not so helpful. What the Community Care Cooperative of Massachusetts (a managed care organization of 18 different federally qualified health centers) taught me is that you also need an IT infrastructure.

 

So I recognize this patient in front of me has a need of housing. It’s difficult for me just to send them down the street to the community-based organization that can give them a housing voucher. So they actually built an IT platform that makes that referral electronic and automatically pulls down the resources, so that when the patient arrives, the community-based organization has the money, they can address the need, and communicate back to the practice that the patient showed up; this is what we did for them; please reinforce this at their next visit.

 

That small managed care organization has pulled down 80% of the funds in that bank in Massachusetts to meet social needs, because they have created this IT platform that has helped those community-based organizations flourish. So there are a number of things that will have to flow on from adjusting payments, and we can’t expect these practices to suddenly start solving social needs without that kind of support.

 

JL: I’m so glad you mentioned that because once you reallocate the funding and the resources, there’s all the care delivery and the community-based touch points and services. What do you think is the pragmatic, maybe optimal, way of figuring out what works there?

 

BP: My best example of this is when the Agency for Healthcare Research and Quality launched EvidenceNOW – research designed to help practices on what is the coaching, what is the support that a practice needs to change its operations. [The Agency] also funded an evaluation program that came alongside it called ESCALATES that was run out of the Oregon Health Sciences University. That parallel process of not just evaluating, but learning what is it that practices are struggling with, what are the innovators doing, it actually helped make that whole process work better and gave the agency not just a sense of “did this work?”, but “what does it take to make it work?” I thought that was just such a brilliant example of how evaluation can actually improve implementation. And I would love to see some of these demonstrations have that kind of parallel evaluation facilitation function, so that we don’t just deem them failures because it takes 3 to 5 years normally for these things to set up, and they don’t have that level of support.

 

AB: I would only add, we don’t historically have the patience to build continuous learning systems out of stacked pilots, even with the wonderful work that CMMI has been doing. We rarely see continuity over multiple pilots and learning. In primary care, we had the good fortune of having CPC+ become CPCI, so that gave you a little more length of time. And to Bob’s point, it really showed in some of the areas where the 1-, 2-, and 3-year evaluations weren’t showing the kind of utilization cost outcomes success that were perhaps hoped for; that with the few more additional years and the addition of qualitative audits, really going into interviewing the clinicians and the patients, who despite not showing some of those early cost utilization metrics, were [showing] this is really working. Having one set of benchmarks, having the kind of support superstructure and the ability to learn from each other, is really working and helping. And back to something we lack historically – our patience – it takes building out that continuous learning system model. Pilots have to be part of something bigger, so that you start to stack together the lessons from each and have tracking over time.

 

Then finally, your accountability mechanisms. You have the ability if you adjust payments, as consensus emerged from our panels and our groups, [for accountability over] how resources are spent, which can also yield great information on the variability on how those resources are spent that helps to lead, when you pair with claims data and the qualitative side, to a better understanding of how one system and its unique way of doing things can inform another; how those are scalable or not scalable. You can use the accountability mechanisms to feed the evaluation mechanisms too.

 

JL: If some of our listeners want to learn more about this topic, either foundational reading on this, what’s more recent, or they want to learn more about the convenings, what are some things that they can check out?

 

BP: Josh, I would point back to some articles that we published in Health Affairs and the American Journal of Public Health that lay out this argument and point to other countries as examples of what we’re talking about, and how to adjust payment for social risk and the importance of it. We’ll be sharing links to those with you.

 

The initial convening, we had a Health Affairs blog that came out in June of 2021 that I’ll share with you that sets up the conversation for the workshop series. Back in January, we had a Commonwealth To The Point blog that came out to summarize the workshop and then a more complete summary that came out in the Health Affairs Forefront. There’ll be a JAMA Internal Medicine piece that sets up the “how much” to adjust payments in order to meet needs [published since the interview was recorded].

 

AB: I would just add specifically to look at the November 2016 Health Affairs in that how other countries use deprivation indices and why the U.S. desperately needs one as a foundational step to understanding particularly the UK and New Zealand approaches. For the technically curious reader, the University of Wisconsin, the creator of the Area Deprivation Index, and the Robert Graham Center, which created the Social Deprivation Index, both have dedicated web pages where if you want to dive in and understand both the strengths [and] the weaknesses involved in using small area deprivation indices to see what they’re built around would be a great resource.

 

BP: With Stanford University and the U.S. Census Bureau, we have a project that we anticipate over the next two years will produce one or more gold standard social deprivation indices at very small areas with Census set up to be the steward of those. Again, to try and prevent a flood of proprietary measures from coming into this space.

 

AB: With Robert Wood Johnson’s support, that we hope will broadly help us disseminate this work.

 

JL: Bob Phillips, Andrew Bazemore thank you again for joining today. More importantly, thank you for the work you have done and you’re continuing to do in this space. I really enjoyed this conversation and I trust our listeners will as well.

 

BP: Our pleasure, thank you for having us Josh.

 

AB: Thanks so much Josh.