Social Determinants of Health: The Role of the Health Plan

  • This page as PDF


Tune into our fourth episode in the Avalere Health Essential Voice podcast series focused on social determinants of health (SDOH). In this segment, our experts discuss what health plans should know about SDOH data, specifically, the different types of data, what to do with them, and how to use them to fairly assess the impact of social risks on health outcomes.
Please note: This is an archived post. Some of the information and data discussed in this article may be out of date. It is preserved here for historical reference but should not be used as the basis for business decisions. Please see our main Insights section for more recent posts.
“Not adjusting quality measures for social risk factors fails to address racial inequities in healthcare. If anything, it makes those disparities worse.” Christie Teigland


Natascha Dixon Edelin , Principal, Associate Immersion Program

Natascha Dixon Edelin partners with clients to navigate dynamic market changes, understand their strategic opportunities and implement innovative payment and delivery models.

Guest Speaker
Christie Teigland , VP, Research Science and Advanced Analytics, Inovalon
Christie Teigland, PhD, leads the design and implementation of studies focused on comparative effectiveness, predictive analytics, and performance measure development and testing.

This interview was originally published as a podcast. The audio is no longer available, but you can read the transcript below. For updates on our newly released content, visit our Insight Subscription page.


Natascha: Hello and welcome to another episode in the Avalere Health Essential Voice series focused on social determinants of health (SDOH). My name is Natascha Dixon-Edelin and I’m an Associate Principal in the Center for Healthcare Transformation at Avalere.

I am joined today by Dr. Christie Teigland, who is a Principal in the health economics and advanced analytics practice at Avalere. She leads studies focused on performance measure development and evaluation, comparative effectiveness, and predictive analytics. Christie serves on the National Quality Forum (NQF) Disparities Standing Committee and is co-chair of the NQF Scientific Methods Panel that evaluates measures for NQF endorsement. She is Avalere’s resident expert when it comes to analyzing SDOH data. Christie, it is great to have you here. Thank you so much for joining me.

Christie: Natascha, I am happy to participate in this important discussion. I’ve been focusing on the impact of social risk factors on health outcomes for more than a decade and it’s one of my passions. This is a timely discussion because this issue is getting more attention than ever as health disparities have come to light during the COVID-19 crisis.

Natascha: While it’s timely, the topic of social risks and needs can be overwhelming. We are often asked, where should health plans start? At Avalere, we start by focusing on data.

During today’s podcast, we’re going to walk through what health plans need to know regarding SDOH data, specifically, the different types of data, what to do once you have the data, and how to fairly assess the impact of social risks on health outcomes.

So, Christie, there’s mounting data documenting the link between health outcomes and social determinants. In fact, a Robert Wood Johnson Foundation study highlights that 80% of the factors contributing to our health outcomes are related to social determinants. You can break that 80% down into social and economic factors, personal health behaviors, and physical environment, which means that only 20% of health outcomes are directly impacted by clinical care. We’ve heard these statistics before. Why is it important for health plans to understand the social determinants impacting their beneficiaries?

Christie: Well, Natascha, it’s not only the fact that there is evidence that social risk factors have a huge impact on health outcomes, but also that there has been significant growth in health plan members with these social risk factors. For example, my team evaluated the socioeconomic profile of beneficiaries enrolled in a Medicare Advantage plan, which is the managed part of Medicare. Traditionally, people were enrolled in Medicare fee-for-service, but in 2020, there were over 24 million people enrolled in a Medicare Advantage plan. That’s a third of all Medicare beneficiaries.

We looked at how that population has changed from 2015 to 2017 and it is astounding. There was a 28% growth in Medicare Advantage beneficiaries who live in a neighborhood with unemployment rates above 11%. There was a 28% increase in beneficiaries who live in a neighborhood where 20% of the households live below the federal poverty level, and there was a huge rise in those with less than a high school education.

We also saw differences in the types of people enrolling in Medicare. We saw an 18% increase in individuals enrolling in Medicare under age 65 because they are disabled. That is also a population that tends to have other social risk factors. There was also growth in dual-eligible beneficiaries and growth in those who identify with a racial ethnic minority.

On top of this, we also see that these Medicare beneficiaries have more chronic conditions based on their comorbidity scores. These changes add up to greater risk of worse health outcomes in these populations and it’s going to have a big impact both on Medicare Advantage plans and on the Medicare program itself.

Outside of Medicare, the Census Bureau recently reported that in 2018, 38.1 million Americans were poor. That means 1 in 8 Americans lives below the federal poverty line, which is income of just over $26,000 a year for a family of four. We know these numbers are going to increase with the COVID-19 pandemic because so many people are losing their jobs and their homes.

These trends have major implications for the growing needs that will impact health plans and providers going forward. Understanding these population changes is going to be critical to managing patients in the months and years ahead.

Natascha: A great place for health plans to start evaluating their populations and changes to their populations is by analyzing the data they currently have. Can you walk us through the various types of data that health plans may have access to, and how they can use this data to identify their beneficiaries’ social risks?

Christie: One of the constant barriers to addressing social risk factors is the lack of data on SDOH. Health plans just haven’t traditionally collected this data, but there are other sources, though they all have some limitations. They’re also not standardized so that you can use them across various populations.

One of the obvious places to go is administrative claims data, and your EHR, your electronic health records data, which sometimes has this information.

People have been talking a lot about the use of Z codes from insurance claims to capture SDOH. Z codes are ICD-10 codes used to document social needs. They can capture things like low education, literacy, low income, inadequate housing, lack of social supports, and other social risks. Unfortunately, these codes aren’t used consistently.

In June, I conducted a poll of over 600 health plan participants and asked whether they use Z codes to capture data on social determinants of members. Seventy percent responded that they were unaware of Z codes. They didn’t even know Z codes existed. Only 4% said, “Yes, we not only use them, but we require the use of a selected set of Z codes to capture social determinants.”

One way that health plans can start to capture this data on member populations is to pick some important social risk factors and make sure their providers are capturing them on a consistent basis. I think most social needs data comes from health plans screening their members, either when they come in for an office visit, or through a survey of their membership.

There are some tools available from the National Association of Community Health Centers. The CMS Center for Medicare and Medicaid Innovation has the Accountable Health Communities tool. A list of all available tools is included below. But the problem with using these different screening tools is that the data aren’t consistently captured. They might ask the question differently or use different response ranges. For example, if my tool uses income buckets in $10,000, and yours uses $100,000, we can’t compare those. We can’t unpack that data and really make it comparable. There are some efforts underway to standardize the collection of social determinants data, but nothing’s on the immediate horizon.

Finally, there are some public and private data sources that researchers have used a lot and that health plans and providers are using. One is the Census Bureau data from the American Community Survey. There’s the Medical Expenditure Panel Survey, and others, included below.

One thing that concerns me about these public sources of data is the preciseness of that data. Census data linked at the five-digit zip code level, for example, only gives me 42,000 distinct groups in the nation. Those are pretty broad areas. Think of a five-digit zip code in Manhattan. We know that it has a range of people from very rich to very poor. What happens to the effect of social determinants? It averages out.

Even the American Community Survey, which is often used in published research, is at what they call the census block group level. There are only 221,000 of those.

Avalere uses a proprietary national database that allows us to link our member data at the nine-digit zip code address level. In contrast to the census block group, it represents 30 million neighborhoods. That’s an average of about five households. If you think about the four households around you, you look a lot like them. Research has even found that data aggregated at this near-neighborhood level is very predictive of health behaviors. So, there is data out there to be used and lots of ways to capture that data, but we have a long way to go.

Natascha: Understood. What’s so critical about this is that once health plans can link their members to social risk factors, they can then evaluate the impact those risks have on health outcomes. Let’s provide our listeners with a few examples. Humana did a study demonstrating that individuals with food insecurity are 50% more likely to have diabetes. Another study showed that if you’re lonely, you’re 64% more likely to develop dementia, and four times more likely to visit the ER.

Christie, during our recent Inovalon Client Congress panel discussion, you highlighted a University of Pittsburgh Medical Center (UPMC) study where they evaluated the impact of social determinants on Healthcare Effectiveness Data and Information Set (HEDIS) gaps. Can you tell us a little more about how a study like this could be a benefit to a health plan?

Christie: Sure. This is an interesting study because the researchers at UPMC selected a set of socioeconomic characteristics—poverty, less than a high school education, age, living alone, unemployed, and having no personal vehicle—and associated each with the likelihood of getting certain preventive screenings and tests that are important to HEDIS quality gaps. These are going to impact your quality outcomes and your performance measures, the likelihood you’re going to get your breast cancer screening, your colon cancer screening, or your diabetes care screenings.

They found that having less than a high school education was the strongest determinant. It impacted almost every one of those measures evaluated. The other factors also impacted not just one of those outcomes, but several. What this tells us is that looking solely at income or dual status, which is often done by CMS and others, is not enough. There are many social determinants that need to be accounted for if we’re going to fully understand patient risk.

Unfortunately, due to the lack of access to data that we’re describing, all too often we see measure developers, like CMS in their five-star rating system, focusing only on dual status as a proxy for low income, and not considering all these other important social risk factors.

Natascha: That’s a perfect segue into talking about the data you’ve used on some of your analysis at Avalere. Let’s talk about the analysis you’ve done that’s specifically relevant for health plans. We can start with an analysis we worked on together where a health plan wanted to evaluate the impact of income on health outcomes.

Christie: Yeah, this study was really eye opening for all of us because, through data analytics and using the granular data I was talking about on SDOH linked to the individual plan members, we were able to identify a group of non-dual-eligible beneficiaries. They didn’t qualify for Medicaid, but they were poor. These low-income, non-dual-eligible individuals had similar outcomes to dual-eligible individuals. In some cases, they had worse outcomes than dual-eligible individuals because they didn’t qualify for Medicaid, despite being poor. They didn’t have the extra benefits and support available to dual-eligible individuals.

This is a real-world example of why it’s critical to look beyond dual status as the only determinant of social risk. There are poor people who are right on that cusp who just didn’t qualify for Medicaid and they may have worse outcomes than dual-eligible individuals.

Just to give you a different perspective from the health plan perspective, we’re currently working on several projects with life sciences organizations who are now getting very interested in evaluating the impact of SDOH on health outcomes. We’re looking at patients with conditions like hepatitis C, severe mental illness, and liver cancer, and really trying to understand the differential outcomes based on social needs. It is going to help not only these drug manufacturers, but health plans and payers and providers determine how to best treat those patients, how to develop early interventions aimed at preventing adverse events like hospital readmissions, and ultimately improving quality of care.

Natascha: Those are some really great examples and I think they help our listeners understand what you can do once you have the data and you’re able to link it to the risk factors that impact your members.

So, given the impact of SDOH outcomes, it’s a bit surprising that the Assistant Secretary for Planning and Evaluation (ASPE) recently recommended in its second report to Congress, “Social Risk Factors and Performance in Medicare’s Value-Based Purchasing Program,” that quality measures should not be adjusted for social risk factors such as lack of housing, food insecurity, and those that we’ve discussed today. What’s interesting is that ASPE does not dispute that social risks impact health outcomes. In fact, they agree, and yet they do not agree that measures should be adjusted for social risk. Christie, what does this mean for health plans and for providers?

Christie: Well, this was a surprise to all of us in the quality measure world. Adjusting quality measures for social risk factors is critical if we’re going to have accurate and fair value-based payment systems, performance measurement systems like the CMS five-star rating programs.

Think about what we do now. It’s very common to adjust quality measures for age and for chronic conditions, like cancer and heart disease. Adjusting for social risk is needed to avoid unfairly penalizing hospitals, health plans, physician practices, and providers for patient characteristics that are beyond their control. We have plenty of evidence now that shows that they affect health outcomes just like old age or chronic conditions do.

What happens, Natascha, is that without this adjustment, providers that care for larger numbers of patients who have these social risk factors, and who might provide the same quality of care as providers who have fewer patients with those social risk factors, will look like they are providing worse care. They might, in fact, be providing exceptional care, but that’s masked if we don’t adjust for social determinants. So, bottom line, accurate and fair reporting on quality is essential if we’re going to use it for public reporting for value-based payment systems and for consumers to make the right choices.

I’ll take that a step further. Value-based payment systems that don’t adjust for social risk factors can actually perpetuate health disparities, can systematically increase inequities. Unfair penalties for those providers that care for the sickest and the most marginalized patients, such as hospitals penalized for high readmissions or a Medicare Advantage plan in the five-star rating system that doesn’t receive a bonus payment, can take resources away from the communities that need them the most. Not adjusting quality measures for social risk factors fails to address racial inequities in healthcare. If anything, it makes those disparities worse.

Natascha: Given that, can you help our listeners better understand the commonly expressed concern that adjusting for social risk factors will mask or excuse poor care? I know that you don’t agree, but can you help us understand that?

Christie: This is this has been a long-posed argument, and I do disagree with it. First, let’s speak from a technical perspective. As an econometrician who has worked on the technical aspects of quality measure specifications and testing, validation, and reliability for more than two decades, and serving on the National Quality Forum Scientific Methods panel, I can say there exist valid statistical methods to adjust for both clinical and social risk that do not adjust away real quality differences.

It’s always good to have examples. Let’s consider an adjustment for age in a measure of hospital readmissions. No one would argue that people who are aged 85 and older are more likely to be readmitted to the hospital. So, if you have a disproportionate number of patients who are 85 years and older, your performance is going to look worse, unless you adjust for age. That’s a given. We’ve established that. No one argues that adjusting for age masks poor quality, or excuses poor care. Providers aren’t going to suddenly provide worse care to old people because there’s an adjustment in the readmissions measure for age.

Why then, is it so different to adjust for lack of stable housing, or access to healthy food, or the ability to follow discharge instructions? These social risk factors also put people at higher risk for readmission. They’re not under the direct control of the provider, and without adjusting for those factors, we’re actually rewarding the providers that serve younger, wealthier patients at the expense of those who are serving the most vulnerable. Those providers serving younger, wealthier patients may be providing worse care to their patients, but that would also not show up unless we have these social risk factors included.

Risk adjustment models level the playing field. Again, let’s imagine we have two providers, and they’re taking care of two patients who look the same. They’ve got the same chronic conditions, they’re the same age, they’re the same gender, and they live in the same area of the country. They have all the same characteristics except one of those patients is living below the federal poverty level, and one is very wealthy. Now imagine those two providers give both of those patients the same quality of care, but the low-income patient ends up being readmitted due to factors relating to his or her living situation. They didn’t have a stable home to go to, or they didn’t have healthy food. Whatever the reason, if we don’t adjust for the added risk of living in poverty, we’re likely to think that that low-income patient got worse care, but that’s not true. So, without adjustment, the measure is not accurately reflecting the quality of care that patient received.

Now, let’s assume those two patients are identical in every way. They even have the same social determinants of health, but one of those providers does in fact provide a lower quality of care. If we adjust for all the characteristics, including the social risk factors, that poor quality is going to be reflected in their quality measure rate. These are what these statistical models do. The provider giving worse care is going to have a lower score than the other provider because they have worse outcomes compared to the other provider. How do you do with patients who look like this compared to everybody else? That’s why it does not mask quality. It does not excuse poor care. It actually helps us better identify it.

Natascha: Awesome, and thank you so much for that really thorough explanation. I know that this is something we’re going to be talking about for a long time to come, but today we are at the end of our podcast. So, Christie, I want to thank you for being here and for your insights on SDOH data. I know that they’re invaluable to our listeners. And I want to thank you all for tuning in to Avalere Health Essential Voice. If you’re interested in contacting either Christie or me, you can simply go to

Tools List

From beginning to end, our team synergy
produces measurable results. Let's work together.

Sign up to receive more insights about Health Equity
Please enter your email address to be notified when new Health Equity insights are published.

Back To Top