We've joined our RTI Health Solutions colleagues under the RTI Health Solutions brand to offer an expanded set of research and consulting services.
Social Determinants of Health Data Is Underutilized
Intro: In this video, Drs. Lisa Lines and Denise Clayton talk about social determinants of health data. Follow along as they outline the current challenges healthcare leaders face and discuss new approaches to produce better health outcomes.
This is the first in a four-part series of videos about social determinants of health. Other topics in this interview series include:
- Using LSI to Improve Social Determinants of Health Data
- Key Findings from Social Determinants of Health Analyses
- Use a Composite Score of Data to Find the Best Intervention Strategy
Denise Clayton: Hello and welcome to the first episode of the RTI Health Advance series on social determinants of health. My name is Denise Clayton and I’m a health economist at RTI Health Advance. I’m joined today by my colleague Lisa Lines, a Senior Health Services Researcher. (0:36)
In today's episode we'll be talking about the challenges of social determinants of health data. (0:43)
To get us started, Lisa, can you tell us a little bit about the current challenges around risk adjustment and quality measurement with respect to social determinants of health and the impacts of these challenges? (0:53)
Payment models and quality measures don’t adequately account for social determinants of health
Lisa Lines: Sure Denise. As we move from paying for volume to paying for value, value-based payment and risk adjustment become even more important. But current formulas for risk adjustment and performance measures or quality measures don't take many social determinants of health into account. (1:16)
What we mean by social determinants are the conditions in which people are born, age, grow, live, and work. It's about educational quality and attainment, healthcare access, community, safety, and social context, economic stability, food insecurity, the neighborhood and built environment, and housing environmental quality. These often have an independent effect over and above the individual level and social determinants of health so it's important to take these into account. (1:56)
When we don't, it can lead to unintended consequences where practices with lower risk patients get rewards and those with worse off patients lose out. A lot of providers feel they're being penalized for factors outside of their control. Payers and networks have incentives to enroll lower risk members so a lack of good data on social determinants of health can bias interventions toward lower risk populations and potentially result in less benefit. (2:23)
Social determinants of health data challenge: existing indices use a limited number of SDOH variables
Denise Clayton: Lisa, you mentioned that payers and networks have the incentive to enroll lower risk members. That leaves high risk members with lower access to care, which is really the last thing they need. What is interesting is that typical risk adjustment formulas use basic demographics like age and sex, and they use the diagnoses or hierarchical condition categories better known as HCCs. (2:50)
They may include an indicator for dual eligible but none of those social determinants of health you mentioned are incorporated either individually or in some aggregated kind of measure. Can you talk about the current options for area-based indices and how well they predict health outcomes? (3:09)
Lisa Lines: There's not a lot of great data on this because it's an evolving field but there are a few indices in common use. The social deprivation index has seven variables taken from the census. The social vulnerability index has 15 variables again that are census based and the area deprivation index has 17 variables. (3:36)
That's pretty much it. The area deprivation index is not valid at the zip code level because you need nine-digit zip codes. That's the census block group level. If you're looking for something to use with zip code level data, you’re really stuck with just two indices that don't capture the nuance and variation. (3:56)
And none of these indices do a great job of explaining health outcomes in terms of what we've examined so far. At best, they capture maybe 65% of the variation but often much less. That's why we created the Local Social Inequity score (LSI). We draw on over 150 variables and they're all related to health based on the conceptual model. (4:16)
Denise Clayton: Can you say a little bit more about what that score is? What's the level and what goes into it? (4:22)
Lisa Lines: The LSI is a measure explaining health outcome disparities or inequities in small geographic areas using predictors related to social factors. The geographic levels that we have a score at are the census tract. That's bigger than the census block group but smaller than the zip code tabulation area. The zip code tabulation area is another level we have. It's not as great because when you get to larger geographic distances you end up with a lot of variation. And averages across a large population are less useful than small areas, but we do have the scores available at the county level, the zip code level, and the census tract. (5:08)
How to make the most of social determinants of health variables
Denise Clayton: We talked about having a conceptual model here. Why is it important to have a conceptual framework when you're doing machine learning? Having 150 variables sounds great, but it's important that we're not just throwing in everything but the kitchen sink. (5:28)
Lisa Lines: Yes. When dealing with machine learning, having a conceptual model is important. Otherwise, you risk introducing issues where you don't really know what the score is built for. You don't really understand why the variables are there. Everything needs to be selected very carefully and spatial auto correlation needs to be considered. You need a good team with health geographers and data scientists involved and people from social worker and justice and education. It's really an undertaking to bring something like this into being. (6:15)
Denise Clayton: I can appreciate that. From the work we've done looking at the outcomes it's impressive that you bring those people together. (6:28)
We've looked at life expectancy at birth and infant mortality and in those cases the LSI explains more of the variation than the other area level indices available. The LSI explains at least 75% of the variation and oftentimes more than 90% of the variation with these outcomes. It's exciting work and can be applied in a lot of different ways to improve health outcomes. (6:54)
Thank you, Lisa, for joining me today. These insights will help us improve health outcomes. Stay tuned for more episodes in our social determinants of health series. And if you'd like to learn more you can contact us at RTIHealthAdvance.org. We're happy to talk with you. (7:17)
Learn about health equity solutions from RTI Health Advance.
Stay up-to-date on our latest thinking. Subscribe to receive blog updates via email.