We recently had the chance to sit down for a virtual coffee with Sarah DeSilvey to chat about her recent ICD-10 application, which proposed new “z-codes” for food insecurity on behalf of BCBS of VT and Yale School of Nursing. Sarah is also a practicing nurse practitioner, a food insecurity policy advocate, and the author of numerous issue briefs related to coding for food insecurity. She brings a multi-faceted, interdisciplinary approach to her work in social determinants of health (SDOH)
Your work spans clinical practice, health policy, informatics, and much more. Can you tell us a little bit about your background, and the work that led you to propose new z-codes to the CDC?
Sure — I come from a long line of doctors, but I studied philosophy in college, and then moved to Montana where I was a farmer for 12 years. Then, I encountered nursing, knew it was my calling, and I moved back home to Vermont and studied to become a nurse practitioner. Vermont is a small state, and I learned quickly that you have to be a bit of a polymath. So in addition to being a full-time rural clinician, I’m the quality improvement (QI) consultant for my hospital, and serve as the liaison to our state’s ACO.
Ever since I’ve been a clinician I’ve focused on addressing poverty and food insecurity in clinical practice. (As a farmer, I realized that I just couldn’t feed everyone.) That mission is how I first became involved with my colleagues at Children’s Healthwatch (CHW) and the Food Research and Action Center (FRAC). Initially, we connected on process and policy initiatives. However, along the polymath path, they knew I had been working QI and informatics for my home hospital, trying to build reporting to measure intelligent practice. In 2017 FRAC asked me to write the food insecurity coding brief and FRAC, Children’s HealthWatch and SIREN and I collaborated on that piece over the following year. Along the way, we wrote code for every open sourced language we could including LOINC codes for the Hunger Vital Sign, and a SNOMED code for food insecurity.
At the same time, my work with the ACO had me asking questions about how to accurately represent social risk in claims within value-based systems. Once we realized that ICD-10-CM was a coding set that could be modified through application to the National Center for Health Statistics we gathered support and decided to submit the ICD-10-CM application with a host of national allies.
Can you describe what you’re pursuing in your doctorate work?
Fundamentally, it concerns developing the language and tools to address the social determinants in clinical practice. We’ve entered this moment where we can see those issues playing out in clinical practice every day, but we face these system-level limitations in how to name and address them.
So, the ICD-10-CM application became the first wing of my doctoral project. The aim is to build the language that lets us take a more granular approach to documenting food insecurity, sharing care, researching best practices and calculating risk within value-based care. I had the honor of working really closely with wise ICD-10-CM experts at the American Academy of Family Physicians, Blue Cross Blue Shield and AHIMA to make sure the application matched ICD structure.
This ties into the second aim, which involves expanding the intervention language to reflect how we care for folks who experience food insecurity in the US — programs like WIC, SNAP, summer and school meals, home-delivered meals, and medically tailored meals. The goal is to incorporate them into our documentation to better inform care and care management
Lastly, beyond just expanding codes, I am working with fellow content experts to develop initial diagnostic considerations toward the development of diagnostic criteria for food insecurity. Our current process for food insecurity assessment lies heavily on screening but in truth, in clinic with all other needs, we assess concerns from patient story and data by applying diagnostic criteria. And although we have a clear conceptual definition for food insecurity we do not have published criteria. We are working on this. For example, “It has been a rough month, my husband lost his job, thank heavens the kids eat free at school, but we don’t have much food once they get home. We have not eaten well in a long time.” This is, fundamentally, food insecurity without the need to apply a screener to assess it.
I see diagnostic integrity correlating with data integrity. Along these lines, I’ve also been developing a thought framework that imagines the three domains of screening, assessment, and data as evolving into their better selves when filtered through standards of ethics and evidence. In essence, ‘screening’ becomes the utilization of validated tools, ‘assessment’ becomes the application of diagnostic criteria, and ‘data’ becomes consistent, equivalent, interoperable terminology.
How would you explain why these new codes for social determinants of health are so important?
These codes are language — that’s a distinction that needs to be made. For the patient and the clinician it’s simply about the language used to communicate. The question for clinicians becomes, “Can I name this thing I see? Can I share what I have seen with others?” The language that we use at this system level is vetted and curated by whatever terminology exists. If we do not have the language the concepts of poverty are just hidden within the clinical space and that is an ethical problem.
I do want to say that we need to speak about protective factors too. As much as we need to say that a family is homeless, or unable to afford gas to get to appointments, to speak to the full stories of our patients we also need to be able to speak to the things that protect them from slipping further into risk such as the faith that fosters resilience or family that prevents social isolation. These are part of the story we aim to communicate too.
Can you comment on some of the business implications that are playing out here, for both clinical practitioners, and broader enterprise level data and analytics efforts?
Sure. I see it in some sense as code-switching — thinking about these concepts as trojan horses to enable broader change. Sometimes we can convince folks to care about poverty via a social justice and ethics lens alone. However, in the US health care is an industry. As an industry, it speaks in a language of data and finance. Our effectiveness at communicating social needs within a data and finance lens depends on the terms we seek. If the concerns of poverty cannot be communicated in these terms, we are limited in our capacity to effect change.
Furthermore, on the data side, we have to be thinking about where this all fits in with broader enterprise level shift towards value-based health. This goes back to my initial thought as a representative of my state’s ACO. For example, the Johns Hopkins ACG model is commonly used by hospitals and health systems to assess risk using a clinical lens, but it’s largely missing social risk just because the codes are not there. This ties into payment: We can use the ACG or even hierarchical condition codes to allocate care management payments or bundled payments, for example, but if poverty and other social determinants are not included, even though we know the impact they have on outcomes, we are failing to allocate resources effectively.
Let’s talk about what these new codes WON’T solve for. Do you have a wishlist for actions that health plans, systems, and state/local officials can take to address SDOH in their communities?
Well, there is A LOT these codes won’t be able to solve for. All they are is information — information is not a solution. Information is a way to communicate the need for a solution, but that solution itself still involves a need to build more intelligent systems to prevent and address food insecurity in the first place. If we look at the bigger picture — of course this is not going to solve the broader macroeconomic forces that perpetuate inequity. At the broader level, I’m 100% guided in my work by the World Health Organization’s Conceptual Framework for Action on SDOH. The framework states there is an ethical mandate for the clinician to advocate for policies that address the root causes of the inequity we see.
As far as a wishlist, it goes back to reasons that I got into this work: How might we turn some of the massive resources of the industry into ways to address deeper drivers of health? I think this work with ICD-10 codes is about informing how the healthcare system includes social risk and determinants into broader population-level medical or clinical risk analysis. And then, the question is, how do we apply this information to address upstream causes of risk? That involves collaborating and funding community partners. It does not mean taking over the provision of these services, but rather supporting them — collaborating with food banks, shelters, other community based organizations (CBOs) to create effective, preventative communities of care.
You’re involved the food insecurity wing of the GRAVITY project. Can you talk about what this new initiative is for someone who might not know a lot about health informatics? What do you hope to accomplish through this group, say a year from today?
The GRAVITY project is a national effort that aims to develop the essential data components and data sets to address three key social determinants, food insecurity, housing instability and inadequate housing, and transportation needs. The project is driven by the amazing teams at SIREN and EMI Advisors and is funded by a grant from the Robert Wood Johnson Foundation.
SIREN has been the nexus of a national conversation on developing social needs terminology for some time. In 2017 they convened a very inspiring group of leaders in this health and terminology space in DC to discuss what was needed to develop the capacity to care for social needs and share information about social needs. The Gravity Project is the outcome of that gathering. The hope is to leverage the momentum in this space, gather stakeholders, and with the help of EMI develop a few key products for each concept: good solid use cases to represent the social needs (basically good stories that capture the data needs,) data elements, a consensus statement on needed terminology for each concept, and implementation guides. That is a lot! It is slated to run weekly over the next year. I have been asked to assist with spearheading the food insecurity conversation along with colleagues at the American Academy of Nutrition and Dietetics.