I saw an interesting tweet recently that got me thinking about the paradox that sometimes exists in our most highly-tuned healthcare interventions: sometimes the very tools deployed to help people can end up making things worse.
Let me say that I admire anyone who attempts to alleviate healthcare problems and the upstream negative social determinants of health (SDOH) that contribute to them, so this post is not meant as a criticism. Rather, I want to explore some of the sophisticated tools used by researchers to gain a better understanding of how they work, and also to examine potential gaps that might unintentionally lead to imprecise results that can, in turn, affect how healthcare is delivered in the real world.
Accounting for Social Risk
It’s conventional wisdom nowadays that devoting an equal amount of healthcare resources to every person in the U.S. would not result in favorable health outcomes for all involved. There are many reasons for this, but the two primary factors include 1) people at the lower end of the socioeconomic spectrum suffer disproportionately poor health outcomes due to being faced with a broader array of negative SDOH, and thus require a higher investment of financial resources to stay healthy, and 2) socioeconomically disadvantaged people often have less access to quality healthcare.
As important as ready access to quality care is, however, it’s only one of many factors that contribute to a person’s health. According to a recent report issued by the Assistant Secretary for Planning and Evaluation’s Office of Health Policy, “Studies estimate that clinical care impacts only 20 percent of county-level variation in health outcomes, while social determinants of health (SDOH) affect as much as 50 percent. Within SDOH, socioeconomic factors such as poverty, employment, and education have the largest impact on health outcomes.”
Put another way, although the care people receive in clinical settings is important, it plays a relatively small part in their overall quality of life. To make up for this gap, some experts propose adjusting value-based performance measures to account for social risk factors. But critics point out that this approach often results in forcing vulnerable populations to accept lower standards of care and doesn’t really fix the problem. Another approach, which is currently being taken up by public (and some private) insurers is to adjust payments to healthcare providers who administer care to underserved groups, a concept called social risk adjustment.
How Social Risk Adjustment Works
Adjusting payments has to be done very carefully, because doing so without considering every nuance can counter-intuitively reinforce health inequities. This point was made by Robert Saunders, PhD, during a recent webinar convened by the Patient-Centered Primary Care Collaborative, when he presented evidence that people living in underserved areas (as measured by a social vulnerability index, a concept I’ll cover a little further down) often face obstacles to accessing healthcare.
This reality can lead to a situation where, as Dr. Saunders notes, “healthcare utilization may not be as high as it needs to be,” and people’s current social needs aren’t being met by the healthcare system. In Dr. Saunders’ example, what this means is that Medicare expenditures for such a population aren’t always closely associated with their reported need. Dr. Saunders added the following:
“If you use the traditional approach to risk adjustment where we plug in variables into a model looking at the past, so looking at the association between that factor and healthcare utilization, you may artificially lower the amount of money that is going to populations that may have high social needs, so you have to be very thoughtful about how you use it.”
There’s also the challenge of paying practitioners during a given year for social needs addressed in a previous year, which can result in a misalignment of resources. When physicians aren’t given the tools they need to mitigate patients’ social needs, it can lead to burnout. In resource-strapped areas this can be especially true, and it can lead to dire consequences. So how can such a circumstance be avoided?
To a large extent, insurers can improve the situation by dispersing prospective payments, also known as capitation payments, that more accurately anticipate patients’ needs from year to year. This practice is already in use by payers like the Centers for Medicare & Medicaid Services (CMS), over two-thirds of whose Medicaid beneficiaries are enrolled in risk-based managed care plans. In this program, CMS provides upfront payments to healthcare plans for anticipated costs of contracted services, the latter of which are influenced by each beneficiary’s risk factors. Similarly, the benchmarks Medicare Advantage utilizes for rate-setting are adjusted for medical risk. According to the Better Medicare Alliance, “CMS risk-adjusts the capitated payments to Medicare Advantage plans based on an enrollee’s ‘risk score’ – a measure of the expected costs associated with a person’s care.”
An example of social risk payment adjustments can be found in the Maryland Primary Care Program (MDPCP) HEART Payment Playbook. The Maryland Department of Health (MDH) collaborated with CMS to create a healthcare transformation program called the MDPCP. The MDPCP HEART payment program offers additional support to MDPCP participants who provide care to Medicare enrollees who are contending with complex social and environmental factors. The program “aims to improve health outcomes and lower costs in this targeted group of high-need individuals” by collecting social needs screening data and using these funds to develop and implement interventions.
Tools to Measure Deprivation
As I alluded to above, there is a substantial amount of money in play when it comes to government agencies and insurers allocating extra resources to help those most in need. So how do entities tasked with distributing this money decide where to spend it? Enter the deprivation index, a tool that helps assess a geographic population’s risk based on a range of factors.
Before we launch into an examination of particular social deprivation measurement tools, I’d like to start with a definition to give us proper context. The Centers for Disease Control and Prevention (CDC) defines any “locally sensitive” area deprivation index as being “an ideal measure to identify and screen for the health care and social services needs and to advance the integration of social determinants of health with clinical treatment and disease prevention.”
There are a range of deprivation indices in existence, but no consensus on exactly how many. According to one 2022 study there are “sixty deprivation indices in seventeen countries,” while, during a recent Health Affairs podcast episode, guest Dr. Meera Kotagal identified 44 distinct deprivation indices. In my research, the index most often used to measure social needs is the Area Deprivation Index (ADI) introduced by the University of Wisconsin (UW) back in 2018.
In the years since its inception, the UW index has been “refined, adapted, and validated” to the Census Block Group (CBG) neighborhood level by a team at the University of Wisconsin-Madison. Given that Census blocks are “the smallest geographic area for which the Bureau of the Census collects and tabulates decennial census data,” some researchers consider it to be the superior index.
Others, however, have pushed back on the effectiveness on the UW ADI. Rounding back to the tweet I cited earlier in this post and the study to which it links, the authors used the UW ADI to examine social risk among Medicare beneficiaries and found that “community-level social risk explained little variation in health care spending, was negatively correlated with spending conditional on demographics and clinical characteristics, and was poorly correlated with self-reported social risk factors.”
Crucially, the authors stated that “Conditional on demographic and clinical characteristics, ADI remained significantly associated with spending, but the direction of the association reversed, with every 1-point increase in ADI associated with a $11.08 decrease in spending.” Echoing Dr. Saunders here, the authors seem to be saying that just because a given population is in greater need of medical services, that doesn’t mean they will automatically seek medical care — and this factor should inform any deprivation index attuned to a small geographic area. People may stay away from medical practitioners because of perceived racism or any number of other factors or, as mentioned above, they may experience physical or geographical challenges when trying to access medical care.
In any discussion of deprivation indices, it’s important to note the spectrum they cover: while some focus on wide geographical areas like one designed specifically for the state of Utah, others, like the first iteration of the U.S. Census Bureau’s Multidimensional Deprivation Index (MDI), have examined deprivation on the county level. Interestingly, the MDI was revised in 2021 using the UW ADI to allow the Bureau to investigate deprivation in a more fine-grained way because, as the authors acknowledged, “even mid-sized counties have a significant amount of heterogeneity that is not captured by county-level measures.”
Although differences exist across these and other indices, what many of them have in common is that they calculate socioeconomic disadvantage using data from the American Community Survey (ACS). The U.S. Census Bureau website states that “The American Community Survey (ACS) is an ongoing survey that provides vital information on a yearly basis about our nation and its people.”
The survey, which is distinct from the census in that it is much more up to date, collects dozens of indicators – including some the census excludes like education, employment, internet access, and transportation. Several prominent ADIs utilize what’s called ACS 5-Year Estimates, which are estimates representing data collected over a five-year period of time. According the U.S. Census Bureau, the main advantage of using multi-year estimates “is the increased statistical reliability of the data for less populated areas and small population subgroups.” Also of note, the five-year estimates are available down to the block group level for the entire country.
It should also be noted that different indices incorporate varying numbers of indicators into their formulation; for example, the UW ADI “includes factors for the theoretical domains of income, education, employment, and housing quality,” while the Robert Graham Center’s social deprivation index (SDI) uses seven demographic characteristics collected by the ACS.
Deprivation Indices in Action
So why all the hubbub about deprivation indices? Sure they provide insight into where socially vulnerable people live, but how do they affect day-to-day life? As with many topics having to do with social deprivation, the U.S. Census Bureau provides us with an answer: “Local communities depend on information from the American Community Survey, as well as the decennial census, to decide where schools, highways, hospitals, and other important services are needed. The data collected through the American Community Survey and the 2020 Census help determine how to distribute more than $675 billion of federal spending each year.”
Clearly these tools, as advanced as they are, still have a way to go in accounting for the specific needs of individual patients. As Dr. Saunders of Duke advises, they should be used with a great deal of care. Looking forward, it will be interesting to see how these tools progress in the era of artificial intelligence as it becomes more important than ever to keep patients at the center of the care process.