In Depth Analysis

Why We Created a New Vulnerability Index Specific to COVID-19

COVID-19 poses unique challenges to communities that are not captured in the CDC’s Social Vulnerability Index.

As the COVID-19 pandemic began spreading rapidly across the United States, we recognized that the disease’s impact would be felt differently for communities across the country. Surely some places would prove more vulnerable than others — not only to the spread of the infection itself, but also to lasting social and economic damage.

To guide our work, we sought out an existing way to measure community-level vulnerability during a pandemic.

We thought we’d found what we were looking for in the Centers for Disease Control’s Social Vulnerability Index (SVI). The SVI is a validated metric that scores each county and census tract in the US on its vulnerability to natural or human-caused disasters. It is composed of four themes that make up a community’s social vulnerability: socioeconomic factors, household composition & disability, minority status & language, and housing type & transportation. The SVI is used by policy-makers and public health officials to guide their emergency responses.

However, on further inspection, we realized that the SVI, as valuable as it was for certain types of emergencies, failed to capture two factors we saw as crucial to a community’s vulnerability in the face of COVID-19: rates of pre-existing health conditions (e.g. diabetes, heart conditions, and respiratory problems) that are known to increase COVID-19 mortality risk, and the state of the community’s healthcare system.

Given the reported disproportionate number of deaths among people both with chronic illness and stories in Italy of overwhelmed healthcare systems, we questioned whether the SVI could adequately identify vulnerability during this crisis.

To identify more robust vulnerability scores, Surgo Foundation developed the COVID-19 Community Vulnerability Index (CCVI). The CCVI incorporates epidemiological and health-care systems themes alongside the four themes already captured by the SVI, to calculate COVID-19 specific vulnerability scores.

We’ve used our new index to produce a wealth of insights into the pandemic in the US over the past couple of weeks. But as the CCVI has been taken up by policymakers across the country, we thought it was important to step back and demonstrate why the CCVI adds value and provides insights that the SVI alone could not.

Here’s what makes the CCVI a stronger index:

1. Vulnerability has different dimensions geographically.

The different types of vulnerability captured by the CCVI and the SVI are not evenly distributed across the US. For example, certain communities whose vulnerability is driven by socioeconomic factors may be less vulnerable epidemiologically (i.e., they have lower rates of chronic conditions and/or a lower population density).

The six maps below show which counties are most vulnerable because of each specific theme, including the four themes that are common to both the SVI and CCVI, and the two that are specific to the CCVI (epidemiological and healthcare system). In the maps below, we show counties with individual theme scores above 0.8 (in red) to highlight those communities most vulnerable because of that specific theme.

The graphs demonstrate that the six vulnerability themes by no means overlap geographically. For example, counties that are vulnerable to COVID-19 because of minority status or language are almost entirely distinct from those that are vulnerable based on epidemiological factors.

Unexpectedly, the Northeast of the US, which we previously reported as being one of the least vulnerable regions to COVID-19, lights up when focusing on epidemiological risk factors alone. This tells us that while many counties across the US have some vulnerability, it is the particular combination of those vulnerabilities that makes a community more susceptible to negative outcomes from the COVID-19 pandemic. We need to consider all six themes together, because focusing on just social vulnerability alone is only part of the picture.

2. CCVI vulnerability scores differ from SVI vulnerability scores.

The CCVI adds a great deal of nuance to the social vulnerabilities captured by the SVI. When we include epidemiological and healthcare-system themes, we see many counties across the US diverging substantially in their CCVI and SVI scores — in effect creating two sets of vulnerable communities.

First, we acknowledge that many counties have more or less the same score on the two indices as shown in the graph below — they’re the ones on or around the diagonal line, which indicates perfect agreement between the CCVI and SVI. But we’re interested in the outliers — the points far above the line, which represent counties that have a high COVID-specific vulnerability (as defined by the CCVI), but a low social vulnerability (as defined by the SVI). The points falling far below the diagonal line represent the opposite — high social vulnerability and low COVID-specific vulnerability. Identifying the reasons behind these divergent scores sheds light on how the CCVI is better equipped to highlight COVID-specific vulnerability.

As a first step, let’s map the counties that have different vulnerabilities according to each index. We subtract the SVI from the CCVI to create one metric comparing the two indices. If the difference between the 2 indices is greater than +0.2 (i.e., 20-percentage-points), that county is designated as having a higher CCVI score compared with the SVI. A score less than -0.2 (i.e., a negative 20-percentage-point difference) is defined as a higher SVI score compared with the CCVI.

The map shows that in many counties (the ones shaded grey), the CCVI and SVI scores are similar or identical. For instance, many southern states such as Alabama, Louisiana, and Mississippi are both socially vulnerable, and also vulnerable based on COVID-specific indicators. Therefore, most counties in these states have both a high CCVI as well as a high SVI score.

Despite these similarities, 22% of counties across the US differ substantially (i.e., by more than 20 percentage points) between their CCVI and SVI scores. About 11% of all counties are specifically vulnerable to COVID-19 but not socially vulnerable, while another 11% of counties are socially vulnerable but not vulnerable to COVID-19 specifically.

The counties with a higher CCVI score (blue counties) are well dispersed throughout the country. Counties with a higher SVI score (red counties) are also dispersed across the country, but a good proportion congregate along the west coast. The higher CCVI scores imply that indicators beyond social vulnerability are imperative when considering the impacts of an infectious disease outbreak like the current pandemic.

Let’s compare two counties on opposite ends of the spectrum: Clark County, NV (home to Las Vegas) has a higher CCVI than SVI, and San Bernardino County, CA has a higher SVI than CCVI.

The two counties neighbor each other in America’s southwest and have almost identical population sizes (~2.2 million for each) and similar socioeconomic factors.

Despite the similarities, Clark County (CCVI = 0.72) is considered much more vulnerable to COVID-19 than San Bernardino County, CA (CCVI = 0.22). Why might this be? Clark County’s epidemiological theme score is around 1.0 compared to San Bernardino, which scores only 0.07 in this category. These differences indicate that Clark County, NV has a much higher incidence of pre-existing conditions as well as a higher population density than San Bernardino County. Clark County’s vulnerability to COVID-19 is also heightened by its healthcare system’s theme score of 0.42 (San Bernardino’s is less than 0.01).

What do these differences in vulnerability mean in practice? The impact of increased vulnerability of Clark County’s higher COVID-19 vulnerability than its neighbor’s can be felt when comparing the rate of growth in their COVID cases. Despite having similar population sizes and socioeconomic status, Clark County experienced 15 total doubling days (i.e., days where the number of COVID cases doubled), compared with only 10 experienced in San Bernardino. Doubling days refers to how quickly the virus is spreading throughout a community. The more doubling days a county has, the quicker that community’s health system can become overwhelmed, which is particularly worrying for a place like Clark County where the heath system is at a higher level of vulnerability than its neighbor. The CDC’s SVI did not pick up on these key differences between these two counties that have been struck so differently by COVID-19, demonstrating clear value added by the CCVI.

3. Why are we seeing these differences? The answer lies in COVID-specific epidemiological and healthcare-system themes.

The fact that some counties have a higher CCVI score than SVI score indicates that epidemiological and healthcare-system themes add an important piece to the puzzle that is missing from the SVI.

The bar graph below shows the average score of each theme in counties with a higher CCVI score. In counties with a higher CCVI score, the themes that are present in both indices (the first four on the bar graph) score around 0.4 on average, which means they would not be considered “vulnerable” under the SVI. But the epidemiological and healthcare-system themes average around 0.8. This divergence indicates that these epidemiological and healthcare system factors are necessary to identify more communities that are vulnerable to this outbreak.

All themes present in the CCVI are important to understand a community’s full range of vulnerabilities. We believe that the epidemiological and healthcare-system themes add critical information when determining vulnerability to COVID-19, as opposed to the themes exclusively addressing social vulnerability (i.e., higher SVI counties). For evidence to support this statement, we just need to open any news article covering the devastating effects of COVID-19 on people’s health and on our healthcare system.

With places like Louisiana reporting that the majority of COVID-related deaths stem from patients with underlying conditions (i.e., epidemiological factors) and stories from New York about overwhelmed healthcare systems (i.e., healthcare systemic factors), it is abundantly clear that these factors are critical to assessing vulnerability to COVID-19. This makes the CCVI, an enhancement of the SVI for the current pandemic, a better tool to help policy-makers target resources to the communities most vulnerable in the face of COVID-19. It is now up to those leaders to respond accordingly.

This work was made possible by everyone at the Surgo Foundation, including but not limited to (in alphabetical order): Yael Caplan, Vincent Huang, Rahul Joseph, Hannah Kemp, Tich Mangono, Sema Sgaier, Peter Smittenaar, Nick Stewart, and Staci Sutermaster.

Technical Notes:

  • Want to find out how we created the CCVI? All methodology can be found on our website here.
  • Want to analyze the numbers for yourself? The CCVI data is publicly available at census, county, and state level (link).
  • The SVI can be found here.
  • Questions, concerns, comments? Get in touch at covid19@surgofoundation.org.

Written by

Surgo brings precision to solutions that save and improve lives by integrating behavioral science and artificial intelligence surgofoundation.org

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store