A paper by Ng, Zhang and Kowal (2025) use the American Community Survey (ACS) data from 2016–2022 were used to construct Social Vulnerability Index (SVI). 

SVI is calculated from 16 social determinants of health (SDoH) variables that are grouped into 4 themes: socioeconomic status, household characteristics, racial and ethnic minority status, and housing type & transportation (see my blog post on SVI for more details).  The ACS data is available at the county and the ZIP code tabulation area (ZCTA) levels.  The authors created crosswalks to estimate SVI at the 3-digit and 5-digit ZIP code levels. The methodology is explained in more detail as follows:

The ZCTA numerator (e.g., people in a geographic unit below 150% of the poverty line) and denominator (e.g., people residing in the geographic unit) values were aggregated to the ZIP-5 and ZIP-3 level to obtain totals representative of the geography of interest. With all numerators and denominators now representing the correct geography, the percentages of each of the 16 variables were calculated for counties, ZIP-5s, and ZIP-3s by dividing each numerator by the corresponding denominator. Each of the 16 variables were then ranked by percentile across all geographies, resulting in a number between 0 and 1 for each geographic area and SDoH variable, which aligns with the scale used to report the SVI (ranging from 0 [least vulnerable] to 1 [most vulnerable]).

This approach was validated against published SVI estimates
from 2014 CDC data at the census tract and county levels.

The authors then demonstrate the utility of this data by linking SVI estimates at 3-digit ZIP code level to health insurance claims data.  They summarize the utility this offers as follows:

…SVI at the ZIP-3 level can be used with a large US health plan claims database, which has a variable for patient ZIP-3. If health plan claims are linked with a specific disease cohort, SVI enrichment could also be used to explore differences between patients living in more vulnerable areas vs. less vulnerable areas, such as the types of treatments prescribed, treatment adherence, and healthcare resource utilization. Crucially, this approach can be used for any kind of data that include a geographic identifier for patients.

You can read the full paper here.