• Social Deprivation Index (SDI)

    Background

    Addressing social determinants of health (SDOH) is critical to achieving the Triple Aim of lower costs, improved care, and better population health. Yet, it still needs to be determined which social determinants of health best direct clinical and community health interventions and guide adjustments to quality measures and payments. We developed Social Deprivation Index (SDI) to quantify levels of disadvantage across small areas, evaluate their associations with health outcomes, and address health inequities. This measure of social deprivation, in combination with other indicators, has potential application in identifying areas that need additional health care resources.

    The Social Deprivation Index measures were initially developed by Butler et al. (2012) using 2005–2009 American Community Survey (ACS) 5-year estimates and calculated at the Primary Care Service Areas (PCSA).  This measure was updated 1) with more recent ACS data (5-year estimates) and 2) using additional areas: counties, census tract, Zip Code Tabulation Areas (ZCTA), and PCSAs. 

    SDI Defined

    SDI is a composite measure of area level deprivation based on seven demographic characteristics collected in the American Community Survey and used to quantify the socio-economic variation in health outcomes.

    Methodology

    The final SDI is a composite measure of seven demographic characteristics collected in the American Community Survey (ACS): percent living in poverty, percent with less than 12 years of education, percent single-parent households, the percentage living in rented housing units, the percentage living in the overcrowded housing unit, percent of households without a car, and percentage nonemployed adults under 65 years of age. The SDI measure was calculated at the four geographic areas: county, census tract, aggregated Zip Code Tabulation Area (ZCTA), and Primary Care Service Area (PSCA, v 3.1). Census tracts are semi-permanent county subdivisions. ZCTA’s are generalized U.S. Postal Service Zip Codes. PCSA’s are small areas of aggregated census tracts and were created by Dartmouth University for the Health Resources and Service Administration. PCSA’s represent the travel patterns of Medicare patients to primary care services. To construct the PCSA-level SDI, all the ACS population measures were first extracted at the census tract level and then summarized at the PCSA level. The SDI measures for the county, census tract, and ZCTA levels were developed directly from ACS population measures. SDI scores are available for all counties, census tracts, ZCTA’s, and PCSA’s.

    Based on similar international and national indices, we started with a more extensive list of 14 candidate measures available in the ACS. We converted each of these measures into centiles to create a standard scale for easy interpretation of results. We used factor analysis methods to create the SDI. Factor analysis is a statistical technique to investigate the relationship between a group of observed variables and latent variable underlying a concept. It allows exploration of the association among measures by reducing many related variables into the underlying latent factors. The output from the factor analysis model is a factor loading for each variable. The factor loading represents the correlation strength between the variables that comprise the factor and the factor itself. The factor loadings can be interpreted as regression coefficients; the higher the factor loading score, the greater the variation explained by that variable.

    To simplify the model, we restricted to only include measures with factor loadings greater than 0.60, thus excluding percent black and percent in high-needs age groups. We constructed a final SDI measure based on weighted factor loading scores for each measure. We also examined the difference between models that include either nonemployed (those not seeking work) or unemployed (those actively seeking work). The factor loading for the nonemployed model was slightly higher than for the unemployed model. These initial results were reproduced for each period and each geography.

     

    SDI Component Description

    SDI Component Formula

    Percent Population Less Than 100% FPL

    (Population < 0.99 FPL) / (Total Population)

    Percent Population 25 Years or More With Less Than 12 Years of Education

    (Population < 12 years of education) / (Total Population)

    Percent Non-Employed for Population 16-64 years

    (Not in Labor Force + Unemployed Between 16-64 Years) / (Civilian + Not in Labor Force between 16-64 years)

    Percent Households Living in Renter-Occupied Housing Units

    (Renter Occupied) / (Owner Occupied + Renter Occupied)

    Percent Households Living in Crowded Housing Units

     (Tenure by Occupants Per Room - (Owner Occupied + Renter Occupied)) / (Total Occupied Housing Units)

    Percent Single Parent Families With Dependents < 18 years

    (Single Parent Households With Dependent Children < 18 Years) / (Total Families)

    Percent Households With No Vehicle

    (Households Without a Vehicle) / (Total Occupied Housing Units)

    **Tenure by Occupants Per Room indicates that the number of occupants per room is ≥ 1.01

    Available SDI Datasets for Download

    SDI List of Variables and Sequence Number Table ID from ACS for 2015-2019 (.csv file) and a link to the previous SDI years is available below:

    • 2012 SDI (comprised of 2008-2012 ACS data)
    • 2015 SDI (comprised of 2011-2015 ACS data)
    • 2016 SDI (comprised of 2012-2016 ACS data)
    • 2017 SDI (comprised of 2013-2017 ACS data)
    • 2018 SDI (comprised of 2014-2018 ACS data)

    Factor Loadings at Different Geographies: The most current year is the 2019 SDI (comprised of 2015-2019 ACS data).

    • The 2019 SDI at the County Level (derived from the 2015-2019 ACS 5-Year Summary Files) – Includes the raw proportions and centile scores for all the measures used in creating SDI, SDI raw and SDI score (.csv file).
    • The 2019 SDI at the Census Tract (derived from the 2015-2019 ACS 5-Year Summary Files) – Includes the raw proportions and centile scores for all the measures used in creating SDI, SDI raw and SDI score (.csv file).
    • The 2019 SDI at the PCSA level (derived from the 2015-2019 ACS 5-Year Summary Files) –   Includes the raw proportions and centile scores for all the measures used in creating SDI, SDI raw, and SDI score (.csv file).
    • The 2019 SDI at the ZCTA level (derived from the 2015-2019 ACS 5-Year Summary Files) – Includes the raw proportions and centile scores for all the measures used in creating SDI, SDI raw, and SDI score (.csv file).

    A link to all the previous years of factor loadings at different geographies is available as well:

    • 2012 SDI (comprised of 2008-2012 ACS data)
    • 2015 SDI (comprised of 2011-2015 ACS data)
    • 2016 SDI (comprised of 2012-2016 ACS data)
    • 2017 SDI (comprised of 2013-2017 ACS data)
    • 2018 SDI (comprised of 2014-2018 ACS data)

    Citation

    Citation: Social deprivation index (SDI). Robert Graham Center - Policy Studies in Family Medicine & Primary Care. (2018, November 5). Retrieved November 29, 2021, from https://www.graham-center.org/rgc/maps-data-tools/sdi/social-deprivation-index.html

    References

    1. Butler DC, Petterson S, Phillips RL, Bazemore AW. Measures of Social Deprivation That Predict Health Care Access and Need within a Rational Area of Primary Care Service Delivery. Health Services Research. 2013;48(2 Pt 1):539-559. doi:10.1111/j.1475-6773.2012.01449.
      https://www.ncbi.nlm.nih.gov/pubmed/22816561

    2. Philips RL, Liaw W, Crampton P. How Other Countries are Using Deprivation Indices. Health Affairs, 2016: 35 (11):1991-1998
      http://content.healthaffairs.org/content/35/11/1991.abstract

    Publications Using SDI

    1. Liaw W. Krist AH, Tong ST, Sabo R. et al. Living in “Cold Spot” Communities Is Associated with Poor Health and Health Quality. J Am Board Fam Med. 2018; 31:342-350
    2. Bevan GH, Nasir K, Rajagopalan S, Al-Kindi S. Socioeconomic Deprivation and Premature Cardiovascular Mortality in the United States. InMayo Clinic Proceedings 2022 Mar 15. Elsevier.
    3. Bevan GH, Josephson R, Al-Kindi SG. Socioeconomic deprivation and heart failure mortality in the United States. Journal of Cardiac Failure. 2020 Dec 1;26(12):1106-7.
    4. Cottrell EK, O'Malley JP, Dambrun K, Park B, Hendricks MA, Xu H, Charlson M, Bazemore A, Shenkman EA, Sears A, DeVoe JE. The Impact of Social and Clinical Complexity on Diabetes Control Measures. J Am Board Fam Med. 2020 Jul-Aug;33(4):600-610. doi: 10.3122/jabfm.2020.04.190367. PMID:
    5. Patel SA, Krasnow M, Long K, Shirey T, Dickert N, Morris AA. Excess 30-Day Heart Failure Readmissions and Mortality in Black Patients Increases With Neighborhood Deprivation. Circ Heart Fail. 2020 Dec;13(12):e007947. doi: 10.1161/CIRCHEARTFAILURE.120.007947. Epub 2020 Nov 9. PMID: 33161734; PMCID: PMC8164383.
    6. Griggs, Stephanie, Christine Horvat Davey, Quiana Howard, Grant Pignatiello, and Deepesh Duwadi. 2022. "Socioeconomic Deprivation, Sleep Duration, and Mental Health during the First Year of the COVID-19 Pandemic" International Journal of Environmental Research and Public Health 19, no. 21: 14367. https://doi.org/10.3390/ijerph192114367
    7. Cottrell EK, Hendricks M, Dambrun K, et al. Comparison of Community-Level and Patient-Level Social Risk Data in a Network of Community Health Centers. JAMA Netw Open. 2020;3(10):e2016852. doi:10.1001/jamanetworkopen.2020.16852
    8. Lucas JA, Marino M, Giebultowicz S, Fankhauser K, Suglia SF, Bailey SR, Bazemore A, Heintzman J. Mobility and social deprivation on primary care utilization among pediatric patients with asthma. Fam Med Community Health. 2021 Jul;9(3):e001085. doi: 10.1136/fmch-2021-001085. PMID: 34244305; PMCID: PMC8278882.
    9. Green BB, Larson AE, Huguet N, Angier H, Valenzuela S, Marino M. High Blood Pressure Reduction, Health Insurance Status, and Social Deprivation Index in US Community Health Centers. AJPM Focus. 2022 Dec 1;1(2):100018.
    10. Andrew W Bazemore, Erika K Cottrell, Rachel Gold, Lauren S Hughes, Robert L Phillips, Heather Angier, Timothy E Burdick, Mark A Carrozza, Jennifer E DeVoe, “Community vital signs” : incorporating geocoded social determinants into electronic records to promote patient and population health , Journal of the American Medical Informatics Association, Volume 23, Issue 2, March 2016, Pages 407–412, https://doi.org/10.1093/jamia/ocv088

    County-Level Social Deprivation Index

    SDI Frequently Asked Questions (FAQs)

    The SDI score is a composite measure of area-level deprivation constructed from seven demographic and household characteristics collected in the ACS (American Community Survey):

    1. percent living in poverty
    2. percent with less than 12 years of education
    3. percent single-parent household
    4. percent living in the rented housing units
    5. percent living in the overcrowded housing units
    6. percent of households without a car
    7. percent non-employed adults under 65 years of age

    Furthermore, the SDI quantifies the levels of deprivation across smaller areas and indicates the extent a community is socially disadvantaged. The relationship between the SDI score and the severity of deprivation is direct, positive, and linear; as the SDI score increases, the severity of deprivation increases too.

    We used factor analysis to create the SDI datasets at various levels of geography: county, census tract, ZCTA (Zip Code Tabulation Areas) and PCSA (Primary Care Service Areas).

    Factor analysis is a statistical method that describes variability between a group of observed, correlated variables and a latent variable underlying a concept. Factor analysis allows for exploring the association among measures by consolidating many related variables into the underlying latent factors.

    The SDI can be used in two ways: first, to quantify levels of disadvantage across geographical areas in the United States; second, to evaluate associations between deprivation and health outcomes to allocate resources most effectively and achieve optimal health equity for everyone, especially those who are in impoverished communities. 

    A centile is defined as the seven raw SDI component measures where the SDI itself is divided into 100 equal parts. Centiles are a common scale for an easy and effective interpretation of the scores and comparison across neighborhoods.

    The SDI raw measure is derived from the seven components that the SDI measures using factor analysis.

    The SDI score indicates the extent of social disadvantage in a community.

    The SDI ranking classifies the geographic areas by the level of deprivation. 

    Each year, the SDI is created at four geographic levels: county, census tract, ZCTA (Zip Code Tabulation Areas), and PCSA (Primary Care Service Areas). 

    No, the SDI can’t be requested at a different geographic level. Currently, the SDI is developed at these four geographies only. However, technical assistance with the SDI methodology can be provided to develop the SDI at geographies other than the ones mentioned above.

    There are seven SDI datasets available for download on the Robert Graham Center website for all the years starting from 2012 through 2020 except 2 years: 2013, 2014. The 2012 SDI was created from the 2008-2012 American Community Survey (ACS) 5-Year summary file data. Similarly, the 2015 SDI was created from the 2011-2015 ACS 5-Year summary file data. Furthermore, the year of the SDI reflects the last year of the ACS 5-year data that was extracted to construct that year’s SDI. 

    The 2016-2020 ACS data was used to construct the most recent 2020 SDI. The 2020 SDI will be updated on the Robert Graham Center website shortly, a year after the ACS 5-year summary files is released. 

    Citation: Social deprivation index (SDI). Robert Graham Center - Policy Studies in Family Medicine & Primary Care. (2018, November 5). Retrieved November 29, 2021, from https://www.graham-center.org/rgc/maps-data-tools/sdi/social-deprivation-index.html

    For any assistance or concerns, please send us an email at policy@aafp.org.