Measuring Concordance and Discordance between Individual Characteristics and Neighborhood-level Social Determinants of Health

doi:10.21203/rs.3.rs-2593711/v1

Download PDF

Research Article

Measuring Concordance and Discordance between Individual Characteristics and Neighborhood-level Social Determinants of Health

https://doi.org/10.21203/rs.3.rs-2593711/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background:

People have both individual and neighborhood level social determinants that may confer health risks or health benefits. While these are often aligned, this is not always the case. An affluent individual may live in a neighborhood with a high median income (concordance) or an affluent individual may live in a neighborhood with a low median income (discordance).

Methods:

We propose a method to operationalize the concepts of concordance and discordance between individual-level and neighborhood-level social determinants and provide two examples. We first consider a factor that is measured with a continuous variable–household income. We then consider race/ethnicity—a categorical variable. To illustrate the proposed method, we apply our operational definitions to a small dataset that contains self-reported individual data paired with geocoded publicly available neighborhood-level data based on residential address.

Results:

In our cohort, 38% of participants lived in neighborhoods that were racial/ethnically concordant; 23% of participants in neighborhood that were racial/ethnically discordant, and 39% of participants in neighborhoods that were Neutral.

Conclusion:

We propose an approach to defining the concepts of concordance and discordance of individual characteristics with neighborhood-level social determinants. Our proposed approach will enable more robust investigations of social determinants of health and their impact on health. Such investigations are needed to inform the growing use of social determinants of health data from various sources for a variety of purposes.

Social determinants of health (SDoH)

Health Disparities

Individual Risk Factors

Neighborhood Risk Factors

The importance of social determinants of health (SDoH) in improving health outcomes and eliminating health disparities is widely recognized.(1) SDoH are conditions in the environments in which people are born, live, learn, work, play, worship, and age that affect a wide range of health, functional, and quality-of-life outcomes and risks.(2) SDoH operate at both the individual and the neighborhood levels; personal factors (e.g. race, income, education) interact with neighborhood factors (e.g. environmental toxins, violence, food access)(3) to impact health.(4, 5, 6)

A large body of literature exists on the association between individual- or patient-level SDoH and health outcomes, and there is a growing literature on the impact of neighborhood SDoH on health.(7) Studies generally either use individual-level data to make inferences about those with similar race/ethnicity, education or income, or they use neighborhood-level data to predict health risks for those who live in that area. However, people have both individual and neighborhood level social determinants that may confer health risks or health benefits. While these are often aligned, this is not always the case. An affluent individual may live in a neighborhood with a high median income (concordance) or an affluent individual may live in a neighborhood with a low median income (discordance). In this latter case, a person would experience the health benefits of their individual income as well as the health risks of living in an area with a low median income. We hypothesize that those with concordant individual and neighborhood level characteristics have different health outcomes than those with discordant individual and neighborhood level characteristics. In order to test such hypotheses, operationalized definitions of concordance and discordance are required.

Here, we propose a method to operationalize the concepts of concordance and discordance between individual-level and neighborhood-level social determinants and provide two examples. We first consider a factor that is measured with a continuous variable – household income. We then consider race/ethnicity, a categorical variable. To illustrate the proposed method, we apply our operational definitions to a small dataset that contains self-reported individual data paired with geocoded publicly available neighborhood-level data based on residential address. We anticipate that the risk profile created by self-reported individual data will differ from that created by neighborhood-level data and those patterns of concordance and discordance will be demonstrated.

Data Source

PRIORity (Predicting Risk and Investigating Outcomes using Patient-Reported and Community-level Social Determinants Data in Vulnerable Populations) is a prospective observational cohort study of randomly selected, high-risk (at least one chronic medical condition) adult Emergency Department (ED) patients who reside in New York City (New York City). A 36-item survey was administered by research staff, including participants’ address. (Survey is available as Appendix—Priority Enrollment Survey). For purposes of our example, we use self-reported household income and self-reported race/ethnicity from this cohort.

Definition of neighborhood

Neighborhood was operationalized as a walkable distance along pedestrian accessible networks in a ¼ mile radius around each participant’s address. This definition is a more realistic and nuanced description of “neighborhood” that is more reflective of a participant’s lived experience than more commonly used definitions such as a simple Euclidean buffer (e.g., a circle with a ¼ mile radius) or simple containment (the census tract or ZIP code in which the home address is located).(8)

Calculation of neighborhood variables

All spatial methods were performed with ArcGIS Pro 3.0 (ESRI, Redlands, CA) and other methods with RStudio.(9) Of the 150 survey participants, two (1.3%) did not have a valid home addresses in NYC resulting in an analytic sample of 148 which were geocoded using ESRI’s world geocoding services. American Community Survey 2020 5-year estimates of median household income (MHHI) and race/ethnicity by census block group (CBG) in NYC were acquired from Integrated Public Use Microdata Series (IPUMS) National Historic geographic information system (GIS) (10) and spatialized. Pedestrian-accessible routes were identified by filtering the LION dataset (Linear Integrated Ordered Network) from NYC Department of City Planning.(11) Pedestrian-accessible network buffers were then created by measuring ¼ mile (~ 400m) along the network from each participant’s home location. Race/ethnicity for each participant’s “neighborhood” was then calculated by using areal weighting,(12) meaning that the area of the portion of each CBG which is intersected by the buffer is calculated, and the ratio of each intersected CBG area to total CBG area is used to weight the population counts (e.g., if CBG “A” has 25% of its area within the buffer, we assume that 25% of its population is also within the buffer). The neighborhood MHHI was then calculated using population-weighted means based on the areal weighting results (e.g., if CBG “A” has 100 residents within the buffer and a MHHI of $10,000, and CBG “B” has 50 residents within the buffer and a MHHI of $40,000, the population-weighted mean MHHI would be ((100*$10,000) + (50*$40,000)) / (100 + 50) = $20,000). Areal- and population-weighting, in combination with the utilization of pedestrian-accessible network buffers, aid in reducing the impact of edge effect, the modifiable area unit problem, and other sources of geospatial-related error.(13) Participant survey data includes race/ethnicity categories and household income (HHI) presented in $20,000 intervals.

Calculation of concordance / discordance for categorical variable

To compare participant race/ethnicity with neighborhood-level characteristics, the majority race for each neighborhood was calculated (i.e., > 50% of one race/ethnicity). If there was no majority, it was coded as “No Majority.” Concordance was defined as when the individual’s race/ethnicity is the same as that of the majority group in the neighborhood and discordance when they were different from the majority group. Participants were coded as “Neutral” if there was no majority in the neighborhood.

Calculation of concordance / discordance for continuous variable

To compare participant-reported HHI and neighborhood-level MHHI, quintiles were calculated for NYC based on CBG-level data. The break values for the quintiles (Q1, Q2, Q3, Q4, Q5) were then modified to match the nearest break values in the survey data, resulting in < $40,000 (Q1), $40,000 to < $60,000 (Q2), $60,000 to < 80,000 (Q3), $80,000 to < $100,000 (Q4), and >= $100,000 (Q5). Concordance was defined as individual and neighborhood HHI being in the same quintile (e.g., both participant and neighborhood HHI are in Q1), discordance when there is more than one quintile between individual and neighborhood HHI (e.g., the individual’s HHI is in Q1, but the neighborhood HHI is in Q3, Q4, or Q5). Participants were coded as “Neutral” when their HHI quintile was only one class away from the neighborhood (e.g., participant in Q1 and neighborhood in Q2).

Table 1a shows distribution of participant HHI and corresponding neighborhood MHHI. Over 40% of participants reported household incomes of less than $20,000, and over 60% reported less than $40,000, the upper limit of the 1st quintile in the analysis. Note that no participants reported HHIs in the (Q5) 5th quintile (>$100,000). Table 1b shows the relationship between participant and neighborhood race/ethnicity. The majority of participants were Hispanic/Latino (48.0%) or non-Hispanic (NH) Black (39.2%).

Table 1a

Individual and neighborhood level household income (n = 147*)
Participant HHI	Quintile	n (%)	Neighborhood MHHI (mean (SD))
< 20k	Q1	60 (40.8)	48,960 (27,225)
20-40k	Q1	31 (21.1)	48,532 (27,260)
40-60k	Q2	17 (11.6)	48,560 (18,929)
60-80k	Q3	7 (4.8)	70,816 (46,146)
80-100k	Q4	15 (10.2)	95,127 (42,787)
Chose not to answer	NA	17 (11.6)	59,158 (23,091)

* Three participants were excluded (2.0%). Two had invalid addresses, and one was in an area without reliable area-level MHHI data.

Table 1b

Individual and neighborhood level race/ethnicity (n = 148*)
Participant race/ethnicity		Neighborhood race/ethnicity (mean (SD))
Participant race/ethnicity	n (%)	% Hispanic	% NH Black	% NH Asian	% NH Other	% NH White
Hispanic/Latino	71 (48.0)	47.13 (16.96)	28.75 (13.93)	7.09 (7.47)	2.49 (1.61)	14.54 (16.57)
NH Black	58 (39.2)	40.72 (17.80)	35.85 (17.50)	5.71 (5.43)	2.95 (2.20)	14.76 (14.65)
NH Other	6 (4.1)	39.57 (20.83)	21.77 (19.48)	13.02 (11.80)	2.95 (2.18)	22.70 (19.48)
NH White	13 (8.8)	26.79 (22.53)	10.91 (11.69)	10.58 (8.04)	3.57 (2.58)	48.17 (26.45)

* Two participants were excluded (1.3%) due to invalid addresses.

Table 2a shows the relationship between participant reported HHI and neighborhood level MHHI by quintile. Overall, nearly 20% of participants live in neighborhoods discordant with their individual HHI. Table 2b shows the relationship between participant race/ethnicity and neighborhood racial/ethnic composition. Over 22% of participants lived in neighborhoods with a different racial/ethnic majority than their own, being driven mostly by NH Black participants, 31% of whom live in predominantly Latino/Hispanic neighborhoods.

Table 2a

Concordance and Discordance of household income (n = 130*)
Individual HHI Quintile	Neighborhood MHHI Quintile (N (row %))					Concordant (N (row %))	Neutral (N (row %))	Discordant (N (row %))
Individual HHI Quintile	Q1	Q2	Q3	Q4	Q5	Concordant (N (row %))	Neutral (N (row %))	Discordant (N (row %))
Q1	36 (39.56)	39 (42.86)	5 (5.49)	5 (5.49)	6 (6.59)	36 (39.56)	39 (42.86)	16 (17.58)
Q2	8 (47.06)	4 (23.53)	3 (17.65)	2 (11.76)	0 (0)	4 (23.53)	11 (64.71)	2 (11.76)
Q3	2 (28.57)	2 (28.57)	1 (14.29)	1 (14.29)	1 (14.29)	1 (14.29)	3 (42.86)	3 (42.86)
Q4	0 (0)	4 (26.67)	2 (13.33)	1 (6.67)	8 (53.33)	1 (6.67)	10 (66.67)	4 (26.67)
Total	46 (35.38)	49 (37.69)	11 (8.46)	9 (6.92)	15 (11.54)	42 (32.31)	63 (48.46)	25 (19.23)

* Twenty participants (13.3%) were excluded due to choosing not to answer the income question (n = 17), invalid addresses (n = 2), or unreliable neighborhood MHHI data (n = 1)

Table 2b

Concordance and Discordance of race/ethnicity (n = 148*)
Individual Race/Ethnicity	Neighborhood Majority Race/Ethnicity (N (row %))				Concordant (N (row %))	Neutral (N (row %))	Discordant (N (row %))
Individual Race/Ethnicity	Hispanic/Latino	NH Black	NH White	No Majority	Concordant (N (row %))	Neutral (N (row %))	Discordant (N (row %))
Hispanic/Latino	34 (47.89)	4 (5.63)	4 (5.63)	29 (40.85)	34 (47.89)	29 (40.85)	8 (11.27)
NH Black	18 (31.03)	14 (24.14)	2 (3.45)	24 (41.38)	14 (24.14)	24 (41.38)	20 (34.48)
NH Other	3 (50)	1 (16.67)	0 (0)	2 (33.33)	0 (0)	2 (33.33)	4 (66.67)
NH White	2 (15.38)	0 (0)	9 (69.23)	2 (15.38)	9 (69.23)	2 (15.38)	2 (15.38)
Total	57 (38.51)	19 (12.84)	15 (10.14)	57 (38.51)	57 (38.51)	57 (38.51)	34 (22.97)

* Two participants were excluded (1.3%) due to invalid addresses.

This manuscript proposes an approach to defining the concepts of concordance and discordance of individual characteristics with neighborhood-level social determinants. In addition to the American Community Survey, there are an ever-increasing number of publicly available sources of neighborhood-level data. The approach described here to compare self-reported household income to neighborhood-level median household income could be used for any bi-level SDoH that is measured continuously including annual income, net worth, years of education, age, household size, years at same address. The described approach for race/ethnicity could be used for preferred language, country of origin, employment status, home ownership, health insurance, and housing type. In our example, we defined the threshold for concordance at > 50%. While we believe this is an appropriate cut-off for defining concordance with respect to race/ethnicity, other thresholds could be selected, and lower thresholds may be appropriate for other social determinants.

Since neighborhoods have multiple social determinants, individuals may live in neighborhoods that are concordant on some domains but discordant on others. There may be observable patterns of concordance and discordance that reflect specific sociodemographic phenomena such as early-stage gentrification, (white individuals with higher income living in largely non-white, lower income neighborhoods; or late-stage gentrification, (lower income Black individuals living in neutral, higher income neighborhoods); or social preference, (residential settlement patterns based on factors such as race, ethnicity, national origin, immigration status, or age). Given that race/ethnicity is related to many social risks, we posit that there may be combinations of concordance and discordance that form social phenotypes that are associated with specific health risks and outcomes. The proposed method of operationalizing concordance and discordance will facilitate exploration of these patterns and their impact on health.

The data reported here are meant only to illustrate the proposed operational definitions of concordance and discordance. Our small cohort is skewed towards individuals with lower HHIs, reflecting their disproportionate use of the emergency department. Much larger cohorts with paired individual- and neighborhood-level data are needed to accurately describe the patterns of concordance and discordance between individual and neighborhood characteristics and their impact on health outcomes.

People have both individual and neighborhood-level social determinants, and these may not be aligned. We propose a method of defining and operationalizing the concepts of concordance and discordance between individual characteristics and neighborhood-level social determinants. Our proposed approach will enable more robust investigations of SDoH and their impact on health. Such investigations are needed to inform the growing use of SDoH data from various sources for a variety of purposes.

Census block group (CBG)

Emergency Department (ED)

Geographic information system (GIS)

household income (HHI)

Integrated Public Use Microdata Series (IPUMS)

LION (Linear Integrated Ordered Network) Dataset

Median household income (MHHI)

New York City (New York City)

Non-Hispanic (NH)

Social determinants of health (SDoH)

Ethics approval and consent to participate.

Once a patient is determined to be eligible for the study, written informed consent was obtained. Our study was approved by the Icahn School of Medicine Institutional Review Board (IRB).

Consent for publication

Not Applicable

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare that they have no competing interests.”

Funding

There is no source of funding for the research reported.

Authors' contributions

KN and AM analyzed and interpreted the study data concordance and discordance of household income, race/ethnicity. KN, AM, and LDR were major contributors in writing the manuscript. All authors read and approved the final manuscript.

Acknowledgements

Not Applicable

Williams DR, Jackson PB. Social sources of racial disparities in health. Health Aff (Millwood). 2005;24(2):325-34.
Promotion OoDPaH. Healthy People 2020: Social Determinants of Health [Available from: https://www.healthypeople.gov/2020/topics-objectives/topic/social-determinants-of-health.
Office of Disease Prevention and Health Promotion. Healthy People 2020: Social Determinants of Health, June 13th 2019. Available from: https://www.healthypeople.gov/2020/topics-objectives/topic/social-determinants-of-health.
Adler NE, Rehkopf DH. U.S. disparities in health: descriptions, causes, and mechanisms. Annual review of public health. 2008;29:235-52.
Bor J, Cohen GH, Galea S. Population health in an era of rising income inequality: USA, 1980-2015. Lancet (London, England). 2017;389(10077):1475-90.
Daly M, Boyce C, Wood A. A social rank explanation of how money influences health. Health psychology : official journal of the Division of Health Psychology, American Psychological Association. 2015;34(3):222-30.
Bouchard ME, Kan K, Tian Y, Casale M, Smith T, De Boer C, et al. Association Between Neighborhood-Level Social Determinants of Health and Access to Pediatric Appendicitis Care. JAMA Netw Open. 2022;5(2):e2148865.
Maroko AR, Hopper K, Gruer C, Jaffe M, Zhen E, Sommer M. Public restrooms, periods, and people experiencing homelessness: An assessment of public toilets in high needs areas of Manhattan, New York. PloS one. 2021;16(6):e0252946.
Team R. RStudio: Integrated Development Environment for R. RStudio. Boston, MA: PRC; 2022.
Steven Manson JS, David Van Riper, Tracy Kugler, and Steven Ruggles. IPUMS National Historical Geographic Information System. Minneapolis, MN: IPUMS; 2022.
LION Single Line Street Base Map [Available from: https://www1.nyc.gov/site/planning/data-maps/open-data/dwn-lion.page.
Maroko A, Maantay J, Pérez Machado RP, Barrozo LV. Improving Population Mapping and Exposure Assessment: Three-Dimensional Dasymetric Disaggregation in New York City and São Paulo, Brazil. Papers in Applied Geography. 2019;5(1-2):45-57.
Stewart Fotheringham A, Rogerson PA. GIS and spatial analytical problems. International Journal of Geographical Information Systems. 1993;7(1):3-19.

AppendixPriorityEnrollmentSurvey.pdf

Download PDF

Version 1

posted

You are reading this latest preprint version

Measuring Concordance and Discordance between Individual Characteristics and Neighborhood-level Social Determinants of Health

Status:

Version 1

Abstract

Background

Methods

Results

Discussion

Limitations

Conclusion

List of Abbreviations

Declarations

References

Supplementary Files

Status:

Version 1