Racialized environments and syringe services program implementation: County-level factors

Abstract

Objective:

Racialized health inequities in substance use-related harms might emerge from differential access to syringe service programs (SSPs). To explore this, we examined the association between county-level racialized environments, other factors, and (1) SSP presence, and (2) per capita syringe and (3) naloxone distribution.

Methods:

2021 US National Survey of SSP data (n=295/412;72% response rate) was used to identify SSP presence and the sum of syringes and naloxone doses distributed in 2020 by county. Study measures included racial residential segregation (RRS; i.e., divergence and dissimilarity indexes for Black:Non-Hispanic White & Hispanic:Non-Hispanic White) and covariates (i.e., demographic proportions, urban/suburban/rural classifications, 2020 US presidential Republican vote share, and overdose mortality from 2019). We used logit Generalized Estimating Equations to determine factors associated with county-level SSP presence, and zero inflated negative binomial regression models to determine factors associated with per capita syringe and naloxone distribution.

Results:

SSPs were reported in 9% (283/3106) of US counties. SSP presence was associated with higher divergence and dissimilarity indexes, urban and suburban counties, higher opioid overdose mortality, and lower 2020 Republican presidential vote share. Per capita syringes distributed was associated with lower RRS (divergence and Hispanic:White dissimilarity), lower racially minoritized population proportions and rural counties, while per capita naloxone distribution was associated with lower Hispanic and “other” population proportions, and rural counties.

Conclusions:

Racialized environments are associated with SSP presence but not the scope of those programs. Preventing HIV and HCV outbreaks, and overdose deaths requires addressing community level factors that influence SSP implementation and accessibility.

1. INTRODUCTION

1.1. Racialized inequities in health risk associated with injection drug use

Injection drug use has been associated with HIV and HCV transmissions, overdose deaths, skin and soft tissue infections, and infective endocarditis (Collier et al., 2018; Larney et al., 2016). Due to changes in drug use preferences, contamination of the illicit drug supply, and changing routes of administration, the US has seen negative outcomes related to injection-related drug use increase in recent years (Ciccarone, 2021). Along with the ongoing national crisis of overdose deaths (Post et al., 2022), the US is experiencing nationwide epidemics of HCV (Zibbell et al., 2017), soft tissue and skin infections (Ciccarone et al., 2016), and infective endocarditis (Kadri et al., 2019; McCarthy et al., 2020). Further, HIV outbreaks are becoming more common (Lyss et al., 2020).

Racially minoritized people (e.g., American Indian, Alaskan Native, Black people, Hispanic people, Pacific Islander), appear to be disproportionately impacted by these multiple epidemics. For instance, in the latest analysis of overdose deaths in the US, increases in per population rates were highest among people who are American Indian, Alaskan Native, and Black (Han et al., 2022; Lee and Singh, 2023). Black people are overrepresented among those with recent HIV and HCV diagnoses, and Hispanic people are overrepresented among those with HIV (Bradley et al., 2020; Centers for Disease, 2020). These ongoing national epidemics are largely preventable if people who inject drugs are provided with access to key harm reduction equipment, such as sterile syringes (for preventing HIV and HCV transmission) and naloxone (for reversing overdoses), and HIV preexposure prophylaxis (Chen et al., 2024; Tonin et al., 2024).

1.2. Racialized inequities in access to harm reduction services

Studies indicate that harm reduction services (e.g., syringe services programs [SSPs] and naloxone distribution, are often less available to racially minoritized populations (Hollander et al., 2021; Nguyen et al., 2022; Sledge et al., 2022). In the latest National HIV Behavioral Surveillance study, investigators found that SSP use among Black people who inject drugs has declined since 2015 (Handanagic et al., 2021). In local studies, racially minoritized people who inject drugs have been found to have less access to naloxone (Kinnard et al., 2021; Nolen et al., 2022b) and overdose reversal training (Kim et al., 2021). However, in Massachusetts investigators found that municipalities with higher proportion of Black people had higher naloxone coverage rates (Nolen et al., 2022a) yet overdose deaths increased significantly among Black people there during the COVID-19 pandemic (Zang et al., 2023). Examining the availability of harm reduction services at the community level is warranted to better understand barriers to these services and to address persistent racialized inequities in health outcomes related to injection drug use.

1.3. Political determinants of harm reduction service availability

Political controversy has surrounded the implementation of harm reduction services in the US for decades (Des Jarlais et al., 1995). In the lone national study of SSP implementation, AIDS activists were identified as critical to SSP implementation in the 1990s (Tempalski, 2007). This finding has been supported by local studies on SSP presence during this period (Bluthenthal, 1998; Downing et al., 2005; Kochems et al., 1996; Shaw, 2006; Sherman and Purchase, 2001; Wieloch, 2002). Lastly, some studies have found that need – measured as HIV infections among people who inject drugs or drug overdose mortality – did not predict implementation during the 2000s (Tempalski, 2007; Tempalski et al., 2007) or after (Lambdin et al., 2023; Stanton et al., 2022). Research indicates that this pattern has resulted in fewer programs in suburban and rural areas for the first 3 decades of the HIV epidemic (Des Jarlais et al., 2015; Welch-Lazoritz et al., 2017), but we do not know if these placing dynamics contribute to less access to harm reduction services for racially minoritized populations.

Availability of public health prevention interventions and medical services has long been associated with political party preferences. Democratic party affiliation has been associated with higher levels of support for substance use treatment (Cook and Worcman, 2019) while Republican party affiliation was associated with less support for policies like the Affordable Care Act (of 2010 https://www.healthcare.gov/glossary/affordable-care-act/) (Metzl, 2019). One study examining the adoption of legal access to naloxone found that states with more political conservative ideology and higher percentages of evangelical Protestants were slower to allow access to naloxone (Bohler et al., 2021). In another study, respondents with negative attitudes towards Black people were less likely to support government funding for Medicaid, services for people living with opioid use disorder, and even naloxone distribution (Pyra et al., 2022). Assessing the association between political preferences and access to harm reduction services appears to be warranted.

1.4. Structural racism, racialized environments and racialized inequities in health

Structural racism is a set of interlocking historical and contemporary cultural and institutional beliefs and practices that diminish the life opportunities of people who have been racially minoritized in the US (Bailey et al., 2017). Structural racism through racial residential segregation (RRS) and economic inequality that derive from historic practices create racialized environments where residents are vulnerable to social, economic and health disadvantages (Cooper et al., 2016a; Smith et al., 2022). These practices include redlining in the federal home mortgage program, slum removal programs, exclusionary real restate practices, educational racial segregation and other practices that consistently diminish the life opportunities and increase vulnerability to harms among racially minoritized populations in these areas (Bailey et al., 2017; Hanlon, 2011; Kraus et al., 2024Fullilove, 2001 #6724). Studies have found racialized environments are strongly associated with health inequities including preterm births (Krieger et al., 2020; Scally et al., 2018), asthma, (Nardone et al., 2020), child health (Wang et al., 2022), and HIV (Wallace, 1988, 1990, 1991). Among people who inject drugs, at least one study has found that Black people were more likely to reside in RRS areas and to have less access to harm reduction services due to restrictive laws (Cooper et al., 2016b).

In the following, we examine if racialized environments as measured by the Divergence Index and Dissimilarity Index are associated with the availability and accessibility of harm reduction services at the county level in the US while controlling for other factors.

2. METHODS

2.1. Data Sources

Data on SSPs come from the National Survey of Syringe Service Programs (NSSSP). Briefly, this survey involved identifying and recruiting SSP programs in the US. We proactively contacted, searched and followed-up with SSPs from different sources including the North American Syringe Exchange Network (NASEN)’s directory, NASEN’s Buyers Club, state and county public health departments, social media platforms, webinars, conferences, and regional and national networks of SSPs among others. From this database (n=412), SSPs operating within the United States were recruited to complete an online survey using Voxco© (Voxco, Montreal, Quebec, Canada) from February to June 2021 in collaboration with NASEN. A total of 295 SSPs responded to the survey (72% response rate). SSP directors were emailed up to three times asking them to participate in the online survey. For SSPs that did not respond, we conducted individual follow-up with programs via email and/or phone calls. SSPs were offered a $75 honorarium if they completed the survey. The 2021 survey gathered information relevant to 2020 on county where the SSP operated, and numbers of syringes and naloxone doses distributed. Other domains, not discussed in this paper, included funding (Facente et al., 2024), syringe coverage (Tookes et al., 2024), and access to other health services (Lambdin et al., 2022) among other items. Study procedures were approved by the Institutional Review Board of RTI International.

County-level population data was obtained from the American Community Survey (ACS) 2016–2020 5-year estimates (Bureau, 2021). These estimates capture the average population characteristics between 2016–2020 and represent the most precise estimates available, particularly for rural counties. County-level Dissimilarity Indices were obtained from the Opioid Environment Policy Scan Data Warehouse (Paykin et al., 2022).

County-level urbanicity was obtained from the 2013 National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme (Ingram and Franco, 2014), which is the most up-to-date urban-rural classification scheme available from the NCHS. County-level data on Republican vote share for the 2020 presidential election was obtained from the Massachusetts Institute of Technology Election Data and Science Lab County Presidential Election Return 2000–2020 (Lab, 2018).

Finally, we used the 2019 restricted-use, county-level, multiple cause of death files from the National Vital Statistics System (NVSS) (Statistics, 2021a) and the 2019 Vintage Postcensal Bridged Race Population data from NVSS (Statistics, 2021b) to calculate opioid overdose mortality rates.

2.2. Outcomes

The SSP database and National Survey of Syringe Service Program (NSSSP) survey data were geocoded to 2020 Census Bureau cartographic county boundaries (Bureau, 2022); the geocoded location represents the county in which the SSP is headquartered. We constructed three county-level outcome measures. The first outcome measure is whether an SSP operated in the county. This measure was constructed from our geocoded sampling frame of n=412 SSPs known to be operating when the NSSSP was disseminated; this measure is not dependent on whether an SSP completed the survey. Counties with no known SSPs were assigned a value of 0; counties with at least one SSP was assigned a value of 1. The second outcome measure is county-level per capita syringe distribution. This measure was constructed from the n=295 SSPs that responded to the survey. To calculate syringe distribution per 100,000 county population, we summed the total number of syringes distributed across SSPs within a county and divided this by the county population size (Bureau, 2021). In the main analyses, SSPs with missing syringe distribution data were assigned a value of 0 and counties without SSPs were assigned a value of 0. Prior national surveys of SSPs have found that SSPs that do not participate tend to be smaller programs (CDC, 2010; Des Jarlais et al., 2015). To align with this, we conducted sensitivity analyses that imputed missing SSP-level syringe distribution data at the 25th percentile of syringe distribution; county-level per capita estimates were recalculated with imputed data. The final outcome measure is county-level per capita naloxone distribution. This measure was constructed from the n=295 SSPs that responded to the survey. To calculate naloxone distribution per 100,000 county population, we summed the total number of naloxone doses distributed across SSPs within a county and divided this by the county population size (Bureau, 2021). In the main analyses, SSPs with missing naloxone distribution data were assigned a value of 0 and counties without SSPs were assigned a value of 0. In sensitivity analyses, SSPs with missing naloxone distribution data were imputed at the 25^th percentile of naloxone distribution and county-level per capita estimates were recalculated based on imputed data.

2.3. Exposures

The main exposure of interest is RRS as measured by the Divergence (Roberto, 2016) and Dissimilarity Indexes, which we regard as measures of structural racism and racialized environments. The Divergence Index is a spatially explicit measure of RRS that accounts for all racial groups simultaneously, providing a holistic value for the level of segregation in a community. It compares the relative proportions of racial groups at smaller [census tract] and larger [county] geographies, by examining the degree of “divergence” between the two geographies (Institute, 2021). The index equals 0 when no segregation is identified and 1 when segregation is complete. We constructed this measure using racial and ethnicity data from the 2016–2020 ACS (Bureau, 2021) and the R segregation package (https://github.com/arthurgailes/rsegregation). The Dissimilarity Index is the most widely used measure of segregation (Massey and Denton, 1988; Paykin et al., 2022) for Black and non-Hispanic White; Hispanic and non-Hispanic White. The Dissimilarity Index measures the percentage of the racially minoritized group’s population that would have to change residence for each county to have the same percentage of that group as the county overall. The index ranges from 0 to 1, where higher values correspond to higher levels of RRS. Recognizing that the pathways through which racial and ethnic representation operate are different from RRS pathways (White, Kellee and Borrell, Luisa N., 2011), we constructed measures of racial and ethnic composition from the 2016–2020 ACS (Bureau, 2021) as percent of non-Hispanic White [NHWhite], non-Hispanic Black [NHBlack], non-Hispanic Asian [NHAsian], non-Hispanic Other [NHOther], and Hispanic. All exposures were standardized with a mean=0 and standard deviation (standard deviation [SD]) =1.

2.4. Covariates

We constructed a three-tier, county-level measure of urbanicity from the NCHS Urban-Rural Classification Scheme (Ingram and Franco, 2014). Urban counties are located in the 53 metropolitan areas with at least a million people; about 31% of people residing in the US live in these counties. Suburban and small metro counties include those outside the core metro areas, referred to as “large fringe metro,” “medium metro” and “small metro” counties in the NCHS classification system. About half of US residents (55%) live in suburban counties. Rural counties are located in non-metropolitan areas. With a median population size of 16,535, only 14% of US residents live in them.

Using data from the MIT Election Science Lab (Lab, 2018), we calculated the percent of county votes for the Republican presidential candidate in the 2020 election. For each county, the numerator was the number of votes for the Republican presidential candidate and the denominator was the total number of votes. This measure may represent norms for using public resources, support for harm reduction, and/or economic distress (Monnat and Brown, 2017).

To account for underlying need for an SSP within a county, we lagged overdose deaths by one year, calculating 2019 county-level opioid-related overdose deaths per 100,000 county population from NVSS data (Statistics, 2021a, b). We included deaths across the entire age spectrum and the following ICD-10 multiple cause of death codes to capture any opioid-related mortality: T40.0, T40.1, T40.2, T40.3, T40.4, T40.6. We employed empirical Bayesian spatial smoothing to improve the precision of the crude mortality rates by borrowing strength from other observations (see [Saunders et al., 2023] for details). All covariates were standardized with a mean=0 and SD=1.

2.5. Analytic Strategy

Analyses were conducted for N=3,106 counties in the contiguous US (98.8% of all counties). We applied a two-step analytic approach to assess whether structural racism predicted (1) SSP presence, (yes/no), (2) number of syringes distributed per 100,000 county resident, and (3) number of naloxone doses distributed per 100,000 county resident.

First, we used multivariable logit Generalized Estimating Equations (GEE) to model SSP presence as a function of RRS (Outcome= SSP presence, (yes/no)). Four GEE models were estimated such that each RRS measure was modeled separately. Counties with no SSPs were used as the reference group. We used a state-level exchangeable correlation matrix to account for clustering of counties within states (Zeger et al., 1988). Results are represented as adjusted odds ratios (aOR) with 95% confidence intervals (CI).

Second, we used zero inflated negative binomial (ZINB) models to analyze the impacts of RRS on the rate of syringe distribution per 100,000 and the rate of naloxone doses distributed per 100,000 county population. Four models for each outcome (number of syringes distributed per 100,000 and number of naloxone doses distributed per 100,000) were estimated such that each RRS measure was modeled separately. ZINB models simultaneously model two separate distributions: count data using a negative binomial model (number of syringes distributed/number of naloxone doses distributed) and a preponderance of zeros using a logit model (no syringes or naloxone distributed because a county does not have an SSP or a county has an SSP, but survey data is not available). We only present the negative binomial models, with results represented as adjusted incident rate ratios (aIRR) and 95% CI. In sensitivity analyses, we additionally included state ID as a fixed effect to account for counties clustered within states. The models were not substantively different from the main models and are not presented below.

The following variables were included as covariates in all models: urbanicity, the percent of Republican presidential candidate vote share in 2020, which we consider as a proxy for greater likelihood of opposition to harm reduction strategies, and the rate of opioid-related overdose mortality. For brevity, the aOR or aIRR for covariates are presented separately from the RRS measures for each outcome.

We conducted sensitivity analyses to assess potential bias from SSP-level missingness in the syringe and naloxone distribution data. As noted above, we recalculated the county-level per capita syringe and naloxone distributions by imputing missing SSP-level data at the 25^th percentiles of each measure. We chose the 25^th percentile based on prior national SSP surveys reporting that SSPs that did not respond tended to be smaller scale programs (CDC, 2010; Des Jarlais et al., 2015). Using the new imputed outcome measures, we estimated ZINB models for each RRS measure. All variables with an p<0.05 were considered significant. Data preparation were conducted in R version 4.1.1 (Team, 2021). Analyses were conducted in SAS Enterprise Guide version 7.15 (SAS Institute, Inc., 2017).

3. RESULTS

SSPs were located in 9% of counties in the contiguous US (Table 1). The average numbers of syringes distributed within counties was 10,099 per 100,000 population (SD=74,981) and the average number of naloxone doses distributed was 112 per 100,000 population (SD=2,130). The levels of RRS across counties varied greatly, with some counties experiencing no segregation and others experiencing near total segregation. For example, the maximum level of RRS identified by the Dissimilarity Index- Black & non-Hispanic White residents = 0.96 (mean=0.39; SD=0.19); similar maximum values were found for the other RRS and racial and ethnic composition measures. The average county Republican presidential election share was 56% (SD=23). Opioid overdose mortality rates in 2019 ranged from 1.22 to 110.34 per 100,000 residents by county (mean=13.07; SD=8.06).

Table 1:

Characteristics of US counties, 2020 (n=3,106)

Presence of Syringe Service Program, n (%)	283 (9.11%)
Syringe Distribution per 100,000 County Population, mean (sd)*	10,099 (74,981)
Naloxone Dose Distribution per 100,000 County Population, mean (sd)	112 (2,130)
% Voted for Republican in 2020 Presidential election, mean (sd)	56 (23)
Opioid-related mortality, 2019, mean (sd)	13.07 (8.06)
Racial Residential Segregation
Divergence Index, mean (sd)	0.06 (0.07)
Dissimilarity Index- Black & Non-Hispanic White, mean (sd)	0.39 (0.19)
Dissimilarity Index- Hispanic & Non-Hispanic White, mean (sd)	0.30 (0.16)
Racial and Ethnic Proportional Composition
% Non-Hispanic White, mean (sd)	76.12 (19.87)
% Non-Hispanic Black, mean (sd)	8.95 (14.42)
% Non-Hispanic Asian, mean (sd)	1.31 (2.44)
% Non-Hispanic Other, mean (sd)	4.00 (2.60)
% Hispanic, mean (sd)	9.63 (13.98)
Urbanicity
Rural, n (%)	1,947 (63%)
Suburban, n (%)	1,092 (35%)
Urban, n (%)	67 (2%)

^*sd = standard deviation

3.1. GEE models predicting the presence of an SSP in a county

Overall, counties with higher levels of RRS had statistically significantly higher odds of having an SSP present (Figure 1). For each SD increase in the Divergence Index, there was a 33% increase in the odds of a county having an SSP (aOR=1.33, 95% CI: 1.22–1.51). Similar results were found for the Black-White Dissimilarity Index (aOR=1.34, 95% CI: 1.18–1.52) and the Hispanic-White Dissimilarity Index (aOR=1.22, 95% CI: 1.12–1.33). County-level racial and ethnic proportion was not significantly associated with the presence of an SSP.

Figure 1.

Racial Residential Segregation, Proportional Composition and Syringe service program Presence (SSP)

aOR = adjusted odds ratio; LCL=95% lower confidence limit; UCL=95% upper confidence limit; NH = non-Hispanic; Each racial residential segregation measure and the proportional racial composition measure was modeled separately. Each model adjusted for urbanicity, the percent of county residents that voted for the Republican presidential candidate in 2020, and the rate of opioid-related overdose mortality.

3.2. ZINB models predicting the rate of syringes distributed per 100k county population

For per capita syringes distributed, RRS was inversely associated with syringes distributed (Figure 2). For each SD increase in the Divergence Index, the rate of syringe distribution decreased by 33% (aIRR=0.67, 95% CI: 0.54–0.83). Similar results were found for the Hispanic-White Dissimilarity Index (aIRR=0.65, 95% CI: 0.48–0.88). Counties with higher proportions of non-Hispanic Black residents distributed fewer syringes (aIRR=0.69, 95% CI: 0.49–0.97).

Figure 2.

Racial Residential Segregation, Proportional Composition, and per Capita Syringe Distribution

aIRR = adjusted incident rate ratio; LCL=95% lower confidence limit; UCL=95% upper confidence limit; NH = non-Hispanic; Each racial residential segregation measure and the proportional racial composition measure was modeled separately. Each model adjusted for urbanicity, the percent of county residents that voted for the Republican presidential candidate in 2020, and the rate of opioid-related overdose mortality.

3.3. ZINB models predicting the rate of naloxone dose distribution per 100k county population

County-level RRS was not significantly associated with rates of naloxone distribution (Figure 3). Counties with higher proportions of people classified as “Other” and non-Hispanic White distributed fewer naloxone doses (%NH Other poulations - aIRR=0.60, 95% CI: 0.49–0.74; %Hispanic - aIRR=0.97, 95% CI: 0.95–0.98).

Figure 3.

Racial Residential Segregation, Proportional Composition and per Capita Naloxone Distribution

aIRR = adjusted incident rate ratio; LCL=95% lower confidence limit; UCL=95% upper confidence limit; NH = non-Hispanic; Each racial residential segregation measure and the proportional racial composition measure was modeled separately Each model adjusted for urbanicity, the percent of county residents that voted for the Republican presidential candidate in 2020, and the rate of opioid-related overdose mortality.

3.4. Association between community characteristics and SSP Implementation

We also observed associations between other community characteristics and SSP presence and per capita syringe and naloxone distribution (Table 2). After accounting for other community-level factors, urban and suburban counties had significantly higher odds of having an SSP (urban aOR=16.71, 95% CI=7.89, 34.42; suburban aOR=1.95, 95% CI=1.51, 2.50), but lower levels of syringe distribution (urban aIRR=0.03, 95% CI=0.01, 0.11; suburban aIRR=0.10, 95% CI=0.05, 0.10) and naloxone distribution (urban aIRR=0.01, 95% CI=0.00, 0.04; suburban aIRR=0.05, 95% CI=0.02, 0.10). Counties with a higher proportion of Republican votes in the 2020 presidential election had significantly lower odds of having an SSP (aOR=0.51, 95% CI=0.33, 0,78), Counties with higher opioid overdose mortality rates had significantly higher odds of having an SSP (aOR=1.40, 95% CI=1.25, 1.57), but this indicator of need was not associated with per capita syringe or naloxone distribution.

Table 2.

Community Characteristics and Syringe Service Program (SSP) implementation, per capita syringe distribution and per capita naloxone distribution.

	SSP Implementation			Per Capita Syringe Distribution			Per Capita Naloxone Distribution
	aOR	95% CI	p-value	aIRR	95% CI	p-value	aIRR	95% CI	p-value
Urbanicity (ref=rural)
Suburban	1.95	(1.51, 2.50)	<.0001	0.10	(0.05, 0.18)	<.0001	0.05	(0.02, 0.10)	<.0001
Urban	16.71	(7.89, 34.42)	<.0001	0.03	(0.01, 0.11)	<.0001	0.01	(0.0, 0.04)	<.0001
% Voted Republican in 2020
Presidential Election	0.51	(0.33, 078)	0.002	0.77	(0.52, 1.14)	0.957	1.08	(0.71, 1.64)	0.729
Opioid overdose mortality	1.40	(1.25, 1.57)	<.0001	1.10	(0.96, 1.27)	0.097	1.17	(0.99, 1.38)	0.058

aOR = adjusted odds ratio; aIRR = adjusted incidence rate ratio; CI=confidence interval; Models are additionally adjusted for proportional racial composition.

3.5. Sensitivity analyses

We found no substantive differences in per capita syringe or naloxone distribution when we imputed missing SSP-level data at the 25th percentiles of each measure (Table 3).

Table 3.

Estimates from Main- and Sensitivity-Analyses of Racial Residential Segregation, Proportional Composition and per Capita Syringe and Naloxone Distribution.

	Syringe Distribution per 100k County Population				Naloxone Distribution per 100k County Population
	aIRR	95% Confidence Limits		Pr > |Z|	aIRR	95% Confidence Limits		Pr > |Z|
Divergence Index
Main Analysis	0.67	0.54	0.83	0.000	0.99	0.73	1.35	0.954
Sensitivity Analysis, 25th pctl	0.66	0.54	0.81	<.0001	0.95	0.71	1.27	0.745
Dissimiliarty Index NH Black
Main Analysis	0.84	0.58	1.21	0.348	0.64	0.33	1.25	0.196
Sensitivity Analysis, 25th pctl	0.82	0.58	1.16	0.268	0.63	0.35	1.12	0.116
Dissimiliarty Index Hispanic
Main Analysis	0.65	0.48	0.88	0.005	0.82	0.50	1.33	0.419
Sensitivity Analysis, 25th pctl	0.66	0.49	0.89	0.006	0.80	0.51	1.25	0.326
% NH Black
Main Analysis	0.69	0.49	0.97	0.031	1.07	0.71	1.60	0.754
Sensitivity Analysis, 25th pctl	0.71	0.51	0.97	0.033	1.04	0.70	1.54	0.852
% NH Asian
Main Analysis	0.89	0.79	1.01	0.066	0.92	0.81	1.05	0.204
Sensitivity Analysis, 25th pctl	0.89	0.80	0.99	0.038	0.90	0.80	1.02	0.103
% NH Other
Main Analysis	0.86	0.65	1.13	0.284	0.60	0.49	0.74	<.0001
Sensitivity Analysis, 25th pctl	0.93	0.74	1.15	0.488	0.77	0.63	0.94	0.011
% Hispanic
Main Analysis	1.01	0.99	1.02	0.424	0.97	0.95	0.98	0.000
Sensitivity Analysis, 25th pctl	1.00	0.99	1.02	0.641	0.97	0.95	0.98	<.0001

Pctl = percentile; aIRR = adjusted incident rate ratio; LCL=95% lower confidence limit; UCL=95% upper confidence limit; NH = non-Hispanic; Each racial residential segregation measure and the proportional racial composition measure was modeled separately. Each model adjusted for urbanicity, the percent of county residents that voted for the Republican presidential candidate in 2020, and the rate of opioid-related overdose mortality.

4. DISCUSSION

Our finding that counties with higher levels of RRS were associated with higher odds of having an SSP tracks closely with other studies that have found RRS to be related to higher levels of community resilience, mutual aid, and cultures of care that respond to immediate community needs (White, K. and Borrell, L. N., 2011). Due to legacies of neighborhood disinvestment, these types of community-based approaches do not rely on other institutions, including federal, state or county agencies, to address community needs. Rather, movements are built by passionate, dedicated groups of individuals who build ground-up responses that address the urgent needs of the community—as was the case for SSP implementation in the early and mid-1990s (Bluthenthal, 1998; Henman et al., 1998; Sherman and Purchase, 2001; Wieloch, 2002).

We also found that counties with lower levels of RRS had higher per capita syringe distributions. Per capita material distribution is largely a function of financial and human resources of organizations operating within the county. Prior studies have documented pathways in which higher levels of RRS contribute to health inequities, including neighborhood disinvestment, socioeconomic disadvantage, and inadequate economic opportunities (Williams et al., 2019). It follows logically that SSPs operating in places with higher RRS could also have access to fewer financial resources, compromising organizations’ abilities to staff and build programs that achieve greater per capita syringe distribution.

Our findings also showed that counties with a higher proportion of Black residents had lower per capita syringe distribution. These findings are particularly concerning because Black people who use opioid and stimulant drugs remain at elevated risk for nearly all the ailments common to populations of people who use drugs (Bradley et al., 2020; Friedman et al., 2021). Studies indicate SSPs contribute to better access to syringes (to prevent infectious disease transmission) among people who use drugs (Dunleavy et al., 2017; Grau et al., 2002; Lambdin et al., 2020). These findings align with other studies showing that residents in communities with a higher concentration of Black residents have restricted access to quality health care and higher prevalence of other factors that shape racialized health inequities (Fullilove 2001; Wang et al., 2022). In addition, it could be that support for SSPs is lower, especially considering the extensive toll of criminalization of drug use and racialized policing on Black communities (Alexander, 2010). It is also possible that Black people who use drugs might be less likely to engage with SSPs due to fear of police encounters and harassment (Beletsky et al., 2015; Bluthenthal et al., 1997; Davis et al., 2005). More research to understand the impact of racialized policing and the criminalization of drug use as well as the structural determinants of SSP accessibility in these communities is warranted.

SSP presence also varied by Republican presidential vote share. The political determinants of health—how voting, policies, and government operations influence health—has been studied extensively in other health areas (Dawes, 2020). Recent examples demonstrate how political dynamics have led to health crises in communities with a high proportion of Black residents, including the water crises in Flint, MI, and Jackson, MS (Mizelle, 2023; Zahran et al., 2017). There is an ongoing debate about how Republican presidential vote share contributes to the poorer health outcomes even while acknowledging that significant differences exist for specific types of mortality (Paul et al., 2022; Rodríguez et al., 2022) and overall mortality (Hamamsy et al., 2021; Warraich et al., 2022). Our findings indicate that one contribution of Republican voting might be the determination to not implement evidence-based interventions that are unsupported by Republican voters. Other investigations have shown that Republican-controlled state legislatures delayed implementation of Medicaid expansion under the Affordable Care Act (Rocco et al., 2020), increased access to firearms (Cherney et al., 2022), and most recently, declining to accept HIV care and prevention funding from the federal government (Sasani, 2023). More research is needed to understand how to overcome Republican tendencies to prefer policies that contribute to ongoing and persistent epidemics, and chronic health conditions..

We found that rural counties were consistently associated with higher per capita syringe distribution and naloxone distribution. We attribute this finding to lower population counts, not to more generous syringe and naloxone distribution policies. Indeed, rural SSPs appear to be more vulnerable to local and state regulations that limit access to syringes and naloxone (Allen et al., 2019; Batty et al., 2023; Palayew et al., 2023) and increased distances and transportation barriers to reaching services as demonstrated in prior studies (Romo et al., 2023; Thakarar et al., 2021).

Study results should be considered in light of potential methodological limitations. Most, but not all known SSPs, contribute data to the NSSSP database. Further, there may have been other SSPs operating in the United States unknown to us. We proactively reached out to SSPs identified from a variety of sources, but the possibility of other SSPs, especially newer, unsanctioned, or smaller scale programs, exist. Data collected in the NSSSP is self-reported and thus may be subject to social desirability and recall bias. Regarding the latter, questionnaire items are limited to the previous year in an effort to reduce recall-based bias.

Our unit of analysis for this study was counties in the contiguous U.S. Counties represent geographic boundaries that do not correspond to populations either in terms of size or characteristics. For our outcome variables, this could be problematic since the impact of an SSP in a county could have a wide range based on coverage of the priority population. We attempted to control for this by examining syringes and naloxone distributed per capita. However, distance to SSP distribution sites has been known to be strongly associated with use patterns and effects, so even this mitigation approach is limited (Rockwell et al., 1999; Whiteman et al., 2020).

There are also limitations related to the independent variables. First, RSS is a measure designed for urban locales. Most of the counties in the US are not urban, where this measure is more precise. Similarly, many public health decisions and programs are implemented at a state-level and not at the county-level. However, as it regards SSPs, states, including states that have more recently allowed SSPs, have left the final decision regarding SSP implementation to county and local governments (Bluthenthal et al., 2008 & https://www.lac.org/assets/files/final-SSP-issue-brief.pdf). Lastly, as noted above, the non-urban character of most counties means that the associations with Republican presidential voting patterns may be overestimated. For instance, in California the presidential Republican overall vote share was 32%, but the average Republican vote share by county was 43%. However, our results align well with other findings indicating that endorsement of conservative ideology typically is inversely associated with support for preventive approaches to substance use-related health problems (Bohler et al., 2021; Cloud et al., 2018).

5. Conclusion

We found that communities with higher RRS were more likely to have SSPs, and once established, these organizations were subject to structural racist forces that limited access to preventive services and drive health inequities in these communities. It is imperative that we understand and address the community-level dynamics that restrict the critical functions SSPs can have in providing access to key harm reduction services to improve the health and well-being of people who use drugs.

Highlights.

• Racialized inequities in substance use-related harms persist

• Harm reduction services are critical to addressing overdose mortality

• Diverse county-level factors contribute to harm reduction services availability

• Racial residential segregation is associated with harm reduction services availability

• Political health determinants must be addressed to optimize harm reduction services

Data Statement:

Data will be made available to other investigators 1 year following the completion of this study.