County-level social vulnerability and nonfatal drug overdose emergency department visits and hospitalizations, January 2018–December 2020

Abstract

Background:

Nonfatal drug overdoses (NFODs) are often attributed to individual behaviors and risk factors; however, identifying community-level social determinants of health (SDOH) associated with increased NFOD rates may allow public health and clinical providers to develop more targeted interventions to address substance use and overdose health disparities. CDC’s Social Vulnerability Index (SVI), which aggregates social vulnerability data from the American Community Survey to produce ranked county-level vulnerability scores, can help identify community factors associated with NFOD rates. This study aims to describe associations between county-level social vulnerability, urbanicity, and NFOD rates.

Methods:

We analyzed county-level 2018–2020 emergency department (ED) and hospitalization discharge data submitted to CDC’s Drug Overdose Surveillance and Epidemiology system. Counties were ranked in vulnerability quartiles based on SVI data. We used crude and adjusted negative binomial regression models, by drug category, to calculate rate ratios and 95% confidence intervals comparing NFOD rates by vulnerability.

Results:

Generally, as social vulnerability scores increased, ED and hospitalization NFOD rates increased; however, the magnitude of the association varied across drugs, visit type, and urbanicity. SVI-related theme and individual variable analyses highlighted specific community characteristics associated with NFOD rates.

Conclusions:

The SVI can help identify associations between social vulnerabilities and NFOD rates. Development of an overdose-specific validated index could improve translation of findings to public health action. The development and implementation of overdose prevention strategies should consider a socioecological perspective and address health inequities and structural barriers associated with increased risk of NFODs at all levels of the social ecology.

1. Introduction

Over 106,000 drug overdose deaths were reported in the U.S. in 2021 (Spencer et al., 2022). Prior research found that in 2018, for each fatal overdose reported, there were nearly ten nonfatal drug overdoses (NFODs) treated in an emergency department (ED) (National Center for Health Statistics, 2021; Pickens et al., 2021). People who have a NFOD often experience medical complications (Office of the Assistant Secretary for Planning and Evaluation. 2019) and are at increased risk of subsequent overdose (Suffoletto and Zeigler, 2020). NFODs also represent a substantial economic cost in the U.S., where opioid use disorder was estimated to cost $470 billion in 2017 (Florence et al., 2021). While many public health interventions focus on individual behaviors associated with overdose, previous research highlights the impact that social and institutional inequities, as well as social determinants of health (SDOH), have on community-level overdose rates (Amaro et al., 2021; Boardman et al., 2001; Galea and Vlahov, 2002; Shiels et al., 2019). Identifying community-level SDOH that influence risk of NFOD can inform upstream population-level prevention strategies and along with individual-level factors be used to develop multi-faceted overdose interventions.

There is increasing interest in identifying and addressing community health disparities in overdose-related outcomes. One recent analysis found that county-level socioecological characteristics were associated with higher opioid dispensing rates (Cremer et al., 2021). Additionally, investigators have examined the relationship between NFOD and patient race and ethnicity (Hasegawa et al., 2014; Khatri et al., 2021) and insurance status (Grossman and Hamman, 2020); however, comprehensive analysis of the associations between SDOH and NFODs is limited. The Centers for Disease Control and Prevention (CDC) Social Vulnerability Index (SVI) tool can facilitate analysis of associations with SDOH (Centers for Disease Control and Prevention. 2018). In recent years, the SVI has informed public health planning and evaluation of health equity in COVID-19 vaccination campaigns (Barry et al., 2021), HIV diagnosis (Dailey et al., 2022), and adult obesity (An and Xiang, 2015).

Identifying correlations between NFOD, social vulnerability, and urbanicity could highlight community-level social influences on NFOD rates and inform substance use and overdose prevention and response strategies to target identified disparities. We address an important gap in the literature by examining the association between county-level NFOD rates across the U.S. and a wide range of social factors. Vulnerability indices have also been analyzed in relation to nonfatal and fatal drug-related outcomes using data from emergency medical services (EMS), vital statistics, and electronic health records. Two studies have used ED or hospitalization data; these included data from a limited number of EDs in a single state and included all “drug-involved” visits (Sauer et al., 2021) or a general drug overdose category (Cobert et al., 2020). Our present study expands on these results by including over 2300 facilities across 23 states and more granular drug category data. Our first aim was to understand how the SVI correlates with county population-based rates of NFOD ED visits and hospitalizations for all drugs, all opioids, heroin, and all stimulants during 2018–2020 using discharge data from the CDC Drug Overdose Surveillance and Epidemiology (DOSE) System (A.M. Vivolo-Kantor et al., 2021). Our second aim was to explore differences in associations between social vulnerability and NFODs treated in EDs and those requiring hospitalization, which has yet to be described in the literature.

2. Methods

2.1. CDC’s DOSE system

CDC’s DOSE System leverages health record data to identify, track, and respond to emerging trends in NFODs within and among 47 states and the District of Columbia (hereafter, states) receiving funding for participation (A.M. Vivolo-Kantor et al., 2021). Twenty-five states submitted aggregate discharge data to CDC for ED visits only (N=4 states), hospitalizations only (N=3 states), or both (N=18 states) on a quarterly basis, with a one-quarter time lag (e.g., data from October–December 2019 were submitted in April 2020) for the time period January 1, 2018, through December 31, 2020. Discharge data on over 90% of all emergency visits (ED and hospitalization) within each included state are reported. Data include counts of NFOD-related ED visits and hospitalizations for four drug categories (all drugs, all opioids, heroin, and all stimulants), by county of patient residence, based on International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) discharge diagnosis codes submitted for medical billing (A. Vivolo-Kantor et al., 2021). State reporting by county of patient residence excludes nonresident NFOD visits. Additionally, patients initially treated in an ED and subsequently admitted to the hospital were excluded from the ED visit dataset. Case definitions were nested; for example, an individual experiencing a heroin overdose would also be categorized as an all opioid overdose and an all drug overdose. Self-harm NFODs, which were identified by ICD-10-CM codes for intentional/self-harm poisonings, were excluded from analysis.

Two of the 25 states had incomplete county-level data and were excluded from analyses due to incomplete reporting (Nebraska and Oklahoma), leaving 20 states (1449 counties) submitting ED data and 19 states (1279 counties) submitting hospitalization data (Fig. 1).

Fig. 1.

States (N=25) Participating in CDC’s DOSE System That Report Emergency Department and Hospitalization Discharge Data, 2018–2020. States that report discharge data but were excluded from analysis due to incomplete reporting (n=2) are marked as “data incomplete”. Twenty states submitted ED discharge data, and 19 states submitted hospitalization discharge data.

2.2. CDC/Agency for Toxic Substances and Disease Registry (ATSDR) social vulnerability index

We used the SVI to measure county-level social vulnerability, which comprises 15 ACS variables used to rank U.S. counties from 0 to 1 (low to high relative vulnerability), with an overall composite rank score, four SVI theme scores, and: (1) Socioeconomic Status (individuals below poverty level, unemployment among civilians aged ≥16 years, per capita income, adults aged ≥25 years with no high school diploma), (2) Household Composition and Disability (persons aged ≥65 years, persons aged ≤17 years, population living with disability, single-parent households), (3) Minority Status and Language (population identifying as members of a racial or ethnic group other than White, non-Hispanic, speak English “less than well”), and (4) Housing Type and Transportation (multi-unit housing, mobile homes, crowded housing, group quarters, and vehicle access). 2018 SVI data were obtained from CDC/Agency for Toxic Substances and Disease Registry’s (ATSDR’s) Geospatial Research, Analysis, and Services Program’s (GRASP’s) publicly available database, which represents the most current sociodemographic census data available for U.S. populations at the time of analysis. SVI methodological documentation, including calculation of relative vulnerability, is available from GRASP (Centers for Disease Control and Prevention, 2018). In addition to the SVI variables, we examined percentage uninsured, a variable included with the SVI data but not incorporated into the index, and urbanicity. Patient county of residence urbanicity was dichotomized as rural (referent; non-core counties) and urban (large metropolitan counties, medium/small metropolitan counties, and micropolitan urban clusters), based on the National Center for Health Statistics’ 2013 Urban-Rural Classification Scheme for Counties (Centers for Disease Control and Prevention, 2013). We classified overall SVI, themes, and individual variable scores into quartiles (using national percentile cutoffs) for our analyses (Q1=lowest vulnerability to Q4=highest vulnerability).

2.3. Data analysis

2.3.1. NFOD rates

Annual and quarterly NFOD rates were calculated per 100,000 county population for each of the four NFOD categories by visit type (i.e., ED visit, hospitalization). Annual population denominators for 2018–2020 were obtained from the U.S. Census Vintage 2020 county population estimates (United States Census Bureau, 2021). Quarterly county population denominators were estimated through linear extrapolation from the U.S. Census-reported 2018–2020 annual population estimates for each county. Chi-square tests were used to determine significance of annual rate changes between 2018 and 2020; p-values <0.05 were considered statistically significant. Analysis was completed using SAS (v9.4; SAS Institute, Cary, NC).

2.3.2. Negative binomial regression models

We calculated rate ratios (RRs) utilizing a generalized estimating equations (GEE) approach to identify associations between county-level social vulnerability (defined as overall SVI, SVI theme, and individual variables) and NFODs. We accounted for low counts with a negative binomial distribution and population offset; the GEE approach allowed for the assessment of county data correlated over time in three-month aggregates. Using correlation matrices to assess multicollinearity between predictor variables, collinearity was identified between income per capita and percentage of population below the poverty level. These variables were removed from the adjusted regression analysis examining individual variables. We report unadjusted RRs, adjusted rate ratios (aRRs), and 95% confidence intervals (CIs) comparing the rates of NFOD by drug category in the lowest vulnerability counties (quartile 1) with higher vulnerability counties in quartiles 2, 3, and 4. We adjusted the overall SVI model for urbanicity, theme models for the other SVI themes and urbanicity, and individual variable models for the other individual variables (SVI individual variables, proportion uninsured and urbanicity). Analyses were performed in R 4.0.4 (RStudio Team, 2021) using geeM and MASS packages.

3. Results

3.1. ED visits

From January 2018 through December 2020, 537,549 all drug NFOD ED visits were reported to DOSE by the 20 states reporting ED visits. During this time, all drug NFOD rates increased 6.4% (115.7 per 100,000 to 123.2); opioid-involved overdoses increased 31.0% from 43.2 to 56.6, and heroin-involved overdoses increased 4.4% from 27.1 to 28.3; all changes were significant. Quarterly all drug, all opioid-, heroin-, and all stimulant-involved NFOD rates demonstrated seasonality, where the highest rates for each drug category were reported July–September, and the lowest rates were reported October–March during the study period (Fig. 2).

Fig. 2.

Rates of Nonfatal Drug Overdose per 100,000 population, by quarter, 2018–2020 (a) Emergency Department Visits, 20 states, and (b) Hospitalizations, 19 states. Emergency department data from 20 states were included in this analysis: AK, CA, CO, GA, HI, IA, IL, IN, KS, KY, MI, MN, MO, MS, NY, OR, RI, SC, UT and WI. Numerator (drug overdose emergency department visits by patient county of residence) data come from CDC’s Drug Overdose Surveillance and Epidemiology (DOSE) system (Centers for Disease Control and Prevention, 2020). Quarterly county population denominators were estimated through linear extrapolation from the U.S. Census-reported 2018–2020 annual population estimates for each county in the 2020 Vintage Population Estimates (https://www.census.gov/programs-surveys/popest/technical-documentation/research/evaluation-estimates/2020-evaluation-estimates/2010s-counties-total.html). (b) Hospitalization data from 19 states were included in this analysis: AK, CO, DE, GA, HI, IL, IN, KS, KY, MN, MO, MS, NC, NY, OR, RI, SC, UT and WA. Numerator (drug overdose hospitalizations by patient county of residence) data come from CDC’s Drug Overdose Surveillance and Epidemiology (DOSE) system.(Centers for Disease Control and Prevention, 2020) Quarterly county population denominators were estimated through linear extrapolation from the U.S. Census-reported 2018–2020 annual population estimates for each county in the 2020 Vintage Population Estimates (https://www.census.gov/programs-surveys/popest/technical-documentation/research/evaluation-estimates/2020-evaluation-estimates/2010s-counties-total.html).

All drug, all opioid-, heroin-, and all stimulant-involved NFOD rates increased as overall social vulnerability increased; however, the highest rates were found among the Q3 vulnerability counties for overdoses involving all drugs (aRR= 1.44, 95% CI: 1.42–1.46), all opioids (aRR=1.55, 1.51–1.59), and heroin (aRR= 1.61, 1.53–1.70). In contrast, all stimulant-involved NFOD rates were highest among the most vulnerable counties (Q4 aRR=1.75, 1.66–1.84). A similar pattern was seen in Theme 1 (SES) and Theme 4 (housing and transportation) (Table 1). NFOD rates for all four drug categories were highest in the counties with highest Theme 2 (household and disability) vulnerability, and rates in all categories were lowest in counties with highest Theme 3 (minority status) vulnerability, indicating that counties with larger minority populations had lower ED visit NFOD rates.

Table 1.

Associations between county-level social vulnerability and rates of nonfatal drug overdoses involving all drugs, all opioids, heroin, and all stimulants among emergency departments in 20 states, 2018–2020.^a

SVI Theme or variable and Quartile	All Drugs (n=537,549)		All Opioids (n=218,174)		Heroin (n=121,387)		All Stimulants (n=35,794)
	RR (95% CI)b	aRR (95% CI)c	RR (95% CI)b	aRR (95% CI)c	RR (95% CI)b	aRR (95% CI)c	RR (95% CI)b	aRR (95% CI)c
Overall SVId
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.22 (1.20,1.24)	1.23 (1.21,1.25)	1.23 (1.21,1.26)	1.29 (1.27,1.31)	1.26 (1.23,1.29)	1.34 (1.31,1.38)	1.28 (1.21,1.35)	1.27 (1.20,1.35)
Q3	1.43 (1.41,1.46)	1.44 (1.42,1.46)	1.51 (1.47,1.55)	1.55 (1.51,1.59)	1.55 (1.47,1.64)	1.61 (1.53,1.70)	1.66 (1.59,1.74)	1.66 (1.58,1.74)
Q4 (highest vulnerability)	1.30 (1.28,1.33)	1.33 (1.30,1.36)	1.11 (1.03,1.20)	1.23 (1.14,1.33)	0.90 (0.82,0.99)	1.03 (0.93,1.13)	1.78 (1.70,1.88)	1.75 (1.66,1.84)
Theme 1: SVI related to SESe SVI Theme 1
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.32 (1.30,1.34)	1.25 (1.23,1.27)	1.41 (1.38,1.44)	1.44 (1.41,1.48)	1.46 (1.42,1.50)	1.49 (1.45,1.53)	1.40 (1.36,1.44)	1.25 (1.21,1.28)
Q3	1.49 (1.47,1.52)	1.36 (1.33,1.38)	1.68 (1.65,1.71)	1.76 (1.71,1.81)	1.76 (1.68,1.84)	1.85 (1.76,1.93)	1.63 (1.57,1.69)	1.35 (1.28,1.42)
Q4 (highest vulnerability)	1.33 (1.30,1.37)	1.22 (1.17,1.27)	1.10 (1.01,1.21)	1.25 (1.15,1.36)	0.84 (0.72,0.97)	0.96 (0.84,1.10)	1.86 (1.80,1.93)	1.47 (1.38,1.56)
Percent civilian unemployed
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.36 (1.32,1.39)	1.23 (1.20,1.26)	1.64 (1.56,1.73)	1.46 (1.37,1.57)	1.85 (1.74,1.98)	1.68 (1.54,1.83)	1.66 (1.54,1.79)	1.17 (1.11,1.24)
Q3	1.47 (1.43,1.50)	1.26 (1.23,1.28)	1.87 (1.80,1.94)	1.58 (1.49,1.68)	2.11 (2.01,2.22)	1.87 (1.74,2.01)	1.61 (1.51,1.72)	1.28 (1.20,1.36)
Q4 (highest vulnerability)	1.50 (1.47,1.53)	1.35 (1.33,1.37)	1.64 (1.56,1.72)	1.68 (1.60,1.77)	1.66 (1.54,1.79)	2.04 (1.91,2.19)	1.87 (1.76,1.99)	1.43 (1.32,1.55)
Percent with no high school diploma
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.33 (1.31,1.35)	1.20 (1.19,1.22)	1.48 (1.44,1.53)	1.44 (1.38,1.50)	1.58 (1.52,1.65)	1.57 (1.48,1.67)	1.43 (1.37,1.49)	1.18 (1.14,1.22)
Q3	1.39 (1.37,1.41)	1.30 (1.27,1.33)	1.56 (1.52,1.60)	1.72 (1.66,1.78)	1.54 (1.49,1.59)	1.93 (1.81,2.05)	1.54 (1.49,1.59)	1.24 (1.18,1.31)
Q4 (highest vulnerability)	1.19 (1.16,1.22)	1.13 (1.10,1.17)	0.98 (0.90,1.08)	1.25 (1.19,1.33)	0.70 (0.62,0.79)	1.13 (1.07,1.21)	1.62 (1.56,1.68)	1.20 (1.12,1.28)
Theme 2: SVI related to household composition & disability statusf SVI Theme 2
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.18 (1.16,1.19)	1.11 (1.09,1.12)	1.13 (1.11,1.15)	1.06 (1.04,1.08)	1.11 (1.09,1.13)	1.05 (1.02,1.08)	1.30 (1.26,1.34)	1.17 (1.13,1.21)
Q3	1.25 (1.23,1.27)	1.13 (1.11,1.15)	1.11 (1.09,1.12)	0.99 (0.97,1.01)	1.06 (1.01,1.11)	0.97 (0.93,1.02)	1.48 (1.42,1.56)	1.21 (1.15,1.27)
Q4 (highest vulnerability)	1.35 (1.32,1.37)	1.21 (1.18,1.24)	1.18 (1.12,1.24)	1.11 (1.06,1.15)	1.10 (1.01,1.19)	1.15 (1.09,1.20)	1.68 (1.60,1.75)	1.26 (1.18,1.35)
Percent aged 65 or older
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.13 (1.12,1.15)	0.96 (0.94,0.98)	1.02 (0.99,1.04)	0.84 (0.82,0.86)	1.03 (0.99,1.06)	0.83 (0.79,0.87)	1.22 (1.17,1.27)	0.97 (0.92,1.02)
Q3	1.03 (1.02,1.05)	0.87 (0.85,0.90)	0.85 (0.83,0.86)	0.68 (0.65,0.71)	0.81 (0.76,0.85)	0.62 (0.57,0.67)	1.19 (1.13,1.26)	0.90 (0.84,0.96)
Q4 (highest vulnerability)	0.95 (0.93,0.98)	0.85 (0.82,0.88)	0.71 (0.68,0.75)	0.61 (0.57,0.64)	0.62 (0.59,0.66)	0.53 (0.48,0.58)	1.03 (0.97,1.10)	0.81 (0.73,0.89)
Percent aged 17 or younger
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.10 (1.08,1.12)	1.04 (1.02,1.06)	1.13 (1.09,1.17)	0.98 (0.95,1.02)	1.20 (1.16,1.24)	1.00 (0.96,1.04)	1.20 (1.13,1.26)	1.07 (1.01,1.13)
Q3	1.12 (1.09,1.15)	1.03 (1.00,1.05)	1.21 (1.15,1.27)	0.93 (0.88,0.98)	1.29 (1.23,1.36)	0.95 (0.87,1.02)	1.13 (1.08,1.18)	1.01 (0.96,1.07)
Q4 (highest vulnerability)	0.97 (0.95,1.00)	0.97 (0.95,0.98)	0.95 (0.91,0.98)	0.80 (0.77,0.84)	0.93 (0.89,0.97)	0.83 (0.76,0.89)	0.97 (0.93,1.01)	0.96 (0.92,1.01)
Percent older than aged 5 with a disability
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.28 (1.26,1.30)	1.16 (1.14,1.18)	1.33 (1.30,1.36)	1.22 (1.18,1.26)	1.40 (1.37,1.44)	1.20 (1.16,1.24)	1.38 (1.33,1.43)	1.12 (1.07,1.18)
Q3	1.35 (1.34,1.37)	1.26 (1.24,1.29)	1.36 (1.33,1.40)	1.39 (1.33,1.46)	1.37 (1.33,1.41)	1.37 (1.32,1.44)	1.49 (1.43,1.55)	1.15 (1.07,1.23)
Q4 (highest vulnerability)	1.35 (1.33,1.37)	1.28 (1.24,1.32)	1.17 (1.11,1.24)	1.36 (1.30,1.42)	1.01 (0.92,1.12)	1.34 (1.28,1.42)	1.82 (1.72,1.93)	1.32 (1.23,1.41)
Percent single-parent households
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.21 (1.18,1.24)	1.14 (1.11,1.17)	1.37 (1.30,1.43)	1.19 (1.14,1.25)	1.55 (1.49,1.61)	1.30 (1.23,1.38)	1.22 (1.17,1.27)	1.14 (1.09,1.20)
Q3	1.41 (1.37,1.46)	1.21 (1.18,1.24)	1.72 (1.65,1.80)	1.30 (1.24,1.36)	2.09 (2.02,2.17)	1.48 (1.40,1.56)	1.60 (1.53,1.68)	1.34 (1.27,1.42)
Q4 (highest vulnerability)	1.40 (1.35,1.44)	1.18 (1.15,1.22)	1.62 (1.53,1.70)	1.17 (1.11,1.23)	1.89 (1.79,2.00)	1.24 (1.15,1.33)	1.59 (1.53,1.65)	1.22 (1.16,1.29)
Theme 3: SVI related to racial/ethnic minority statusg SVI Theme 3
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	0.99 (0.98,1.00)	0.97 (0.96,0.98)	1.01 (0.99,1.04)	0.94 (0.91,0.97)	1.04 (0.99,1.08)	0.91 (0.88,0.95)	0.92 (0.88,0.97)	0.95 (0.90,1.00)
Q3	0.97 (0.95,0.99)	0.93 (0.91,0.95)	1.02 (0.99,1.04)	0.90 (0.88,0.93)	1.03 (0.99,1.06)	0.85 (0.81,0.88)	0.80 (0.75,0.85)	0.83 (0.78,0.90)
Q4 (highest vulnerability)	0.95 (0.93,0.96)	0.89 (0.87,0.91)	1.00 (0.90,1.00)	0.86 (0.82,0.91)	0.93 (0.88,0.97)	0.77 (0.72,0.82)	0.84 (0.78,0.90)	0.81 (0.75,0.88)
Percent minority
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.08 (1.06,1.10)	1.04 (1.01,1.06)	1.08 (1.06,1.10)	1.06 (1.02,1.10)	1.29 (1.25,1.34)	1.03 (0.97,1.09)	0.95 (0.90,1.00)	1.03 (0.98,1.09)
Q3	1.10 (1.07,1.12)	0.97 (0.94,1.00)	1.10 (1.07,1.12)	0.87 (0.81,0.93)	1.28 (1.22,1.33)	0.78 (0.71,0.85)	0.92 (0.86,0.98)	0.93 (0.85,1.01)
Q4 (highest vulnerability)	0.96 (0.94,0.98)	0.80 (0.78,0.83)	0.96 (0.94,0.98)	0.68 (0.64,0.73)	0.85 (0.81,0.89)	0.52 (0.47,0.57)	0.89 (0.83,0.96)	0.77 (0.69,0.87)
Percent speaking English “less than well”
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	0.94 (0.93,0.96)	0.97 (0.95,0.99)	0.95 (0.91,0.99)	0.97 (0.93,1.01)	0.91 (0.87,0.96)	0.91 (0.86,0.97)	0.86 (0.80,0.91)	0.94 (0.88,1.00)
Q3	0.98 (0.97,0.99)	1.03 (1.01,1.05)	1.10 (1.07,1.13)	1.13 (1.09,1.18)	1.12 (1.07,1.17)	1.11 (1.06,1.16)	0.77 (0.73,0.81)	0.97 (0.91,1.03)
Q4 (highest vulnerability)	0.91 (0.89,0.93)	1.01 (0.97,1.05)	0.94 (0.90,0.98)	1.00 (0.95,1.06)	0.89 (0.83,0.95)	0.89 (0.85,0.95)	0.80 (0.75,0.85)	1.10 (1.01,1.20)
Theme 4: SVI related to housing type & transportationh SVI Theme 4
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.16 (1.14,1.19)	1.09 (1.06,1.11)	1.15 (1.12,1.18)	1.06 (1.03,1.09)	1.15 (1.11,1.19)	1.07 (1.03,1.12)	1.23 (1.17,1.29)	1.10 (1.05,1.15)
Q3	1.22 (1.21,1.24)	1.12 (1.10,1.13)	1.24 (1.21,1.26)	1.09 (1.08,1.11)	1.28 (1.23,1.32)	1.18 (1.14,1.22)	1.29 (1.22,1.37)	1.15 (1.08,1.22)
Q4 (highest vulnerability)	1.20 (1.19,1.22)	1.09 (1.06,1.11)	1.12 (1.08,1.16)	1.00 (0.97,1.02)	1.08 (1.03,1.12)	1.02 (0.98,1.06)	1.41 (1.34,1.49)	1.20 (1.13,1.28)
Percent living in multiunit structures
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.07 (1.04,1.11)	1.05 (1.03,1.08)	1.22 (1.17,1.26)	1.05 (1.02,1.09)	1.38 (1.30,1.47)	1.03 (0.97,1.1)	1.00 (0.92,1.09)	1.06 (0.98,1.15)
Q3	1.06 (1.03,1.09)	1.08 (1.05,1.11)	1.20 (1.16,1.25)	1.03 (0.99,1.07)	1.47 (1.37,1.58)	1.03 (0.97,1.11)	0.85 (0.79,0.91)	1.02 (0.94,1.11)
Q4 (highest vulnerability)	1.05 (1.02,1.09)	1.10 (1.07,1.14)	1.41 (1.35,1.48)	1.14 (1.08,1.20)	1.85 (1.68,2.03)	1.20 (1.10,1.30)	0.77 (0.72,0.82)	1.05 (0.94,1.17)
Percent living in mobile homes
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.07 (1.06,1.09)	0.99 (0.97,1.01)	0.96 (0.93,0.98)	0.93 (0.90,0.97)	0.89 (0.86,0.93)	0.89 (0.85,0.93)	1.18 (1.15,1.22)	1.09 (1.05,1.13)
Q3	1.06 (1.05,1.08)	0.91 (0.88,0.93)	0.89 (0.86,0.92)	0.80 (0.77,0.83)	0.79 (0.73,0.85)	0.72 (0.67,0.77)	1.30 (1.25,1.34)	1.05 (0.97,1.14)
Q4 (highest vulnerability)	1.12 (1.08,1.15)	1.04 (1.00,1.08)	0.86 (0.78,0.94)	0.97 (0.87,1.07)	0.63 (0.54,0.72)	0.82 (0.71,0.94)	1.58 (1.53,1.63)	1.17 (1.08,1.26)
Percent living in crowded homes
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.07 (1.05,1.08)	0.99 (0.97,1.01)	1.16 (1.13,1.19)	1.01 (0.99,1.03)	1.14 (1.11,1.18)	1.02 (0.98,1.06)	1.04 (1.00,1.07)	0.96 (0.91,1.00)
Q3	1.08 (1.05,1.11)	0.97 (0.94,1.00)	1.08 (1.05,1.12)	0.99 (0.96,1.01)	1.01 (0.96,1.05)	0.99 (0.94,1.04)	1.07 (1.02,1.13)	0.90 (0.85,0.96)
Q4 (highest vulnerability)	0.98 (0.96,0.99)	0.95 (0.91,0.98)	0.89 (0.87,0.92)	0.89 (0.85,0.92)	0.73 (0.69,0.77)	0.84 (0.77,0.92)	0.96 (0.91,1.01)	0.88 (0.81,0.95)
Percent with no vehicle
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.12 (1.09,1.16)	1.03 (1.00,1.06)	1.08 (1.05,1.11)	1.00 (0.96,1.05)	1.08 (1.03,1.12)	1.02 (0.98,1.07)	1.16 (1.11,1.22)	0.98 (0.94,1.03)
Q3	1.22 (1.20,1.26)	1.01 (0.99,1.04)	1.19 (1.16,1.22)	0.96 (0.93,0.99)	1.21 (1.18,1.25)	1.02 (0.98,1.05)	1.35 (1.26,1.45)	0.99 (0.93,1.05)
Q4 (highest vulnerability)	1.32 (1.29,1.35)	1.11 (1.08,1.14)	1.35 (1.31,1.39)	1.16 (1.12,1.21)	1.45 (1.40,1.51)	1.33 (1.27,1.39)	1.59 (1.51,1.67)	1.13 (1.08,1.19)
Percent living in group quarters
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.09 (1.07,1.10)	1.05 (1.03,1.06)	1.06 (1.04,1.08)	1.02 (0.99,1.05)	1.10 (1.07,1.13)	1.02 (0.98,1.06)	1.18 (1.13,1.23)	1.07 (1.02,1.12)
Q3	1.10 (1.09,1.12)	1.02 (1.01,1.04)	1.03 (1.01,1.05)	0.92 (0.90,0.95)	1.04 (1.02,1.06)	0.85 (0.83,0.86)	1.25 (1.18,1.32)	1.13 (1.06,1.20)
Q4 (highest vulnerability)	1.08 (1.06,1.10)	0.95 (0.93,0.97)	0.99 (0.97,1.01)	0.79 (0.75,0.82)	1.03 (0.99,1.07)	0.73 (0.68,0.79)	1.28 (1.21,1.35)	1.07 (1.01,1.14)
Additional Variables Percent uninsured
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.06 (1.05,1.08)	0.98 (0.96,0.99)	1.02 (0.99,1.06)	0.94 (0.91,0.97)	0.92 (0.88,0.96)	0.87 (0.83,0.91)	1.05 (1.02,1.08)	0.95 (0.93,0.98)
Q3	1.05 (1.03,1.08)	0.94 (0.91,0.96)	0.97 (0.91,1.04)	0.88 (0.83,0.93)	0.81 (0.75,0.87)	0.80 (0.74,0.85)	1.17 (1.11,1.23)	1.01 (0.96,1.06)
Q4 (highest vulnerability)	0.91 (0.88,0.94)	0.82 (0.80,0.84)	0.67 (0.61,0.73)	0.64 (0.60,0.69)	0.47 (0.41,0.55)	0.55 (0.48,0.62)	1.09 (1.02,1.17)	0.90 (0.86,0.95)
Urbanicityi
Rural	*	*	*	*	*	*	*	*
Urban	1.08 (1.06,1.11)	1.04 (1.01,1.07)	1.49 (1.45,1.53)	1.30 (1.25,1.35)	1.87 (1.76,1.99)	1.46 (1.39,1.53)	0.83 (0.80,0.85)	0.92 (0.88,0.95)

Note. CI=confidence interval; RR= rate ratio; aRR= adjusted rate ratio. Boldface indicates statistical significance (p<0.05).

^aThe overall SVI, each of four SVI themes, and 16 social vulnerability variables were categorized into quartiles. Emergency department data from 20 states were included in this analysis: AK, CA, CO, GA, HI, IA, IL, IN, KS, KY, MI, MN, MO, MS, NY, OR, RI, SC, UT and WI. Numerator (drug overdose emergency department visits by patient county of residence) data come from CDC’s Drug Overdose Surveillance and Epidemiology (DOSE) system.(Centers for Disease Control and Prevention, 2020) County population denominators for 2018–2020 were obtained from the 2020 Vintage Population Estimates (https://www.census.gov/programs-surveys/popest/technical-documentation/research/evaluation-estimates/2020-evaluation-estimates/2010s-counties-total.html).

^bFor the overall SVI, each SVI theme, each of the included 13 social vulnerability variables, percent uninsured, and urbanicity we used negative binomial regression to calculate unadjusted rate ratios (95% CIs) comparing the rates of drug overdoses among counties in quartiles 2, 3, and 4 of the SVI variable to counties in quartile 1 of the SVI variable (referent; designated with *).

^cCollinearity was identified between theme 1 social vulnerability variables; therefore, the percent below poverty level and per capita income variables were removed from the individual variable analysis. For the overall SVI, each SVI theme, and the remaining 13 social vulnerability variables we used negative binomial regression to calculate adjusted rate ratios (95% CIs) comparing the rates of drug overdoses in counties with SVI values in quartiles 2, 3, and 4 to counties with SVI values in quartile 1 (referent; designated with *). Point estimates for overall SVI were adjusted for urbanicity. Point estimates for each SVI theme are adjusted for the other three SVI themes (i.e., we used one model with four SVI theme variables: SES theme, household composition/disability theme, racial/ethnic minority status theme, and housing type/transportation theme) and urbanicity. Point estimates for each social vulnerability variable are adjusted for the other included social vulnerability variables and urbanicity.

^dThe overall SVI includes 15 variables in the index: percent below poverty level, percent unemployed among civilians aged ≥16 years, per capita income, adults aged ≥25 years with no high school diploma, percent aged 65 or older, percent aged 17 or younger, percent older than age 5 with a disability, percent single-parent households, percent identifying as a race or ethnicity other than White, non-Hispanic, percent speaking English “less than well,” percent living in multiunit structures, percent living in mobile homes, percent living in crowded homes (i.e., >1 occupant per room), percent with no vehicle, and percent living in group quarters (i.e., a group living arrangement that is owned or managed by an entity or organization providing housing and/or services for the residents including treatment facilities, skilled nursing facilities, group homes, correctional facilities, and facilities for people experiencing homelessness).

^eIncludes percent below poverty level, percent unemployed among civilians aged ≥16 years, per capita income, adults aged ≥25 years with no high school diploma.

^fIncludes percent population aged 65 or older, percent population aged 17 or younger, percent population older than age 5 with a disability, and percent of households.

^gIncludes percent identifying as a race or ethnicity other than White, non-Hispanic and percent speaking English “less than well.”

^hIncludes percent living in multiunit structures, percent living in mobile homes, percent living in crowded homes, percent with no vehicle, and percent living in group quarters.

ⁱUrbanicity data were obtained from the 2013 National Center of Health Statistics Rural-Urban Classification Scheme for Counties (https://www.cdc.gov/nchs/data_access/urban_rural.htm). Patient counties of residence were classified as non-core were coded as rural; otherwise, counties were coded as urban. The ED visit dataset included 1449 counties (935 rural, 514 urban).

Among ED visits, many socioecological factors were found to be associated with higher NFOD rates for all categories. Counties with higher percentages of unemployment, adults with no high school diploma, people with a disability, or single parent households had higher NFOD rates involving all drugs, all opioids, heroin, and all stimulants. Counties with the highest percentage population with no vehicle had higher NFOD rates than counties with the lowest percentage; other quartiles had equivalent or slightly lower NFOD rates compared to Q1. In general, lower NFOD rates were associated with counties where a greater proportion of people were older, uninsured, or living in crowded housing. As county minority population increased, ED NFOD rates decreased or remained equivalent for heroin-involved and stimulant-involved NFOD. Compared to counties with the lowest percentage of residents identifying as part of a racial or ethnic minority group, counties with the next lowest percentage of residents identifying as part of a racial or ethnic minority group (Q2) were marginally associated with higher all drug and all opioid-involved NFOD rates; however, counties with the highest percentage of residents identifying as part of a racial or ethnic minority group (Q3 and Q4) were associated with lower or equivalent rates of all drug or all opioid-involved NFOD rates compared to Q1.

Stimulant-involved NFODs displayed several inverse associations compared to the other drug categories. All drug, all opioid-, and heroin-involved NFOD rates were highest in urban counties and in those counties with the highest percentage of multi-unit housing, and lowest in counties with the highest percentage of younger people and highest percentage of population living in group quarters. Additionally, greater percentage of mobile home housing was associated with lower or equivalent All drug, all opioid-, and heroin-involved NFOD rates. All stimulant-involved NFOD rates were slightly lower in urban counties and slightly higher in counties with more group quarters and mobile home housing. Unlike the results that were found with other drug categories, percentage multi-unit housing and population ≤17 years generally were not associated with all stimulant-involved NFOD rates.

3.2. Hospitalizations

From January 2018 through December 2020, 182,274 all drug NFOD hospitalizations were reported to DOSE by the 19 states in our hospitalizations analysis. The all drug NFOD rates among patients decreased 7.4% during this period (55.6 per 100,000 to 51.5). All opioid- and heroin-involved NFOD rates remained stable at 18.7 and 5.6, respectively, and all stimulant-involved NFOD rates increased 4.2% from 10.8 to 11.2; however, the decrease in the all-drug NFOD rate was the only significant change among hospitalized NFODs. Like ED visits, quarterly NFOD hospitalization rates fluctuated seasonally across drug categories. The highest rates were reported April–September, and the lowest rates were reported October–March each year (Fig. 2).

Similar to NFOD ED visit rates, increased overall county-level social vulnerability was associated with higher NFOD hospitalization rates involving all drugs (Q4 aRR=2.30, 2.14–2.48), all opioids (Q4 aRR=1.86, 1.70–2.04), heroin (Q3 aRR=2.09, 1.81–2.40), and all stimulants (Q4 aRR=3.60, 3.35–3.86) (Table 2). Among hospitalizations, higher county-level vulnerability for SVI themes 3 and 4 (minority status and housing and transportation, respectively) was associated with increased NFOD rates for all four drug categories. Theme 1 (SES) NFOD hospitalization rates increased with increasing vulnerability across all drug categories in Q2 and Q3 vulnerability quartiles, however compared to the least vulnerable counties (Q1), hospitalization rates in the most vulnerable counties (Q4) only increased for all drug NFODs. Additionally, Q4 all opioid-involved and all stimulant-involved NFOD rates were equivalent to those in Q1, while all heroin-involved NFOD rates were lower in Q4 counties. Conversely all drug NFOD rates increased as vulnerability increased For Theme 2 (household and disability), the most vulnerable counties had equivalent rates to the least vulnerable, except all stimulant-involved NFOD, for which rates were higher among the most vulnerable counties. Twenty-one percent of NFOD hospitalizations involved stimulants, while only 6.7% of NFOD ED visits involved stimulants. Heroin-involved NFODs made up a smaller proportion of hospitalizations (10.2% of NFODs) compared to ED visits (22.6%).

Table 2.

Associations between county-level social vulnerability and rates of nonfatal drug overdoses involving all drugs, all opioids, heroin, and all stimulants among hospitalizations in 19 states, 2018–2020.^a

SVI Theme or variable and Quartile	All Drugs (n=182,274)		All Opioids (n=61,957)		Heroin (n=18,617)		All Stimulants (n=38,171)
	RR (95% CI)b	aRR (95% CI)c	RR (95% CI)b	aRR (95% CI)c	RR (95% CI)b	aRR (95% CI)c	RR (95% CI)b	aRR (95% CI)c
Overall SVId
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.25 (1.22,1.27)	1.50 (1.42,1.58)	1.25 (1.22,1.29)	1.48 (1.36,1.61)	1.19 (1.12,1.27)	1.54 (1.33,1.78)	1.46 (1.38,1.54)	2.06 (1.89,2.24)
Q3	1.51 (1.48,1.55)	1.81 (1.73,1.89)	1.52 (1.46,1.57)	1.75 (1.61,1.91)	1.62 (1.50,1.76)	2.09 (1.81,2.40)	1.98 (1.87,2.09)	2.76 (2.54,2.99)
Q4 (highest vulnerability)	1.66 (1.62,1.70)	2.30 (2.14,2.48)	1.58 (1.50,1.66)	1.86 (1.70,2.04)	1.46 (1.33,1.59)	1.78 (1.50,2.10)	2.45 (2.35,2.55)	3.60 (3.35,3.86)
Theme 1: SVI related to SESe SVI Theme 1
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.28 (1.25,1.30)	1.26 (1.20,1.30)	1.26 (1.23,1.30)	1.25 (1.17,1.33)	1.35 (1.29,1.42)	1.32 (1.25,1.40)	1.50 (1.46,1.54)	1.29 (1.16,1.42)
Q3	1.46 (1.44,1.48)	1.34 (1.30,1.40)	1.53 (1.48,1.58)	1.32 (1.23,1.41)	1.65 (1.53,1.78)	1.53 (1.42,1.65)	1.71 (1.66,1.76)	1.18 (1.03,1.36)
Q4 (highest vulnerability)	1.46 (1.44,1.48)	1.46 (1.30,1.60)	1.51 (1.46,1.57)	1.09 (0.99,1.20)	1.19 (1.08,1.32)	0.79 (0.67,0.94)	2.13 (2.06,2.20)	1.05 (0.92,1.21)
Percent civilian unemployed
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.40 (1.36,1.45)	1.66 (1.52,1.82)	1.39 (1.32,1.46)	1.51 (1.32,1.72)	1.64 (1.55,1.73)	1.62 (1.33,1.97)	1.57 (1.50,1.64)	1.53 (1.36,1.72)
Q3	1.52 (1.48,1.56)	1.55 (1.40,1.70)	1.55 (1.47,1.64)	1.58 (1.38,1.81)	2.02 (1.91,2.13)	1.82 (1.56,2.13)	1.86 (1.77,1.95)	1.47 (1.30,1.66)
Q4 (highest vulnerability)	1.76 (1.72,1.80)	1.84 (1.69,2.02)	1.72 (1.63,1.81)	1.68 (1.46,1.93)	2.34 (2.20,2.49)	2.35 (1.87,2.96)	2.37 (2.31,2.43)	1.68 (1.38,2.06)
Percent with no high school diploma
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.29 (1.27,1.32)	1.27 (1.19,1.36)	1.34 (1.32,1.37)	1.39 (1.25,1.55)	1.38 (1.33,1.42)	1.45 (1.34,1.57)	1.35 (1.31,1.39)	1.18 (1.05,1.33)
Q3	1.38 (1.36,1.41)	1.37 (1.24,1.51)	1.44 (1.39,1.49)	1.39 (1.22,1.58)	1.50 (1.39,1.61)	1.42 (1.23,1.64)	1.49 (1.45,1.53)	1.04 (0.89,1.22)
Q4 (highest vulnerability)	1.50 (1.48,1.53)	1.36 (1.21,1.53)	1.52 (1.47,1.57)	1.49 (1.26,1.76)	1.22 (1.11,1.35)	1.69 (1.33,2.15)	1.77 (1.70,1.84)	1.02 (0.86,1.21)
Theme 2: SVI related to household composition & disability statusf SVI Theme 2
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.19 (1.17,1.20)	1.10 (1.10,1.10)	1.21 (1.18,1.25)	1.08 (1.02,1.14)	0.95 (0.92,0.99)	0.89 (0.82,0.97)	1.08 (1.05,1.11)	1.04 (0.92,1.18)
Q3	1.29 (1.26,1.32)	1.10 (1.00,1.20)	1.33 (1.28,1.37)	1.16 (1.09,1.23)	1.04 (0.99,1.10)	0.67 (0.57,0.79)	1.22 (1.17,1.27)	0.95 (0.82,1.10)
Q4 (highest vulnerability)	1.40 (1.38,1.42)	1.07 (1.00,1.20)	1.35 (1.31,1.40)	0.98 (0.91,1.06)	1.03 (0.95,1.12)	0.87 (0.74,1.02)	1.54 (1.51,1.58)	1.19 (1.10,1.28)
Percent aged 65 or older
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.15 (1.13,1.16)	0.91 (0.86,0.97)	1.12 (1.09,1.15)	0.83 (0.76,0.91)	1.00 (0.95,1.04)	0.96 (0.81,1.13)	1.07 (1.04,1.10)	0.96 (0.86,1.06)
Q3	1.03 (1.01,1.05)	0.69 (0.64,0.76)	1.02 (0.98,1.05)	0.60 (0.53,0.67)	0.75 (0.71,0.79)	0.45 (0.34,0.59)	0.92 (0.89,0.95)	0.56 (0.46,0.68)
Q4 (highest vulnerability)	0.97 (0.94,1.00)	0.58 (0.51,0.65)	1.01 (0.97,1.06)	0.53 (0.46,0.60)	0.78 (0.72,0.85)	0.85 (0.57,1.28)	0.78 (0.74,0.82)	0.50 (0.39,0.64)
Percent aged 17 or younger
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.07 (1.03,1.11)	0.99 (0.90,1.09)	1.04 (1.01,1.08)	0.97 (0.87,1.09)	1.13 (1.06,1.20)	0.85 (0.72,1.01)	1.05 (0.99,1.12)	0.95 (0.85,1.05)
Q3	1.12 (1.10,1.14)	0.93 (0.84,1.02)	1.08 (1.05,1.11)	0.75 (0.71,0.81)	1.02 (0.96,1.08)	0.62 (0.53,0.72)	1.13 (1.07,1.20)	0.77 (0.68,0.87)
Q4 (highest vulnerability)	0.99 (0.97,1.02)	0.80 (0.71,0.90)	0.94 (0.91,0.98)	0.62 (0.58,0.67)	0.86 (0.81,0.91)	0.53 (0.45,0.63)	0.98 (0.92,1.04)	0.72 (0.62,0.84)
Percent older than aged 5 with a disability
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.25 (1.23,1.28)	1.34 (1.30,1.38)	1.29 (1.26,1.32)	1.53 (1.46,1.61)	1.20 (1.15,1.24)	1.32 (1.14,1.53)	1.23 (1.20,1.26)	1.44 (1.32,1.57)
Q3	1.34 (1.32,1.37)	1.55 (1.45,1.67)	1.37 (1.34,1.41)	1.75 (1.62,1.90)	0.99 (0.93,1.05)	1.86 (1.30,2.65)	1.33 (1.29,1.36)	1.76 (1.48,2.09)
Q4 (highest vulnerability)	1.42 (1.38,1.46)	1.84 (1.65,2.06)	1.43 (1.38,1.48)	1.92 (1.75,2.12)	0.77 (0.71,0.83)	0.74 (0.38,1.43)	1.38 (1.32,1.45)	1.69 (1.28,2.23)
Percent single-parent households
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.19 (1.15,1.23)	1.20 (1.08,1.32)	1.19 (1.12,1.25)	1.36 (1.22,1.52)	1.21 (1.13,1.29)	1.49 (1.28,1.73)	1.13 (1.07,1.18)	1.26 (1.10,1.44)
Q3	1.28 (1.24,1.32)	1.20 (1.09,1.31)	1.33 (1.27,1.39)	1.37 (1.22,1.54)	1.55 (1.46,1.64)	1.54 (1.28,1.84)	1.37 (1.31,1.43)	1.47 (1.29,1.68)
Q4 (highest vulnerability)	1.45 (1.40,1.51)	1.08 (0.96,1.22)	1.43 (1.34,1.53)	1.16 (0.99,1.36)	1.53 (1.42,1.65)	1.18 (0.89,1.57)	1.78 (1.70,1.86)	1.39 (1.19,1.63)
Theme 3: SVI related to racial/ethnic minority statusg SVI Theme 3
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.00 (0.98,1.03)	1.05 (1.00,1.10)	1.04 (0.99,1.09)	1.28 (1.16,1.42)	1.19 (1.06,1.34)	1.22 (0.75,1.98)	1.01 (0.96,1.07)	1.33 (1.07,1.65)
Q3	1.10 (1.07,1.12)	1.32 (1.20,1.40)	1.12 (1.09,1.15)	1.88 (1.70,2.07)	1.32 (1.22,1.43)	2.74 (1.70,4.41)	1.10 (1.05,1.16)	2.21 (1.64,2.97)
Q4 (highest vulnerability)	1.14 (1.12,1.16)	1.38 (1.30,1.40)	1.15 (1.12,1.18)	2.47 (2.24,2.73)	1.82 (1.66,1.99)	5.79 (3.84,8.72)	1.43 (1.38,1.49)	3.87 (2.80,5.35)
Percent minority
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.09 (1.06,1.11)	1.23 (1.13,1.33)	1.09 (1.06,1.11)	1.40 (1.15,1.69)	1.36 (1.19,1.54)	1.66 (0.81,3.36)	1.11 (1.05,1.18)	1.29 (0.95,1.76)
Q3	1.19 (1.16,1.22)	1.33 (1.21,1.46)	1.19 (1.16,1.22)	1.66 (1.36,2.02)	1.77 (1.61,1.94)	2.36 (1.17,4.72)	1.23 (1.17,1.31)	1.51 (1.08,2.12)
Q4 (highest vulnerability)	1.22 (1.20,1.25)	1.05 (0.95,1.16)	1.22 (1.20,1.25)	1.39 (1.13,1.71)	2.00 (1.80,2.21)	2.12 (1.11,4.05)	1.73 (1.66,1.81)	1.65 (1.18,2.30)
Percent speaking English “less than well”
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	0.97 (0.94,1.00)	1.17 (1.08,1.27)	0.99 (0.96,1.03)	1.19 (1.06,1.33)	1.06 (0.99,1.13)	0.79 (0.48,1.30)	0.98 (0.95,1.01)	1.07 (0.92,1.24)
Q3	1.04 (1.03,1.05)	1.44 (1.32,1.57)	1.09 (1.06,1.12)	1.59 (1.37,1.86)	1.30 (1.22,1.39)	1.47 (0.96,2.24)	1.06 (1.04,1.08)	1.36 (1.17,1.57)
Q4 (highest vulnerability)	1.03 (1.01,1.05)	1.39 (1.25,1.55)	1.08 (1.05,1.11)	1.69 (1.46,1.95)	1.74 (1.61,1.87)	2.57 (1.82,3.63)	1.21 (1.19,1.23)	1.70 (1.43,2.02)
Theme 4: SVI related to housing type & transportationh SVI Theme 4
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.15 (1.12,1.18)	1.16 (1.10,1.20)	1.15 (1.11,1.20)	1.10 (1.00,1.21)	1.24 (1.16,1.32)	1.78 (1.56,2.03)	1.31 (1.26,1.37)	1.49 (1.28,1.73)
Q3	1.23 (1.20,1.26)	1.25 (1.20,1.30)	1.19 (1.15,1.24)	1.15 (1.04,1.27)	1.50 (1.44,1.56)	2.09 (1.82,2.41)	1.55 (1.46,1.64)	1.92 (1.67,2.22)
Q4 (highest vulnerability)	1.36 (1.33,1.39)	1.50 (1.40,1.60)	1.25 (1.21,1.29)	1.22 (1.11,1.34)	1.68 (1.60,1.76)	2.07 (1.79,2.41)	1.95 (1.86,2.04)	2.33 (1.98,2.74)
Percent living in multiunit structures
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.06 (1.04,1.09)	1.44 (1.31,1.58)	1.13 (1.06,1.20)	1.79 (1.43,2.24)	1.34 (1.18,1.53)	1.72 (0.96,3.08)	1.06 (1.00,1.13)	1.82 (1.34,2.49)
Q3	1.00 (0.97,1.02)	1.65 (1.54,1.76)	0.99 (0.94,1.04)	1.92 (1.60,2.29)	1.45 (1.25,1.69)	2.24 (1.28,3.92)	0.90 (0.84,0.97)	2.13 (1.51,3.00)
Q4 (highest vulnerability)	1.02 (1.00,1.05)	1.95 (1.80,2.12)	1.01 (0.97,1.06)	2.68 (2.23,3.22)	1.87 (1.60,2.19)	3.85 (2.10,7.05)	1.13 (1.08,1.18)	4.57 (3.33,6.27)
Percent living in mobile homes
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	0.98 (0.96,1.00)	0.85 (0.80,0.90)	0.99 (0.95,1.02)	0.75 (0.70,0.80)	0.71 (0.66,0.76)	0.56 (0.44,0.70)	0.91 (0.89,0.94)	0.96 (0.85,1.08)
Q3	1.06 (1.04,1.08)	0.81 (0.74,0.87)	1.09 (1.07,1.12)	0.71 (0.64,0.79)	0.78 (0.74,0.84)	0.59 (0.44,0.77)	0.91 (0.87,0.96)	0.81 (0.72,0.90)
Q4 (highest vulnerability)	1.26 (1.24,1.29)	1.10 (1.00,1.22)	1.28 (1.24,1.32)	0.84 (0.77,0.92)	0.70 (0.63,0.78)	0.69 (0.45,1.05)	1.27 (1.22,1.32)	1.14 (0.93,1.40)
Percent living in crowded homes
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.09 (1.06,1.12)	1.07 (0.99,1.15)	1.14 (1.10,1.17)	1.22 (1.10,1.35)	1.13 (1.05,1.22)	1.23 (1.01,1.49)	1.23 (1.17,1.29)	1.33 (1.15,1.55)
Q3	1.24 (1.20,1.28)	1.16 (1.06,1.26)	1.25 (1.21,1.29)	1.27 (1.14,1.41)	1.25 (1.14,1.37)	1.03 (0.77,1.37)	1.41 (1.35,1.48)	1.39 (1.23,1.58)
Q4 (highest vulnerability)	1.15 (1.12,1.18)	0.99 (0.92,1.06)	1.12 (1.07,1.17)	0.96 (0.85,1.07)	1.33 (1.21,1.46)	0.77 (0.58,1.02)	1.42 (1.34,1.49)	1.28 (1.14,1.44)
Percent with no vehicle
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.11 (1.09,1.14)	1.10 (1.03,1.17)	1.09 (1.05,1.13)	0.92 (0.82,1.04)	1.11 (1.06,1.16)	0.85 (0.71,1.03)	1.21 (1.15,1.27)	1.13 (1.02,1.24)
Q3	1.21 (1.19,1.24)	0.99 (0.93,1.05)	1.13 (1.10,1.16)	0.73 (0.68,0.79)	1.27 (1.18,1.36)	0.94 (0.78,1.13)	1.48 (1.42,1.54)	1.20 (1.07,1.36)
Q4 (highest vulnerability)	1.42 (1.40,1.45)	1.20 (1.11,1.30)	1.32 (1.28,1.36)	1.04 (0.91,1.18)	1.84 (1.76,1.92)	1.22 (1.02,1.47)	2.05 (1.97,2.12)	1.95 (1.63,2.34)
Percent living in group quarters
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.04 (1.02,1.06)	1.01 (0.94,1.08)	1.01 (0.99,1.04)	0.98 (0.88,1.09)	1.45 (1.39,1.51)	1.38 (1.27,1.50)	1.20 (1.15,1.26)	1.23 (1.08,1.40)
Q3	1.11 (1.08,1.14)	1.03 (0.94,1.12)	1.09 (1.04,1.14)	1.00 (0.88,1.13)	1.56 (1.50,1.62)	1.41 (1.27,1.58)	1.42 (1.38,1.47)	1.13 (0.98,1.29)
Q4 (highest vulnerability)	1.00 (0.98,1.02)	0.81 (0.75,0.87)	0.94 (0.92,0.97)	0.59 (0.52,0.66)	1.16 (1.10,1.23)	0.67 (0.53,0.86)	1.14 (1.08,1.20)	0.60 (0.49,0.73)
Additional Variables Percent uninsured
Q1 (Lowest vulnerability)	*	*	*	*	*	*	*	*
Q2	1.08 (1.05,1.12)	0.92 (0.85,0.99)	1.11 (1.07,1.15)	1.00 (0.90,1.10)	0.90 (0.85,0.95)	0.94 (0.84,1.06)	1.06 (1.01,1.12)	1.04 (0.86,1.25)
Q3	1.26 (1.23,1.29)	0.98 (0.91,1.06)	1.32 (1.27,1.37)	1.14 (1.03,1.27)	1.21 (1.15,1.28)	1.03 (0.91,1.17)	1.51 (1.45,1.58)	1.56 (1.38,1.76)
Q4 (highest vulnerability)	1.22 (1.18,1.26)	0.87 (0.78,0.96)	1.26 (1.19,1.32)	0.97 (0.83,1.13)	0.74 (0.66,0.82)	0.51 (0.42,0.62)	1.37 (1.31,1.43)	1.09 (0.91,1.30)
Urbanicityi
Rural	*	*	*	*	*	*	*	*
Urban	1.13 (1.11,1.16)	1.99 (1.87,2.12)	1.15 (1.10,1.21)	2.54 (2.24,2.87)	2.10 (1.88,2.35)	9.65 (3.69,25.19)	1.23 (1.18,1.29)	3.28 (2.38,4.50)

Note. CI=confidence interval; RR= rate ratio; aRR= adjusted rate ratio. Boldface indicates statistical significance (p<0.05).

^aThe overall SVI, each of four SVI themes, and 16 social vulnerability variables were categorized into quartiles based on ranked relative vulnerability. Hospitalization data from 19 states were included in this analysis: AK, CO, DE, GA, HI, IL, IN, KS, KY, MN, MO, MS, NC, NY, OR, RI, SC, UT and WA. Numerator (drug overdose hospitalizations by patient county of residence) data come from CDC’s Drug Overdose Surveillance and Epidemiology (DOSE) system.(Centers for Disease Control and Prevention, 2020) County population denominators for 2018–2020 were obtained from the 2020 Vintage Population Estimates (https://www.census.gov/programs-surveys/popest/technical-documentation/research/evaluation-estimates/2020-evaluation-estimates/2010s-counties-total.html).

^bFor the overall SVI, each SVI theme, each of the included 13 social vulnerability variables, percent uninsured, and urbanicity, we used negative binomial regression to calculate unadjusted rate ratios (95% CIs) comparing the rates of drug overdoses among counties in quartiles 2, 3, and 4 of the SVI variable to counties in quartile 1 of the SVI variable (referent; designated with *).

^cCollinearity was identified between theme 1 social vulnerability variables; therefore, the percent below poverty level and per capita income variables were removed from the individual variable analysis. For the overall SVI, each SVI theme, and the remaining 13 social vulnerability variables we used negative binomial regression to calculate adjusted rate ratios (95% CIs) comparing the rates of drug overdoses in counties with SVI values in quartiles 2, 3, and 4 to counties with SVI values in quartile 1 (referent; designated with *). Point estimates for overall SVI were adjusted for urbanicity. Point estimates for each SVI theme are adjusted for the other three SVI themes (i.e., we used one model with four SVI themes: SES theme, household composition/disability theme, racial/ethnic minority status theme, and housing type/transportation theme) and urbanicity. Point estimates for each social vulnerability variable are adjusted for the other social vulnerability variables and urbanicity.

^dThe overall SVI includes 15 variables in the index: percent below poverty level, percent unemployed among civilians aged ≥16 years, per capita income, adults aged ≥25 years with no high school diploma, percent aged 65 or older, percent aged 17 or younger, percent older than age 5 with a disability, percent single-parent households, percent identifying as a race or ethnicity other than White, non-Hispanic, percent speaking English “less than well,” percent living in multiunit structures, percent living in mobile homes, percent living in crowded homes (i.e., >1 occupant per room), percent with no vehicle, and percent living in group quarters (i.e., a group living arrangement that is owned or managed by an entity or organization providing housing and/or services for the residents including treatment facilities, skilled nursing facilities, group homes, correctional facilities, and facilities for people experiencing homelessness).

^eIncludes percent below poverty level, percent unemployed among civilians aged ≥16 years, per capita income, adults aged ≥25 years with no high school diploma.

^fIncludes percent aged 65 or older, percent aged 17 or younger, percent older than age 5 with a disability, and percent single-parent households.

^gIncludes percent identifying as a race or ethnicity other than White, non-Hispanic and percent speaking English “less than well.”

^hIncludes percent living in multiunit structures, percent living in mobile homes, percent living in crowded homes, percent with no vehicle, and percent living in group quarters.

ⁱUrbanicity data were obtained from the 2013 National Center of Health Statistics Rural-Urban Classification Scheme for Counties (https://www.cdc.gov/nchs/data_access/urban_rural.htm). Patient counties of residence were classified as non-core were coded as rural; otherwise, counties were coded as urban. The hospitalization dataset included 1279 counties (805 rural, 474 urban).

Our analysis of the individual SVI variables found that counties with higher percentage of people with a disability, single-parent households, higher levels of unemployment, and population living in multiunit structures or crowded housing generally had higher NFOD hospitalization rates involving all drugs, all opioids, heroin, and all stimulants. While increased county population adults with no high school diploma was associated with increased all drug, all opioid-, and heroin-involved NFOD rates, this was not associated with stimulant-involved NFOD rates.

Lower NFOD hospitalization rates were associated with higher percentages of senior or youth populations. Additionally, counties with the highest percentage of population living in group quarters or uninsured (Q4) were associated with lower or equivalent NFOD rates across the four drug categories. Lower NFOD rates were not observed in the other quartiles, where NFOD rates were equivalent or higher. Unlike NFOD ED visits, counties reporting the highest percentage (Q4) of people from racial and ethnic minority groups were associated with higher all drug, all opioid-, heroin-, and all stimulant-involved NFOD rates. While the percentage of the population reporting to speak English “less than well” does not exhibit as consistent a pattern, NFOD hospitalization aRRs were generally higher than NFOD ED aRRs. Additionally, NFOD hospitalization rates were more strongly associated with urbanicity than NFOD ED visit rates were. Compared to rural counties, NFOD hospitalization rates in urban counties were higher for those involving all drugs (aRR=1.99, 1.87–2.12), all opioids (aRR=2.54, 2.24–2.87), heroin (aRR=9.65, 3.69–25.19) and all stimulants (aRR=3.28, 2.38–4.50).

4. Discussion

Our analysis found that all drug and opioid-involved NFOD ED visit rates increased significantly between 2018 and 2020, while all drug NFOD hospitalization rates decreased, and opioid-involved NFOD hospitalization rates remained level. All stimulant-involved NFOD hospitalization rates were higher than stimulant-involved NFOD ED visit rates, a pattern not seen for other categories. This could indicate that stimulant-involved NFODs may require advanced medical care involving hospitalization, particularly when co-involving other drug categories such as opioids (Pickens et al., 2021). NFOD rates showed consistent seasonality, wherein the highest rates were observed annually April–September; this should be taken into consideration when developing overdose response activities to ensure that programmatic activities are scaled seasonally or focus on spring and summer months to have the highest impact.

As NFOD rates continue to increase across the U.S., it is important that public health agencies build multilevel intervention approaches that focus on community-level risk factors associated with higher NFOD rates. In this study, associations between NFODs, community socioecological variables, and urbanicity highlighted differences between those treated in an ED and discharged and those admitted to a hospital. Many associations between SDOH and NFODs were weaker among ED visits compared to hospitalizations. People who experience NFODs may seek treatment via EMS and refuse transport to a facility; these incidents are not captured in DOSE and may cause an artificial decrease in ED visit NFOD rates. As NFOD rates with EMS contact and no transport continue to increase, EMS NFOD data will be critical to ensure appropriate, targeted interventions are developed, especially in communities where EMS NFOD encounters are highest (Casillas et al., 2022).

Compared to rural counties, NFOD rates were higher in urban counties for opioid- and heroin-involved ED visits as well as for hospitalizations across all four drug categories. Rural NFOD hospitalization rates may be influenced in part by rural hospital closures that could limit access to medical care (Kaufman et al., 2016). Additional research to identify community-level risk factors that differ by urbanicity would inform development of geographically targeted interventions. We found that higher unemployment rates were associated with higher NFOD rates across drug categories, which aligns with previous research; however, without racial and ethnic disaggregation, results may mask differences between people from racial and ethnic minority groups (Frankenfeld and Leslie, 2019).

A higher percentage of a county’s population identifying as members of a minority group or speaking English “less than well” was generally associated with higher hospitalization NFOD rates, even when controlling for urbanicity and SES. These findings were not observed for NFOD ED visits; NFOD ED visit rates were generally lower or equivalent in counties with a higher percentage of people from racial and ethnic minority groups, and those counties in the highest vulnerability quartile (Q4) had the smallest aRRs for all four drug categories. NFOD ED rates among people speaking English “less than well” did not demonstrate a consistent pattern across quartiles; however, NFOD hospitalization rates were highest among the counties with the highest percentage of persons who speak English “less than well”. These results indicate that during the study period, people living in counties with the largest racial and ethnic minority populations were at greater risk of NFOD hospitalization, while the inverse was true for NFOD ED visits. Systemic barriers and institutional racial and ethnic biases and potentially related differences in patient treatment may decrease trust in healthcare providers and discourage or delay non-White persons from accessing care and lead to more severe health outcomes, including higher rates of fatal overdose (Kariisa, 2022). Prior research has found that when compared to White persons, drug use among people from racial and ethnic minority groups was more likely to be perceived as a criminal or moral failing than a public health crisis (Mendoza et al., 2019). Regardless of drug use history, racial and ethnic disparities in clinical care negatively impact the healthcare experiences of people from racial and ethnic minority groups, who can face longer wait times in EDs for injuries requiring analgesics (Epps et al., 2008), be viewed as less emergent (Vigil et al., 2016), and face more distrust from physicians (Moskowitz et al., 2011) when compared to White persons. More research is needed to understand healthcare provider traits that may influence NFOD treatment outcomes by race and ethnicity.

Interestingly, in many adjusted models, we identified the highest rates of NFOD within the second most vulnerable quartile (Q3), despite previous research findings that have indicated drug overdoses rates increase with increasing socioeconomic vulnerability (Pear et al., 2019). This could be attributed to the fact that overdoses among higher vulnerability populations may be more likely to result in fatality as highlighted by Kariisa et al. or be treated with naloxone in the field and refuse transport to an emergency facility as identified by Slavova et al. (Kariisa, 2022; Slavova et al., 2020). It is also possible that these findings reflect differences in access to emergency care rather than NFOD rate differences. Uninsured or underinsured individuals may less frequently seek facility-based medical treatment, regardless of need, due to financial concerns, which is supported by our finding of lower NFOD rates for all drug categories in counties with the highest proportion of uninsured individuals (Q4) compared to the second highest (Q3). Similarly, in their analysis of opioid-involved EMS encounters Casillas et al. found that counties with the highest rates of uninsured population had lower rates of opioid-involved EMS compared to the second highest quartile. Interventions that ensure affordability of medical services, such as expansion of Medicaid, have been associated with increased opioid-involved ED visits covered by Medicaid and may help address this disparity (Grossman and Hamman, 2020). To gain a comprehensive understanding of how SDOH impact overall overdose risk, it is necessary for future research to examine the association between SDOH and all overdose outcomes. This includes overdoses treated by EMS in the field, nonfatal overdoses treated in emergency facilities, and fatal overdoses.

Increased overall county-level social vulnerability was associated with increased NFOD rates; however, the magnitude of the association between visit type (ED vs. hospitalization) and drug category differed. Associations between NFOD rates and individual community variables within SVI themes highlight the complexity of using a vulnerability index to characterize the impact of discrete community SDOH. Our study also found divergent relationships between individual factors within Themes 2 (household and disability) and 4 (housing and transportation) and NFOD rates which, when summed, created the appearance of a weak or null association at the theme level. If the SVI is only applied at the overall or theme level, this could overlook critical vulnerabilities associated with higher NFOD rates and miss important community characteristics that are important to address through prevention strategies.

4.1. Limitations

With respect to limitations, this study focused on county-level SVI characteristics without inclusion of individual-level sociodemographic characteristics, which may modify associations with substance use outcomes (Boardman et al., 2001). While we evaluated the association between county percentage of people from any racial and ethnic minority group and NFOD, identified associations may vary between racial and ethnic minority groups as well as by sex or age and warrant further research. Improvements to demographic data in overdose surveillance sources would allow public health to better address disparities. Additionally, while the SVI includes many relevant sociodemographic variables, it was designed for disaster management and does not include several indicators that may contribute to substance use and overdose inequities. These include mental health, incarceration, and healthcare provider access, as well as individual racial and ethnic groups (Centers for Disease Control and Prevention, 2021). Development and validation of a drug overdose-specific index that includes additional demographic, prescription drug monitoring program, criminal justice, and mental health variables could better characterize geographic vulnerability and magnitude of associations between SDOH and NFOD. The overall and theme-level vulnerability scores for each county are summations of the theme level and individual variable scores, respectively; however, variables may have differential impact on NFOD rates.

This study period comprises years leading up to and including the onset of the COVID-19 pandemic, which in 2020 was associated with significant decreases in all-cause ED visits and increases in overdose-related ED visits and overdose-related deaths, as compared to 2019 and early 2020 (Hartnett et al., 2020; Holland et al., 2021; Kariisa, 2022). The main analysis did not stratify NFOD rates by year, and therefore, the impact of the pandemic on associations between 2020 NFOD rates and SVI characteristics is not described. Additionally, the pandemic may have impacted community-level social vulnerability, as trends in community characteristics, such as proportion of adults who were unemployed, housing stability, and food security rapidly shifted across the United States (Budget and Policy Priorities, 2021).

Finally, DOSE leverages administrative discharge diagnosis-coded data not originally collected for surveillance purposes (i.e., primary purpose is for medical billing); thus, counts from this data source may not provide accurate NFOD cost estimates. NFOD cost estimates of the four drug categories are reported to DOSE as aggregate counts; therefore, polysubstance use and analysis of drug categories not reported to DOSE (e.g., benzodiazepine, cocaine) could not be examined in this analysis. Furthermore, an ICD-10-CM code for methamphetamine poisoning was not available during our study period, while an ICD-10-CM code for fentanyl poisoning was only introduced during the last quarter of our study period; hence, these types of overdoses could not be analyzed specifically. While included states provided data from over 90% of their healthcare facilities, this analysis was not nationally representative, and some states report only ED data (n=4) or only hospitalization data (n=3). Therefore, differences between ED and hospitalization findings may be influenced by these geographic differences. While we did adjust for urbanicity in our analysis, we did not account for spatial geographic clustering. States exclude from county-level discharge data ED visits and hospitalizations for individuals who were not a resident of the state, which may artificially decrease NFOD rates within the patient’s county of residence if they reside in another state included in the analysis.

4.2. Public health implications

As nonfatal and fatal overdoses continue to rise, more community-level interventions that address risk factors at all socioecological levels are needed. The findings of this study highlight community-level characteristics associated with increased NFOD rates including urbanicity, common types of housing (e.g., multi-unit, group quarters), population demographics including minority status, disability, single-parent households, and unemployment and how the magnitude of these associations differs between ED visit and hospitalization NFOD. Increasing public awareness of inequities associated with increased overdose rates in all communities may decrease stigma and raise support for programs that address the underlying disparities and barriers associated with increased overdose rates. This could include social support resources to address economic instability or providing access to harm reduction resources (e.g., naloxone, fentanyl test strips), as well as improving linkage to substance use disorder treatment as part of NFOD follow-up care.

Additionally, it is essential that overdose response strategies address SDOH that contribute to racial and ethnic disparities in NFOD rates. Public health and healthcare providers can work collectively to identify and address systemic inequities in both communities and health systems that contribute to substance use and overdose among racial and ethnic minorities. Interventions could include partnerships with community- and faith-based organizations to develop culturally sensitive messaging and to identify community members to serve as trusted liaisons, helping to destigmatize drug use.

Finally, ongoing surveillance and analysis of community-level risk factors for NFODs as well as associated patient demographic data are needed to drive public health program development and evaluate changes needed as the socioecological characteristics of communities change over time. Public health and clinicians can work together to ensure standardized, complete patient data are collected and reported to inform and evaluate work to decrease overdose rates and minimize disparities.

Abbreviations:

ACS

American Community Survey

ATSDR

Agency for Toxic Substances and Disease Registry

CDC

Centers for Disease Control and Prevention

CI

confidence interval

DOSE

Drug Overdose Surveillance and Epidemiology

ED

emergency department

EMS

emergency medical services

GRASP

Geospatial Research, Analysis, and Services Program

ICD-10-CM

International Classification of Diseases, 10th Revision, Clinical Modification

NFOD

nonfatal overdose

RR

rate ratio

aRR

adjusted rate ratio

SDOH

social determinants of health

SES

socioeconomic status

SVI

social vulnerability index