Effect of the Communities That HEAL Intervention on Overdose Education and Naloxone Distribution: A Cluster-Randomized, Wait-List Controlled Trial

Abstract

Objectives. To determine whether the Communities That HEAL (CTH) intervention is effective in increasing naloxone distribution compared with usual care.

Methods. The HEALing (Helping to End Addiction Long-Term) Communities Study (HCS) is a cluster-randomized, parallel-arm, wait-list controlled implementation science trial testing the impact of the CTH intervention on increasing the use of evidence-based practices to lower opioid-related overdose deaths. Communities (n = 67) highly impacted by opioid overdose in Kentucky, Massachusetts, New York, and Ohio were allocated to CTH intervention (n = 34) or wait-list comparison (usual care; n = 33) arms. The primary outcome for this study was the number of naloxone units distributed in HCS communities during the comparison period (July 1, 2021‒June 30, 2022), examined using an intent-to-treat negative binomial regression model.

Results. Naloxone distribution was 79% higher in the CTH intervention versus usual care arm (adjusted relative rate = 1.79; 95% confidence interval = 1.28, 2.51; P = .001; adjusted rates of naloxone distribution 3378 vs 1884 naloxone units per 100 000 residents), when controlling for urban‒rural status, state, baseline opioid-related overdose death rate, and baseline naloxone distribution rate.

Conclusions. The CTH intervention increased naloxone distribution compared with usual care in communities highly impacted by the opioid crisis.

Trial Registration. ClinicalTrials.gov identifier: NCT04111939. (Am J Public Health. 2025;115(1):83–94. https://doi.org/10.2105/AJPH.2024.307845)

The public health crisis of drug overdose continues unabated in the United States. Nearly 107 000 drug overdose deaths occurred in 2021, an almost 5-fold increase since 2001.^1,2 Deaths are driven by synthetic opioids, primarily fentanyl, now ubiquitous in the illicit drug supply. Accelerating disparities in opioid-related overdoses are evident, with American Indian/Alaska Native and Black populations most impacted.³ The continued escalation and widening disparities in opioid overdose demand innovative approaches in public health response to reduce drug overdose mortality.

The HEALing (Helping to End Addiction Long-Term) Communities Study (HCS) is a cluster-randomized, parallel-arm, wait-list controlled implementation science trial testing the impact of the Communities That HEAL (CTH) intervention on increasing the use of evidence-based practices (EBPs) to lower opioid-related overdose deaths.^4,5 The CTH intervention has 3 components: (1) a coalition-driven community-engagement process to select and support the implementation of strategies to facilitate the uptake of EBPs⁶; (2) the Opioid-overdose Reduction Continuum of Care Approach (ORCCA), a menu of strategies and technical assistance guide for implementing EBPs of overdose education and naloxone distribution (OEND), medication for opioid use disorder (MOUD), and prescription opioid safety⁷; and (3) communication campaigns designed to reduce stigma and increase awareness of and demand for OEND and MOUD.⁸

Overdose education and naloxone distribution, an EBP that decreases opioid-related overdose deaths, is a harm-reduction strategy that was focused initially on people who use drugs and bystanders who were most likely to witness and respond to opioid overdoses.⁹ As the opioid crisis escalated, expanding access to OEND became a focus of US health policy interventions. Naloxone access laws were implemented by states to allow naloxone dispensing from pharmacies via standing order or direct prescribing,¹⁰ and federal leaders encouraged OEND for a wider array of community members.¹¹ Despite this expansion of OEND, modeling data show in most states there is insufficient access to naloxone,¹² particularly for rural populations¹³ and among Black and Hispanic persons.^14,15

As efforts to reduce opioid-related overdose mortality through expanded OEND continue, it is imperative to identify effective strategies for increasing OEND. In this study, we examined the effect of the CTH intervention on OEND in 67 communities highly impacted by opioid overdose. We conducted analyses to evaluate 3 hypotheses:

• Hypothesis 1: CTH intervention communities will distribute, overall, significantly higher rates of naloxone units compared to usual care.

• Hypothesis 2: CTH intervention communities will distribute, via community-based organizations and events, significantly higher rates of naloxone units compared to usual care.

• Hypothesis 3: CTH intervention communities will distribute, via pharmacy dispensing, significantly higher rates of naloxone units compared to usual care.

METHODS

The HCS is a cluster-randomized, parallel-arm, wait-list controlled implementation science trial testing the impact of the CTH intervention on increasing the use of EBPs to lower opioid-related overdose deaths.

Participants

The study consisted of 67 communities (48 counties and 19 cities or towns) in Kentucky, Massachusetts, New York, and Ohio. To participate in the study, communities had to be both highly impacted by opioid-related overdose deaths and willing to partner with behavioral health, criminal justice, and health care settings to implement MOUD and OEND. Within each state, at least 30% of the communities had to be classified as rural, and across each state, a threshold of at least 150 opioid-related overdose deaths (with rural communities accounting for 15% of deaths) and an overdose death rate of at least 25 per 100 000 people was required, based on 2016 data.

Randomization

Study arm allocation was assigned using a covariate-constrained randomization procedure balancing urban‒ rural classification, opioid-related overdose death rate, and community population size within each state using a macro described by Greene.^4,16 The HCS Data Coordinating Center Study System Developer ran the randomization code, assigned communities to the study arms, and notified the states of the assignment. Thirty-four communities were allocated to wave 1 (CTH intervention arm), and 33 communities were allocated to wave 2 (waitlist control [usual care] arm). The study follows the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials.¹⁷

Intervention

HCS communities allocated to wave 1 received the CTH intervention, an adaptation of the Communities That Care model,¹⁸ from January 2020 through June 2022. An iterative 7-step process guided community coalitions in the creation of data-driven action plans to reduce opioid-related overdose deaths.⁶ As part of the action plan, community coalitions selected strategies from the ORCCA to increase community OEND. Specifically, the CTH required community coalitions to choose at least 1 active OEND strategy, which involved proactive delivery to at-risk individuals and their social networks and delivery in high-risk venues including behavioral health, criminal justice, and health care sectors.⁷ Passive OEND strategies were optional and included making OEND more accessible by referral (e.g., a health care provider refers someone to a pharmacy for OEND) or self-request (e.g., a person requests OEND at a pharmacy or community event) and increasing naloxone availability for immediate use in overdose hotspots (e.g., installation of opioid emergency boxes containing naloxone). Community coalitions could also choose an optional strategy of increasing capacity for first-responder naloxone administration. The unique strategy‒ sector‒venue triad combinations used to assess strategy implementation and the strategies initially selected for implementation by wave 1 communities in their action plans have been reported in detail.¹⁹ The actual strategies to promote the uptake of OEND that were actively implemented (i.e., strategies that reached the service delivery stage of implementation) by communities during the intervention period are reported herein.

Outcomes

The primary outcome (hypothesis 1) evaluated in this study was the total number of naloxone units distributed in communities. Secondary outcomes included the number of naloxone units distributed through community programs (hypothesis 2) and the number of naloxone units dispensed by retail pharmacies (hypothesis 3). The overall primary outcome for the HCS trial is opioid-related overdose deaths and was evaluated separately.⁵ We assessed outcomes during a 12-month comparison period from July 1, 2021, through June 30, 2022, to allow time for community selection and implementation of strategies to promote OEND.

Total naloxone distribution was measured by the sum of the number of naloxone units distributed through community programs with support by state and federal funding (including HCS funding) and the number of naloxone units dispensed by retail pharmacies located within the communities, as previously described.²⁰ Data were captured from the IQVIA Xponent database for retail pharmacy dispensing and from administrative data from state agencies and HCS study records for community program distribution (see Supplementary Methods, pp. 1–2, available as a supplement to the online version of this article at https://ajph.org). One unit of naloxone equals 1 carton or kit containing 2 doses. Per the ORCCA, all naloxone distributed was accompanied by overdose education; education could be accomplished through a variety of approaches including, for example, in-person training or the use of training videos or educational pamphlets.

Statistical Analysis

The HCS was designed to have high power (> 99%) to detect a 40% reduction for the overall primary outcome, opioid-related overdose deaths, with a total sample size of 67 communities randomized into wave 1 (n = 34) and wave 2 (n = 33).⁵

For each community-level outcome of distribution counts and total community population size during 2020, we estimated the comparison of rates between wave 1 and wave 2 communities during the comparison period by fitting a negative binomial regression model adjusting for urban‒rural status; state (KY, MA, NY, OH); community-level opioid-related overdose death rate (defined across the study as the number of overdose deaths per 100 000 adults) for January 1, 2019, through December 31, 2019 (baseline); and the community-level rate (per 100 000 residents) of the outcome variable at baseline.²¹ We used the natural log of the community population as the offset. We used small, sample-adjusted, empirical standard errors for all inference testing.^22,23

For each outcome, we calculated the marginal adjusted rates of wave 1 and wave 2 communities, adjusted relative rate (RR; wave 1 rate over wave 2 rate), and associated 95% confidence intervals (CIs) and P values. We conducted these analyses in both the intent-to-treat and per-protocol (which excludes 1 community that withdrew from the study before the intervention) populations. The results were consistent for the 2 approaches; thus, only the intent-to-treat analyses are presented.

For 2 potential effect modifiers of the CTH intervention, state and urban‒rural status, we fitted separate models. Each model included fixed effects for the stratification variable as well as an interaction between wave and the stratification variable. For state, urban‒rural status, and overall, we computed means and standard deviations for community-level naloxone units, population size, and rates per 100 000 residents by randomized group and outcome. In addition, we evaluated the comparisons of the raw rates between wave 1 and wave 2 communities. To account for multiple comparisons arising from these subgroup analyses, we controlled the false discovery rate for all effect modification interaction tests between wave and the stratification variable as well as pairwise tests between levels of stratification variables using the Benjamini‒Hochberg correction.²⁴ We considered results statistically significant for P values < .05.

The statistical analysis plan was prespecified. We conducted all analyses using SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

All randomized communities (n = 67) were included in the intent-to-treat analysis (16 each from KY, MA, and NY, and 19 from OH). The CONSORT diagram is shown in Appendix Figure A (available as a supplement to the online version of this article at https://ajph.org). Baseline (January 2019‒December 2019) characteristics of wave 1 (n = 5 595 837 residents) and wave 2 (n = 4 833 976 residents) communities are summarized in Table 1. The 2 study arms were balanced on urban‒rural status and opioid-related overdose death rate at baseline. Overall, baseline mean rates of total naloxone distribution were similar between wave 1 (mean = 1229.9; SD = 1109.9) and wave 2 (mean = 1320.8; SD = 1196.3) communities. When stratified by state, baseline mean rates of total naloxone distribution were similar between wave 1 and wave 2 for Kentucky and New York but differed by nearly 50% in Massachusetts (mean = 1342.9; SD = 1075.3 vs mean = 2187.4; SD = 1846.0) and Ohio (mean = 1219.2; SD = 1363.6 vs mean = 655.1; SD = 423.9).

TABLE 1—

Baseline Demographic Characteristics of Communities Participating in the HEALing Communities Study, by Intervention Wave: 4 US States, January 2019‒December 2019

	Intervention Wave		Overall
	Wave 1	Wave 2
Randomized communities, no. (%)
Overall	34	33	67
Kentucky	8 (23.5)	8 (24.2)	16 (23.9)
Massachusetts	8 (23.5)	8 (24.2)	16 (23.9)
New York	8 (23.5)	8 (24.2)	16 (23.9)
Ohio	10 (29.4)	9 (27.3)	19 (28.4)
Urban‒rural classification, no. (%)
Urban	19 (55.9)	19 (57.6)	38 (56.7)
Rural	15 (44.1)	14 (42.4)	29 (43.3)
Total population,a no.
Overall	5 595 837	4 833 976	10 429 813
Kentucky	782 758	1 044 354	1 827 112
Massachusetts	439 910	447 385	887 295
New York	1 357 580	1 247 888	2 605 468
Ohio	3 015 589	2 094 349	5 109 938
Rate of opioid-related overdose deaths, mean (SD)b
Overall	38.4 (22.8)	37.4 (21.0)	37.9 (21.8)
Kentucky	43.2 (16.6)	37.5 (21.8)	40.4 (19.0)
Massachusetts	46.0 (14.2)	54.1 (23.0)	50.1 (18.9)
New York	19.3 (5.6)	29.3 (12.8)	24.3 (10.8)
Ohio	43.8 (33.0)	29.6 (17.9)	37.1 (27.2)
Rate of total naloxone units distributed, mean (SD)c
Overall	1 229.9 (1 109.9)	1 320.8 (1 196.3)	1 274.7 (1 145.4)
Kentucky	1 039.6 (663.1)	894.3 (422.1)	966.9 (542.2)
Massachusetts	1 342.9 (1 075.3)	2 187.4 (1 846.0)	1 765.2 (1 523.2)
New York	1 320.4 (1 324.8)	1 629.7 (1 003.2)	1 475.1 (1 146.4)
Ohio	1 219.2 (1 363.6)	655.1 (423.9)	952.0 (1 045.6)
Rate of naloxone units distributed through community-based programs, mean (SD)c
Overall	796.3 (907.0)	888.7 (869.2)	841.8 (883.1)
Kentucky	458.3 (252.4)	596.7 (346.7)	527.5 (301.5)
Massachusetts	902.3 (982.7)	1 346.9 (1 106.2)	1 124.6 (1 036.5)
New York	1 178.4 (1 302.2)	1 384.5 (990.6)	1 281.4 (1 122.7)
Ohio	676.2 (799.8)	300.1 (238.4)	498.1 (618.3)
Rate of naloxone units dispensed by retail community pharmacies, mean (SD)c
Overall	433.6 (472.2)	432.1 (722.4)	432.9 (603.7)
Kentucky	581.3 (576.5)	297.6 (119.7)	439.4 (428.1)
Massachusetts	440.6 (357.4)	840.4 (1 413.8)	640.5 (1 017.4)
New York	142.0 (60.3)	245.2 (93.2)	193.6 (92.7)
Ohio	543.0 (590.3)	354.9 (303.0)	453.9 (473.7)

Note. HEAL = Helping to End Addiction Long-Term. The sample size was n = 67 communities. Percentages may not add up to 100 because of rounding. There are no missing data.

^a For communities that represent counties (n = 48 of 67), population estimates are from 2019 Bridged-Race Population Estimates retrieved via https://www.cdc.gov/nchs/nvss/bridged_race.htm on March 17, 2023. For communities that represent units smaller than counties (n = 19 of 67), population estimates are from 2015 to 2019 American Community Survey 5-Year Estimates retrieved via https://data.census.gov/cedsci on March 17, 2023.

^b Rate per 100 000 community residents aged ≥ 18 years.

^c Rate per 100 000 community residents.

Wave 1 CTH intervention communities implemented 254 strategies to increase OEND (Table 2). Variation existed among the number of strategies implemented by state (KY = 104; OH = 58; MA = 54; NY = 38) and urban‒rural status (urban = 145; rural = 109). Communities implemented far more active (68.9%) than passive (28.0%) OEND strategies, with more strategies implemented within high-risk venues (59.4%) than for at-risk individuals (40.6%), although the latter varied considerably by state. Overall, communities implemented more strategies in behavioral health (52.4%) than in criminal justice (24.0%) or health care (23.6%) sectors. Specific examples of the strategies to increase OEND that were implemented by wave 1 communities can be found in Appendix Table A.

TABLE 2—

Actively Implemented Overdose Education and Naloxone Distribution (OEND) Strategies for Wave 1 Communities Participating in the HEALing Communities Study: 4 US States, April 8, 2020‒June 30, 2022

	State				Urban‒Rural Classification		Total
	KY	MA	NY	OH	Urban	Rural
No. of communities	8	8	8	9	18	15	33
No. of OEND strategies implemented	104	54	38	58	145	109	254
Strategy, no. (%)
Active OEND	58 (55.8)	38 (70.4)	28 (73.7)	51 (87.9)	101 (69.7)	74 (67.9)	175 (68.9)
At-risk individualsa	9 (15.5)	19 (50.0)	22 (78.6)	21 (41.2)	40 (39.6)	31 (41.9)	71 (40.6)
High-risk venuesa	49 (84.5)	19 (50.0)	6 (21.4)	30 (58.8)	61 (60.4)	43 (58.1)	104 (59.4)
Passive OEND	44 (42.3)	12 (22.2)	9 (23.7)	6 (10.3)	42 (29.0)	29 (26.6)	71 (28.0)
Referralb	0 (0)	2 (16.7)	0 (0)	1 (16.7)	1 (2.4)	2 (6.9)	3 (4.2)
Self-requestb	37 (84.1)	7 (58.3)	2 (22.2)	2 (33.3)	30 (71.4)	18 (62.1)	48 (67.6)
Naloxone availability for immediate  use in overdose hotspotsb	7 (15.9)	3 (25.0)	7 (77.8)	3 (50.0)	11 (26.2)	9 (31.0)	20 (28.2)
Capacity for first-responder administration	2 (1.9)	3 (5.6)	1 (2.6)	1 (1.7)	1 (0.7)	6 (5.5)	7 (2.8)
Otherc	0 (0)	1 (1.9)	0 (0)	0 (0)	1 (0.7)	0 (0)	1 (0.4)
Sector, no. (%)
Health care	18 (17.3)	15 (27.8)	9 (23.7)	18 (31.0)	34 (23.4)	26 (23.9)	60 (23.6)
Behavioral health	55 (52.9)	35 (64.8)	23 (60.5)	20 (34.5)	76 (52.4)	57 (52.3)	133 (52.4)
Criminal justice	31 (29.8)	4 (7.4)	6 (15.8)	20 (34.5)	35 (24.1)	26 (23.9)	61 (24.0)

Note. HEAL = Helping to End Addiction Long-Term. The sample size was n = 33 communities; n = 1 community randomized to wave 1 withdrew before strategy selection. Summary of unique strategy‒sector‒venue triad combinations that were actively implemented (defined as strategies that reached the service delivery stage of implementation) during wave 1. The most recent entry for each triad is chosen. Duplicate triad entries are rolled to the triad level. Results are based on data pulled from the Opioid-overdose Reduction Continuum of Care Approach (ORCCA) Tracker on March 24, 2023.

^a Percentage out of active OEND strategies.

^b Percentage out of passive OEND strategies.

^c Other strategy: Never Use Alone helpline.

Table 3 presents results of the efficacy outcomes analyses for the 3 hypotheses. The adjusted rate of total naloxone distribution was 79% higher for the wave 1 CTH intervention versus wave 2 usual care arm (adjusted RR = 1.79; 95% CI = 1.28, 2.51; P = .001; adjusted rates of naloxone distribution were 3378 vs 1884 naloxone units per 100 000 residents for CTH vs usual care), when controlling for urban‒rural status, state, baseline opioid-related overdose death rate, and baseline naloxone distribution rate. The rate of naloxone distributed by community programs was 104% higher in CTH versus usual care communities (adjusted RR = 2.04; 95% CI = 1.32, 3.17; P = .002). There were no significant differences in the rates of naloxone dispensed from retail pharmacies (adjusted RR = 0.99; 95% CI = 0.75, 1.30; P = .917).

TABLE 3—

Analyses Testing the Effect of the Communities That HEAL Intervention on the Rate of Each Efficacy Outcome Between Wave 1 and Wave 2 Communities During the Comparison Period: 4 US States, July 1, 2021‒June 30, 2022

Outcome	Adjusted Rate (95% CI)a		Adjusted RR (95% CI)b
	Wave 1	Wave 2
Hypothesis 1: Total no. of naloxone units distributed in communities	3378.1 (2663.1, 4285.0)	1884.4 (1466.9, 2420.7)	1.79 (1.28, 2.51)
Hypothesis 2: No. of naloxone units distributed through community-based programs	2754.1 (2081.9, 3643.3)	1348.4 (962.4, 1889.3)	2.04 (1.32, 3.17)
Hypothesis 3: No. of naloxone units dispensed by retail community pharmacies	524.0 (447.8, 613.1)	531.6 (421.7, 670.2)	0.99 (0.75, 1.30)

Note. CI = confidence interval; HEAL = Helping to End Addiction Long-Term; RR = relative rate. A separate negative binomial model was fit for each hypothesis adjusting for urban‒rural classification (urban, rural), state (KY, MA, NY, OH), baseline opioid-related overdose death rate, and baseline outcome rate using the intention-to-treat population (n = 67 communities). The natural log of the community population was used as the offset. Dispersion parameters for each of the hypotheses’ models are: k = 0.36 (95% CI = 0.26, 0.49) for hypothesis 1; k = 0.93 (95% CI = 0.67, 1.29) for hypothesis 2; and k = 0.25 (95% CI = 0.18, 0.35) for hypothesis 3.

^a Model estimated marginal event rate expressed as per 100 000 community residents.

^b Adjusted RR of wave 1 communities over wave 2 communities.

State and urban‒rural status were evaluated as prespecified potential effect modifiers by fitting separate models testing the interaction between each of these factors and intervention group. Observed mean rates (per 100 000 residents) of total naloxone distribution varied considerably by state, ranging from 9971.6 (SD = 6126.2) in Kentucky to 2255.5 (SD = 2338.7) in Ohio wave 1 communities and 3279.0 (SD = 2171.1) to 1398.3 (SD = 1098.3) in Kentucky and Ohio wave 2 communities, respectively (Table 4). Raw RRs of total naloxone distribution in wave 1 over wave 2 communities varied by state, ranging from 3.04 in Kentucky to 1.18 in New York, although the test of effect modification for state was not statistically significant. Observed mean rates of total naloxone distribution were higher in urban communities (wave 1: 4682.9; SD = 5368.0; wave 2: 2599.0; SD = 1729.4) compared with rural communities (wave 1: 4025.0; SD = 3528.4; wave 2: 1811.5; SD = 1729.8). The raw RRs of total naloxone distribution in wave 1 over wave 2 urban (1.80) and rural (2.22) communities were similar, however, and the test of effect modification for urban‒rural status was not statistically significant. While variability by state and urban‒rural status was also noted for community program distribution and retail pharmacy dispensing, because all interactions were nonsignificant, we conducted no post hoc analyses.

TABLE 4—

Descriptive Statistics for Efficacy Outcomes, Overall and by State and Rurality, During the Comparison Period Using the Intention-to-Treat Population in the HEALing Communities Study: 4 US States, July 1, 2021‒June 30, 2022

	Wave 1 (n = 34)		Wave 2 (n = 33)		Raw Rate Ratioc
	Naloxone Unitsa	Raw Rateb	Naloxone Unitsa	Raw Rateb
Total no. of naloxone units distributed in communities
Overall
Mean (SD)	4 946.8 (5 474.0)	4 392.6 (4 594.4)	3 841.8 (8 530.1)	2 264.9 (1 747.6)	1.94
Median (Q1, Q3)	2 968.0 (823.0, 8 065.0)	2 716.7 (1 557.5, 5 528.2)	1 065.0 (631.0, 2 078.0)	1 937.3 (956.2, 3 388.8)	1.40
State, mean (SD)
Kentucky	6 382.8 (2 854.8)	9 971.6 (6 126.2)	2 977.9 (4 919.0)	3 279.0 (2 171.1)	3.04
Massachusetts	1 606.9 (1 235.0)	3 198.5 (2 227.4)	1 336.4 (1 529.0)	2 219.2 (1 744.9)	1.44
New York	6 153.4 (8 064.6)	2 679.1 (1 705.1)	4 629.1 (6 320.0)	2 271.4 (1 637.5)	1.18
Ohio	5 504.8 (6 194.4)	2 255.5 (2 338.7)	6 137.0 (14 809.2)	1 398.3 (1 098.3)	1.61
Urban‒rural classification, mean (SD)
Urban	7 136.4 (6 251.2)	4 682.9 (5 368.0)	6 039.3 (10 813.7)	2 599.0 (1 729.4)	1.80
Rural	2 173.4 (2 375.0)	4 025.0 (3 528.4)	859.6 (750.1)	1 811.5 (1 729.8)	2.22
No. of naloxone units distributed through community-based programs
Overall
Mean (SD)	4 085.9 (4 732.8)	3 631.3 (3 890.8)	3 029.8 (7 065.2)	1 664.0 (1 577.1)	2.18
Median (Q1, Q3)	2 575.0 (761.0, 6 405.0)	2 346.4 (945.1, 5 054.7)	821.0 (281.0, 1 789.0)	1 263.6 (535.0, 2 176.0)	1.86
State, mean (SD)
Kentucky	5 408.3 (2 318.8)	8 448.1 (4 981.4)	2 048.3 (3 261.1)	2 353.5 (2 155.6)	3.59
Massachusetts	1 388.0 (1 180.5)	2 789.7 (2 194.8)	981.6 (1 137.6)	1 650.2 (1 468.6)	1.69
New York	5 687.6 (7 576.6)	2 427.2 (1 703.0)	3 948.3 (5 569.4)	1 879.8 (1 478.8)	1.29
Ohio	3 904.8 (4 677.1)	1 414.2 (1 483.7)	4 906.4 (12 333.9)	871.5 (924.4)	1.62
Urban‒rural classification, mean (SD)
Urban	5 939.2 (5 444.0)	3 924.3 (4 475.5)	4 810.2 (8 977.4)	1 931.9 (1 600.4)	2.03
Rural	1 738.3 (2 031.4)	3 260.0 (3 108.9)	613.5 (600.8)	1 300.4 (1 526.1)	2.51
No. of naloxone units dispensed by retail community pharmacies
Overall
Mean (SD)	861.0 (1 167.1)	761.4 (1 131.0)	812.0 (1 578.2)	600.9 (430.0)	1.27
Median (Q1, Q3)	335.5 (159.0, 873.0)	386.0 (287.0, 612.4)	289.0 (163.0, 630.0)	551.8 (305.3, 759.5)	0.70
State, mean (SD)
Kentucky	974.5 (917.1)	1 523.5 (1 934.6)	929.6 (1 717.4)	925.6 (514.0)	1.65
Massachusetts	218.9 (119.7)	408.7 (103.3)	354.8 (429.1)	569.0 (421.7)	0.72
New York	465.8 (512.1)	251.9 (61.1)	680.9 (847.4)	391.6 (219.2)	0.64
Ohio	1 600.0 (1 732.4)	841.3 (950.4)	1 230.6 (2 481.0)	526.8 (393.0)	1.60
Urban‒rural classification, mean (SD)
Urban	1 197.2 (1 386.0)	758.5 (1 311.7)	1 229.1 (1 990.2)	667.1 (474.6)	1.14
Rural	435.1 (627.0)	765.0 (896.1)	246.1 (218.4)	511.2 (358.1)	1.50

Note. HEAL = Helping to End Addiction Long-Term.

^a Raw mean (SD) of the community-level number of naloxone units in that group in that wave.

^b Raw mean (SD) of the community-level rates per 100 000 individuals in that group in that wave. Population means used for rate calculations are presented in Table B (available as a supplement to the online version of this article at https://ajph.org).

^c Raw rate ratio calculated as the raw rate in wave 1 over the raw rate in wave 2.

DISCUSSION

As hypothesized, the CTH intervention significantly increased OEND compared with usual care. Increased naloxone distribution was driven by OEND provided by community programs, with no significant effect observed on the rate of naloxone dispensed by retail pharmacies. The CTH intervention offers a flexible way for communities to select strategies to significantly increase OEND, which is critical given access to this life-saving medication is suboptimal in most communities highly impacted by opioid-related overdose deaths.¹² The finding that the CTH intervention resulted in a 104% increase in naloxone distribution by community programs compared with control communities receiving usual care is particularly noteworthy given that the intervention was implemented during a time in which opioid overdose mortality was increasing and calls for increased naloxone access as a public health response grew louder.¹¹ A 2022 report to Congress on State Opioid Response (SOR) grants indicates that from April 1, 2021, to March 30, 2022, grantees reported distributing 2 177 367 naloxone kits.²⁵ Naloxone distribution increased from baseline in the wave 2 usual care communities during the comparison period, in part attributable to SOR-funded naloxone distribution and other ongoing public health efforts that were occurring across all communities regardless of their participation status in the HCS. The policy landscape supporting naloxone distribution in the HCS states has been described in detail.²⁶

The HCS is the largest implementation study in the field of addiction science and the first randomized controlled trial to examine the impact of a community-engaged intervention designed to promote the uptake of EBPs to reduce opioid-related overdose mortality. HCS communities participating in the CTH intervention selected and implemented strategies to increase uptake of OEND as part of a broader data-driven community action plan to reduce opioid-related overdose deaths. The HCS Consortium recently reported that opioid-related overdose deaths were 9% lower in communities implementing the CTH intervention compared with usual care.⁵ While the results did not reach statistical significance, the estimated 483 deaths averted in the intervention communities represents a meaningful public health impact. The HCS was designed to have high power (> 99%) to detect a 40% reduction in opioid-related mortality. The HCS consortium acknowledges that a 40% reduction was an ambitious goal and that the study was likely underpowered to detect smaller reductions in mortality. Ongoing research is evaluating the impact of the CTH on other overdose-related outcomes, including nonfatal overdoses and opioid-related overdoses in combination with stimulants.

In total, wave 1 communities implemented 254 unique OEND strategies, which represented a 40% increase in the number of strategies initially selected and included in action plans.¹⁹ Although the number and type of strategies implemented varied by state, all communities implemented OEND strategies across behavioral health, health care, and criminal justice sectors. More than two thirds of the strategies implemented were active OEND strategies designed to reach people at high risk for opioid overdose and their social networks and be deployed at high-risk venues. This distribution of strategies was expected given that proactively providing OEND to those at high risk for overdose is most effective²⁷ and that active OEND strategies were a required component of the ORCCA.⁷ Common high-risk venues where active OEND strategies were implemented included addiction treatment centers (behavioral health sector), syringe service programs (behavioral health sector), emergency departments (health care sector), and jails and other criminal justice settings (criminal justice sector). The implementation of active OEND strategies in criminal justice settings was especially important during the period of the study that coincided with the COVID-19 public health emergency and expedited release of individuals from carceral settings, a known risk factor for fatal overdose.²⁸ As such, wave 1 communities fast-tracked the implementation of OEND strategies in jails, syringe service programs, and other high-risk venues.

The absence of a significant difference in the rate of naloxone dispensed from retail pharmacies is not completely unexpected. Implementation of passive OEND strategies, which were more commonly pharmacy-based, was optional in the ORCCA and implemented far less often compared with the required active strategies (71 vs 175 actively implemented strategies, respectively). To increase naloxone dispensing from pharmacies during the intervention period, communities would have needed to implement strategies to encourage practitioners to write more prescriptions for naloxone or encourage pharmacists to increase their direct prescribing or initiation of naloxone prescriptions via standing order, collaborative care agreements, or protocols as allowed by state law.¹⁰ The CTH intervention was conducted during the COVID-19 pandemic, when retail pharmacies had adopted an expansive role in testing, vaccinating, and providing COVID-19 treatment. Additional factors, such as insurance formularies, prior authorization requirements, and individual cost-sharing expectations, may have influenced naloxone accessibility in retail pharmacies.²⁶ Finally, research has suggested that pharmacists may not be adequately trained or comfortable dispensing naloxone, and people seeking naloxone may face stigma or discrimination from pharmacy staff.²⁹ Comprehensive efforts are needed to bridge the training gap among pharmacists, combat stigma within the health care community, and raise awareness about naloxone’s vital role in preventing opioid-related overdose deaths.

In March 2023, the US Food and Drug Administration approved the first over-the-counter (OTC) naloxone product, a 4-milligram naloxone nasal spray, which became available at pharmacies and other retail outlets in September 2023.³⁰ While the goal of making naloxone available OTC is increased accessibility, the impact of OTC status on naloxone availability may be limited because of cost. At the time of writing, the manufacturer’s suggested retail price for a 2-dose carton of OTC Narcan Nasal Spray was $44.99,³¹ which may limit accessibility for many unless insurance companies provide coverage for OTC naloxone products. As such, continued free distribution through community organizations will likely be needed to ensure access for individuals at highest risk (e.g., those who are actively using drugs) for whom cost, stigma, and discrimination encountered at the pharmacy may be deterrents.

Although we did not detect any statistically significant differences when assessing effect modification of the CTH intervention by state or urban‒ rural status, we noted differences in observed mean rates of naloxone distribution both at baseline and follow-up. These differences may be important to consider given that previous modeling studies have indicated that increasing naloxone distribution rates as part of a multicomponent intervention with engagement and retention in MOUD treatment is necessary to achieve opioid overdose prevention goals.³² Further investigation into the specific factors that may be associated with increases in naloxone distribution, including the number of strategies, specific strategies deployed, how they were implemented, and other factors such as communication campaign messaging, is warranted and will be the focus of future publications in this implementation science trial. HCS researchers adapted the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM)/Practical, Robust, Implementation, and Sustainability (PRISM) implementation science framework to guide our assessment of contextual factors that affect CTH implementation and outcomes.^33,34 Future analyses of qualitative data from the study may help elucidate the relationships among community context, coalition characteristics, and EBP adoption among partner organizations.

Because this study primarily used administrative data from state agencies to capture OEND, demographic information about individuals who received OEND was not consistently collected. Analysis of available data in one state suggests that the demographic makeup of individuals who received OEND was similar to that of the county where OEND was delivered, although this varied by facility type (e.g., criminal justice vs behavioral health).³⁵ Additional data-capture methodologies were implemented in wave 2 to collect demographic information about OEND recipients consistently across the consortium, and future studies will specifically examine the impact of the CTH on vulnerable populations.

Strengths and Limitations

This study had numerous strengths. Our community-engaged intervention guided community coalitions through a phased, data-driven approach to create community action plans to promote the uptake of EBPs, including OEND. This process allowed communities to build on strengths, address gaps, and develop action plans uniquely tailored to deliver the maximum impact on scaling up naloxone distribution. Our study included a large sample of communities in both urban and rural areas highly impacted by opioid overdose.

Despite these strengths, the study is subject to limitations. We used administrative data from state agencies to capture naloxone distribution by community programs. These data may not capture all naloxone distributed in the community as naloxone purchased with support from private donations, foundations, or locally awarded federal funding may not be captured in state administrative data. For example, Ohio captured community naloxone distributed through Project DAWN (Deaths Avoided With Naloxone); however, during the intervention period, efforts to blanket the state with naloxone were conducted. These surges provided naloxone to some community agencies not affiliated with Project DAWN, resulting in likely undercounting of naloxone distribution in Ohio study communities.

In this study, we were unable to verify if the individuals who received the naloxone were HCS community residents or if individuals received more than 1 unit of naloxone. Retail pharmacy naloxone dispensing data were provided by IQVIA in aggregate with no information on the number of pharmacies in the community that dispensed the naloxone. IQVIA suppression rules preclude the reporting of data for geographic areas with fewer than 4 retail pharmacies, resulting in reporting a weighted mean number of naloxone units dispensed based on population size for 3 Massachusetts communities. Dispensed naloxone prescriptions in IQVIA are attributed to communities based on the dispensing pharmacy address rather than the customer’s residence, which may result in an overcount of naloxone in a community with pharmacies that serve residents of non-HCS communities, or an undercount if there is a pharmacy located outside an HCS community that serves HCS residents.

Public Health Implications

This community-level, cluster-randomized controlled implementation science trial showed that compared with usual care, the CTH intervention significantly increased OEND, an EBP that reduces opioid-related overdose mortality, in US communities highly affected by opioid overdose. The CTH intervention, including the ability to tailor selection and implementation of strategies to promote the uptake of OEND to individual communities, can serve as a model for other communities to adopt as part of ongoing public health response efforts to reduce opioid-related overdose mortality.