Community-Based Cluster-Randomized Trial to Reduce Opioid Overdose Deaths

Abstract

BACKGROUND

Evidence-based practices for reducing opioid-related overdose deaths include overdose education and naloxone distribution, the use of medications for the treatment of opioid use disorder, and prescription opioid safety. Data are needed on the effectiveness of a community-engaged intervention to reduce opioid-related overdose deaths through enhanced uptake of these practices.

METHODS

In this community-level, cluster-randomized trial, we randomly assigned 67 communities in Kentucky, Massachusetts, New York, and Ohio to receive the intervention (34 communities) or a wait-list control (33 communities), stratified according to state. The trial was conducted within the context of both the coronavirus disease 2019 (Covid-19) pandemic and a national surge in the number of fentanyl-related overdose deaths. The trial groups were balanced within states according to urban or rural classification, previous overdose rate, and community population. The primary outcome was the number of opioid-related overdose deaths among community adults.

RESULTS

During the comparison period from July 2021 through June 2022, the population-averaged rates of opioid-related overdose deaths were similar in the intervention group and the control group (47.2 deaths per 100,000 population vs. 51.7 per 100,000 population), for an adjusted rate ratio of 0.91 (95% confidence interval, 0.76 to 1.09; P = 0.30). The effect of the intervention on the rate of opioid-related overdose deaths did not differ appreciably according to state, urban or rural category, age, sex, or race or ethnic group. Intervention communities implemented 615 evidence-based practice strategies from the 806 strategies selected by communities (254 involving overdose education and naloxone distribution, 256 involving the use of medications for opioid use disorder, and 105 involving prescription opioid safety). Of these evidence-based practice strategies, only 235 (38%) had been initiated by the start of the comparison year.

CONCLUSIONS

In this 12-month multimodal intervention trial involving community coalitions in the deployment of evidence-based practices to reduce opioid overdose deaths, death rates were similar in the intervention group and the control group in the context of the Covid-19 pandemic and the fentanyl-related overdose epidemic. (Funded by the National Institutes of Health; HCS ClinicalTrials.gov number, NCT04111939.)

Rates of opioid-related overdose deaths have been a rising scourge in the United States, with an increase in the annual number of deaths from 8050 in 1999¹ to 82,136 in 2022.² This increase was initially attributed to excessive opioid prescribing and more recently to the expansion of fentanyl in the illicit drug supply.³ Fentanyl has become the primary illicit opioid and has contaminated stimulants⁴ and counterfeit pills,⁵ along with escalating the rate of opioid-related overdose deaths disproportionately among Black and Indigenous populations.^6,7

In 2016, the U.S. Surgeon General urged medical professionals to address the opioid crisis through advocacy, stigma reduction, uptake of treatment for opioid use disorder, and safer opioid prescribing.⁸ The National Academy of Medicine echoed this recommendation in 2017.⁹ Both reports highlighted the value of evidence-based practices to prevent or reverse opioid overdoses. These practices include education regarding overdose prevention and naloxone distribution,^10,11 the use of medications (including methadone and buprenorphine) for the treatment of opioid use disorder,^12,13 and safer opioid prescribing, dispensing, and disposal practices (i.e., prescription opioid safety).¹⁴

Despite efforts to advance evidence-based practices in the United States,¹⁵ patient-level and organizational barriers such as stigma¹⁶ and regulations⁹ impede care delivery,^9,17 transportation options,¹⁸ prescriber access,¹⁹ and naloxone availability.²⁰ In the United States, fewer than 1 in 5 persons with opioid use disorder received medications for treatment in 2021,²¹ with even lower rates among Black persons.²² Overdose education at the community level and the availability of naloxone distribution are inadequate to achieve a substantial reduction in opioid-related overdose deaths, according to simulation models.²³

To address these barriers, the National Institutes of Health launched the HEALing (Helping to End Addiction Long-term Initiative) Communities Study (HCS), a large, multistate implementation-science trial focused on substance use.²⁴ The HCS research consortium designed the community-engaged, data-driven Communities That HEAL (CTH) intervention to rapidly scale up strategies to increase access to evidence-based practices and conduct communication campaigns in highly affected communities in Kentucky, Massachusetts, New York, and Ohio, with the primary goal of reducing the rate of opioid-related overdose deaths.²⁵

Methods

TRIAL DESIGN AND OVERSIGHT

The HCS was a multisite, community-level, cluster-randomized, controlled trial to evaluate the CTH intervention in reducing the rate of opioid-related overdose deaths in highly affected communities.²⁵ The trial was grounded in the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) and the Practical, Robust Implementation and Sustainability Model (PRISM) implementation-science framework.²⁶

The trial protocol (available with the full text of this article at NEJM.org) was approved by Advarra, a company providing research review services that served as the single institutional review board for the trial. A waiver of consent and full waiver of Health Insurance Portability and Accountability Act (HIPAA) authorization were granted for secondary data analysis. The data and safety monitoring board, chartered by the National Institute on Drug Abuse, monitored the trial. Data that are reported in this trial are protected by multiple data-use agreements among researchers and agencies providing the data.

The authors vouch for the completeness and accuracy of the data and for the fidelity of the trial to the protocol. Details regarding the authors’ roles in manuscript preparation are provided in Figure S1 in the Supplementary Appendix, available at NEJM.org.

TRIAL POPULATION

All the participating communities (the unit of analysis) had a high baseline rate of opioid-related overdose deaths (≥25 per 100,000 adults); at least 30% of the communities were rural. The demographic characteristics of the trial populations were similar to those in the overall U.S. populations that have been severely affected by fatal opioid overdoses (Fig. S2).

TRIAL PROCEDURES, INTERVENTION, AND RANDOMIZATION

The CTH intervention was adapted from the evidence-based Communities That Care approach that used a five-phase process to support communities in selecting and implementing evidence-based practices to prevent adolescent substance use.²⁷ The CTH intervention used a similar phased-planning process with community coalitions that included local residents who had involvement in the crisis of opioid use disorder through their vocation or personal experience.²⁸ Coalitions assessed the local effect of opioid overdose using data to understand gaps and resources.²⁹ They implemented communication campaigns promoting opioid overdose education and naloxone distribution, the use of medications for opioid use disorder, and messages to reduce stigma through print, digital, radio, outdoor, and social media channels.³⁰ Funds were allocated to the HCS intervention communities to implement strategies for evidence-based practices and communication campaigns, with an average allocation per intervention community of $672,000 in New York, $922,500 in Ohio, $1.69 million in Kentucky, and $1.72 million in Massachusetts.

The HCS investigators have previously described the implementation-science framework²⁶ and the CTH intervention, along with strategies and implementation resources,³¹ community-engaged processes to select and promote adoption and sustainability of evidence-based practices,²⁹ communication campaigns to drive demand and reduce stigma,³⁰ and community dashboards to support data-driven strategy selection.³² A strategy for evidence-based practices was considered to have been implemented if at least one service had been delivered. The number of persons who had been exposed to each strategy was not uniformly assessed. Thus, the pragmatic definition threshold for implementation was the provision of a single service.

We used a covariate-constrained randomization procedure,³³ stratified according to state. Within each state, the intervention and control groups were balanced according to urban or rural classification, rate of opioid-related overdose deaths, and population.²⁵

TIMELINE

The intervention was delivered from January 2020 through June 2022 and later to the previous control communities from July 2022 through December 2023 (Fig. S3). Most implementation of evidence-based practices in the intervention communities occurred from September 2020 through June 2022, except for some early overdose education and naloxone distribution strategies in response to orders for early release of inmates from detention centers related to the coronavirus 2019 disease (Covid-19) pandemic. The comparison period between intervention and control communities was the 12-month period from the beginning of July 2021 through the end of June 2022.²⁸

OUTCOMES

The primary outcome was the number of opioid overdose deaths among community adults, as determined from death certificates with drug overdose as an underlying cause of death (International Classification of Diseases, 10th Revision [ICD-10] codes X40 to X44, X60 to X64, X85, and Y10 to Y14) due to opioids (contributing cause of death, ICD-10 codes, T40.0 to T40.4 and T40.6). In Kentucky, additional death-investigation records determined opioid involvement if specific drugs were absent from death certificates.³⁴ Deaths were attributed to individual communities on the basis of the residence address on the death certificate. A post hoc secondary analysis evaluated the change in rates from 2019 through the comparison period as the outcome.

ADVERSE EVENTS

We used the mean rate of emergency medical service runs for a suspected opioid overdose per 1000 adults to assess potential adverse events and serious adverse events in the community. Adverse events and serious adverse events were reported if a community rate in a single month increased by more than 3 standard deviations above the moving average of the previous 3 months for adverse events or by more than 4 standard deviations for serious adverse events. The number of emergency medical service runs for a suspected opioid overdose was a surrogate measure for opioid-related overdose deaths because of the much shorter lag-time availability for data regarding such runs as compared with deaths (1 to 3 months vs. 6 to 11 months).

STATISTICAL ANALYSIS

We determined that the inclusion of 67 communities would provide the trial with 99% power to detect a prespecified 40% reduction in the rate of opioid-related overdose deaths in the intervention group as compared with the control group and 83% power to detect a 20% reduction in the overdose death rate. We performed analyses using the intention-to-treat principle for the 67 randomized communities. There were no missing data for these analyses. A negative binomial regression analysis modeled the population-averaged rate of opioid-related overdose deaths with the natural log of the adult community population as the offset, after adjustment for state, urban or rural classification, and community-level baseline rates.³⁵ Small sample-adjusted empirical estimates of standard error were used to ensure valid inference.³⁶ The model accounted for clustering by means of the overdispersion parameter; the estimated positive value (i.e., variance exceeding mean) indicated that the negative binomial model was appropriate.

In prespecified exploratory analyses, the primary model was extended to assess the potential interaction of trial group and five variables: state, urban or rural category, age (categorical), sex, and race or ethnic group. We applied multiple-testing adjustments according to the Benjamini–Hochberg procedure³⁷ (using sequential modified Bonferroni correction for multiple hypothesis testing) only to the P values from the five effect-modification tests. We report rate ratio estimates to compare the two groups of communities for each level of the given categorical variables, along with 95% confidence intervals adjusted by a Bonferroni correction on the basis of the number of levels.

We performed a secondary per-protocol analysis with the exclusion of one intervention community that withdrew before the initiation of the intervention. The primary test was two-sided, with a P value of less than 0.05 indicating statistical significance.

A series of post hoc analyses (Table S1) included four sensitivity analyses in which we fit a modified Poisson regression model, which assumed no overdispersion; used log transformation for the baseline rate of opioid-related overdose deaths, which excluded one community with no deaths at baseline; included an interaction between state and urban or rural category; included an interaction between the intervention group and the state, which generated a rate ratio averaged across states; and evaluated combinations of the five variables as listed above. We assessed the goodness of fit for both the negative binomial and Poisson models and determined that the negative binomial provided a better fit than the Poisson model on the basis of plots of deviance residuals (Fig. S4), various fit statistics (Table S2), and the positive dispersion parameter. All the analyses were performed with the use of SAS software, version 9.4 (SAS Institute). The results of secondary and exploratory analyses that were not part of a hierarchical statistical plan and that controlled for multiple comparisons are reported as point estimates with 95% confidence intervals, from which causal inferences should not be made.

RESULTS

BASELINE CHARACTERISTICS

The 67 communities (adult population, 8.2 million) were randomly assigned to either the intervention group (34 communities, including 15 rural) or the control group (33 communities, including 14 rural) (Fig. 1 and Fig. S5). The designated communities included 48 counties, 14 municipalities (11 in Massachusetts and 3 in New York), and rural clusters of municipalities (5 in Massachusetts). The mean (±SD) baseline rate of opioid-related overdose deaths was similar in the intervention group and the control group (38.2±22.8 and 37.1±20.3 per 100,000 population, respectively). Baseline community characteristics are shown in Table 1.^38,39

Figure 1. Enrollment and Randomization.

Shown are the randomization procedures and analyses that were performed in the 67 communities that participated in the HEALing Communities Study (HCS) in Kentucky, Massachusetts, New York, and Ohio. The HCS research consortium designed the Communities That HEAL (CTH) intervention to rapidly scale up strategies to reduce the rate of opioid-related overdose deaths. One intervention community was not included in the per-protocol analysis because the community withdrew before the initiation of the intervention.

Table 1.

Characteristics of the 67 Communities at Baseline.^*

Variable	Intervention Communities				Control Communities				Rate Ratio
	Community (N = 67)	Opioid-Related Overdose Deaths	Population†	Death Rate	Community (N = 67)	Opioid-Related Overdose Deaths	Population†	Death Rate
	no. (%)	no.		no./100,000 population	no. (%)	no.		no./100,000 population
All communities	34 (50.7)	1771	4,439,170	39.9	33 (49.3)	1546	3,772,336	41.0	0.97
Opioid-related overdose deaths
Mean no.	—	—	—	38.2±22.8	—	—	—	37.1±20.3	—
Median no. (IQR)	—	—	—	35.2 (21.6–49.3)	—	—	—	32.7 (23.6– 48.6)	—
State
Kentucky	8 (23.5)	260	617,841	42.1	8 (24.2)	361	815,764	44.3	0.95
Massachusetts	8 (23.5)	172	359,314	47.9	8 (24.2)	201	356,545	56.4	0.85
New York	8 (23.5)	256	1,101,497	23.2	8 (24.2)	318	976,069	32.6	0.71
Ohio	10 (29.4)	1083	2,360,518	45.9	9 (27.3)	666	1,623,958	41.0	1.12
Urban or rural category
Urban	19 (55.9)	1538	3,793,353	40.5	19 (57.6)	1377	3,242,663	42.5	0.95
Rural	15 (44.1)	233	645,817	36.1	14 (42.4)	169	529,673	31.9	1.13
Age group — yr
18 to 34	—	531	1,334,880	39.8	—	508	1,178,210	43.1	0.92
35 to 54	—	848	1,353,341	62.7	—	751	1,180,392	63.6	0.98
≥55	—	392	1,750,949	22.4	—	287	1,413,734	20.3	1.10
Sex
Male	—	1235	2,133,827	57.9	—	1071	1,825,776	58.7	0.99
Female	—	536	2,305,343	23.3	—	475	1,946,560	24.4	0.95
Race or ethnic group‡
Non-Hispanic White	—	1374	3,229,233	42.5	—	1119	2,750,369	40.7	1.05
Non-Hispanic Black	—	267	728,037	36.7	—	278	545,357	51.0	0.72
Hispanic	—	115	281,329	40.9	—	130	322,654	40.3	1.01
Non-Hispanic other	—	12	200,571	6.0	—	17	153,956	11.0	0.54
Missing data	—	3			—	2

^*Plus–minus values are means ±SD. Percentages may not total 100 due to rounding. IQR denotes interquartile range.

^†For communities that represent counties (48 of 67), population estimates are from 2020 Bridged-Race Population Estimates.³⁹ For communities that represent units smaller than counties (19 of 67), population estimates are from the 2017–2021 American Community Survey 5-Year Estimates.⁴⁰

^‡Race or ethnic group was reported on death certificates.

IMPLEMENTATION OF EVIDENCE-BASED PRACTICES

Intervention communities implemented 615 evidence-based practice strategies (254 involving overdose education and naloxone distribution, 256 involving the use of medications for opioid use disorder, and 105 involving prescription opioid safety). However, only 235 of the 615 strategies (38%) were implemented before the comparison period: 43% involving overdose education and naloxone distribution, 36% involving the use of medications for opioid use disorder, and 32% involving prescription opioid safety. Details regarding the implementation of these strategies are provided in Tables S3, S4, and S5.

RATES OF OPIOID OVERDOSE DEATHS

Unadjusted means and rate ratios for the number of opioid-related overdose deaths during the 12-month comparison period are reported according to trial group, state, urban or rural category, age, sex, and race or ethnic group (Table 2).⁴⁰ The overall unadjusted rate ratios of opioid-related overdose deaths comparing the intervention group with the control group were 0.94 (on the basis of the mean unadjusted rates across communities) and 0.82 (on the basis of the sum of unadjusted rates across communities). In the primary adjusted analyses, the rate of opioid-related overdose deaths was 47.2 per 100,000 population for the intervention communities and 51.7 per 100,000 population for the control communities (rate ratio, 0.91; 95% confidence interval [CI], 0.76 to 1.09; P= 0.30) (Table 3). The results of the per-protocol analysis were similar to those in the primary analysis (Table 3). There was no evidence of effect modification according to state, urban or rural category, age, sex, or race or ethnic group (Table 4), although there was substantial variability among states and among urban and rural locations.

Table 2.

Unadjusted Rate of Opioid-Related Overdose Deaths (Intention-to-Treat Population).^*

Variable	Intervention Communities		Control Communities		Unadjusted Rate Ratio
	Opioid-Related Overdose Deaths	Population	Opioid-Related Overdose Deaths	Population
number of persons
All communities	65.3±98.1	130,563.8±200,088.0	69.6±143.8	114,313.2±201,417.3	0.94
State
Kentucky	48.9±46.5	77,230.1±80,938.9	76.1±168.7	101,970.5±202,045.3	1.11
Massachusetts	25.1±22.2	44,914.3±26,559.3	30.1±24.7	44,568.1±33,628.8	0.74
New York	59.0±73.9	137,687.1±140,779.9	67.9±74.6	122,008.6±106,012.1	0.73
Ohio	115.6±156.3	236,051.8±322,922.0	100.4±223.0	180,439.8±325,173.2	1.24
Urban or rural category
Urban	100.2±120.1	199,650.2±248,385.2	106.8±182.1	170,666.5± 252,614.9	0.96
Rural	21.1±19.4	43,054.5±19,075.4	19.1±14.8	37,833.8±23,733.0	0.92
Age group — yr
18 to 34	17.6±25.6	39,261.2±60,440.2	19.5±42.1	35,703.3±69,802.6	0.88
35 to 54	32.4±48.8	39,804.1±60,321.8	33.7±70.3	35,769.5±65,163.6	0.97
≥55	15.3±24.7	51,498.5±79,908.4	16.4±32.3	42,840.4±67,485.8	0.97
Sex
Male	44.9±68.2	62,759.6±94,370.9	48.5±98.4	55,326.5±96,853.5	0.89
Female	20.4±30.1	67,804.2±105,743.9	21.1±45.6	58,986.7± 104,579.4	1.08
Race or ethnic group
Non-Hispanic White	46.6±61.6	94,977.4±128,763.1	46.6±91.2	83,344.5±135,601.7	0.97
Non-Hispanic Black	13.6±30.6	21,412.9±55,089.6	16.2±46.9	16,526.0±45,296.8	0.96
Hispanic	4.0±8.0	8,274.4±13,482.1	5.4±10.8	9,777.4±17,743.5	0.77
Non-Hispanic other	0.9±2.2	5,899.1±9,757.0	1.2±2.9	4,665.3±11,990.9	0.79
Missing data	0.3±0.8		0.4±0.9

^*Plus–minus values are means ±SD. Listed are data for adults (≥18 years of age) during the comparison period from July 1, 2021, to June 30, 2022. For communities that represent counties (48 of 67), population information is drawn from 2020 Bridged-Race Population Estimates.³⁹ For communities that represent units smaller than counties (19 of 67), population information is drawn from 2017–2021 American Community Survey 5-Year Estimates.⁴⁰ Figure S2 in the Supplementary Appendix provides additional data regarding the total number of opioid-related overdose deaths in community subgroups.

Table 3.

Adjusted Rate of Opioid-Related Overdose Deaths, According to Population.^*

Population	Adjusted Rate in Intervention Communities (95% CI)	Adjusted Rate in Control Communities (95% CI)	Adjusted Rate Ratio (95% CI)	P Value†
no./100,000 population
Intention-to-treat analysis	47.2 (41.8–53.2)	51.7 (44.9–59.6)	0.91 (0.76–1.09)	0.30
Per-protocol analysis	47.2 (41.8–53.3)	51.6 (44.8–59.4)	0.91 (0.77–1.09)	—

^*Listed are results obtained during the comparison period from July 1, 2021 through June 30, 2022. Data were analyzed by means of a negative binomial model after adjustment for trial group, state (Kentucky, Massachusetts, New York, and Ohio), urban or rural category, and baseline rate of opioid-related overdose deaths. The natural log of the community population size during the comparison period is used as an offset.

^†P values are reported only for the primary analysis and for additional prespecified analyses that controlled for multiple testing.

Table 4.

Adjusted Rate of Opioid-Related Overdose Deaths (Intention-to-Treat Population).^*

Variable	Adjusted Rate in Intervention Communities (95% CI)	Adjusted Rate in Control Communities (95% CI)	Adjusted Rate Ratio (95% CI)
no./100,000 population
State
Kentucky	59.8 (44.5–80.4)	59.3 (35.5–99.2)	1.01 (0.56–1.81)
Massachusetts	45.0 (30.7–65.9)	52.2 (38.0–71.8)	0.86 (0.54–1.37)
New York	46.4 (32.1–67.0)	53.1 (41.7–67.7)	0.87 (0.56–1.35)
Ohio	39.3 (31.7–48.7)	43.0 (28.9–64.1)	0.91 (0.58–1.44)
Urban or rural category
Urban	48.2 (41.5–55.9)	50.3 (41.6–60.9)	0.96 (0.76–1.21)
Rural	45.1 (35.3–57.7)	54.4 (41.4–71.5)	0.83 (0.57–1.20)
Age — yr
18 to 34	46.2 (39.8–53.5)	51.3 (42.4–62.1)	0.90 (0.72–1.13)
35 to 54	69.9 (58.2–83.9)	76.6 (64.8–90.5)	0.91 (0.73–1.15)
≥55	33.6 (27.8–40.7)	39.3 (28.9–53.6)	0.86 (0.59–1.23)
Sex
Male	60.6 (50.6–72.6)	69.2 (58.4–82.1)	0.88 (0.71–1.08)
Female	34.1 (29.0–40.0)	37.2 (30.2–45.8)	0.91 (0.71–1.18)
Race or ethnic group
Non-Hispanic White	45.4 (36.6–56.3)	47.6 (38.0–59.5)	0.95 (0.72–1.26)
Non-Hispanic Black	70.3 (52.4–94.2)	77.1 (54.9–108.4)	0.91 (0.59–1.40)
Hispanic	39.3 (24.4–63.3)	46.4 (31.6–68.0)	0.85 (0.46–1.57)
Non-Hispanic other	16.5 (7.5–36.4)	28.2 (13.6–58.7)	0.59 (0.20–1.68)

^*Listed are results obtained during the comparison period from July 1, 2021 through June 30, 2022. A separate negative binomial model was fit for each test of effect modification. Each model included the fixed effects of trial group, state, urban or rural category, baseline opioid-related overdose death rate, stratification variable (age, sex, and race or ethnic group), and a two-way interaction between the trial group and the stratification variable.

ADVERSE EVENTS

The risks of adverse events and serious adverse events (as defined according to the number of opioid-related emergency medical service runs) were similar in the two groups (Table S6). None of the adverse events or serious adverse events were attributed to the trial intervention.

PROTOCOL DEVIATIONS AND NONCOMPLIANCE REPORTS

Five protocol deviations were reported with respect to consent procedures. One noncompliance report was related to a lack communication with the institutional review board regarding revisions to the trial protocol (Fig. S6).

DISCUSSION

The HCS was a large implementation-science trial focused on substance use involving 67 communities with high rates of opioid-related overdose deaths in four states. During the comparison year (July 2021 through June 2022) across communities, 4517 deaths were linked to opioid overdoses. Despite the breadth and depth of the intervention, the risk of opioid-related overdose death was similar in the intervention group and the control group. Numerous factors may have tempered the effect of the intervention. Of these factors, three warrant discussion: the duration and intensity of the implementation, the Covid-19 pandemic, and the evolving illicit drug market.

First, the evidence-based practices were addressed by a complex array of strategies for high-risk populations in health care, behavioral health, and criminal legal sectors. After the strategy selection process, only 10 months preceded the comparison period to establish agency partnerships and implement evidence-based practices.²⁸ The time frame was insufficient to initiate many strategies for evidence-based practices, which often required recruiting new staff from an increasingly scarce health care workforce, changing clinical practice workflows, or developing new interagency collaborations to introduce services (e.g., in criminal legal settings). Ultimately, 615 strategies were implemented, which suggests the effectiveness of the CTH intervention in mobilizing communities to pursue the adoption of evidence-based practices. However, greater penetrance of the practices through continuous quality improvement and longer time horizons may be needed to observe the full effect of the CTH intervention on opioid-related overdose deaths.⁴¹

Second, the intervention launch (in January 2020) occurred 2 months before the Covid-19 shutdown, which severely disrupted systems targeted by the CTH intervention. Through a protocol amendment, 6 months were added for implementation, but the initiation of activities of the organizations was still delayed. Hence, the pandemic reduced the capacity of communities to implement the CTH intervention, which mitigated the potential for the intervention to affect the rate of opioid-related overdose deaths.

Third, the change in the illicit drug market may have reduced the effectiveness of the intervention, because fentanyl became a more prevalent opioid and a more commonly used adulterant in stimulants⁴ and counterfeit pills.⁵ Both fentanyl and, increasingly, xylazine⁴² posed new challenges for opioid-overdose treatment. We do not know whether surges in fentanyl use over time were similar across communities.

The embrace of evidence-based practices by coalitions in the intervention communities varied considerably in part because of competing challenges posed by the Covid-19 pandemic, the fentanyl epidemic, and continuing stigma related to drug use — all of which affected willingness and capacity to respond to the opioid crisis. The CTH intervention had a major effect on the implementation of evidence-based practices (615 of 806 that were selected) and communication campaigns (with 165 implemented). Failure to observe significant reductions in opioid-related overdose deaths may have resulted from the temporal lags between organizational implementation and the servicing of a sufficiently large number of affected persons. In retrospect, the prespecified 40% reduction in opioid-related overdose deaths was clearly ambitious. The trial may have been underpowered to detect substantially smaller yet clinically meaningful differences.

The trial has several strengths. The coalitions used timely data to inform the selection of many strategies for evidence-based practices and communication. Heterogeneous coalitions, including members with lived experience who were working with university and state public health partners, enhanced the decision-making process and established partnerships for future sustainability. The randomized, controlled trial design involving 67 communities that were balanced with respect to key covariates addressed threats to internal validity. Finally, coalitions created a sense of urgency and developed partnerships among community agencies, persons with lived experience, public health practitioners, and government officials to address the opioid crisis.

Several trial limitations are noteworthy. First, despite randomization, differences among states may have played a role in observed outcomes. For example, the number of residents in HCS communities in the state with the lowest population (Massachusetts, with 715,859 residents) was less than 20% of that in the state with the highest population (Ohio, with 3,984,476 residents). In addition, the steep rise in opioid-related overdose deaths during the Covid-19 pandemic was not uniform across states. For example, from January 2020 through June 2022, Kentucky had an increase in deaths of 77% as compared with an increase of 15% in Massachusetts.² These differences led to variation in the absolute number of persons who could benefit from the allocated trial resources. It is possible that secular trends and contamination by the inadvertent exposure of control communities to evidence-based practice strategies may have attenuated the effect of the CTH intervention. For example, control communities could access non-HCS funds to address the opioid epidemic. These funds became more available in response to increases in opioid-related overdose deaths during the Covid-19 pandemic and may have resulted in activities that could have masked an intervention effect. Also, adults in control communities that were proximal to intervention communities may have been exposed to CTH evidence-based practices and communication campaigns. Another limitation is that the HCS did not consistently assess the number of persons who were affected by the strategies that were implemented in the intervention communities.

HCS investigators examined the potential of the community-engaged, data-driven CTH intervention to reduce the rate of opioid-related overdose deaths in highly affected communities. Intervention communities implemented hundreds of strategies to expand opioid overdose education and naloxone distribution, the use of medications for opioid use disorder, and safety measures for prescription opioid use, as well as communication campaigns to support these efforts. Although there were no significant between-group differences in the rate of opioid-related overdose deaths, the trial showed that the CTH community-engaged intervention, with its leveraging of community coalitions and a data-driven approach, can bring about meaningful progress in implementing evidence-based practices.

Appendix

The authors’ full names and academic degrees are as follows: Jeffrey H. Samet, M.D., M.P.H., Nabila El-Bassel, Ph.D., T. John Winhusen, Ph.D., Rebecca D. Jackson, M.D., Emmanuel A. Oga, M.D., M.P.H., Redonna K. Chandler, Ph.D., Jennifer Villani, Ph.D., M.P.H., Bridget Freisthler, Ph.D., Joella Adams, Ph.D., Arnie Aldridge, Ph.D., Angelo Angerame, B.S., Denise C. Babineau, Ph.D., Sarah M. Bagley, M.D., Trevor J. Baker, M.S., Peter Balvanz, M.P.H., Carolina Barbosa, Ph.D., Pharm.D., Joshua Barocas, M.D., Tracy A. Battaglia, M.D., M.P.H., Dacia D. Beard, M.P.H., M.B.A., Donna Beers, M.S.N., R.N.-B.C., Derek Blevins, M.D., Nicholas Bove, M.P.H., Carly Bridden, M.A., M.P.H., Jennifer L. Brown, Ph.D., Heather M. Bush, Ph.D., Joshua L. Bush, Ph.D., Ryan Caldwell, Katherine Calver, Ph.D., Deirdre Calvert, M.S.W., Aimee N.C. Campbell, Ph.D., Jane Carpenter, M.P.H., Rachel Caspar, M.A., Deborah Chassler, M.S.W., Joan Chaya, M.A., Debbie M. Cheng, Sc.D., Chinazo O. Cunningham, M.D., Anindita Dasgupta, Ph.D., James L. David, M.S., Alissa Davis, Ph.D., Tammy Dean, B.S., Mari-Lynn Drainoni, Ph.D., Barry Eggleston, M.S., Laura C. Fanucchi, M.D., M.P.H., Daniel J. Feaster, Ph.D., Soledad Fernandez, Ph.D., Wilson Figueroa, Ph.D., Darcy A. Freedman, Ph.D., M.P.H., Patricia R. Freeman, Ph.D., Caroline E. Freiermuth, M.D., M.H.S., Eric Friedlander, B.A., Kitty H. Gelberg, Ph.D., Erin B. Gibson, M.P.H., Louisa Gilbert, Ph.D., LaShawn Glasgow, Dr.P.H., Dawn A. Goddard-Eckrich, Ed.D., Stephen Gomori, M.S., Dawn E. Gruss, M.Ed., Jennifer Gulley, M.P.H., B.S.N., R.N., Damara Gutnick, M.D., Megan E. Hall, M.P.H., Nicole Harger Dykes, Pharm.D., Sarah L. Hargrove, M.S., Kristin Harlow, Ph.D., Aumani Harris, M.P.H., Daniel Harris, Ph.D., Donald W. Helme, Ph.D., JaNae Holloway, M.S.P.H., Juanita Hotchkiss, M.S.W., Terry Huang, Ph.D., M.P.H., Timothy R. Huerta, Ph.D., Timothy Hunt, Ph.D., M.S.W., Ayaz Hyder, Ph.D., Van L. Ingram, Tim Ingram, M.S., Emily Kauffman, D.O., M.P.H., Jennifer L. Kimball, B.A., Elizabeth N. Kinnard, M.S., Charles Knott, M.P.A., Hannah K. Knudsen, Ph.D., Michael W. Konstan, M.D., Sarah Kosakowski, M.P.H., Marc R. Larochelle, M.D., M.P.H., Hannah M. Leaver, M.B.A., Patricia A. LeBaron, M.S., R. Craig Lefebvre, Ph.D., Frances R. Levin, M.D., Nikki Lewis, M.P.H., Nicky Lewis, Ph.D., Michelle R. Lofwall, M.D., David W. Lounsbury, Ph.D., Jamie E. Luster, M.P.H., Michael S. Lyons, M.D., M.P.H., Aimee Mack, M.P.H., Katherine R. Marks, Ph.D., Stephanie Marquesano, J.D., Rachel Mauk, Ph.D., Ann Scheck McAlearney, Sc.D., Kristin McConnell, M.S., Margaret L. McGladrey, Ph.D., Jason McMullan, M.D., Jennifer Miles, C.P.A., Rosie Munoz Lopez, M.P.H., Alisha Nelson, M.B.A., Jessica L. Neufeld, M.P.H., Lisa Newman, M.P.H., Trang Q. Nguyen, M.D., Dr.P.H., Edward V. Nunes, M.D., Devin A. Oller, M.D., Carrie B. Oser, Ph.D., Douglas R. Oyler, Pharm.D., Sharon Pagnano, M.P.H., Theodore V. Parran, M.D., Joshua Powell, M.S., Kim Powers, B.S.W., William Ralston III, M.D., Kelly Ramsey, M.D., M.P.H., Bruce D. Rapkin, Ph.D., Jennifer G. Reynolds, M.P.H., Monica F. Roberts, Pharm.D., Will Robertson, Peter Rock, M.P.H., Emma Rodgers, M.P.H., Sandra Rodriguez, M.P.H., Maria Rudorf, M.A., Shawn Ryan, M.D., Pamela Salsberry, Ph.D., M.P.H., Monika Salvage, M.Ag., Nasim Sabounchi, Ph.D., Merielle Saucier, B.A., Caroline Savitzky, M.S.W., Bruce Schackman, Ph.D., Elizabeth Schady, B.A., Eric E. Seiber, Ph.D., Aimee Shadwick, M.A., Abigail Shoben, Ph.D., Michael D. Slater, Ph.D., Svetla Slavova, Ph.D., Drew Speer, Dr.P.H., Joel Sprunger, Ph.D., Laura E. Starbird, Ph.D., Michele Staton, Ph.D., Michael D. Stein, M.D., Danelle J. Stevens-Watkins, Ph.D., Thomas J. Stopka, Ph.D., M.H.S., Ann Sullivan, Ph.D., Hilary L. Surratt, Ph.D., Rachel Sword Cruz, M.S.W., M.P.H., Jeffery C. Talbert, Ph.D., Jessica L. Taylor, M.D., Katherine L. Thompson, Ph.D., Nathan Vandergrift, Ph.D., Rachel A. Vickers-Smith, Ph.D., Deanna J. Vietze, M.S., Daniel M. Walker, Ph.D., M.P.H., Alexander Y. Walley, M.D., Scott T. Walters, Ph.D., Roger Weiss, M.D., Philip M. Westgate, Ph.D., Elwin Wu, Ph.D., April M. Young, Ph.D., Gary A. Zarkin, Ph.D., and Sharon L. Walsh, Ph.D.

The authors’ affiliations are as follows: Boston Medical Center, Boston University Chobanian and Avedisian School of Medicine, Department of Medicine, Section of General Internal Medicine, and Boston University School of Public Health (J.H.S.), Boston Medical Center and Boston University Chobanian and Avedisian School of Medicine, Departments of Medicine and Pediatrics (S.M.B.), Boston Medical Center, Section of General Internal Medicine (T.J.B., P.B., D. Beers, C. Bridden, K.C., J. Carpenter, E.B.G., A. Harris, S.K., Nikki Lewis, R.M.L., M.R., M. Saucier, R.S.C.), Boston University Chobanian and Avedisian School of Medicine (T.A.B.), Boston University School of Public Health, Department of Health Law, Policy and Management (D.D.B., M.D. Stein), Massachusetts Department of Public Health, Bureau of Substance Addiction Services (D. Calvert), Boston University School of Social Work (D. Chassler), Boston University School of Public Health, Department of Biostatistics (D.M.C.), Boston University Chobanian and Avedisian School of Medicine, Department of Medicine, Section of Infectious Diseases, and Boston University School of Public Health, Department of Health Law, Policy and Management (M.-L.D.), Massachusetts HCS Community Advisory Board (J.L.K., K.P.), Boston Medical Center, Section of Infectious Diseases (E.N.K., C.S.), Boston Medical Center and Boston University Chobanian and Avedisian School of Medicine, Department of Medicine, Section of General Internal Medicine (M.R. Larochelle, J.L.T., A.Y.W.), Boston Medical Center, Department of Medicine (H.M.L.), Massachusetts Department of Public Health, Registry of Vital Records and Statistics (S.P.), Tufts University School of Medicine, Department of Public Health and Community Medicine (T.J.S.), and McLean Hospital, Division of Alcohol, Drugs, and Addiction, and Harvard Medical School, Department of Psychiatry (R.W.) — all in Boston; the Social Intervention Group, School of Social Work, Columbia University (N.E.-B., A. Dasgupta, J.L.D., A. Davis, K.H.G., L. Gilbert, D.A.G.-E., D.E.G., J. Hotchkiss, T. Hunt, J.L.N., E.R., S. Rodriguez, E.W.), New York HCS Community Advisory Board (A. Angerame, R. Caldwell, S.M., K.M., J.P., K.R., W.R., M. Salvage), Columbia University Irving Medical Center, Department of Psychiatry (D. Blevins, A.N.C.C., F.R.L., E.V.N.), Albert Einstein College of Medicine, Department of Epidemiology and Population Health (N.B., D.G., D.W.L., B.D.R.), Montefiore Medical Center (J. Chaya), New York State Office of Addiction Services and Supports (C.O.C.), City University of New York (T. Huang, N.S.), Weill Cornell Medicine, Department of Population Health Sciences (B.S.), and the New York Office of Mental Health (A. Sullivan), New York, and the New York State Department of Health, Albany (T.Q.N., E.S.) — all in New York; the Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati College of Medicine (T.J.W.), University of Cincinnati Corrections Institute (T.D.), University of Cincinnati College of Medicine, Department of Emergency Medicine (C.E.F., J. McMullan), University of Cincinnati Medical Center, Emergency Medicine (N.H.D.), University of Cincinnati College of Medicine, Department of Environmental and Public Health Sciences (T.I.), Case Western Reserve University School of Medicine, Department of Family Medicine and Community Health (T.V.P.), Brightview Health (S. Ryan), and University of Cincinnati College of Medicine, Department of Psychiatry and Behavioral Neuroscience (J.S.), Cincinnati, the College of Medicine (R.D.J., S.F., K.H., J.E.L., M.S.L.) and the College of Social Work (B.F.), Center of Health Outcomes and Policy Evaluation Studies (W.F.), Department of Family and Community Medicine (T.R.H., A.S.M., D.M.W.), College of Public Health and Translational Data Analytics Institute (A. Hyder), Department of Emergency Medicine (E.K.), Ohio Colleges of Medicine Government Resource Center (A.M., R.M.), One Ohio Foundation (A.N.), College of Public Health (P.S., E.E.S., A. Shoben), Recovery Ohio (A. Shadwick), and the School of Communication (M.D. Slater), Ohio State University, Columbus, Case Western Reserve University School of Medicine, Department of Population and Quantitative Health Sciences (D.A.F.), and Case Western Reserve University School of Medicine, Clinical and Translational Science Institute (M.W.K.), Cleveland, and Brown County Mental Health and Addiction Services, Georgetown (D.J.V.) — all in Ohio; RTI International, Research Triangle Park, NC (E.A.O., J.A., A. Aldridge, D. Babineau, C. Barbosa, R. Caspar, B.E., L. Glasgow, S.G., M.E.H., J. Holloway, C.K., P.A.L., R.C.L., L.N., N.V., G.A.Z.); the National Institute on Drug Abuse, Bethesda, MD (R.K.C., J.V.); University of Colorado School of Medicine, Divisions of General Internal Medicine and Infectious Diseases, Aurora (J.B.); University of North Texas Health Science Center, Fort Worth (S.T.W.); Kentucky Office of Drug Control Policy, Frankfort (V.L.I.), University of Kentucky, College of Public Health (H.M.B.), University of Kentucky, Kentucky Injury Prevention Research Center (J.L. Bush, S.L.H ), University of Kentucky College of Medicine, Department of Medicine, Division of Infectious Diseases, Center on Drug and Alcohol Research (L.C.F.), University of Kentucky, Department of Pharmacy Practice and Science (P.R.F., D.H., D.R.O.), Commonwealth of Kentucky, Cabinet for Health and Family Services (E.F., K.R.M.), University of Kentucky, Department of Communication (D.W.H., Nicky Lewis), University of Kentucky, Department of Behavioral Science (H.K.K.), University of Kentucky College of Medicine, Center on Drug and Alcohol Research (M.R. Lofwall, S.L.W.), University of Kentucky, Department of Health Management and Policy and Center for Innovation in Population Health (M.L.M.), University of Kentucky, Substance Use Research Priority Area (J. Miles, M.F.R., P.R., D.S.), University of Kentucky College of Medicine, Internal Medicine (D.A.O.), University of Kentucky, Department of Sociology (C.B.O.), University of Kentucky (B.D.R.), University of Kentucky, Department of Biostatistics (S.S., P.M.W.), University of Kentucky, Dr. Bing Zhang Department of Statistics (K.L.T.), University of Kentucky, Department of Behavioral Science (M. Staton, H.L.S.), University of Kentucky, Center for Health Equity Transformation (D.J.S.-W.), University of Kentucky College of Medicine, Institute for Biomedical Informatics (J.C.T.), and University of Kentucky, Department of Epidemiology and Environmental Health (R.A.V.-S., A.M.Y.), Lexington, and the Commonwealth of Kentucky, Clark County Health Department, Winchester (J.G.) — all in Kentucky; Purdue University, Department of Psychological Sciences, Lafayette, IN (J.L. Brown); University of Miami Miller School of Medicine, Department of Public Health Sciences, Miami (D.J.F.); Oak Ridge Associated Universities (ORAU), Health Communications, Marketing, and Promotion Program, Oak Ridge, TN (J.G.R.); and University of Pennsylvania School of Nursing, Department of Family and Community Health, Philadelphia (L.E.S.).