Do Communities Implementing the Communities That HEAL Intervention Have Significantly Lower Rates of High-Risk Opioid Prescribing and Dispensing?

Abstract

Introduction

Prescription opioids can contribute to risk for opioid use disorder and overdoses. Improving prescription opioid safety is a critical component in reducing opioid risks. This report aims to determine whether communities randomized to the Communities That HEAL (CTH) intervention have significantly different rates of prescription opioid safety measures.

Study Design

A multisite, 2-arm, community-level, cluster randomized, unblinded, wait-list controlled comparison trial designed to assess the effectiveness of the CTH intervention in reducing opioid-related overdose deaths among community residents 18 years of age or older (adults).

Setting/Participants

Sixty-seven (67) communities in Kentucky, Massachusetts, New York, and Ohio. Participants were communities in this study.

Intervention

The Communities That Heal intervention consists of multiple dimensions: a coalition-driven community engagement process to select and support implementation of evidence-based practices; the Opioid-overdose Reduction Continuum of Care Approach, a compendium of evidence-based practices and technical assistance resources organized under overdose education and naloxone distribution, medication for opioid use disorder, and prescription opioid safety menus; and communication campaigns intended to reduce opioid use disorder stigma and raise awareness and demand for naloxone and medication for opioid use disorder.

Main Outcome(s) and Measure(s)

The main outcome was the number of adults with new incident high-risk opioid prescribing episodes after at least a 45-day washout. Other outcomes included the number of opioid-naïve adults with new opioid prescriptions limited to a 7-day supply, number of adults who received opioid prescriptions from multiple prescribers or pharmacies, and number of locations providing drug take-back services. Outcomes were assessed from July 2021 to June 2022.

Results

There was no statistically significant difference in the adjusted rates for new incident high-risk opioid prescribing per 100,000 adults during the comparison period between intervention (1,094.48; 95% CI=1,063.15; 1,126.74) and wait-list control communities (1,121.90; 95% CI=1,079.62; 1,165.84). The adjusted relative rate comparing intervention to wait-list control communities was 0.98 (95% CI=0.93, 1.02; p-value=0.296). Similarly, there were no statistically significant differences between intervention and wait-list control communities for the other outcomes.

Conclusions

Although no statistically significant differences were found in prescription opioid safety measures between study arms, improvement in these measures during the comparison period for both study arms suggested that there may have events outside the trial, such as published revised Center for Disease Control and Prevention clinical practice guidelines for prescribing opioids, that may have impacted study outcomes.

HIGHLIGHTS

• •The Communities That HEAL intervention was well implemented in diverse community settings.
• •For high-risk prescribing episodes, intervention and control groups did not differ.
• •Both groups had improvements, suggesting that outside factors impacted study outcomes.
• •Revised Center for Disease Control and Prevention opioid guidelines may have affected outcomes during the trial.

INTRODUCTION

The opioid crisis claimed over 80,000 lives in the U.S. in 2021¹ and is currently being fueled by overdoses due to fentanyl and stimulants.^2,3 Although synthetic opioids, primarily illicit fentanyl, currently account for most opioid-related overdose deaths, nearly 90% of individuals who misused opioids in 2021 reported prescription pain reliever misuse only.⁴ There is an established link between nonmedical or high-risk prescription opioid use and subsequent illicit drug use, opioid use disorder (OUD), and overdose death.5, 6, 7 Although overall opioid prescribing in the U.S. has been decreasing for a decade owing to increased education on prescribing guidelines and regulations, strategies to promote safe opioid prescribing and address high-risk prescribing practices must continue to reduce OUD and opioid-related overdose deaths.

Key measures for high-risk prescribing practice include the following: dose, with 90 morphine milligram equivalents per day commonly used as a threshold for high risk^8,9; concomitant use of an opioid and another central nervous system depressant, specifically benzodiazepines; and initiation of therapy with extended-release or long-acting opioid formulations.^10,11 The duration of initial opioid use is associated with a transition to long-term use, with nearly half of individuals who fill an initial opioid prescription of more than 30 days remaining on opioids for at least 1 year.¹²

Even short-duration opioid prescriptions carry risk for both individuals and communities. Approximately 15% of individuals who fill a 7-day opioid prescription remain on the drug for at least a year.¹² Numerous studies suggest that acute opioid prescriptions are written for an unnecessarily high quantity or long duration,13, 14, 15, 16 and unused opioids present a risk to households and communities.17, 18, 19 Increasing the availability of controlled substance disposal locations in a community may help mitigate this risk because patients are more likely to properly dispose of unused opioids when disposal kiosks are conveniently located.²⁰ Finally, the use of multiple prescribers or multiple pharmacies is associated with an increased likelihood of opioid-related overdose,21, 22, 23 and strategies to reduce multiple-provider episodes may also reduce opioid harms, including overdose deaths.^24,25

The HEALing (Helping to End Addiction Long-term) Communities Study (HCS) used a cluster RCT to test the impact of the Communities That HEAL (CTH) intervention on the use of evidence-based practices (EBPs) to reduce opioid-related overdose deaths among community residents aged ≥18 years (adults).²⁶ The CTH intervention consists of multiple dimensions: a coalition-driven community engagement process to select and support implementation of EBPs²⁷; the Opioid-overdose Reduction Continuum of Care Approach (ORCCA), a compendium of EBPs and technical assistance resources organized under overdose education and naloxone distribution, medication for OUD (MOUD), and prescription opioid safety menus²⁸; and communication campaigns intended to reduce OUD stigma and raise awareness and demand for naloxone and MOUD.²⁹

The focus of this analysis was to compare the rate of adults with new incident high-risk opioid prescribing (defined in detail below) in intervention with that of wait-list control communities during the comparison period. Secondary analyses included comparing the rate of opioid-naïve adults with new opioid prescriptions limited to a 7-day supply, the rate of individuals who received opioid prescriptions from multiple prescribers or pharmacies, and the rate of locations providing drug take-back services.

METHODS

The HCS consisted of a multisite, 2-arm, community-level, cluster randomized, unblinded, wait-list controlled comparison trial designed to evaluate the CTH intervention’s effectiveness in reducing opioid-related overdose deaths among adult residents in highly affected communities.26, 30 The intervention communities received the CTH intervention from January 2020 through June 2022. The timeframe for EBP implementation was effectively September 1, 2020 through June 30, 2022. Due to the time needed to work with communities and organizations to expand delivery of EBPs, a lag was expected from when the CTH intervention was introduced into a community to when its effect on key outcomes would be observed. Therefore, outcomes were measured from July 1, 2021 to June 30, 2022, defined as the comparison period.

This study’s aims were to determine whether communities randomized to the CTH intervention have significantly different rates of prescription opioid safety measures as compared to wait-list control communities.

This study protocol (Pro00038088) was approved by Advarra Inc., the single Institutional Review Board (sIRB) selected to oversee the HCS. The study was granted a Waiver of Consent and a Full Waiver of HIPAA Authorization for secondary data analysis from the sIRB (Advarra, 10/25/2019, MOD00521925). The DSMB, chartered by NIDA, is an independent group charged with monitoring the safety of participants and the quality of the data.

Study Population

The Intention-to-Treat (ITT) population includes adult residents 18 years of age or older in 67 communities in Kentucky (KY), Massachusetts (MA), New York (NY), and Ohio (OH). Communities consist of counties (N=48), or cities and towns (N=19). Communities across four states (Kentucky, Massachusetts, New York, Ohio) were selected to participate based on the following eligibility criteria established by the National Institute on Drug Abuse: (1) expressed willingness to address the implementation of MOUD and overdose education and naloxone distribution; (2) expressed willingness to develop partnerships across health care, behavioral health, and justice settings for EBPs to address opioid misuse, OUD, and overdoses; (3) within each state, ≥30% of selected communities were rural; and (4) across the HCS communities in each state, ≥150 opioid-related overdose fatalities (at least 15% occurring in rural communities) and a rate of ≥25 opioid-related overdose fatalities per 100,000 people, based on 2016 data. The communities are either counties (NY, KY, OH) or cities/towns (MA). Additional state-specific criteria were applied to further refine selection. KY selected counties with: 1) a syringe service program (marker of community readiness); 2) a jail; 3) ≥1 buprenorphine-waivered provider; and 4) ≥5 opioid overdose deaths in 2017. The demographic characteristics of the trial populations, as reported in previous studies, were similar to those of U.S. populations most affected by the opioid crisis.³¹ Further details are provided in the protocol paper.²⁶

Study Procedures and Intervention

The HCS Consortium published papers focused on study components³² including: an introduction;³³ the study protocol;²⁶ the ORCCA,²⁸ which is a compendium of EBPs and resources designed to facilitate EBP implementation; community engagement to support the adoption and sustainability of EBPs;²⁷ health communication campaigns to drive demand for EBPs and reduce stigma;²⁹ and community dashboards to support data-driven adoption of EBPs.³⁴

The HCS compiled a comprehensive list of strategies and associated resources for implementing evidence-based practices known to impact opioid overdose reduction and gave each community the flexibility to decide what strategies to include in their action plan based on a data informed and community driven approach to meet goals, and consider viability and impact.^26,35

All intervention communities were required to implement at least one strategy to improve prescription opioid safety through: (1) safer opioid prescribing for acute pain across varied healthcare settings, (2) safer opioid prescribing for chronic pain, or (3) safer opioid dispensing. Safer opioid disposal practices were considered an optional strategy. For example, various guidelines and academic detailing services were made available for safer opioid prescribing for acute pain that may have been included in a community’s action plan. Similarly, various national guidelines, pain educational toolkits, and online free curriculum were made available to target safer opioid prescribing for chronic pain. Safer opioid dispensing was promoted through numerous academic detailing and educational programs targeting pharmacists, while drug take back programs may have been chosen for implementation. A data-driven approach was used to guide the selection and implementation of strategies related to prescription opioid safety.

Allocation to study arm was conducted by the HCS Data Coordinating Center (DCC) using a covariate constrained randomization procedure35, 36, 37, 38 stratified by state. Within each state, study arms were balanced on community characteristics including urban/rural classification, opioid-related overdose death rates, and community population. Thirty-four (34) communities were allocated to the CTH intervention and 33 were allocated to wait-list control. The wait-list control communities were not provided with specific intervention strategies but were also not discouraged from implementing their own strategies. Each community had an equal probability to be allocated to either study arm.²⁶

Measures

Two sources of administrative data, the Prescription Drug Monitoring Program (PDMP) and Drug Enforcement Administration (DEA), provide the outcome measures used in these analyses.³⁹ The PDMP provided data on dispensed prescriptions for opioids and benzodiazepines and the DEA and state law enforcement agencies provided data on drug take-back services. To protect confidentiality, PDMP data were subject to suppression rules (counts of 1-4 in MA and OH, 1-5 in KY and NY). The ORCCA Tracker (ORCCAT) served as the data source for actively implemented EBPs. Actively implemented is defined as any prescribers/pharmacists/health systems participating in opioid safety education for providers and/or patients or technical assistance, or at least one physical location has conducted a drug take-back strategy.

The main outcome for this analysis was the number of adults with new incident high-risk opioid prescribing (after a washout period of at least 45 days) based on meeting at least one of four criteria: (a) opioid-naïve individual receiving opioid prescriptions for more than 30 days (continuous opioid receipt with gaps of no more than 7 days); (b) opioid-naïve individual receiving an initial prescription with extended-release or long-acting opioid formulation; (c) an incident of prescribed high-dose opioid (average daily dose of 90 mg morphine milligram equivalent (MME) or more) over three calendar months; or d) an incident of overlapping opioid and benzodiazepine prescriptions for at least 30 days over three calendar months. Other outcomes included: number of opioid-naïve adults receiving new opioid prescriptions limited to a 7-day supply; number of adults who received opioid prescriptions from four or more prescribers or four or more pharmacies in a three-month period; and number of locations providing drug take-back services.

Statistical Analysis

The HCS was designed to have high power (>99%) to detect a 40% reduction in the primary outcome for the HCS trial, number of opioid-related overdose deaths.²⁶ The outcomes presented in this manuscript were secondary outcomes of the HCS trial and so no sample size calculations were performed.

Descriptive statistics were used to summarize the baseline characteristics of communities and the baseline rate of each outcome (by study arm and overall). Prescription opioid safety strategies implemented in intervention communities were also summarized by state and by urban/rural classification.

Each community-level outcome measured during the comparison period in the ITT population was analyzed using negative binomial or Poisson (used when there was no evidence of overdispersion) regression with robust, empirical, sandwich SE estimates and small-sample adjustments.⁴⁰ Each model included the following fixed effects: study arm (intervention or wait-list control), state, urban/rural classification, the baseline opioid-related overdose death rate, and the natural log of the baseline rate of the outcome. The rate of locations providing drug take-back services per 100,000 adults during the baseline period was not available for all communities, therefore, analyses of this outcome did not adjust for the baseline rate. The natural log of the relevant community population size during the comparison period was also included in the model as an offset term. For each outcome, the adjusted rate (and 95%) was reported for each study arm, as well as the adjusted relative rate (and 95% CI and p-value) comparing intervention to wait-list control communities. To minimize the number of comparisons reported in this manuscript, no formal comparisons between the intervention and wait-list control communities were performed for the four criteria defining the number of adults with new incident high-risk opioid prescribing.

Pre-specified sub-group analyses (state, urban/rural classification, sex, and age) were also performed for each outcome by including an interaction term between study arm and a stratification variable representing the sub-group of interest as a fixed effect in the model. If needed, a covariance structure for residual effects was also specified in the model. While some of the data required for sub-group analyses met suppression rules within each state, data use agreements allowed suppressed data in KY and NY to be used for analytic purposes. However, suppressed data in MA and OH were not available and so any strata with suppressed data were excluded from the analyses. To control the false discovery rate, the Benjamini-Hochberg procedure⁴¹ was applied to all p-values corresponding to all interaction tests and pairwise tests that were performed in sub-group analyses. All analyses assumed a 2-sided significance level of 0.05 and were conducted using SAS Version 9.4.

RESULTS

Thirty-four and 33 communities were randomized to intervention and wait-list control, respectively, and included in the ITT population (Figure 1). One selected community assigned to the intervention withdrew after randomization, resulting in 66 of 67 randomized communities included in the per-protocol population. Baseline characteristics of intervention (n=4,439,170 adults) and wait-list control (n=3,772,336 adults) communities are summarized in Table 1.^42,43 Communities were randomized evenly to study arms by state and urban/rural classification. The mean baseline rate of opioid-related overdose deaths per 100,000 adults was similar among intervention and wait-list control communities (38.2 [SD=22.8] and 37.1 [SD=20.3], respectively). The population of adults was similar in age and sex distribution among the 2 study arms. Overall, 51.8% of residents were female, 30.6% were aged 18–34 years, 30.9% were aged 35–54 years, and 38.5% were aged 55 years or older.

Figure 1.

CONSORT flow diagram: HEALing Communities in KY, MA, NY, and OH.

KY, Kentucky; MA, Massachusetts; NY, New York; OH, Ohio.

Table 1.

Baseline Demographic Characteristics of 67 Communities Participating in the HEALing Communities Study, 2019

Characteristic	Intervention		Overall
	Intervention	Wait-list control
Number of randomized communities, n	34	33	67
State, n (%)
KY	8 (23.5%)	8 (24.2%)	16 (23.9%)
MA	8 (23.5%)	8 (24.2%)	16 (23.9%)
NY	8 (23.5%)	8 (24.2%)	16 (23.9%)
OH	10 (29.4%)	9 (27.3%)	19 (28.4%)
Urban/rural classification, n (%)
Urban	19 (55.9%)	19 (57.6%)	38 (56.7%)
Rural	15 (44.1%)	14 (42.4%)	29 (43.3%)
Population aged ≥18 yearsa
Total	4,439,170	3,772,336	8,211,506
Mean (SD)	130,563.8 (200,088.0)	114,313.2 (201,417.3)	122,559.8 (199,385.0)
Age, years,an (%)
18–34	1,334,880 (30.1%)	1,178,210 (31.2%)	2,513,090 (30.6%)
35–54	1,353,341 (30.5%)	1,180,392 (31.3%)	2,533,733 (30.9%)
≥55	1,750,949 (39.4%)	1,413,734 (37.5%)	3,164,683 (38.5%)
Sex,an (%)
Male	2,133,827 (48.1%)	1,825,776 (48.4%)	3,959,603 (48.2%)
Female	2,305,343 (51.9%)	1,946,560 (51.6%)	4,251,903 (51.8%)
Rate of opioid overdose deaths, mean (SD)b	38.2 (22.8)	37.1 (20.3)	37.7 (21.4)
Rate of individuals with new incident high-risk opioid prescribing, mean (SD)b	1,419.7 (617.9)	1,417.6 (451.0)	1,418.7 (538.0)
Rate of individuals with new acute opioid prescriptions limited to a 7-day supply, mean (SD)c	87,853.3 (3,161.2)	88,608.1 (2,617.8)	88,225.1 (2,909.3)
Rate of individuals who received opioid prescriptions from multiple prescribers or pharmacies, mean (SD)b	495.6 (154.9)	505.3 (163.3)	500.4 (157.9)
Rate of individuals with an incident opioid prescribing episode duration of more than 30 days, mean (SD)b	904.0 (433.4)	902.6 (319.4)	903.3 (378.7)
Rate of individuals with an incident opioid prescribing episode with extended-release or long-acting opioid formulation, mean (SD)b	103.4 (34.4)	102.8 (32.9)	103.1 (33.4)
Rate of individuals with an incident high-dose opioid prescribing, mean (SD)b	216.5 (98.3)	230.0 (97.5)	223.2 (97.4)
Rate of individuals with an incident overlapping opioid and benzodiazepine prescriptions for at least 30 days over 3 calendar months, mean (SD)b	392.8 (180.7)	380.3 (158.9)	386.6 (169.2)

Note: Percentages may not add up to 100 owing to rounding.

^aFor communities that represent counties (n=48 of 67), population estimates are from 2020 Bridged-Race Population Estimates retrieved through https://www.cdc.gov/nchs/nvss/bridged_race.htm on July 2, 2024. For communities that represent units smaller than counties (n=19 of 67), population estimates are from 2017–2021 American Community Survey 5-year estimates retrieved through https://data.census.gov/cedsci on July 2, 2024.

^bRate per 100,000 individuals aged ≥18 years calculated as the observed number of events as measured from January 2019 to December 2019 divided by the observed community population size of individuals aged ≥18 years measured multiplied by 100,000. For communities that represent counties (48 of 67), population size was drawn from 2020 Bridged-Race Population Estimates.⁴² For communities that represent units smaller than counties (19 of 67), population size was drawn from 2017–2021 American Community Survey 5-year averages.⁴³

^cRate per 100,000 individuals aged ≥18 years with 1 or more new opioid analgesic episodes calculated as the observed number of individuals aged ≥18 years with new acute opioid prescriptions limited to a 7-day supply measured from January 2019 to December 2019 within a community divided by the observed community population size of individuals aged ≥18 years with 1 or more new opioid analgesic episodes measured from January 2019 to December 2019 multiplied by 100,000. In Wave 1 communities, the total number of individuals aged ≥18 years with 1 or more new opioid analgesic episodes measured from January 2019 to December 2019 was 632,395 with a mean (SD) of 18,614.6 (29,218.1). In Wave 2 communities, the total number of individuals aged ≥18 years with 1 or more new opioid analgesic episodes measured from January 2019 to December 2019 was 568,200 with a mean (SD) of 17,218.2 (31,751.0). In all communities, the total number of individuals aged ≥18 years with 1 or more new opioid analgesic episodes measured from January 2019 to December 2019 was 1,201,095 with a mean (SD) of 17,926.8 (29,762.85).

KY, Kentucky; MA, Massachusetts; NY, New York; OH, Ohio.

Prescription opioid safety strategies implemented overall, by state, and by urban/rural classification are summarized in Table 2. During the 22-month implementation timeframe, intervention communities implemented 105 prescription opioid safety strategies. The majority of the strategies were implemented in the healthcare sector (86 of 105, 81.9%). Communities were required to implement at least 1 ORCCA strategy from across 3 submenus, with communities implementing 37 safer opioid prescribing strategies for acute pain (35.2%), 15 safer opioid prescribing strategies for chronic pain (14.3%), and 18 strategies for safer opioid dispensing (17.1%). Thirty-five (33.3%) of the implemented strategies were the optional strategy of prescription drug drop-box/mail-back programs. Only 34 of the 105 strategies (32.4%) were implemented prior to the start of the comparison period. Notably, some venues had substantially less uptake in prescription opioid safety strategies such as ambulatory surgery clinics and healthcare inpatient services, suggesting missed opportunities for impacting care (Table 3).

Table 2.

Implemented Prescription Opioid Safety Strategies by State and Urban/Rural Classification for 33 Intervention Communities Participating in the HEALing Communities Study

Safety strategy, sector, venue	State				Urban/rural classification			Total
	KY	MA	NY	OH	Rural		Urban
Number of communities	8	8	8	9	15		18	33
Number of prescription opioid safety Strategies implemented	38	9	36	22	47		58	105
Strategy, n (%)a
Safer opioid prescribing for acute pain across varied healthcare settings	15 (39.5%)	1 (11.1%)	19 (52.8%)	2 (9.1%)	18 (38.3%)		19 (32.8%)	37 (35.2%)
Safer opioid prescribing for chronic pain	7 (18.4%)	2 (22.2%)	2 (5.6%)	4 (18.2%)	7 (14.9%)		8 (13.8%)	15 (14.3%)
Safer opioid dispensing	8 (21.1%)	0 (0%)	1 (2.8%)	9 (40.9%)	10 (21.3%)		8 (13.8%)	18 (17.1%)
Prescription drug drop-box/mail back programs	8 (21.1%)	6 (66.7%)	14 (38.9%)	7 (31.8%)	12 (25.5%)		23 (39.7%)	35 (33.3%)
Sector, n (%)b
Health care	38 (100.0%)	5 (55.6%)	26 (72.2%)	17 (77.3%)	40 (85.1%)		46 (79.3%)	86 (81.9%)
Behavioral health	0 (0%)	3 (33.3%)	7 (19.4%)	4 (18.2%)	5 (10.6%)		9 (15.5%)	14 (13.3%)
Criminal justice	0 (0%)	1 (11.1%)	3 (8.3%)	1 (4.5%)	2 (4.3%)		3 (5.2%)	5 (4.8%)
Venue,cn (%)d
Criminal justice, community supervision	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	1 (2.1%)		0 (0%)	1 (1.0%)
Criminal justice, other	0 (0%)	0 (0%)	1 (2.8%)	1 (4.5%)	1 (2.1%)		1 (1.7%)	2 (1.9%)
Health care, emergency department	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	1 (2.1%)		0 (0%)	1 (1.0%)
Health care, health department	0 (0%)	0 (0%)	2 (5.6%)	1 (4.5%)	1 (2.1%)		2 (3.4%)	3 (2.9%)
Health care, pharmacy	16 (42.1%)	1 (11.1%)	9 (25.0%)	11 (50.0%)	20 (42.6%)		17 (29.3%)	37 (35.2%)
Health care, outpatient	14 (36.8%)	3 (33.3%)	4 (11.1%)	0 (0%)	9 (19.1%)		12 (20.7%)	21 (20.0%)
Health care, ambulatory surgery	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	0 (0%)		1 (1.7%)	1 (1.0%)
Health care, dental clinics	8 (21.1%)	0 (0%)	2 (5.6%)	0 (0%)	4 (8.5%)		6 (10.3%)	10 (9.5%)
Health care, other	0 (0%)	0 (0%)	7 (19.4%)	5 (22.7%)	5 (10.6%)		7 (12.1%)	12 (11.4%)
First responder stations	0 (0%)	2 (22.2%)	1 (2.8%)	0 (0%)	0 (0%)		3 (5.2%)	3 (2.9%)
Addiction treatment and recovery facilities, medical	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	0 (0%)		1 (1.7%)	1 (1.0%)
Addiction treatment and recovery facilities, nonmedical	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	0 (0%)		1 (1.7%)	1 (1.0%)
Mental/behavioral health treatment facilities, medical	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	1 (2.1%)		0 (0%)	1 (1.0%)
Community-based social service agencies, other	0 (0%)	1 (11.1%)	2 (5.6%)	4 (18.2%)	3 (6.4%)		4 (6.9%)	7 (6.7%)
Other	0 (0%)	2 (22.2%)	2 (5.6%)	0 (0%)	1 (2.1%)		3 (5.2%)	4 (3.8%)

^aThe number of prescription opioid safety strategies implemented by type across communities within each state, within rural or urban communities, and across all communities. The percentage is calculated using the ratio between the number of prescription opioid safety strategies implemented by type divided by the total number of prescription opioid safety strategies implemented across communities within each state, within rural or urban communities, and across all communities and multiplying this ratio by 100.

^bThe number of prescription opioid safety strategy implemented by sector across communities within each state, within rural or urban communities, and across all communities. The percentage is calculated using the ratio between the number of prescription opioid safety strategy implemented by sector divided by the total number of prescription opioid safety strategies implemented across communities within each state, within rural or urban communities, and across all communities and multiplying this ratio by 100.

^cNo communities implemented strategies in the following venues: criminal justice, jails and syringe service programs; health care, inpatient, mental/behavioral health treatment facilities, nonmedical; community-based social service agencies, homeless shelters; community-based social service agencies, halfway houses; hotline responding to service requests.

^dThe number of prescription opioid safety strategies implemented by venue across communities within each state, within rural or urban communities, and across all communities. The percentage is calculated using the ratio between the number of prescription opioid safety strategies implemented by venue divided by the total number of prescription opioid safety strategies implemented across communities within each state, within rural or urban communities, and across all communities and multiplying this ratio by 100.

KY, Kentucky; MA, Massachusetts; NY, New York; OH, Ohio.

Table 3.

Implemented Strategies From ORCCAT Menu 3: Prescription Opioid Safety Strategies for n=33^a Wave 1 Communities Participating in the HEALing Communities Study^b

Venue	Prescription opioid safety strategy, n (% total)				Total
	Safer opioid prescribing for acute pain across varied healthcare settings	Safer opioid prescribing for chronic pain	Safer opioid dispensing	Prescription drug drop-box/mail-back programs
Total	37 (35.2%)	15 (14.3%)	18 (17.1%)	35 (33.3%)	105
Criminal justice, jails	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Criminal justice, community supervision	0 (0%)	0 (0%)	0 (0%)	1 (100.0%)	1
Criminal justice, other	0 (0%)	0 (0%)	0 (0%)	2 (100.0%)	2
Syringe service programs	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Health care, emergency department	1 (100.0%)	0 (0%)	0 (0%)	0 (0%)	1
Health care, health department	0 (0%)	1 (33.3%)	0 (0%)	2 (66.7%)	3
Health care, pharmacy	6 (16.2%)	3 (8.1%)	16 (43.2%)	12 (32.4%)	37
Health care, inpatient services	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Health care, outpatient clinics	9 (42.9%)	10 (47.6%)	0 (0%)	2 (9.5%)	21
Health care, ambulatory surgery	1 (100.0%)	0 (0%)	0 (0%)	0 (0%)	1
Health care, dental clinics	10 (100.0%)	0 (0%)	0 (0%)	0 (0%)	10
Health care, other	7 (58.3%)	1 (8.3%)	2 (16.7%)	2 (16.7%)	12
First responder stations	0 (0%)	0 (0%)	0 (0%)	3 (100.0%)	3
Addiction treatment and recovery facilities, medical	0 (0%)	0 (0%)	0 (0%)	1 (100.0%)	1
Addiction treatment and recovery facilities, nonmedical	0 (0%)	0 (0%)	0 (0%)	1 (100.0%)	1
Mental/behavioral health treatment facilities, medical	1 (100.0%)	0 (0%)	0 (0%)	0 (0%)	1
Mental/behavioral health treatment facilities, nonmedical	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Community-based social services agencies, homeless shelters	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Community-based social services agencies, half-way houses	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Community-based social service agencies, other	1 (14.3%)	0 (0%)	0 (0%)	6 (85.7%)	7
Hotline (phone or internet) responding to service requests	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
Other	1 (25.0%)	0 (0%)	0 (0%)	3 (75.0%)	4

Note: Presented is a summary of unique strategy–sector–venue triad combinations that were implemented at any time during Wave 1. The most recent entry for each triad is chosen. Duplicate triad entries are rolled to the triad level.

^an=1 community randomized to Wave 1 withdrew prior to strategy selection.

^bResults based on data pulled on March 27, 2023.

ORCCAT, Opioid-overdose Reduction Continuum of Care Approach Tracker.

Among intervention communities, the mean rate of adults with new incident high-risk opioid prescribing per 100,000 adults was 1,419.7 (SD=617.9) and 1,146.5 (SD=471.9) during the baseline period (Table 1^42,43) and comparison period (Table 4^42,43), respectively. Among wait-list control communities, the mean rate of adults with new incident high-risk opioid prescribing per 100,000 adults was 1,417.6 (SD=451.0) and 1,176.7 (SD=362.9) during the baseline and comparison periods, respectively. Descriptive statistics of this rate during the comparison period are also provided in Table 4^42,43 by study arm within state, urban/rural classification, sex, and age. Breaking this rate down into each of its 4 components also showed a similar distribution among study arms during the baseline period (Table 1^42,43) and comparison period (Table 4^42,43). In adjusted analyses, the rate of adults with new incident high-risk opioid prescribing per 100,000 adults during the comparison period was 1,094.48 (95% CI=1,063.15; 1,126.74) and 1,121.90 (95% CI=1,079.62; 1,165.84) in intervention and wait-list control communities, respectively, resulting in an ARR comparing intervention with wait-list control communities of 0.98 (95% CI=0.93, 1.02) (p=0.296) (Table 5).

Table 4.

Descriptive Statistics of Outcomes Measured During the Comparison Period in Intervention and Wait-List Control Communities in the Intention-To-Treat Population by State and Urban/Rural Classification

Main and other outcomes	Group	Intervention				Wait-list control				Relative raw rate
		n (%)a	Mean (SD) number of events across communitiesb	Mean (SD) population size across communitiesb	Mean (SD) raw rate of events across communitiesc	n (%)a	Mean (SD) number of events across communitiesc	Mean (SD) population size across communitiesc	Mean (SD) raw rate of events across communitiesc
Individuals with new incident high-risk opioid prescribing	Overall	34 (100.0%)	1,289.6 (1,762.7)	130,563.8 (200,088.0)	1,146.5 (471.9)	33 (100.0%)	1,209.7 (2,006.9)	114,313.2 (201,417.3)	1,176.7 (362.9)	0.97
	State
	KY	8 (100.0%)	902.0 (602.9)	77,230.1 (80,938.9)	1,541.0 (721.6)	8 (100.0%)	1,408.8 (2,791.6)	101,970.5 (202,045.3)	1,505.9 (451.9)	1.02
	MA	8 (100.0%)	407.8 (186.2)	44,914.3 (26,559.3)	967.1 (299.1)	8 (100.0%)	401.3 (297.8)	44,568.1 (33,628.8)	961.7 (267.7)	1.01
	NY	8 (100.0%)	1,305.4 (1,408.8)	137,687.1 (140,779.9)	953.6 (169.9)	8 (100.0%)	1,137.0 (832.3)	122,008.6 (106,012.1)	1,021.2 (224.1)	0.93
	OH	10 (100.0%)	2,292.7 (2,756.2)	236,051.8 (322,922.0)	1,128.9 (345.1)	9 (100.0%)	1,816.0 (2,738.0)	180,439.8 (325,173.2)	1,213.4 (247.0)	0.93
	Urban/rural classification
	Urban	19 (100.0%)	1,866.6 (2,199.7)	199,650.2 (248,385.2)	1,003.6 (275.8)	19 (100.0%)	1,741.8 (2,526.0)	170,666.5 (252,614.9)	1,099.0 (307.0)	0.91
	Rural	15 (100.0%)	558.9 (280.9)	43,054.5 (19,075.4)	1,327.6 (603.6)	14 (100.0%)	487.6 (323.7)	37,833.8 (23,733.0)	1,282.1 (415.7)	1.04
	Sex
	Male	33 (97.1%)	526.2 (702.0)	63,275.4 (95,785.5)	990.9 (440.8)	32 (97.0%)	495.9 (792.1)	56,475.3 (98,174.6)	1,009.0 (350.3)	0.98
	Female	34 (100.0%)	772.2 (1,071.1)	67,804.2 (105,743.9)	1,308.2 (510.3)	33 (100.0%)	722.2 (1,226.4)	58,986.7 (104,579.4)	1,333.5 (399.6)	0.98
	Missing	16 (47.1%)	0.8 (1.3)	NA	NA	16 (48.5%)	0.5 (1.3)	NA	NA	NA
	Age, years
	18–34	32 (94.1%)	47.8 (67.3)	41,373.4 (61,726.0)	123.9 (53.7)	29 (87.9%)	60.0 (135.7)	39,706.3 (73,687.4)	143.4 (65.3)	0.86
	35–54	32 (94.1%)	296.6 (361.0)	41,864.2 (61,633.4)	862.8 (461.3)	29 (87.9%)	315.3 (532.3)	39,544.7 (68,768.3)	889.4 (305.2)	0.97
	≥55	33 (97.1%)	985.5 (1,365.8)	52,648.5 (80,861.2)	2,093.1 (776.6)	32 (97.0%)	896.4 (1,408.3)	43,942.3 (68,263.4)	2,120.9 (598.8)	0.99
Individuals with new acute opioid prescriptions limited to a 7-day supply	Overall	34 (100.0%)	15,114.1 (24,143.8)	16,859.1 (26,558.7)	88,406.4 (3,999.5)	33 (100.0%)	14,036.6 (25,972.8)	15,608.2 (28,760.4)	89,371.7 (2,875.1)	0.99
Individuals who received opioid prescriptions from multiple prescribers or pharmacies	Overall	34 (100.0%)	531.0 (803.6)	130,563.8 (200,088.0)	421.3 (124.7)	33 (100.0%)	476.6 (818.6)	114,313.2 (201,417.3)	432.2 (139.0)	0.97
locations providing drug take-back services	Overall	34 (100.0%)	13.5 (19.4)	130,563.8 (200,088.0)	12.6 (7.8)	33 (100.0%)	8.5 (9.5)	114,313.2 (201,417.3)	11.2 (6.6)	1.13
Individuals with an incident opioid prescribing episode duration of more than 30 days	Overall	34 (100.0%)	854.4 (1,228.2)	130,563.8 (200,088.0)	736.5 (309.4)	33 (100.0%)	821.5 (1,452.3)	114,313.2 (201,417.3)	766.6 (259.4)	0.96
Individuals with an incident opioid prescribing episode with extended-release or long-acting opioid formulation	Overall	34 (100.0%)	87.1 (140.6)	130,563.8 (200,088.0)	72.0 (30.2)	33 (100.0%)	76.7 (107.6)	114,313.2 (201,417.3)	78.9 (28.7)	0.91
Individuals with an incident high-dose opioid prescription	Overall	34 (100.0%)	192.4 (246.3)	130,563.8 (200,088.0)	177.9 (85.5)	33 (100.0%)	177.7 (264.5)	114,313.2 (201,417.3)	177.8 (70.1)	1.00
Individuals with an incident overlapping opioid and benzodiazepine prescription for at least 30 days over 3 calendar months	Overall	34 (100.0%)	311.5 (388.5)	130,563.8 (200,088.0)	301.4 (146.4)	33 (100.0%)	268.1 (402.9)	114,313.2 (201,417.3)	293.5 (115.6)	1.03

^aNumber of communities (%) by group and study arm that are included in the calculation of the descriptive statistics that summarize events, population, and raw rate. Communities that have suppressed event and/or population data or do not report event and/or population data for missing levels of a stratification variable are not included in the calculation of the descriptive statistics that summarize events, population, and raw rate.

^bMean (SD) of the community-level events/population by group and study arm. For all outcomes except individuals with new acute opioid prescriptions limited to a 7-day supply, population size was drawn from 2020 Bridged-Race Population Estimates⁴² for communities that represent counties (48 of 67) or drawn from 2017–2021 American Community Survey 5-year averages⁴³ for communities that represent units smaller than counties (19 of 67). For individuals with new acute opioid prescriptions limited to a 7-day supply, population size was the number of adults with 1 or more new opioid analgesic episodes in the community during the comparison period.

^cMean (SD) of the community-level raw rates by group and study arm. The raw rate of individuals with new acute opioid prescriptions limited to a 7-day supply is calculated per 100,000 adults with 1 or more new opioid analgesic episodes. All other raw rates are calculated per 100,000 adults.

KY, Kentucky; MA, Massachusetts; NY, New York; OH, Ohio.

Table 5.

Adjusted Rate and Relative Rates of Each Outcome Measured During the Comparison Period Between Intervention and Wait-List Control Communities in the Intention-To-Treat Population

Main and other outcomes	Intervention	Wait-list control	Adjusted relative rate (95% CI)a	p-value
	Adjusted rate (95% CI)	Adjusted rate (95% CI)
Number of individuals with new incident high-risk opioid prescribingb,c	1,094.48 (1,063.15; 1,126.74)	1,121.90 (1,079.62; 1,165.84)	0.98 (0.93, 1.02)	0.296
Number of individuals with new acute opioid prescriptions limited to a 7-day supplyd	8,8822.7 (8,8425.2; 8,9221.9)	8,8922.0 (8,8530.0; 8,9315.8)	1.00 (0.99, 1.00)	0.676
Number of individuals who received opioid prescriptions from multiple prescribers or pharmaciesb,e	411.93 (392.99, 431.78)	410.08 (390.09, 431.08)	1.00 (0.94, 1.07)	0.893
Number of locations providing drug-take back servicesf,g	12.38 (10.46, 14.65)	9.92 (8.19, 12.02)	1.25 (0.96, 1.62)	0.094

^aAdjusted relative rate is the ratio of the adjusted rate in intervention communities to the adjusted rate in wait-list control communities.

^bResults obtained from a negative binomial model adjusting for study arm, state (KY, MA, NY, and OH), urban/rural classification, baseline opioid overdose death rate, and natural log of the baseline rate of the outcome. The natural log of the community population size in the most recent year available is used as an offset. Adjusted rates are expressed per 100,000 adults.

^cEstimated dispersion parameter, k=0.005; 95% CI=0.003, 0.008.

^dResults obtained from a Poisson model adjusting for study arm, state (KY, MA, NY, and OH), urban/rural classification, baseline opioid overdose death rate, and natural log of the baseline rate of the outcome. The natural log of the community size of individuals aged ≥18 years with 1 or more new opioid analgesic episodes during the comparison period is used as an offset. Adjusted rates are expressed per 100,000 adults with 1 or more new opioid analgesic episodes.

^eEstimated dispersion parameter, k=0.011; 95% CI=0.006, 0.017.

^fResults obtained from a negative binomial model adjusting for study arm, state (KY, MA, NY, and OH), urban/rural classification, and the baseline opioid overdose death rate. The natural log of the community population size in the most recent year available is used as an offset. Adjusted rates are expressed per 100,000 adults.

^gEstimated dispersion parameter, k=0.075; 95% CI=0.034, 0.163.

KY, Kentucky; MA, Massachusetts; NY, New York; OH, Ohio.

Among intervention communities, the mean rate of adults with new acute opioid prescriptions limited to a 7-day supply (per 100,000 adults with 1 or more new opioid analgesic episodes) was 87,853.3 (SD=3,161.2) and 88,406.4 (SD=3,999.5) during the baseline period (Table 1^42,43) and comparison period (Table 4^42,43), respectively. Among wait-list control communities, the mean rate of adults with new acute opioid prescriptions limited to a 7-day supply (per 100,000 adults with 1 or more new opioid analgesic episodes) was 88,608.1 (SD=2,617.8) and 89,371.7 (SD=2,875.1) during the baseline and comparison periods, respectively. In adjusted analyses, the rate of adults with new acute opioid prescriptions limited to a 7-day supply (per 100,000 adults with 1 or more new opioid analgesic episodes) during the comparison period was 88,822.7 (95% CI=88,425.2; 89,221.9) and 88,922.0 (95% CI=88,530.0; 89,315.8) in intervention and wait-list control communities, respectively, resulting in an ARR of 1.00 (95% CI=0.99, 1.00) (p=0.676) (Table 5).

Among intervention communities, the mean rate of adults who received opioid prescriptions from multiple prescribers or pharmacies per 100,000 adults was 495.6 (SD=154.9) and 421.3 (SD=124.7) during the baseline period (Table 1^42,43) and comparison period (Table 4^42,43), respectively. Among wait-list control communities, the mean rate of adults who received opioid prescriptions from multiple prescribers or pharmacies per 100,000 adults was 505.3 (SD=163.3) and 432.2 (SD=139.0) during the baseline and comparison periods, respectively. In adjusted analyses, the rate of adults who received opioid prescriptions from multiple prescribers or pharmacies per 100,000 adults during the comparison period was 411.93 (95% CI=392.99, 431.78) and 410.08 (95% CI=390.09, 431.08) in intervention and wait-list control communities, respectively, resulting in an ARR of 1.00 (95% CI=0.94, 1.07) (p=0.893) (Table 5).

The mean rate of locations providing drug take-back services per 100,000 adults during the comparison period was 12.6 (SD=7.8) and 11.2 (SD=6.6) in intervention and wait-list control communities, respectively. In adjusted analyses, the rate of locations providing drug take-back services per 100,000 adults during the comparison period was 12.38 (95% CI=10.46, 14.65) and 9.92 (95% CI=8.19, 12.02) in intervention and wait-list control communities, respectively, resulting in an ARR of 1.25 (95% CI=0.96, 1.62) (p=0.094) (Table 5).

Across all outcomes, adjusted analyses of the per-protocol population (Figure 1) were similar to those conducted in the ITT population (results not reported). There was also no evidence to suggest that the effect of the CTH intervention on each outcome differed by state, urban/rural classification, sex, or age (results not reported).

DISCUSSION

Implementation of the CTH intervention did not result in statistically significant differences in the rate of safer opioid prescribing, dispensing, and disposal between intervention and wait-list control communities. Factors external to the HCS and limitations of the HCS itself probably contributed to these findings.

Efforts to improve prescription opioid safety contributed to significant declines in opioid prescribing before the CTH implementation began. Between 2006 and 2020, the national opioid dispensing rate fell from 72.4 to 43.2 prescriptions per 100 persons,44, 45 a trend attributable to a variety of initiatives implemented to address the opioid crisis. For example, since 2014, there has been a steady increase in the use of PDMPs among prescribers, from about 61 million times to 910 million times in 2020.⁴⁶ Although evidence of PDMP efficacy for preventing opioid-related harms is mixed, mandatory PDMP use (in place in all 4 HCS states prior to the CTH intervention) has been shown to reduce opioid prescribing and dispensing.^47,48 The 2016 CDC Guideline for Prescribing Opioids for Chronic Pain was also associated with a significant acceleration in the decline in the overall rate of opioid prescribing, rate of high-dosage opioid prescriptions, and percentage of patients with overlapping benzodiazepine and opioid prescriptions.^49,50 By 2020, 36 states, including all 4 HCS states, had enacted laws limiting the duration or dosage of opioid prescriptions.⁴⁸ Insurers, pharmacy benefit managers, and pharmacies also adopted similar policies in an attempt to limit inappropriate opioid dispensing. Taken together, these changes contributed to substantial reductions in opioid prescribing and improved prescription opioid safety prior to the initiation of the CTH intervention.

The goals, structure, and timeline of the HCS may have also contributed to the lack of significant difference in the rate of prescription opioid safety measures between study arms. HCS community coalitions selected and implemented EBP strategies with a primary goal of reducing opioid-related overdose deaths. The ORCCA required each community to implement just 1 strategy for the prescription opioid safety EBP (compared with 3 for MOUD),²⁸ which represented primary prevention in the continuum of care. The range of prescription opioid safety strategies subsequently implemented within each state (from 9 to 38) may reflect varying priority of this EBP among communities. Across intervention communities, prescription opioid safety represented just 24.1% of strategies chosen (compared with 40.8% for overdose education and naloxone distribution and 35.1% for MOUD).⁵¹

The timing of the CTH intervention and the comparison period also posed challenges. The intervention occurred from January 2020 through June 2022, coinciding with the peak years of the coronavirus disease 2019 (COVID-19) pandemic and response. Implementation efforts may have been hampered by limitations on in-person interactions in healthcare settings. Notably, only 32% of prescription opioid safety strategies were implemented prior to the start of the comparison period. More importantly, healthcare professionals were faced with unprecedented demands on their time and attention and likely did not have the capacity to engage fully with HCS. With the comparison period occurring during the final year of the intervention, any behavior change that was spurred by prescription opioid safety efforts in intervention communities may not have been in place long enough to be captured in the analysis.

Strategy implementation took different forms across intervention communities. Examples of frequently implemented community efforts to improve opioid prescribing and dispensing included continuing education and consulting programs, staff training and guideline implementation in healthcare facilities, and academic detailing for prescribers and pharmacists. Although these strategies are all commonly used to influence practitioners, changing their behavior is difficult and may require more complex and sustained structural interventions.⁵¹ Safer opioid disposal was an optional strategy in the ORCCA. Community selections included funding the installation of disposal drop boxes in pharmacies, distributing home-disposal packets at community events, and promoting Drug Enforcement Administration Drug Take-Back Days. Although these all encourage safer opioid disposal, the strategies vary in their impact on the HCS measure of prescription drug drop-box/mail-back programs. The study measured the number of drug take-back boxes but not the number of drugs collected. This might be a critical variable in assessing the impact of this strategy.

Furthermore, the authors did not assess how well matched a strategy was to the sector or venue targeted for implementation. For example, strategies that focus on acute pain opioid prescribing might not be relevant in some behavioral health settings where there are few prescribers. However, to mitigate this potential, coalitions engaged community sites and venues to examine which prescription opioid safety strategy was most appropriate given their perceived knowledge gaps and needs of the community setting. On the basis of Table 3, this seems to have been mostly successful. For example, pharmacies and dental clinics appropriately adopted safer opioid-dispensing strategies relevant to their missions. However, future research might specifically evaluate landscape data-informing strategy selection and its association with specific prescription opioid safety goals to assess the effectiveness of particular strategies and the success of their active implementation. Furthermore, studies are needed to evaluate prescription opioid safety strategies and their association with changes in services delivery and patient outcomes.

Although there are numerous possible reasons for the lack of differences in the rate of prescription opioid safety measures between intervention and wait-list control communities, it does not diminish the importance of HCS communities’ success in selecting and implementing 105 strategies that promote prescription opioid safety. Although opioids are indicated for both acute and chronic pain conditions, avoiding or delaying the initiation of opioid prescribing when nonopioid alternatives might be effective makes good clinical sense.^52,53 There is a clear recognition in the Centers for Disease Control and Prevention guidelines⁵⁴ that the initiation of opioids needs to be done in a careful, individualized manner. However, overzealous refusal to prescribe opioids or discontinuation of opioids to those chronically prescribed opioids may lead to harmful consequences such as severe opioid withdrawal with initiation of illicit opioid use and suicidal ideation/suicide.55, 56, 57, 58, 59 Thus, adoption of safer opioid prescribing is best accomplished by focusing on primary and secondary prevention efforts as was done for the HCS.

Limitations

An overarching limitation of this paper was that the sample size selected for the HCS trial was determined to detect a 40% reduction in the rate of opioid-related overdose deaths between the intervention and wait-list control communities rather than any of the rates presented in this manuscript. Consequently, the sample size may not have been sufficient to detect differences in these rates between intervention and wait-list control communities. Another limitation of this analysis was the reliance on administrative data sources (state PDMPs and state and federal disposal databases) to assess select prescription opioid safety measures. These data sources limit the ability to evaluate important effect modifiers such as patient race or prescriber specialty. The prescribing and dispensing measures, which account only for new incident high-risk opioid prescriptions, fail to capture the complexity of appropriate opioid use. In practice, prescription opioid safety includes many additional elements such as naloxone coprescribing, risk mitigation and monitoring, patient-centered opioid tapering, and disposal education that would not be evident in this analysis. In addition, extensive efforts to reduce high-risk prescribing were underway before the CTH intervention began and may have reduced the likelihood of detecting differences. Finally, the outcome measure was already on the decline, and the prescription opioid safety strategies chosen by the communities may not have been adequate to produce differential responses in the rates between study arms.

CONCLUSIONS

The strategies implemented by the HCS to promote safer opioid prescribing, dispensing, and disposal in intervention communities did not result in overall improvement in the rate of prescription opioid safety measures compared with that in wait-list control communities. Several external factors such as increased prescriber participation in PDMPs, prescriber adherence to revised Centers for Disease Control and Prevention guidelines, and state policies limiting dosing and duration of opioids all contributed to changes in prescribing patterns prior to the initiation of the HCS. Furthermore, limitations inherent in the HCS and the use of administrative data sources precluded detection of safety approaches not captured by the available data. Nonetheless, the finding that prescription opioid safety improved for all communities enrolled in the HCS bodes well for the overarching goal of reducing opioid-related overdose deaths.