The Association of State Opioid Misuse Prevention Policies With Patient‐ and Provider‐Related Outcomes: A Scoping Review

Abstract

Policy Points.

• This scoping review reveals a growing literature on the effects of certain state opioid misuse prevention policies, but persistent gaps in evidence on other prevalent state policies remain.

• Policymakers interested in reducing the volume and dosage of opioids prescribed and dispensed can consider adopting robust prescription drug monitoring programs with mandatory access provisions and drug supply management policies, such as prior authorization policies for high‐risk prescription opioids.

• Further research should concentrate on potential unintended consequences of opioid misuse prevention policies, differential policy effects across populations, interventions that have not received sufficient evaluation (eg, Good Samaritan laws, naloxone access laws), and patient‐related outcomes.

Context

In the midst of an opioid crisis in the United States, an influx of state opioid misuse prevention policies has provided new opportunities to generate evidence of policy effectiveness that can inform policy decisions. We conducted a scoping review to synthesize the available evidence on the effectiveness of US state interventions to improve patient and provider outcomes related to opioid misuse and addiction.

Methods

We searched six online databases to identify evaluations of state opioid policies. Eligible studies examined legislative and administrative policy interventions that evaluated (a) prescribing and dispensing, (b) patient behavior, or (c) patient health.

Findings

Seventy‐one articles met our inclusion criteria, including 41 studies published between 2016 and 2018. These articles evaluated nine types of state policies targeting opioid misuse. While prescription drug monitoring programs (PDMPs) have received considerable attention in the literature, far fewer studies addressed other types of state policy. Overall, evidence quality is very low for the majority of policies due to a small number of evaluations. Of interventions that have been the subject of considerable research, promising means of reducing the volume and dosages of opioids prescribed and dispensed include drug supply management policies and robust PDMPs. Due to low study number and quality, evidence is insufficient to draw conclusions regarding interventions targeting patient behavior and health outcomes, including naloxone access laws and Good Samaritan laws.

Conclusions

Recent research has improved the evidence base on several state interventions targeting opioid misuse. Specifically, moderate evidence suggests that drug supply management policies and robust PDMPs reduce opioid prescribing. Despite the increase in rigorous evaluations, evidence remains limited for the majority of policies, particularly those targeting patient health–related outcomes.

The united states is in the midst of an opioid overdose crisis. In 2017 there were 70,237 drug overdose deaths in the United States, 47,600 of which were attributable to opioids.1, 2 Prescription opioid medications caused most fatal opioid overdose deaths in the first decade of the 2000s.3 Although today most opioid overdoses involve heroin and illicit fentanyl, many who experience opioid harms were first exposed to opioids via a prescription.1, 4

States have implemented a panoply of preventive measures in recent years to address health consequences associated with opioid misuse and addiction. These state policies target prevention at different levels, from primary prevention of initial exposure to opioids, to secondary prevention to avoid high‐risk opioid exposure, to tertiary prevention to treat individuals with opioid use disorder.5, 6 Table 1 summarizes this array of approaches. While these prevention categories are not mutually exclusive, we place each state policy within a prevention group to facilitate organization of policies based on their chief intent.

Table 1.

State Policies to Curb Opioid Misusea

Stage	Examples of Intervention	Intervention Description
Primary prevention	Continuing medical education requirements	Continuing medical education requirements on pain management or opioid prescribing. These requirements can be tied to licensure.
	Laws related to pain management clinicb	Policies that target inappropriate prescribing from health care facilities that primarily manage and treat chronic pain.
	Opioid prescribing guidelines and prescription formsb	Recommendations to providers around opioid prescribing. Guidance documents vary but typically include opioid selection, dosage, duration, titration, and discontinuation; screening tools; written treatment agreements; and urine drug testing.
Secondary prevention	Anti‐doctor‐shopping laws	Laws and programs that restrict or prohibit patients from seeking or filling multiple opioid prescriptions from different prescribers or dispensers within a short period of time.
	Drug supply managementc	Policies that limit opioid prescribing by restricting quantity or dosage that can be prescribed and/or requiring payer prior authorization before authorizing payment for an opioid prescription.
	Prescription drug monitoring programs (PDMPs)c	An electronic database that collects, monitors, and analyzes controlled‐substance prescribing and dispensing. Laws vary widely but can include which providers and state officials have access to the PDMP; mandatory prescriber and dispenser querying; interstate data sharing; update frequency; schedule of controlled substance monitored; and operating agency.
Tertiary prevention	Naloxone access laws	Policies that increase lay access to naloxone. Laws vary but can include third‐party prescriptions; pharmacist dispensing without a prescription; prescriber, dispenser, and layperson immunity from civil and criminal penalties; and standing‐order provisions.
	Good Samaritan laws	Laws that offer legal protection to individuals who seek emergency help for a drug overdose.
	Policies affecting opioid addiction treatment	Policies that influence access to treatments for opioid addiction, such as residential treatment and medication‐assisted therapy. Policies vary greatly but include mandating or restricting benefit coverage, modifying public funding for treatment, and imposing provider licensing requirements.

Data derived from Haffajee (2016).22

^aThis table includes interventions assessed in the research articles included in the scoping review. It is not exhaustive of all state strategies to address opioid misuse. As is identified in footnotes b and c, we acknowledge that some policies intend to influence multiple prevention categories. However, we use this categorization system to clearly communicate the chief intent of the state policies evaluated.

^bThese interventions can also be considered secondary prevention.

^cThese interventions could be considered primary, secondary, or tertiary intervention because they influence primary exposure to opioids, high‐risk opioid exposure, and treatment access for individuals with an opioid dependence.

Previous studies aggregated evidence from specific interventions7, 8 and integrated strategies in a single review.9, 10 Reviews published in the past two years of prescription drug monitoring program (PDMP) evaluations are inconclusive with regard to PDMP effects on overdose and other outcomes.7, 8 Reviews that synthesize evaluations of multiple interventions published prior to 2016 identified some promising state policies to decrease opioid prescribing, including PDMPs, policies targeting insurance practices, pain clinic regulations, clinical guidelines, and naloxone access laws.9, 10 However, they also highlighted that evidence quality was low and that rigorous evaluations were needed to further investigate policy effects.9, 10 Since the publication of these reviews, state policies have evolved significantly and original empirical evaluations of state interventions have improved in study rigor,6 suggesting that an updated review would provide additional insight into the effects of state policies targeting opioid misuse and overdose.

This scoping review aims to synthesize the available evidence on the effectiveness of prevalent state opioid policies on improving outcomes related to opioid prescribing and dispensing, patient behavior, and patient health. Given the recent increase in the adoption of state opioid policies and interest among policymakers to address the opioid crisis, we hypothesized that the evidence base evaluating these policies would have grown substantially in recent years, offering a clearer sense of policy effects on patient and prescriber outcomes. We also hypothesized that policies would demonstrate more significant effects on the outcomes most closely related to the behavior(s) they target. Specifically, we expected primary and secondary prevention policies to be most associated with changes in outcomes related to opioid prescribing and dispensing and patient behavior, and tertiary prevention strategies to have the greatest impacts on patient health. Finally, we expected that promising policies identified by previous reviews—specifically PDMPs, policies targeting insurance practices, pain clinic regulations, clinical guidelines, and naloxone access laws—would have the largest effects on provider‐ and patient‐related outcomes compared to other state laws.

Methods

We systematically identified and synthesized findings from empirical evaluations of state opioid misuse prevention policies.

Data Sources and Searches

Following consultation with an informationist at the Taubman Health Sciences Library at the University of Michigan, we searched six online literature databases: Cumulative Index to Nursing and Allied Health Literature Complete, Criminal Justice Abstracts, the National Bureau of Economic Research (NBER), PubMed, PsychINFO, and Scopus. We conducted the initial search in PubMed; searches in other databases, with the exception of NBER, were analogous to the original search. In NBER, we searched “opioid” and reviewed all yielded articles for inclusion. We examined references from the selected materials to identify additional articles that met the inclusion criteria. To ensure that we captured all relevant studies, we compared our yielded articles with the evaluations included in the following review papers: Haegerich et al., 2014;9 Beaudoin et al., 2016;10 Finley et al., 2017;8 and Fink et al., 2018.7 We conducted the search in summer 2018 and no additional articles were added after September 1, 2018. All of the resulting citations and abstracts were exported to Mendeley 1.19.1. We did not impose a date restriction on searches. See Online Appendix 1 for terms and the algorithm used in the literature search.

Eligibility Criteria

Inclusion in the scoping review required that the original quantitative research article be written in English and evaluate the effect of a US state policy on a patient‐ or provider‐related outcome (defined below). We defined state policy as a legislative or administrative action, such as a law or regulation, that directly targeted opioid misuse. For example, naloxone access laws are a legislative action in that they intend to affect naloxone access by modifying statutorily who is allowed to prescribe, dispense, and possess naloxone. We also included PDMPs because they are most often established through a formal legislative or regulatory action. We generally excluded state programs that were not triggered by law passage or rulemaking, with the exception of drug supply management policies and opioid prescribing guidelines. While state funded and administered programs play a large part of public strategies to address opioid misuse and overdose, we focused on state initiatives with a policymaking component to inform activities directly relevant to legislative and regulatory policymakers. As a result, we determined that evaluations of state programs not triggered by a law or regulation were generally beyond the scope of this review; other studies have synthesized the evidence on the effects of these programs.9, 11, 12

We included drug supply management policies (eg, quantity and dosing limits, prior authorization restrictions) and opioid prescribing guidelines, both of which can be implemented through informal policymaking, such as bulletins, guidelines, and Medicaid protocols, for three reasons. First, these policies are an important state policy tool in promoting or restricting access to opioids and medications used in the treatment of opioid dependence. Second, state actors, depending on the state, can use their formal policymaking powers to enact these policies and guidelines. Third, it is unclear from the articles included in this section whether state actors enacted the policy through a formal or informal policymaking process.

We required that the original empirical research study assess at least one of the following outcomes: prescribing/dispensing (eg, volume of opioids prescribed or dispensed, opioid dosage prescribed or dispensed), patient behavior (eg, use of multiple providers or pharmacies, diverted opioids), and patient health (eg, fatal and nonfatal overdose, treatment visits). Outcomes classified as opioids prescribed or dispensed include total/monthly/daily opioid prescriptions, dispensed controlled substances, mean per person per month fills, and days supplied. Outcomes classified as opioid dosage prescribed include average and per‐transaction morphine milligram equivalent (MME) dosage; and long‐acting and short‐acting opioid prescriptions.

We excluded qualitative studies, book chapters, review articles, dissertations, editorials, letters to the editor, and purely descriptive studies. We did not place restrictions on sample size or age. Eligible studies were peer‐reviewed or published in Morbidity and Mortality Weekly Report or NBER. Two authors independently reviewed articles for inclusion, while a third author resolved outstanding conflicts regarding study inclusion.

Policies Evaluated

Included articles reviewed nine types of state policy: three primary prevention strategies (ie, continuing medical education requirements, laws related to pain management clinics, and opioid prescribing guidelines); three secondary prevention strategies (ie, anti‐doctor‐shopping laws, drug supply management policies, and PDMPs); and three tertiary prevention strategies (ie, naloxone access laws, Good Samaritan laws, and policies affecting opioid addiction treatment).

Continuing Medical Education Requirements

State continuing medical education requirements for pain management or controlled substances mandate that physicians receive postgraduate training in opioid prescribing, addiction, and/or related topics. As of December 2015, 23 states required at least some physicians to receive training in pain management or controlled‐substance prescribing as a condition of obtaining or renewing their medical license or to specialize in pain management. Only five states required all or nearly all physicians to obtain periodic continuing medical education on topics related to pain management, controlled‐substance prescribing, or substance use disorders.13

Laws Related to Pain Management Clinics

Pain management clinic policies regulate facilities that primarily manage and treat chronic pain by imposing operational, personnel, inspection, and other requirements on the businesses. As of June 2018, 12 states had implemented pain management clinic laws.14, 15

Opioid Prescribing Guidelines

Opioid prescribing guidelines provide recommendations to providers on opioid prescribing practices. Guidelines vary but typically include opioid selection, dosage, duration, titration, and discontinuation; screening tools; written treatment agreements; and urine drug testing. As of July 2017, 41 states had adopted opioid prescribing guidelines for acute or emergency care.16 This domain may include both payor policies embedded in informal regulatory actions (eg, Medicaid prescribing guidelines) and state laws or regulations requiring the development and implementation of prescribing standards. See the section on eligibility criteria inclusion parameters regarding opioid prescribing guidelines.

Anti‐Doctor‐Shopping Laws

Doctor shopping refers to a patient obtaining controlled substances from multiple health care prescribers without the providers’ knowledge of the other prescriptions. All 50 states and the District of Columbia have a general fraud statute, which prohibits patients from obtaining drugs by fraud, deceit, misrepresentation, subterfuge, or concealment of material fact. As of 2012, 20 states also have laws that specifically prohibit patients from withholding from practitioners that they received a controlled substance or prescription order from another prescriber.17

Drug Supply Management Policies

Drug supply management policies limit opioid prescribing by restricting quantity or dosage that can be prescribed, or by imposing prior authorization requirements or fail‐first protocols (whereby insurers require a treatment to be demonstrated as ineffective before they will approve a more expensive treatment). Such restrictions can apply to public programs and/or private plans regulated at the state level. This domain may include both payor policies embedded in informal regulatory actions (eg, Medicaid plan protocols) and state restrictions affecting private and/or public payors enacted through statute or regulation (eg, statutory prohibition of all state‐regulated payors from applying concurrent review to daily buprenorphine formulations). See the section on eligibility criteria inclusion parameters regarding drug supply management policies in the analysis.

PDMPs

A PDMP is an electronic database that tracks controlled‐substance prescriptions dispensed in a state. PDMPs can be used as a clinical tool to help identify patients who may be at risk for adverse consequences associated with high‐risk prescription opioid receipt. Since the 1990s, PDMPs have proliferated across the country; now all states except Missouri have an operational program.18 PDMPs vary in their features, with the most robust PDMPs requiring prescribers to register and query the database before prescribing opioids.

Naloxone Access Laws

Naloxone is an opioid antagonist designed to rapidly reverse opioid overdose. Naloxone access laws are designed to increase access to naloxone among those in a position to administer the medication in the event of overdose. Laws vary but can include the following provisions: (1) third‐party prescriptions, which permit naloxone to be prescribed to third parties who might be in a position to assist others who overdose; (2) provisions that make naloxone available through non‐patient specific prescriptions, such as standing order, collaborative practice agreements, and full prescriptive authority; (3) prescriber immunity provisions, which provide civil or criminal immunity to naloxone prescribers; and (4) lay dispensing provisions, which allow persons not otherwise permitted to dispense prescription medications to dispense naloxone. As of December 2018, all states and Washington, DC, had a naloxone access law: 48 had a third‐party prescribing provision and 44 had a standing‐order provision.19, 20

Good Samaritan Laws

Good Samaritan laws provide legal protection for persons who overdose and bystanders who call emergency authorities during an overdose event. These laws vary in specific criminal protections for drug possession, drug paraphernalia, and parole or probation violation. As of December 2018, 46 states and Washington, DC, had adopted a Good Samaritan law.20, 21

Policies Affecting Opioid Addiction Treatment

This category includes policies that influence access to treatments for opioid addiction, such as residential treatment and medication‐assisted treatment. Policies vary greatly but include mandating or restricting benefit coverage for opioid use disorder, modifying public funding for treatment, or imposing provider licensing requirements. Articles included in this review assess policies related to buprenorphine access, methadone maintenance treatment, and mandated naltrexone therapy.

Data Extraction

We extracted data using a standardized article assessment form that captured the following elements: policy studied, outcome data source, study design, study years, sample, results, and limitations (Online Appendix 2). The limitations extracted focus on information relevant to sampling and covariate inclusion. Two authors independently reviewed 10 randomly selected articles and entered relevant content into the extraction table. The same two authors reviewed the 10 extractions for consistency and to resolve differences. One author then completed article extraction for the other 61 articles, while the other two authors provided feedback on the extraction.

Data Synthesis

Due to heterogeneity in the policies and outcomes evaluated, we performed a qualitative assessment and synthesis. We categorized policies as (1) primary prevention; (2) secondary prevention; and (3) tertiary prevention. Table 1 summarizes these policies but is not an exhaustive list of state strategies to address opioid misuse, overdose, and prescribing; it lists only the state policies assessed in the original empirical articles included in this review.

We categorized articles using the following three‐step procedure. First, we organized studies by research design using a simplified hierarchy adopted from Haffajee (2016) (see Online Appendix 3).22 Although not exhaustive of the different types of study designs used to assess public health legal interventions, the hierarchy aids policymakers in evaluating evidence quality to make policy decisions. Next, we classified studies into three categories based on outcomes evaluated: prescribing and dispensing, patient behavior, and patient health. We included studies that evaluated multiple outcomes in all relevant outcome categories. Finally, we organized studies by policy type evaluated. Similar to outcome categories, we classified studies that evaluated the independent effects of multiple policies in each relevant policy category.

We rated the quality of evidence for each policy/outcome group using a modified Grading of Recommendations, Assessment, Development and Evaluations (GRADE) approach.23, 24 The GRADE framework is a systematic strategy for rating the quality of a body of evidence for synthesis with the following quality grades: high quality—further research is very unlikely to change our confidence in the estimate of the effect; moderate quality—further research is likely to have an important impact on our confidence in the estimate of the effect and may change the estimate; low quality—further research is very likely to have an important impact on our confidence in the estimate of the effect and is likely to change the estimate; very low quality—we are very uncertain about the estimate of the effect.

Our modified GRADE approach employs the following procedure. First, we assigned all policy and outcome groups a low quality of evidence score, as the GRADE approach rates all observational studies a low score and all of our included articles used an observational design. Second, we modified the original GRADE score based on factors that can reduce or increase the quality of evidence. Factors that can reduce the quality of evidence include limitations in study design or execution, result inconsistency, indirectness of evidence, imprecision, and publication bias. Factors that can improve the quality of evidence include effect size and if unaccounted‐for confounding is suspected to strengthen the findings. We automatically assigned a very low quality of evidence score for policy/outcome groups with only one evaluation. We did not assign a GRADE score to outcomes associated with multiple policies because articles within this category evaluate different combinations of policies. Since the GRADE approach rates the quality of evidence across evaluations of the same or very similar interventions, we do not believe that it is appropriate to assign a GRADE score to the synthesized findings of articles evaluating different combined interventions. The GRADE scores assigned for each policy/outcome group are available in Online Appendix 4.

Results

Figure 1 depicts the literature search and selection process; 71 articles met the inclusion criteria. Table 2 provides a summary of the articles included in the review: 10 assessed primary prevention interventions, 42 assessed secondary prevention interventions, and 12 assessed tertiary prevention policies. Studies most frequently evaluated PDMPs (n = 38), followed by combined effects of multiple policies (n = 10) and opioid addiction treatment policies (n = 7). The number of articles by publication year ranged from 41 in 2016‐2018 to 0 between 2001‐2005 (see Online Appendix 5 for a visual depiction of number of articles published annually by policy type).

Figure 1.

Literature Search and Selection Process

Table 2.

Study Characteristics

Characteristic	Number of Studies
Total studies	71
Publication year
1980‐2000	2
2001‐2005	0
2006‐2010	6
2011‐2015	22
2016‐2018	41
Study designa
Interrupted time series with comparison	8
Interrupted time series without comparison	8
Controlled pre‐post	28
Uncontrolled pre‐post	18
Uncontrolled post‐only	0
Cross‐sectional	10
Intervention typeb
Primary prevention	10
Secondary prevention	42
Tertiary prevention	12
Combined effects of multiple policies	10
Interventionc
Anti‐doctor‐shopping laws	2
Continuing medical education requirements	1
Drug supply management	5
Good Samaritan laws	2
Naloxone access laws	3
Opioid prescribing guidelines	5
Laws related to pain management clinics	4
Policies affecting opioid addiction treatment	7
Prescription drug monitoring programs	38
Combined effects of multiple policies	10

The totals from study design, intervention type, and intervention do not sum to 71 because certain studies fall into multiple categories (see footnotes b, c, and d).

^aHaffajee et al., (2018)25 is included in two study design categories: interrupted time series with comparison and controlled pre‐post.

^bKuo et al., (2016)26 and Meara et al., (2016)27 analyzed policies categorized in primary prevention and secondary prevention. Dowell et al., (2016)28 analyzed a primary prevention policy and the combined effects of multiple policies.

^cKuo et al., (2016)26 and Meara et al., (2016)27 are in three intervention categories: anti‐doctor‐shopping laws, laws related to pain management clinics, and prescription drug monitoring programs. Dowell et al., (2016)28 is in two intervention categories: laws related to pain management clinics and combined effects of multiple policies. Rees et al., (2017)29 is in two intervention categories: naloxone access and Good Samaritan laws.

The following sections provide an overall summary of the evidence evaluating each policy. As is detailed later in the paper, contradictory rigorous evaluations on laws related to pain management clinics provide mixed findings on the effects of these policies on prescribing outcomes. Evidence suggests that drug supply management laws and robust PDMPs reduce opioid prescribing and dispensing. Specifically, drug supply management policies reduce prescribing of higher‐risk opioids targeted by the policies, while increasing the frequency of lower‐risk prescriptions. Robust PDMPs with mandatory access provisions were associated with reductions in a variety of opioid prescribing measures, including total prescriptions and number of opioid fills. Across interventions, the quality of evidence on patient health outcomes is insufficient to facilitate conclusions. Of the 19 policy and outcome groups, 13 (68.4%) received a very low quality of evidence score; 5 (26.3%) received a low score; and 1 (5.3%) received a moderate score.

In the subsequent policy results sections, we focus on the most rigorously designed studies, which are more appropriate for causal inference. Studies of weaker design for causal inference are described in Tables 3 to 6. All findings reported are significant at the 0.05 significance level. In other words, findings reported as “no effect” or “no change” were not significant at the 0.05 level. See Appendix 2 for more detailed quantitative results, including effect estimates and confidence intervals.

Table 3.

Primary Prevention

Outcome Type *GRADE Quality of Evidence Score a	Study Design	Number of Studies	Summarized Findings
Continuing medical education requirements
Prescribing/dispensing *Very low due to 1 evaluation and limitations in study design	Uncontrolled pre‐post	1	Decline in high‐dosage opioids dispensed (Katzman et al., 2014)30 Increase in low‐dosage opioids dispensed (Katzman et al., 2014)30 No change in opioid prescriptions filled (Katzman et al., 2014)30
Laws related to pain management clinics
Prescribing/dispensingc *Very low due to inconsistency in results	ITS without comparison	1	Decline in opioids prescribed (Lyapustina et al., 2016)31 Decline in opioid dosage prescribed (Lyapustina et al., 2016)31 Effects concentrated among highest baseline opioid prescribers and highest baseline opioid users (Lyapustina et al., 2016)31
	Controlled pre‐post	3	Decline in long‐term opioid receipt (Meara et al., 2016)27 No change in receipt of high‐dosage or non‐long‐term opioid receipt (Meara et al., 2016)27 No change in prescription opioid dosage dispensed associated with pain clinic law alone (Dowell et al., 2016)28 Decline in schedule II opioids prescribed (Kuo et al., 2016)26 No change in schedule III opioids prescribed (Kuo et al., 2016)26
Patient behavior *Very low due to 1 evaluation	Controlled pre‐post	1	No change in four or more opioid prescribers (Meara et al., 2016)27
Patient health *Low	Controlled pre‐post	2	No change in nonfatal prescription opioid overdose (Meara et al., 2016)27 No change in prescription opioid overdose death rates associated with pain clinic laws alone (Dowell et al., 2016)28 No change in heroin‐related mortality (Dowell et al., 2016)28
Opioid prescribing guidelines
Prescribing/dispensing *Low	ITS with comparison	1	Decline in total opioid prescriptions and total MME per month (Weiner et al., 2017)32 Decline in total prescriptions greater than 3‐day supply and total MME per month per prescription greater than a 3‐day supply (Weiner et al., 2017)32
	Uncontrolled pre‐post	3	Decline in opioids prescribed (Franklin, 2012)33 Decline in high‐dose opioid prescriptions (Garg, 2013; Sullivan, 2016)34, 35 No change in median opioid dose (Sullivan, 2016)35
Patient health *Very low due to 1 evaluation	Uncontrolled pre‐post	1	Increase in methadone poisonings (Fulton‐Kehoe, 2015)36 No change in other prescription opioid poisonings (Fulton‐Kehoe, 2015)36

Abbreviations: ITS, Interrupted time series; MME, morphine milligram equivalent.

^aSee Online Appendix 4 for the modified GRADE Summary of Findings. The GRADE approach automatically rates observational studies a low quality of evidence score. Since all of our included articles use an observational approach, compared to a randomized trial, all policy/outcome pairs are initially given a low quality of evidence score. Policy/outcome groups can then be rated up or down. If the quality of evidence score is moved up or down from the low rating, we provide an explanation following the score.

Table 6.

Multiple Policies

Outcome Typea	Study Design	Number of Studies	Significant Findings
Prescribing/dispensing	ITS with comparison	3	Decline in opioids prescribed (Rutkow et al., 2015b)86 Decline in opioids prescribed by high‐risk providers (Rutkow et al., 2015b;86 Chang et al., 2016b)87 Decline in percentage of high‐risk patients prescribed opioids (Chang et al., 2018b)88 Decline in opioid dosage dispensed (Rutkow et al., 2015b)86 Decline in opioid dosages prescribed by high‐risk prescribers (Chang et al., 2016b;87 Rutkow et al., 2015b)86 Decline in opioid dosage prescribed to high‐risk patients (Chang et al., 2018b;88 Rutkow et al., 2015b)86 No change in opioid dosages prescribed by low‐risk prescribers (Chang et al., 2016b)87 No change in opioid dosage prescribed to low‐risk patients (Chang et al., 2018b)88
	ITS without comparison	2	Decline in daily MEDs per patient for opioid, hydrocodone, oxycodone, methadone, and hydromorphone dispensed (Al Achkar et al., 2018)89 Decline in opioids dispensed in the overall cohort, prior risk of opioid use cohort, and opioid chronic opioid use cohort (Sun, 2017)90 No change in daily MEDs per patient for morphine, fentanyl, oxymorphone, and buprenorphine (Al Achkar et al., 2018)89
	Controlled pre‐post	1	Decline in opioids prescribed (Dowell, 2016)28
Patient behavior	Uncontrolled pre‐post	1	Decline in diversion rates for oxycodone, methadone, and morphine (Surratt et al., 2014b)91 No decline in diversion rates for fentanyl, hydromorphone, and buprenorphine (Surratt et al., 2014b)91
Patient health	ITS with comparison	1	Decline in oxycodone‐related mortality (Delcher et al., 2015b)53
	Controlled pre‐post	2	Decline in prescription‐opioid‐related mortality (Kennedy‐Hendricks et al., 2016b;92 Dowell, 2016)28 Smaller heroin‐related mortality increase than comparison state (Kennedy‐Hendricks et al., 2016b)92
	Uncontrolled pre‐post	1	Decline in overdose mortality due to oxycodone, methadone, hydrocodone, and other opioid analgesics (Johnson et al., 2014b)93 Increase in overdose mortality due to morphine, hydromorphone, and heroin (Johnson et al., 2014b)93

Abbreviations: ITS, interrupted time series; MED, morphine equivalent dose.

^aWe do not provide a GRADE quality of evidence score for multiple policies because each article evaluates different components of the same group of policies or a different combination of policies entirely.

^bArticles evaluating some components or the entire combined effects of the 2010‐2011 Florida interventions.

Continuing Medical Education Requirements

Evidence on statutory or regulatory continuing medical education requirements is extremely limited due to the single evaluation that met our inclusion criteria and thus received a very low quality of evidence score (see Table 3). The one study in this category assessed prescribing behaviors among clinicians before and after a 2012 New Mexico Senate law, which required all health care professional licensing boards to mandate continuing medical education training for the treatment of chronic pain. The authors observed a reduction in high opioid prescription dosages (>100 MME per day) and an increase in moderate opioid prescription dosages (≤40 MME per day). They observed slight increases in the total number of opioid prescriptions filled.30

Laws Related to Pain Management Clinics

Based on available evidence, it is unclear whether laws related to pain management clinics exert a direct, combined, or null effect on opioid prescribing (see Table 3). Only one evaluation, by Lyapustina and colleagues (2016) of the 2010 Texas pain management clinic law, observed reductions in opioids prescribed, including average MME per transaction, total opioid volume (ie, total MME across all transactions), number of opioid prescriptions, and quantity of opioid pills dispensed, following policy implementation.31 However, other studies suggest that laws related to pain management clinics have no direct effect on opioids prescribed. Dowell and colleagues (2016) did not identify an independent association between pain management clinic laws and MMEs prescribed per state resident.28 Evidence from Meara and colleagues (2016) further suggests that laws related to pain management clinics do not affect opioid prescribing. Using a sample of Medicare beneficiaries, the authors observed no association between pain clinic regulations and non‐long‐term opioid receipt and opioid dosage greater than 120 daily MME.27 Further, other rigorous evaluations suggest that the potential effects of pain management clinic laws on opioid prescribing may occur only in combination with other policies. The evaluation conducted by Dowell and colleagues, while not identifying an independent effect of these policies, observed that states with both pain management clinic laws and mandatory provider review of the state PDMP experienced decreases in opioid MME prescribing rate.28 In addition, several evaluations of the 2010‐2011 Florida policies targeting opioid misuse observed PDMPs and pain management clinic policies together were associated with reductions in opioids prescribed. Florida introduced these policies in quick succession (see section on combined effects of multiple policy interventions).53, 86, 87, 88 Given that the initial Florida PDMP implemented on September 1, 2011, was relatively weak, since it did not contain critical provisions, such as registration or use mandates, it is challenging to attribute the entirety of the change in opioid prescribing to the PDMP, and not the combined or singular effect of the pain clinic law and other policies implemented during the same period.94

Two rigorous evaluations suggest that pain clinic laws alone have no effect on patient health outcomes. Dowell and colleagues did not identify an association between pain clinic laws and prescription opioid overdose deaths, heroin overdose deaths, and combined drug overdose deaths.28 However, states with both pain clinic laws and mandatory provider review experienced decreases in prescription opioid overdose deaths and combined drug overdose deaths, but not heroin overdose deaths.28 Meara and colleagues also observed no relationship between pain clinic laws and nonfatal prescription opioid overdose.27

Opioid Prescribing Guidelines

We identified only one rigorous evaluation that observed significant reductions in opioid prescribing behaviors following state opioid guideline implementation (see Table 3). Weiner and colleagues (2017) evaluated the Ohio 2012 emergency physician guidelines that encouraged physicians to check the Ohio PDMP before prescribing controlled medication and urged physicians to limit the quantity of opioids prescribed to no more than a three days’ supply, among other provisions. The guideline was associated with a 12% decrease in the level of statewide total monthly opioid prescriptions.

Anti‐Doctor‐Shopping Laws

Evidence on anti‐doctor‐shopping laws is extremely limited and of very low quality (see Table 4). Only two studies met the inclusion criteria for this category, both of which assessed the independent effects of multiple state opioid prevention policies, including doctor‐shopping restrictions.26, 27 Neither study identified an association between anti‐doctor‐shopping laws and opioid prescribing outcomes.

Table 4.

Secondary Prevention

Outcome Type *GRADE Quality of Evidence Score a	Study Design	Number of Studies	Specific Findings
Anti‐doctor‐shopping laws
Prescribing/dispensing *Very low due to limitations in study design	Controlled pre‐post	2	No change in schedule II or III opioid prescriptions (Kuo et al., 2016)26 No change in receipt of high‐dosage opioids and non‐long‐term opioid receipt (Meara et al., 2016)27
Patient behavior *Very low due to one evaluation	Controlled pre‐post	1	No change in four or more opioid prescribers (Meara et al., 2016)27
Patient health *Very low due to one evaluation	Controlled pre‐post	1	No change in nonfatal prescription opioid overdose (Meara et al., 2016)27
Drug supply management policies
Prescribing/dispensing *Moderate due to magnitude and consistency of effect	Controlled pre‐post	3	Decline in high‐dose opioid prescriptions (Hartung et al., 2018; Keast et al., 2018)37, 38 Increase in low‐dose opioids (Hartung et al., 2018; Keast et al., 2018)37, 38 No change in total opioids or opioid dosage between 61 and 120 MED (Hartung et al., 2018)37 Stringent prior authorization policy associated with a reduction in controlled‐release oxycodone use compared to lenient prior authorization policy (Morden et al., 2008)39
	Uncontrolled pre‐post	1	No change in high‐dose opioids (Riggs et al., 2017)40 Minimal decrease in total daily opioids dispensed (Riggs et al., 2017)40
Patient behavior *Very low due to one evaluation	Controlled pre‐post	2	Decline in multiple pharmacy visits (Hartung et al., 2018)37 Decline in multiple prescriber use among high‐risk opioid users (Keast, 2018)38
Patient health *Very low due to limitations in study design	Controlled pre‐post	1	No change in opioid‐related emergency department visit or hospitalization (Hartung et al., 2018)37
	Cross‐sectional	1	Lower rates of opioid misuse in high and low prior authorization policies compared to no prior authorization policy (Cochran et al., 2017)41 Lower rates of opioid overdose in low prior authorization policy compared to absence of prior authorization policy (Cochran et al., 2017)41
Prescription drug monitoring programs b
Prescribing/dispensing *Low	ITS with comparison	1	Decline in schedule II and III opioids prescribed (Moyo et al., 2017)42 No change in total opioids and schedule IV‐V opioids prescribed (Moyo et al., 2017)42
	Controlled pre‐post	6	Decline in schedule II opioids prescribed (Bao et al., 2016)43 and overall opioid dosage (Brady et al., 2014)44 Decline in oxycodone shipments (Reisman et al., 2009)57
			No change in high‐dosage opioids prescribed (Buchmueller et al., 2018),45 total opioids prescribed (Bao et al., 2016; Buchmueller et al., 2018),43, 45 overall opioid dosage dispensed (Brady et al., 2014; Paulozzi et al., 2017),44, 46 long‐term opioid receipt (Meara et al., 2016)27
	Uncontrolled pre‐post	4	Decline in opioids dispensed (Deyo et al., 2018)38 No change in opioids prescribed (Baehren et al., 2010;47 Landau et al., 201848), controlled substances nor uncontrolled substances (McAllister et al., 2015)49
	Cross‐sectional	3	Higher odds of any analgesic prescription (Simoni‐Wastila et al., 2018)42 Lower opioid and controlled‐release oxycodone prescriptions (Curtis et al., 2006)50 No change in prescription of pain medication or opioids (Lin et al., 2018)51
Patient behavior *Very low due to inconsistency in results	Controlled pre‐post	2	Decline in frequency of two‐plus opioid prescribers and four‐plus new patient visits (Ali et al., 2017;52 Buchmueller et al., 201845) No change in illegitimate opioid source (Ali et al., 2017)52 No change in overlapping claims, five‐plus prescribers, out‐of‐state prescribers and pharmacies (Buchmueller et al., 2018)45
Patient health *Very low due to inconsistency in results	ITS with comparison	2	Decline in oxycodone‐related mortality (Delcher et al., 2015)53 and overall opioid‐related mortality (Patrick, 2016)54
			No change in non‐oxycodone‐ or heroin‐related mortality (Delcher et al., 2015)53
	ITS without comparison	1	Increase in prescription opioid and heroin treatment admissions (Branham et al., 2017)55
	Controlled pre‐post	10	Increase in drug overdose mortality (Li, 2014)56 Decline in past‐year days used of NMPRc and heroin (Ali et al., 2017)52 Decline in inpatient drug rehabilitation admissions (Reisman et al., 2009)57 No change in overall drug overdose mortality or opioid‐related overdose mortality (Nam et al., 2017;58 Paulozzi et al., 2011)46 No change in heroin or prescription opioid overdose mortality (Nam et al., 2017)58 No change in opioid‐related poisonings (Buchmueller et al., 2018)45 No change in prescription‐drug‐ or heroin‐related treatment admissions (Dave et al., 2017),59 emergency department visits involving an opioid (Maughan et al., 2015)60 No change in past‐year NMPRc or past‐year heroin use, abuse/dependence, or initiation (Ali et al., 2017)52 Smaller increase in intentional exposures and opioid treatment admissions (Reifler, 2012)61; and prescription opioid‐related overdose (Pauly, 2018)62

Abbreviations: ITS, interrupted time series; NMPR: Nonmedical prescription pain reliever.

^aSee Online Appendix 4 for the modified GRADE Summary of Findings. The GRADE approach automatically rates observational studies a low quality of evidence score. Since all of our included articles use an observational approach, compared to a randomized trial, all policy/outcome pairs are initially given a low quality of evidence score. Policy/outcome groups can then be rated up or down. If the quality of evidence score is moved up or down from the low rating, we provide an explanation following the score.

^bWe excluded the following studies from Table 4 because they evaluated PDMP provisions, not overall PDMPs, or compared robust to nonrobust PDMPs: Brown et al., 2017;63 Gilson et al., 2011;64 Green et al., 2012;65 Haffajee et al., 2018;25 Kuo et al., 2016;26 Pardo et al., 2016;66 Phillips et al., 2017;67 Rasubala et al., 2015;68 Ringwalt et al., 2015;69 Sigler et al., 1984;70 Suffoletto et al., 2018;71 Sun et al., 2017;72 Wastila et al., 1996;73 Wen et al., 2017;74 and Yarbrough et al., 2018.75 See Online Appendix 2 for a detailed summary of these evaluations.

^cLow‐dose opioids are prescriptions <61 morphine equivalent dose or short‐acting opioids. High‐dose opioids are prescriptions >120 morphine equivalent dose or long‐acting opioids.

Drug Supply Management Policies

Existing evidence suggests that prior authorization laws fulfill their intended effect of limiting access to higher‐risk opioids targeted by the policies (see Table 4). Hartung and colleagues (2018) evaluated a 2012 Oregon Medicaid prior authorization policy that required prior authorization for high‐dose opioid prescriptions; the study demonstrated a decrease in opioid prescriptions above the high‐dosage threshold and an increase in the monthly probability of low‐dosage opioid prescriptions following policy implementation.37 Keast and colleagues (2018) found that a 2008 Oklahoma Medicaid prior authorization policy that required a trial of short‐acting opioids prior to initiating extended release/long‐acting therapy resulted in a reduction in new extended release/long‐acting opioid use among opioid‐naïve patients and regardless of past opioid use. The policy also was associated with an increase in short‐acting opioid use.38

Research by Morden and colleagues (2018) suggests that prior authorization policies of varying stringency have differential effects on controlled‐release oxycodone use.39 The authors compared strict, lenient, and no prior authorization policies using outpatient fee‐for‐service Medicaid prescription claims in 49 states and the District of Columbia. States with prior authorization policies did not differ in controlled‐release oxycodone use from states without prior authorization policies. However, in aggregate, strict Medicaid prior authorization policies were associated with a 34% reduction in controlled‐release oxycodone use.39

Prior authorization policies may be effective at reducing outcomes related to doctor shopping. Two rigorous evaluations observed that prior authorization policies were associated with decreases in multiple pharmacy or prescriber use. Hartung and colleagues observed a small decrease in multiple pharmacy visits following policy implementation.37 Among persons with high‐risk opioid use, Keast and colleagues identified a reduction in multiple prescriber use associated with the 2008 Oklahoma Medicaid policy.38

The evidence on the effect of drug supply management policies on patient health outcomes is extremely limited and of very low quality. The one rigorous evaluation available suggests that a prior authorization policy for high‐dosage prescriptions (>120 MME) had no effect on opioid‐related emergency department visits or hospitalizations.37

Prescription Drug Monitoring Programs

Although studies evaluating PDMPs have mixed results across outcomes, certain PDMP features (specifically, mandatory access provisions) show more promise in reducing opioids prescribed (see Table 4).

PDMPs Overall

Evidence from the most rigorous evaluations suggest that PDMPs have no effect on opioid prescribing overall but may reduce higher‐risk prescribing behaviors. For example, Moyo and colleagues (2017) observed that PDMP implementation is associated with decreases in schedule II and schedule III opioid prescriptions, but has no effect on mean overall MME, total schedule IV, or schedule V opioids dispensed.42 Research by Bao and colleagues (2016) using the National Ambulatory Medical Care Survey suggests that PDMPs reduce schedule II prescriptions, but do not affect total opioid and pain medication prescriptions.43 Other rigorous evaluations suggest that PDMPs have no effect on opioid dosage prescribed. Of the four evaluations that measured opioid dosage before and after PDMP implementation compared to a control group, no study identified a change in opioid dosage following policy implementation.42, 44, 45, 46

The published evidence on the effects of PDMPs on patient health outcomes is also heavily mixed. Thirteen studies evaluated the independent effects of PDMPs on patient health. Outcomes varied greatly by study and included overdose mortality; drug use, misuse, dependence, and initiation; and health care use. Studies considered both illicit (eg, heroin and nonmedical prescription pain reliever use) and legal prescription drug use. Due to the variation in the outcomes considered, and the mixed results across studies that evaluated similar outcomes, more research is needed to clarify the effect of PDMPs on patient‐health‐related measures.

PDMP Features

Recent studies on the adoption of robust PDMP features suggest that PDMP design influences effectiveness, helping to clarify the mixed results on PDMPs overall. Robust PDMPs with mandatory access provisions are associated with decreases in opioid prescribing and reduced doctor‐shopping‐related behaviors, compared to PDMPs without these provisions.

Studies most commonly evaluated mandatory access provisions, which require practitioners to check a PDMP before prescribing or dispensing an opioid. Findings from these evaluations suggest that mandatory access provisions are associated with reductions in opioid prescribing behaviors. For example, Suffoletto and colleagues’ (2018) evaluation of a 2016 Pennsylvania mandatory access provision identified a reduction in the opioid prescribing rate using electronic medical record data from 15 emergency departments in a single health system.71 Buchmueller and colleagues (2018) found that mandatory access provisions were associated with a decline in the probability of receiving opioids.45 Wen and colleagues (2017) found that the effect of mandatory access provisions may actually be explained by the presence of a mandatory registration provision in the Medicaid population, suggesting that further research should explore interactions among features.74

Mandatory access provisions also appear to be associated with reductions in behaviors related to doctor shopping. Two rigorous studies, by Ali and colleagues (2017) and Buchmueller and colleagues (2018), observed that mandatory access provisions were associated with declines in new patient visits,45 multiple prescribers,45, 52 multiple pharmacy visits,45 and overlapping claims,45 but had no effect on social or illegitimate opioid source use.45 Similar to overall PDMPs, results are mixed on the effect of mandatory access provisions on patient health outcomes.

Robust PDMPs, defined as those with multiple provisions (notably, use and registration mandates and delegate access) known or hypothesized to improve the ability of prescribers to use and access PDMPs, also appear to reduce opioid prescriptions. Haffajee and colleagues (2018) used commercial claims data between 2010 and 2014 to examine the effects of four robust PDMPs on overall and high‐risk opioid prescribing compared to results in four similar states without robust PDMPs. The authors observed that robust PDMP implementation was associated with declines in total opioid dosage prescribed and number of opioid fills. Robust PDMPs were less consistently associated with reduced percentage of patients prescribed opioids, with the magnitude and significance of the effects varying by state. The authors also assessed the effect of robust PDMPs on opioid prescriptions filled by three or more prescribers and pharmacists, observing a decrease only in Kentucky, compared to Mississippi, but not in the other state pairs.96

Good Samaritan Laws

Few studies have evaluated Good Samaritan laws and thus, while robust in design, the quality of evidence assessing the effect of these laws on patient health is low (see Table 5). One rigorous evaluation by Nguyen and colleagues (2018) suggests that, consistent with its goals, the 2011 New York Good Samaritan law was associated with increased heroin‐related acute hospital utilization. However, the policy had no effect on non‐heroin opioid‐related visits, supporting the authors’ hypothesis that the law would have a greater effect on heroin‐related overdose than non‐heroin‐related events because the threat of charge and conviction is less salient for non‐heroin cases.76 Conversely, Rees and colleagues’ (2017) research found no association between Good Samaritan laws and opioid‐related mortality.29

Table 5.

Tertiary Prevention

Outcome Type *GRADE Quality of Evidence Score a	Study Design	Number of Studies	Specific Findings
Good Samaritan laws
Patient health *Low	Controlled pre‐post	2	Increase in emergency department and inpatient hospital admissions for opioids and heroin (Nguyen et al., 2018)76 No change in opioid‐related, non‐heroin‐related, or heroin‐related mortality (Rees et al., 2017)29 No change in nonprescription use of prescription pain killers (Rees et al., 2017)29
Policies affecting opioid addiction treatment
Patient health *Very low due to inconsistency in results	ITS without comparison	1	Decline in high‐dose buprenorphine fills following buprenorphine prior authorization policy (Clark et al., 2014)77 Increase in medium‐ and low‐dose fills following buprenorphine prior authorization policy (Clark et al., 2014)77
	Uncontrolled pre‐post	2	Decrease in methadone maintenance enrollment after removal of methadone from Medicaid benefit (Deck et al., 2006)78 Patients who paid out of pocket for methadone treatment more likely to leave care than patients with benefit coverage (Fuller et al., 2006)79
	Cross‐sectional	4	Increase in buprenorphine use associated with state funds to subsidize buprenorphine and state special prescribing requirements (Andrews et al., 2014)80 No change in buprenorphine use associated with state regulating buprenorphine beyond federal standards (Andrews et al., 2014)80 Greater use of opioid addiction treatment in states with Medicaid methadone coverage (Bachhuber et al., 2017)81 Lower relapse rate associated with mandated naltrexone treatment (Merlo et al., 2011)82 Opioid addiction treatment use higher in states with Medicaid coverage than in states with block‐grant coverage or no public coverage (Saloner et al., 2016)83
Naloxone access laws
Prescribing/dispensing *Low	Controlled pre‐post	2	Increase in naloxone prescriptions associated with naloxone access law, lay dispensing, provider immunity (Gertner et al., 2018)84 Increase in naloxone prescriptions associated with standing‐order provision (Gertner et al., 2018; Xu et al., 2018)84, 85 Increase in naloxone prescriptions associated with third‐party provisions (Xu et al., 2018)85 Decrease in naloxone prescriptions associated with third‐party provisions (Gertner et al., 2018)84
Patient health *Very low due to one evaluation	Controlled pre‐post	1	Decrease in opioid‐related and non‐heroin opioid‐related mortality associated with naloxone access laws (Rees et al., 2017)29 Decrease in opioid‐related and non‐heroin opioid‐related mortality associated with naloxone access laws that remove criminal liability for naloxone possession (Rees et al., 201)29 No change in opioid‐related mortality, non‐heroin opioid‐related mortality, and heroin‐related mortality associated with standing order provision (Rees et al., 201)29 No change in heroin‐related mortality associated with naloxone access law, standing order, or removing criminal liability for naloxone possession (Rees et al., 201)29

^aSee Online Appendix 4 for the modified GRADE Summary of Findings. The GRADE approach automatically rates observational studies a low quality of evidence score. Since all of our included articles use an observational approach, compared to a randomized trial, all policy/outcome pairs are initially given a low quality of evidence score. Policy/outcome groups can then be rated up or down. If the quality of evidence score is moved up or down from the low rating, we provide an explanation following the score.

Policies Affecting Opioid Addiction Treatment

Due to variation in the policies evaluated and outcomes considered, we are unable to draw conclusions about the effects of policies influencing opioid addiction treatment (see Table 5). Further, no study included in this category longitudinally evaluated changes in a treatment group compared to a control group, limiting our ability to infer causal policy effects. Of the seven less rigorous studies that met the inclusion criteria in this category, four articles assessed policies related to methadone and suggest that Medicaid coverage restrictions for methadone may be associated with decreased treatment use.78, 79, 81, 83 One rigorous article evaluated policy changes related to buprenorphine access. Clark and colleagues (2014) observed that a 2008 Massachusetts Medicaid policy requiring more frequent prior authorization for higher‐dose buprenorphine prescriptions was associated with a decrease in the percentage of members filling higher dosages as well as an increase in medium‐ and low‐dosage fills.77

Naloxone Access Laws

Few studies have evaluated the effects of state naloxone access laws (see Table 5). Evidence from two rigorous evaluations, Gertner et al. (2018) and Xu et al. (2018), suggests that naloxone access laws increase prescription naloxone dispensing overall.84, 85 Xu et al. found that naloxone access laws are associated with a 79% increase in naloxone prescriptions dispensed per state‐quarter. Xu et al. also found an independent effect of both standing‐order provisions and third‐party prescribing provisions on naloxone prescribing.85 But Gertner et al. found that the presence of a standing‐order provision was the only naloxone access law feature that independently predicted naloxone prescribing; such a provision corresponded to an increase of 33.1 dispensed prescriptions per state‐quarter, or 74% of the average number of naloxone prescriptions dispensed.84

Evidence from the rigorous study by Rees et al. suggests that naloxone access laws reduced overall opioid‐related mortality by 9%. This effect was significant for non‐heroin opioid‐related mortality but not heroin‐related mortality. In addition, the overall effect was limited to naloxone access laws that remove criminal liability for naloxone possession.29

Combined Effects of Multiple Policy Interventions

Ten articles evaluated the combined effect of multiple policies,28, 53, 86, 87, 88, 89, 90, 91, 92, 93 including seven interested in the 2010‐2011 Florida law enforcement, pharmaceutical, and public health interventions (see Table 6).53, 86, 87, 88, 91, 92, 93 Florida state activities during this period included a January 2010 requirement that pain management clinics register with the Florida Department of Health, a July 2011 law that strengthened state regulation of activities by controlled‐substance dispensing entities, and the implementation of the Florida PDMP in October 2011. Overall, the evidence suggests that combined policies corresponded to reductions in opioid prescribing, lower diversion rates for some types of opioid, and potentially fewer prescription opioid overdose fatalities.

Three rigorous evaluations suggest that the combined 2010‐2011 Florida interventions were associated with reductions in opioids prescribed, with effects concentrated among the highest baseline opioid users and prescribers.86, 87, 88 Surratt and colleagues (2014) observed a decline in diversion rates following implementation of the Florida policy interventions. Using data from the Researched Abuse Diversion and Addiction‐Related Surveillance System from 2009 to 2012, the authors identified a decline in average diversion rates for oxycodone, methadone, and morphine. They did not observe a change in diversion rates for fentanyl, hydrocodone, hydromorphone, or buprenorphine.91 One rigorous evaluation found that these policies were associated with reductions in mortality related to prescription opioids. Kennedy‐Hendricks and colleagues (2016) compared drug overdose deaths from 2003 to 2012, observing a reduction in prescription opioid overdose mortality of 0.6 per 100,000 in 2010, 1.8 per 100,000 in 2011, and 3.0 per 100,000 in 2012 in Florida compared to North Carolina.92 Moreover, increases in heroin‐related mortality during this time period were smaller in Florida than in North Carolina.92

Two articles evaluated other state policies containing multiple opioid‐relevant components; results were generally consistent with evaluations of the Florida laws. Sun and colleagues (2017) investigated a 2012 Washington state mandate that required hospitals to implement seven best practices to reduce potentially avoidable emergency department visits by Medicaid beneficiaries, including several mandates that directly or indirectly targeted opioid prescribing.90 The authors observed that the mandates were associated with a small reduction in number of opioid prescriptions dispensed in the overall, prior risky opioid use, and chronic opioid use cohorts. However, there was no overall or subgroup change in MME per dispensed prescription.90 Al Achkar and colleagues (2018) measured the change in total opioids dispensed in Indiana before and after a 2013 opioid prescribing emergency rule that required providers to, for certain patients, (1) evaluate opioid recipients for psychiatric conditions; (2) review patients’ drug prescription history in Indiana's Prescription Electronic Collection and Tracking Program; (3) perform regular drug screenings; and (4) obtain a signed controlled‐substance agreement from the patient.89 The emergency rules were associated with an instantaneous decrease in daily MMEs per patient for all opioids, hydrocodone, oxycodone, methadone, and hydromorphone. No change was observed for morphine, fentanyl, oxymorphone, or buprenorphine.89

Discussion

States can wield a variety of legal tools to address opioid misuse; these tools warrant evaluation to identify the best use of resources in tackling the opioid crisis. Recent research articles add rigor to the body of evidence assessing opioid misuse policies. In contrast with earlier reviews that identified few rigorous empirical evaluations in this area, more than half of our included studies used quasi‐experimental designs helpful for causal inference (eg, interrupted time series or pre‐post test designs compared to a control group).9 Despite recent improvements in methodological rigor overall, the lack of consistent rigor within policy type and outcome groups limits our ability to confirm our second hypothesis, that policies would have the most significant effect on the outcome most closely related to their intent. Only six of our policy and outcome groups did not receive a very low GRADE rating, challenging our ability to synthesize the evidence within policy and outcome groups.

Despite insufficient evaluation of many policies, research has identified several state opioid misuse prevention policies that appear to influence opioid prescribing and dispensing. Evidence on drug supply management policies and robust PDMPs with mandatory access provisions suggests that these policies reduce the volume and dosages of opioids prescribed and dispensed. Specifically, drug supply management policies achieve their intended effect of reducing prescribing of higher‐risk opioids (in terms of formulations, dosages, and quantity) while increasing access to less high‐risk opioid prescriptions. Robust PDMPs with mandatory access provisions are associated with decreases in a variety of opioid prescribing measures, including total prescriptions, number of fills, and dosages. Research comparing robust PDMPs and mandatory access provisions to PDMPs without these provisions observed that the latter were not associated with similar reductions.45 Evidence on the 2010‐2011 Florida policy interventions suggest that a combination of law enforcement, pharmaceutical, and public health approaches (eg, PDMPs and laws related to pain management clinics) effectively reduced opioids prescribed, especially among high‐risk prescribers and users.

Two rigorous evaluations suggest that naloxone access laws increase prescription naloxone dispensing.84, 85 However, several low‐rigor studies published after our article review suggest that many pharmacies fail to supply naloxone despite these laws. For example, researchers observed that only about a quarter of pharmacies dispensed naloxone two years after implementation of a 2016 California naloxone standing order.96 An evaluation of a 2015 Texas naloxone access law with a standing‐order provision observed that nearly 25% of audited pharmacies did not stock naloxone in 2018.97 Future research should investigate barriers to pharmacist naloxone dispensing in states with standing‐order provisions.

We found insufficient evidence regarding the effect of state interventions on patient health–related outcomes across policies. Two or fewer studies evaluated patient health outcomes for all primary and secondary interventions, with the exception of PDMPs. Synthesis of the patient health effects of PDMPs is complicated by the use of varied outcomes, including overdose mortality; drug use, misuse, dependence, and initiation; health care use; and consideration of both illicit (eg, heroin and nonmedical prescription pain reliever use) and licit prescription drug use. Variation in outcomes poses similar challenges for evaluation of mandatory access provision effectiveness.

Future research should concentrate on the effects of tertiary prevention policies on patient health outcomes. Studies assessing policies that influence access to opioid addiction treatment are of low rigor overall; however, initial evidence suggests that policies limiting access to methadone maintenance therapy may be associated with lower treatment use.78, 79, 81, 83 Future investigations should rigorously evaluate variation in state funding for medications used in the treatment of opioid dependence, state‐imposed Medicaid and private payor prohibitions on utilization management applied to medication‐assisted treatment formulations, and policies affecting buprenorphine waiver requirements. Evidence from two rigorous evaluations suggests that Good Samaritan laws may increase hospitalizations, especially for heroin‐related adverse health events, but do not influence opioid‐related mortality.29, 76 However, a controlled pre‐post evaluation by McClellan and colleagues (2018), published after our article review, observed that Good Samaritan laws were associated with reductions in opioid overdose deaths.98 We captured only one study evaluating the effect of naloxone access laws on opioid overdose deaths, which demonstrated decreases in non‐heroin opioid‐related mortality but not heroin‐related morality. The recent study by McClellan and colleagues also identified an association between naloxone access laws and reductions in opioid overdose deaths.98 Unlike the prior study, McClellan and colleagues did not disaggregate opioid overdose deaths by opioid type.98 Future research should further explore the effects of Good Samaritan and naloxone access laws on patient health.

Our review has two main limitations. First, we generally do not review evaluations of state programs not initiated by legislative or administrative actions. This limitation is particularly important when considering the small number of evaluations on naloxone access laws and anti‐doctor‐shopping policies. For example, previous research has identified a positive association between community‐implemented naloxone distribution programs and improved patient health outcomes, such as decreased overdose and increased recovery.99, 100 Further, model‐based studies provide additional evidence that increasing naloxone availability is associated with reductions in overdose mortality.101, 102 Research on anti‐doctor‐shopping programs suggests that these programs reduce multiple prescriber and pharmacy use but may have an unintended consequence of increasing circumvented opioids.103 Although it is beyond the scope of this review to evaluate these programs, they add to the evidence base on what governments can do to address opioid misuse and overdose.

Second, we limited our review to evaluations implemented by US states, thereby excluding relevant evaluations of policies enacted abroad from which the United States could glean insights. Specifically, a robust literature on syringe services programs, which provide sterile equipment to injection drug users, suggests that these policies reduce blood‐borne infections.104, 105, 106

Beyond these limitations, our synthesis suggests a need for future research at the state policy level. First, research should examine policies included (eg, Good Samaritan and naloxone access laws) and absent (eg, opioid prescription limits and state policies affecting opioid dependence treatment among criminal justice populations) from our review that have received insufficient attention. Second, studies on opioid prescribing and dispensing policies should take a holistic perspective regarding policy effects by investigating (or highlighting as a potential limitation) unintended consequences, such as changes in illicit opioid use, underprescribing and clinically inappropriate opioid therapy tapers or discontinuation, and suicide; and differential effects of policies by socioeconomic status, race, ethnicity, and criminal justice involvement. And third, research should evaluate the effects of all policies on patient health outcomes, specifically overdose.

Conclusions

Our scoping review reveals a growing rigorous literature on the effects of state opioid misuse prevention policies on patient and provider outcomes, but persistent gaps in evidence remain. The evidence now more clearly suggests that drug supply management policies and robust PDMPs with mandatory access provisions reduce multiple opioid prescribing and dispensing measures. Despite the increase in rigorous evaluations, the literature on most state opioid misuse prevention policies remains limited, particularly as they relate to patient health outcomes. We recommend future research examine policies that have received insufficient attention, investigate unintended consequences and differential effects across socioeconomic groups, and focus on patient health outcomes.

Supporting information

Online Appendix 1: Search Strategy

Online Appendix 2: Articles Included in Scoping Review

Online Appendix 3: Hierarchy of Types of Public Health Law Research Designs

Appendix 4: GRADE Summary of Findings

Appendix 5: Number of Studies Annually by Intervention Type