PROTOCOL: The effects of cannabis liberalization laws on health, safety, and socioeconomic outcomes: An evidence and gap map

Department of Criminal Justice and Criminology, Andrew Young School of Policy Studies, Georgia State University, Atlanta, Georgia, USA Sol Price School of Public Policy and Schaeffer Center for Health Policy & Economics, University of Southern California, Los Angeles, California, USA College of Education, University of Iowa, Iowa City, Iowa, USA Department of Political Science, Georgia State University, Atlanta, Georgia, USA


| The problem, condition, or issue
Cannabis is one of the most commonly used psychoactive substances globally, with an estimated 3.9% of the world's population aged 15-64 reporting past-year use (United Nations, 2020). Many world regions report higher annual prevalence rates (e.g., North America, 14.6%; Australasia, 10.6%; Western; and Central Europe, 7.8%), with certain countries in these regions documenting significant increases in cannabis use over the last decade (United Nations, 2020).
Cannabis dependence accounts for a small fraction (5.5%) of the overall global burden of disease attributable to alcohol and drugs, but this burden commonly surpasses that of amphetamines in world regions with high rates of cannabis use (Degenhardt et al., 2013).
Against this backdrop, many countries and US states have liberalized their cannabis laws over the past 25 years (Decorte et al., 2020). Between 1996 and year-end 2019, 33 US states plus the District of Columbia enacted medical marijuana laws granting authorized patients legal access to cannabis. 1 Moreover, since 2012, 11 states and the District of Columbia passed recreational marijuana laws legalizing adult use, including retail sales in nine states. These developments have led to a patchwork of state laws regulating access to cannabis through a variety of supply mechanisms, with state legislatures and citizen initiatives continuing to spur both new laws and amendments to existing laws (Chapman et al., 2016;Hoffmann & Weber, 2010;Klitzner et al., 2017;Williams et al., 2016). Indeed, voters in five US states will consider recreational or medical cannabis initiatives during the 2020 election. Dozens of other countries have also expanded legal access to cannabis under a variety of regulatory models, including decriminalization of home cultivation and the establishment of "cannabis social clubs" (Belackova et al., 2020;Decorte et al., 2017Decorte et al., , 2020Fischer et al., 2015;Rehm et al., 2019).
Perhaps most notably, Uruguay became the first country to legalize recreational cannabis in 2013 (Queirolo, 2020), followed by Canada in 2018 (Fischer et al., 2020).
The empirical literature examining the effects of cannabis laws and policies is interdisciplinary and diverse. New research appears almost weekly, with studies examining a wide range of health, safety, and socioeconomic outcomes. Health outcomes measure physical and mental well-being or disease including cardiovascular disease (Abouk & Adams, 2018), opioid overdose (Chan et al., 2020), and suicide (Anderson et al., 2014;Chan et al., 2020). Safety outcomes measure security or risk of harm including crime (Morris et al., 2014), impaired driving (Sevigny, 2018), and vehicular accidents (Salomonsen-Sautel et al., 2014). Socioeconomic outcomes capture social and economic metrics including property values (Burkhardt & Flyr, 2019), labor (Plunk et al., 2016). To date, no concerted efforts have fully scoped this varied and growing literature. Consequently, stakeholders have an incomplete understanding of the effects of cannabis liberalization laws.
This EGM aims to fill this gap by collecting and summarizing the available evidence for decision-makers and identifying opportunities for future knowledge generation.

| Scope of the EGM
This EGM summarizes research investigating the effects of cannabis liberalization reforms on health, safety, and socioeconomic outcomes internationally. This purview encompasses statutory and regulatory provisions governing legal access to cannabis among both specific (e.g., patients with certain medical conditions) and general (e.g., adults 21 and older) populations. For outcomes, the EGM will capture the array of health, safety, and socioeconomic outcomes reported in the literature.

| Conceptual framework of the EGM
This EGM will document how cannabis law reforms affect health, safety, and socioeconomic outcomes across different target populations (e.g., patients, adults). The conceptual framework presented in Figure 1, which draws from legal epidemiology (Burris et al., 2013(Burris et al., , 2016, depicts a causal chain between cannabis laws and associated outcomes.
Cannabis laws reflect the "law on the books" as determined by statutes and regulations. Assessing policy impacts also requires understanding implementation, or the "law on the streets." Path A captures these on-the-ground legal practices that may not be reflected in statute but still impact outcomes in meaningful ways.
Paths B and C depict the effects of laws and legal practices on markets/environments (e.g., cannabis prices and potency, drug arrests) and attitudes/behaviors (e.g., frequency of cannabis use, modes of consumption). Bilateral Path D recognizes that markets/environments and attitudes/behaviors may influence each other, such as the plausible inverse association between cannabis potency and frequency of use due to longer-lasting periods of intoxication. Finally, paths E and F capture population-level health, safety, and socioeconomic impacts as a consequence of intermediate changes in markets/environments and attitudes/behaviors. Although understanding the direct effects of law on health, safety, and socioeconomic outcomes is often a primary concern, highlighting these links informs study inclusion and coding decisions.
The framework differentiates studies examining, for instance, legal allowances for recreational dispensaries (cannabis law) from those examining geographic variation in recreational dispensary density (legal practice). The conceptual framework allows coders to capture these important differences in the EGM. Likewise, intermediate market/environment and attitude/behavior outcomes are conceptualized along the causal chain in the framework, as they can enhance or mitigate health, safety, and socioeconomic consequences, and are thus worthy of documentation and study in their own right. This is particularly true when insufficient time has passed to allow a complete and valid assessment of longer-term outcomes (e.g., cardiovascular or respiratory health). In short, the conceptual framework guides our understanding of where in the causal chain each study's interventions and outcomes are situated, and how they might be incorporated into the EGM.

| Why it is important to develop the EGM
Despite a quarter-century or more of cannabis liberalization reforms across the globe, no EGM exists to summarize the available evidence and knowledge gaps in this policy space. The development of this EGM will fill this need. Importantly, it will document the heterogeneity in target populations (Pacula & Smart, 2017), variation in policy design and implementation (Klieger et al., 2017;Klitzner et al., 2017), and diversity of outcomes across multiple sectors (Fischer et al., 2018;Maslov et al., 2016). Moreover, current research increasingly employs complex methods and statistical models (Choo & Emery, 2017;Hunt & Miles, 2015). Mapping this knowledge base is therefore needed to (i) provide decision-makers with a catalogue of quality evidence on cannabis liberalization reforms and (ii) identify knowledge gaps for prioritizing new primary research and systematic reviews.

| Existing EGMs and/or relevant systematic reviews
To our knowledge, this review will provide the first comprehensive EGM of cannabis liberalization laws and associated policy outcomes.
While there are now several published systematic narrative reviews (Hunt & Miles, 2015;Leung et al., 2018;Smart & Pacula, 2019;Vyas et al., 2018) and meta-analyses (Melchior et al., 2019;Sarvet et al., 2018) in this policy space, they primarily assess the effects of medical and recreational cannabis laws on drug use outcomes among US youth and adults. Researchers have examined a much broader and arguably more salient set of policy outcomes than just cannabis use (Sznitman & Zolotov, 2015;Waddell & Wilson, 2017). Recent conceptual mappings across varied national contexts have identified dozens of performance indicators in the public health, community safety, economic, and children and youth sectors that are relevant for F I G U R E 1 Conceptual framework for understanding the effects of cannabis liberalization laws on health, safety, and socioeconomic outcomes understanding the full scope of cannabis policy impacts (Campeny et al., 2020;Fischer et al., 2018;Maslov et al., 2016).

| OBJECTIVES
The proposed EGM will seek to answer several questions. What is the extant evidence-base on cannabis liberalization policies, and how does this evidence vary across policy types (i.e., medicalization, legalization, decriminalization)? What outcomes are predominant in the available literature? What gaps exist in the literature, and are there clusters of studies that can support future systematic reviews and meta-analyses? How do studies vary by target population, geographic focus, and analytic methods? Along these lines, the objectives of this EGM are to: 1. Develop a comprehensive intervention-outcome framework of cannabis liberalization laws and related health, safety, and socioeconomic outcomes.
2. Map relevant systematic reviews and primary studies within this framework.
3. Summarize the intervention, outcomes, context, study design, study quality, and main findings of included studies.  An EGM is a systematic review tool that provides a visual summary of strength of evidence and knowledge gaps within a particular policy domain (Miake-Lye et al., 2016;O'Leary et al., 2017). EGMs are a recent addition to the systematic reviewer's toolkit (Da Silva et al., 2017;Saran & White, 2018). This EGM will be developed and populated according to the following steps: (i) develop a conceptually-driven intervention-outcome framework, (ii) define study inclusion criteria and establish literature search strategy following established guidelines, (iii) devise a coding instrument for data extraction and critical appraisal of studies, including brief summaries of main findings. Findings will be presented in a graphical and/or tabular formats to better inform the current status of evidence and inquiry, including assessments of study quality and gaps in knowledge (Lum et al., 2011;Snilstveit et al., 2016).

| Population
For this EGM, populations will be defined by subgroups that are reflected in the primary research and provide meaningful cross-study aggregations. Target populations vary with the type of cannabis law.
For instance, recreational cannabis laws apply to the general adult population (typically 21+), whereas medical cannabis laws apply more narrowly to individuals diagnosed with certain medical conditions (e.g., epilepsy, HIV, chronic pain). Because research in this area is often concerned with unintended consequences, such as illicit diversion of cannabis to youth, study populations are generally more encompassing than those defined by statute. Researchers in this area often account for other types of population heterogeneity, including age, gender, race/ethnicity, riskiness of use (e.g., heavy vs. casual users), and criminal justice involvement (Pacula & Smart, 2017).
These various populations will serve as filters in the interactive EGM.

| Interventions
Interventions are a primary dimension of the EGM. Table 1 presents the specific cannabis liberalization policies covered by this EGM. The top level of the taxonomy captures the major type of cannabis law, including both comprehensive and limited (i.e., CBD/low-THC) medical cannabis laws, recreational (i.e., adult access) cannabis laws, and decriminalization and industrial hemp laws that establish a licit supply of consumable cannabis. Regulatory domains and specific policy provisions in the taxonomy are drawn from previous legal mappings in this policy space (Bestrashniy & Winters, 2015;Chapman et al., 2016;Klieger et al., 2017;Klitzner et al., 2017;Pacula et al., 2014;Williams et al., 2016). Note that listed provisions are not relevant to each major law, as regulations concerning qualifying medical conditions, for instance, apply only to medical cannabis laws.

| Comparison
Typically, studies compare cannabis liberalization policies to status quo cannabis prohibition (i.e., "business as usual"). In some cases, the comparator may be an earlier adopted cannabis law. For example, a study may report the effect of recently adopted recreational cannabis laws against previously enacted medical cannabis laws. We will code these comparators to serve as filters in the interactive EGM.

| Outcomes
Outcomes are the second primary dimension of the EGM, which include intermediate (changes in markets and environments, changes in attitudes and behaviors) and final outcomes (health, safety, socioeconomic). Broadly, health outcomes measure physical and mental well-being or disease, safety outcomes measure security or risk of harm, and socioeconomic outcomes capture community and fiscal impacts. Intermediate outcomes reflect changes in markets and environments (e.g., cannabis prices and potency, advertising, and marketing) and changes in attitudes and behaviors (e.g., attitudes toward marijuana, modes of consumption such as vaping), encompassing factors that shift the structural conditions and incentives impacting final outcomes. Table 2 identifies candidate intermediate and final outcomes to be coded. The listed outcomes draw from existing conceptual mappings of potential cannabis policy impacts (e.g., Campeny et al., 2020;Fischer et al., 2018;Maslov et al., 2016). The list will be updated based on studies included in the EGM. Along with expected positive impacts of these laws (e.g., reduced criminal justice costs, greater tax revenues), the purview of the EGM also encompasses the range of potential adverse or unintended outcomes (e.g., drugged driving and explosions/fires related to cannabis production).

| Interventions and outcomes
Interventions are eligible if they increase legal access to cannabis supply. This includes recreational cannabis laws and both comprehensive and limited product (i.e., CBD/low-THC) medical cannabis laws. Only cannabis decriminalization policies that remove criminal penalties for small-scale cannabis cultivation are eligible because they establish a legitimate (or tolerated) supply of cannabis. Lastly, we do not explicitly include industrial hemp laws that regulate and incentivize cannabis production for textiles, biofuel, and other commercial uses. However, these laws can open loopholes for sales of cannabis products meant ostensibly, albeit not legally, for human consumption (e.g., Carrieri et al., 2019). Consequently, we will include studies that examine whether the sale of cannabis products for human consumption increases as a result of these laws.
Some primary studies may investigate cannabis liberalization policies in various policy combinations (e.g., states with both medical and recreational cannabis laws). If the study combines eligible interventions, it will be included in the EGM and identified as a combined policy to avoid misleading comparisons with individually assessed policies. If the combined intervention includes any ineligible intervention (e.g., decriminalization of cannabis use), it will be excluded from the EGM.
In accordance with Figure 1, the EGM aims to capture all intermediate (markets/environments and attitudes/behaviors) and final (health, safety, and socioeconomic) outcomes, including those characterized as adverse. Thus, within this framework, we place no a priori exclusions on specific outcomes. Our purview is purposely broad in scope.

| Types of study designs
Eligible studies employ quasi-experimental designs that control for confounding factors, have a relevant comparator, and/or assess  (Rockers et al., 2015). This EGM focuses on these types of observational designs because they have relatively high internal validity. Lastly, RCTs are eligible, although we do not anticipate finding this design as cannabis policies cannot be feasibly randomized. Studies that examine hypothetical policy scenarios using simulation or forecasting designs will not be included in the EGM, as we restrict our purview to studies analyzing real-world data.

| Treatment of qualitative research
We do not include qualitative research in the EGM.

| Types of settings
Eligible settings include national or subnational jurisdictions that have legalized the supply of cannabis for medical or recreational purposes.
There are no geographic or timeframe restrictions on eligibility.

| Status of studies
Completed and ongoing research will be included in the EGM. For completed research, published and unpublished (e.g., dissertations, working papers, reports) studies will be included. Working papers or preliminary studies that are subsequently published will updated in the EGM as they become available.

| Academic databases
We will search the following indexed bibliographic databases and systematic review registries for eligible studies:

| Gray literature databases
Additionally, we will search the following gray literature databases for eligible studies: T A B L E 3 Search terms and boolean logic Domain Natural language a Controlled vocabulary 1. Cannabis cannabis OR marijuana OR hemp OR "cannabis" OR "cannabis edibles" OR "hemp" OR "marijuana" OR "medical marijuana" AND 2. Policy Change commercializ* OR "cannabis social club*" OR cultivat* OR decrim* OR dispensar* OR grow* OR industr* OR law* OR legal* OR legislat* OR liberaliz* OR medical* OR polic* OR program* OR recreation* OR regulat* OR retail* OR "commercialization" OR "drug control" OR "drug laws & regulations" OR "drug legalization" OR "government policy" OR "government regulation" OR "government regulations" OR "law" & "legalization" OR "legislation" OR "legislation, drug" OR "legislation, medical" OR "liberalization" OR "marijuana dispensaries" OR "marijuana growing" OR "marijuana industry" OR "marijuana laws" OR "marijuana legalization" OR "medical marijuana -law & legislation" OR "medicalization" OR "public policy" OR "public policy (law)" OR "regulations" OR "retail stores" OR "state laws" OR "state regulation" AND 3. Quantitative Methods analysis OR "before-and-after" OR causal OR counterfactual OR "difference-in-difference*" OR experiment* OR "fixed effect*" OR "instrumental variable*" OR longitudinal OR matching OR metaanaly* OR model* OR "panel data" OR "propensity score" OR "propensity match*" OR quasi-experiment* OR regress* OR "regression discontinuity" OR "repeated measures" OR RCT OR "synthetic control" OR "systematic review" OR "time-series" OR "causal modeling" OR "causal models" OR "controlled before-after studies" OR "controlled before and after studies" OR "data analysis" OR "data analysis, statistical" OR "econometric models" OR "fixed effects model" OR "granger causality test" OR "instrumental variables (statistics)" OR "instrumental variable estimation" OR "interrupted time series analysis" OR "longitudinal method" OR "longitudinal studies" OR "meta-analysis" OR "panel analysis" OR "panel data" OR "policy analysis" OR "population-based case control" OR "pretests posttests" OR "pretest-posttest design" OR "propensity score" OR "propensity score matching" OR "statistical analysis" OR "statistical matching" OR "quantitative research" OR "quasi-experimental studies" OR "regression" OR "regression analysis" OR "regression discontinuity design" OR "repeated measures" OR "repeated measures design" OR "time series" OR "time series analysis" a Listed terms will be searched in title, abstract, and keyword fields.

| Contacting researchers and other supplementary approaches
We will distribute copies of the preliminary EGM within our professional networks and to authors of included EGM studies to seek additional studies of potential relevance. Study authors will also search their personal bibliographic libraries for relevant studies.
Finally, we will perform backward and forward citation searching of included studies.

| Screening and selection of studies
Retrieved studies will be independently screened at both titleabstract and full-text levels by two members of the study team.
Specific screening questions for title and abstract review include the following: If both screeners answer "yes" to all three questions, then the study moves to full-text screening. Screening conflicts will be resolved by reaching consensus through arbitration with a third member of the study team.
At the full-text screening stage, studies will be read and reviewed to ensure they clearly report (quasi)experimental results examining the effect of cannabis liberalization policies on relevant intermediate and final outcomes. Studies that do not meet these criteria or are published in other languages will be excluded at this stage.

| Data extraction, coding, and management
Full bibliographic information along with key population, intervention, comparison, and outcome dimensions will be extracted for each study.
We will also code additional study characteristics, including country, data years, publication type, unit of analysis, study design, and study quality. A draft coding guide is included in Appendix A. The PI will facilitate quality control and pilot data extraction. Eligible studies will be independently coded by two members of the study team, with any conflicts arbitrated and resolved by a third member of the study team.

| Quality appraisal
Primary studies and systematic reviews will be assessed for quality using appropriate instruments. Current guidance to researchers is to choose a tool that will provide meaningful appraisals of bias based on included study designs (Quigley et al., 2019). Thus, we will finalize the critical appraisal tool post-screening. Preliminarily, we plan to assess risk of bias for primary studies using either ROBINS-I (Sterne et al., 2016), RoBANS (Kim et al., 2013), or SMS (Sherman et al., 2002) for observational studies and RoB 2 for RCTs (Sterne et al., 2019). For systematic reviews, we will assess the risk of bias using AMSTAR 2 (Shea et al., 2017). All risk of bias instruments will be independently coded by two members of the study team, with any conflicts arbitrated and resolved by a third member of the study team.

| Unit of analysis
The unit of analysis is a systematic review or primary study, represented as a single entry within each cell of the EGM. When multiple primary studies are reported from the same research project or underlying data, these studies will be entered independently into the EGM. When there are multiple reports of a single study, the reports will be combined for presentation in the EGM.

| Planned analyses
The EGM report will provide descriptive graphical and/or tabular summaries of included studies, accompanied by narrative descriptions of the evidence. The report structure will include background, methods, results, and discussion sections. SEVIGNY ET AL.

| 7 of 12
Descriptively, in addition to presenting a conceptual framework and PRISMA flow diagram, an online interactive will be presented with two main dimensions: interventions and outcomes. Additional map dimensions will include the number of primary studies indicated by relative bubble size within the cells of the map, color coding of data points to reflect critical appraisal of study quality, and filtering by key study characteristics. Static maps presented in a report or article will present this information as a basic intervention-outcome matrix, with additional tables and graphs such as bar and pie charts describing characteristics of included studies. Key factors include the following: • Intervention and specific policy provisions

| Presentation
Cannabis liberalization policies and associated health, safety, and economic outcomes are the primary dimensions of the EGM. The following characteristics will be coded from primary studies to allow the map to be interactively filtered: population subgroup, country, data years, study design, comparator, unit of analysis, and risk of bias.

| STAKEHOLDER ENGAGEMENT
The conceptual framework and intervention-outcome categories were developed through a consultative process. Specifically, we solicited feedback from selected researchers and other stake-

ROLES AND RESPONSIBILITIES
The three key competency areas for EGM development (content, methods, and information retrieval) will be managed by the research team as explained below. at Georgia State University will provide information retrieval advice and expertise. Danye N. Medhin and Jared Greathouse will retrieve, screen, and code studies under the supervision of Eric L. Sevigny.

SOURCES OF SUPPORT
The proposed EGM is supported by funding from the National Institute on Drug Abuse (NIH Award No.: R03DA046806).