Search strategy and data sources
This systematic review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) Statement (Page et al. 2021). Studies were identified by searching four electronic databases: MEDLINE (Ovid), EMBASE (Ovid), PsycINFO (EBSCOhost), and Transport Research International Documentation. All databases were initially searched from their inception dates until May 31 2022 and the search was later updated to include studies published up to February 26, 2024. Key terms relating to cannabis use (e.g., cannabis OR marijuana OR THC), attitudes (e.g., attitude* OR belie* OR opinion*), and driving (e.g., driv* OR motor vehicle OR DACU) were combined. The keyword search terms were adapted for each database (see Additional File 1). All articles were uploaded to Covidence, (Covidence 2023) an online platform that streamlines the production of systematic reviews and allows for asynchronous collaboration among reviewers. Two reviewers (B.B. and Y.Y.) independently screened all studies by title and abstract. Full text of the retained studies was independently reviewed by two reviewers (B.B. and D.A.). Disagreements in both stages were resolved by consensus or involving a fourth reviewer (J.B.). The protocol was pre-registered on PROSPERO (CRD42022337260).
Eligibility criteria
Studies that met the following criteria were included: (i) report attitudes or changes in attitudes toward cannabis use and driving, (ii) report qualitative, quantitative, or mixed methods empirical research, (iii) use primary research data, (iv) English-language, and (v) full-text available. Studies reporting secondary research data were excluded but used to identify the original study. Studies that met the following criteria were excluded: (i) attitudes or changes in attitudes toward cannabis use and driving not reported, (ii) narrative non-research reports (e.g., commentaries, position statements), (iii) not in English, (iv) full text not available. Both academic and gray literature (e.g., reports, policy literature) were included. As this review aimed to synthesize all available data on attitudes toward DACU, no studies were excluded based on study design or sample characteristics. Where two or more studies used the same data-set, the report with the largest sample size was retained.
Data extraction and quality assessment
Data from eligible studies were collected in a pre-developed data extraction form by two reviewers (B.B. and D.A.) using Microsoft Excel. Extracted data included: (i) title, (ii) author(s), (iii) year of publication, (iv) location from which subjects were sampled, (v) aim(s) and hypothesis(es), (vi) sample characteristics, including sample size, age, sex/gender, cannabis use history, and inclusion/exclusion criteria, (vii) data collection date(s), (viii) study design/data collection method(s), (ix) outcome measures, (x) data analysis method(s), (xi) control for confounding variables, and (xii) key findings. Both reviewers performed extraction for 10 studies and cross-checked the extracted data to assess agreement. Each reviewer then extracted data for approximately half of the remaining 60 studies.
Methodological quality and risk of bias was assessed by two reviewers (B.B. and D.A.) using the Standard Quality Assessment Criteria for Evaluating Primary Research Papers (SQAC) (Kmet et al. 2004). The SQAC comprises separate checklists for quantitative and qualitative studies; both checklists were used to assess mixed-method studies. A conservative inclusion threshold of 75% was used (Kmet et al. 2004) (no studies were excluded on this basis). Additional File 2 provides a summary of the included studies.
Data synthesis
An inductive thematic synthesis approach (Ryan et al. 2018) was used to identify, analyze, and report patterns across findings from included studies. First, two reviewers (B.B. and D.A.) read the original texts, adding to a bank of free codes as necessary. Decisions to combine, rename, or eliminate these initial codes were reached through discussion. This process yielded 48 initial codes (see Additional File 3). Each code was entered as a row in a table and populated with data from relevant studies. Studies reporting any finding related to the code were added to the table, including those reporting non-significant results. Reviewers then grouped initial codes into a hierarchical structure by identifying similarities between the codes. Analytic themes were generated by synthesizing extracted findings in relation to the research question. Both reviewers reviewed themes and assigned theme names.
