{"paper_id":"24d6fdb4-4b2a-43fa-90e4-d9bfd5b17226","body_text":"Genetic analysis of ancients cannabis indicates the plant cannabis sativa was first cultivated for use as a as a medicinal agent up to 2400 years ago [ 1 ]. From the 1800’s, people in the United States (US), widely used cannabis as a medicine by either prescription or as an over the counter therapeutic [ 2 ]. Yet by the mid-20th century, cannabis use was prohibited in many parts of the developed world with the passing of legislation in the US, the United Kingdom (UK) and various European countries that proscribed its use [ 3 – 6 ]. Since the 2000s, the use of cannabis for medicinal purposes has been decriminalized in many countries including Israel, Canada, Netherlands, United States, United Kingdom, and Australia [ 7 – 9 ]. More recently, new evidence regarding the clinical effect of medical cannabis on the management of symptoms for some conditions [ 10 ] has triggered public interest in cannabis and cannabis-derived products [ 11 , 12 ], resulting in a global trend towards public acceptance, and subsequent legalisation of cannabis for both medicinal and non-medicinal use.\nThere is emerging evidence of cannabis efficacy for childhood epilepsy, spasticity, and neuropathic pain in multiple sclerosis, acquired immunodeficiency syndrome (AIDS) wasting syndrome, and cancer chemotherapy-induced nausea and vomiting [ 13 – 15 ]. Although researchers are investigating cannabis for treating cancer, psychiatric disorders [ 16 ], sleep disorders [ 17 ], chronic pain [ 18 ] and inflammatory conditions such as rheumatoid arthritis [ 19 ], there is currently insufficient evidence to support its clinical use. Scientific studies on emerging therapeutics typically exclude vulnerable populations such as pregnant women, young people, the elderly, patients with multimorbidity and polypharmacy, and this limits the availability of evidence for cannabis effectiveness across these population groups [ 20 ].\nCannabis as medicine is associated with a rapidly expanding industry [ 21 ]. Patient demand is increasing, as is reflected in an increasing number of approvals for prescriptions over time [ 22 ], with one study showing that 61% of Australian GPs surveyed reported one or more patient enquiries regarding medical cannabis [ 23 ]. With this increasing demand, is sophisticated marketing by medicinal cannabis companies that leverages evidence from a small number of studies to promote their products [ 24 , 25 ]. In light of this, concerns regarding patient safety is warranted especially when marketing for some cannabinoid products is associated with inadequate labelling and/or inappropriate dosage recommendations [ 26 ]. These concerns are compounded by the downscheduling of over-the-counter cannabis products which do not require a prescription [ 27 ] and the illicit drug market [ 28 ]. Given this dynamic interplay between marketing, product innovation, regulation, and consumer demand, innovative methods are required to augment existing established approaches to the surveillance and monitoring of emerging and unapproved drugs.\nAlthough there is considerable engagement from many stakeholders to improve the scientific evidence regarding the efficacy and safety of cannabis through randomised controlled trials, many gaps remain in the literature [ 29 ]. Yet data from real-world and patient reported data sources could provide opportunities to address this evidence deficit [ 30 ]. This real-world data can be captured from a variety of sources such as found in routinely collected health care and health services records that include but are not limited to patient generated data from medical, administrative and claims data, as well as patient reported data from surveys, wearable trackers, patient registries, and social media [ 31 – 33 ].\nPeople readily consult the internet when looking for and sharing health information [ 34 , 35 ]. According to 2017 survey of Health Information National Trends, almost 78% of US adults used online searches first to inquire about health or medical information [ 34 ]. Data resulting from these online activities is labelled ‘user generated’ and is increasingly becoming a component of surveillance systems in the health data domain [ 36 ]. Monitoring user-generated data on the web can be a timely and inexpensive way to generated population-level insights [ 37 ]. The collective experiences and opinions shared online are an easily accessible wide-ranging data source for tracking emerging trends–which might be unavailable or less noticeable by other surveillance systems.\nThe objective of this systematic scoping review is to understand the utility of online user generated text in providing insight into the use of cannabis as a medicine. In this review, we aim to systematically search published literature to examine the extent, range, and nature of research that utilises user-generated content to examine cannabis as a medicine [ 38 – 40 ]. The objective of this review is to synthesise primary research that uses social media discourse and internet search engine queries to answer the following questions:\nIn what way, is online user-generated text used as a data source in the investigation of cannabis as a medicine?\nWhat are the aims, data sources, methods, and research themes of studies using online user-generated text to discuss the medicinal use of cannabis?\n\nFor this review, we used an established methodological framework for scoping reviews to inform our methodology and we reported the review in accordance with the PRISMA reporting guidelines [ 38 – 41 ]. Literature database queries were developed for four categories of studies. The first three categories used social media text as a data source, the fourth relied on internet search engine query data. For the first category, the database queries combined words used to describe social media forums, and cannabis-related keywords and general medical-related keywords ( Table 1  Category 1). The second category also included the social media and cannabis-related keywords, but used keywords specific to psychiatric disorders, for which the use of medical cannabis has been described. Our search terms for this second category were informed by a systematic review of medicinal cannabis for psychiatric disorders [ 16 ] ( Table 1  Category 2). The third category included social media and cannabis related keywords but focused on non-psychiatric medical conditions for which cannabis is sometimes used ( Table 1  Category 3). The fourth category included studies using Internet search engine queries as a data source, there were no medical conditions included in these searches ( Table 1  Category 4). A manual search of MEDLINE: Web of Science (1900–2022), Embase: OVID (1974–2022), Web of Science: Core Collection (1900–2022), and Scopus (1996–2022) databases was conducted by SKH and CMH in May 2021 and again in October 2022. The search was limited to English-language studies that were published between January 1974 and April 2022 ( S1 Appendix ).\nThe inclusion criteria for this review were: (i) peer reviewed research studies, (ii) peer reviewed conference papers (iii) studies which used online user-generated text as a data source, and (iv) social media research that was either directly focused on cannabis and cannabis products that have an impact on health, or were health-related studies that found medicinal use of cannabis.\nExclusion criteria comprised: (i) editorials, letters, commentaries, surveys, protocols and book chapters; (ii) studies that used social media for recruiting participants; (iii) studies where the full text of the publication was not available; (iv) conference abstracts (iv) studies primarily focused on electronic nicotine delivery systems adapted to deliver cannabinoids; (v) studies that used bots or autonomous systems as the main data source and (vi) studies that focused exclusively on synthetic cannabis.\nAll studies captured by the search queries listed in  Table 1  were uploaded into excel to enable all duplicates to be removed. Following this all titles and abstracts were reviewed independently and in duplicate by CMH and SKH. Records were excluded based on title and abstract screening as well as publication type.\nThe full text articles that were identified for inclusion following screening process were then independently critiqued by pairs of reviewers using a checklist developed for this study. The purpose of the checklist that we developed for this systematic scoping review was to provide an overall assessment of quality rather than generate a specific score, ( S2 Appendix ). Assessments of quality in each study were based on evidence of relative quality in the aims or objectives, main findings, data collection method, analytic methods, data source, and evaluation and interpretations of the study. CMH and SKH critiqued all articles, and YB and MC each critiqued a selection of studies to ensure each article had been independently reviewed by two researchers. Where initial disagreement existed between reviewers regarding the inclusion of a study, team members met to discuss the disputed article’s status until consensus was achieved.\nAssessments of quality in each study were based on evaluating each study’s aims and objectives, main findings, data collection and analytic methods, data sources, and evaluation and interpretations of the results. Social media studies were included if there were no major biases affecting the internal, external or construct validity of the study [ 42 ]. In doing so, the internal validity of each study was determined by the quality of the data and analytic processes used, the external validity determined by the extent to which the findings can be generalised to other contexts, and the construct validity was ascertained by the extent to which the chosen measurement tool correctly measured what the study aimed to measure ( S3 Appendix ).\n\nOf the 1556 titles identified in the electronic database searches, 859 duplicate articles were removed, 450 were excluded following the screening of title and abstracts and 195 were excluded based on publication type (i.e., survey, letter, comment, abstract). This screening process provided fifty-two potentially relevant full text primary research studies to be included in the review ( Fig 1 ). Of these, five articles were not able to be retrieved, two out of forty-seven articles had initial disagreement. Upon consensus, five were excluded with reasons ( S3 Appendix ) using the quality assessment checklist as described above. This provided forty-two papers for inclusion in this systematic scoping review published between 2014 and 2022. Regarding publication type, the majority were journal articles 40/42 (95.2%), and two were conference-based publications 2/42 (4.8%). Although the first study was published eight years ago, nearly two-thirds 24/42 (57%), have been published over the last four years.  Table 2  provides a summary of each paper that includes author names, publication year, data source, duration of the study, number of collected posts, number of analysed posts, and the coding or labelling approach used.\nAdapted from: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. doi:  10.1136/bmj.n71 . For more information, visit:  http://www.prisma-statement.org/ .\nThe largest manually annotated dataset that contained 47,000 labelled tweets was published by Thompson et al. in 2015 [ 43 ]. This paper was one of 22 studies included in this review (52.4%) that either collected a limited number of data points, or sampled their collected data, and manually coded the data to gain an in-depth understanding of the domain [ 44 – 65 ]. Four of the 42 studies (9.5%) used existing meta-data including Google trends summary data [ 66 – 68 ], and geo-location data [ 69 ]. Two (4.8%) studies used data that was manually coded using crowdsourcing services [ 45 , 49 ], and two (4.8%) used automated coding supplied by a social media analytics company [ 70 , 71 ]. Fifteen of the forty-two studies (35.7%) used automated methods for labelling data, which included the use of machine learning, lexicon, and rule-based algorithms [ 60 , 72 – 85 ]. Automated coding was increasingly used as an analytic tool for social media data on this topic from 2017 onward ( Table 2 ).\nFor the studies that were manually labelled, analysis included the calculation of proportions and trends, and the development of repeating and emergent themes. [ 43 , 44 , 46 , 47 , 49 , 50 , 54 – 56 , 60 , 61 , 64 , 65 , 82 , 86 – 94 ]. Leas, Nobles et al. 2019, Shi, Brant et al. 2019, and Saposnik and Huber 2020, reported on Google trends data that delivered an index of Google search trends over time [ 66 – 68 ]. Daniulaityte, Nahass et al. 2015, processed Twitter data that contained existing geographical fields to identify geolocation at source [ 69 ]. The studies that utilised a large volume of data used advanced computational methods, which included sentiment analysis, topic modelling, and rule-based text mining [ 60 , 72 , 77 – 82 , 85 ]. The use of sentiment analysis in the [ 60 , 80 , 95 ] enabled the analysis of people’s sentiments, opinions, and attitudes. Topic modelling in the [ 74 , 76 , 79 , 80 , 85 ] studies enabled the development of themes via automatic machine learning methods. The use of rule-based text mining such as found in the [ 60 , 72 , 83 , 96 , 97 ] studies enabled the classification of posts into pre-existing health-related categories.\nIn this review, we categorized the forty-two research articles into six broad themes. Themes were based on the research questions motivating the studies, where each paper was classified as belonging a primary theme, based on alignment with the research aims ( Table 3 ).\nNine studies were included in the theme relating to cannabis-related conversations ( Table 4 ). The main keywords used in these studies included general terms such as ‘cannabis’, ‘marijuana’, ‘pot’, and ‘weed’. The major aim of these studies was to either identify topics of conversations regarding cannabis, or to examine the role of normative and valence information in the perception of medicinal cannabis. These studies are included because they reported on conversations around cannabis use for medical purposes, the valence associated with perceptions of health benefits of cannabis, and reports of adverse effects. For example, a study on veterans use of cannabis found that cannabis is used to self-medicate a number of health issues, including Post-Traumatic Stress Disorder (PTSD), anxiety and sleep disorder [ 94 ]. Seven of the studies used Twitter as a data source [ 43 , 73 , 78 , 87 , 96 , 98 , 99 ], one examined the content of YouTube videos about cannabis [ 100 ], one investigated online self-help forums [ 88 ] and another used Reddit data [ 94 ].\nSeven studies were included in the theme relating to the mode of use of cannabis as a medicine ( Table 5 ). These studies collected data using keywords such as ‘vape,’ ‘vaping’, ‘dabbing’, and ‘edibles’. Conversations around modes of use revealed a theme about lacking, seeking, or sharing knowledge about health consequences of the modes of use. Another theme was around the perceived health benefits of cannabis and the various modes of use of cannabinoids that included sleep improvement and relaxation resulting from dabbing oils [ 49 ] or consuming ‘edibles’ [ 50 ]. The findings suggest that for emerging modes of use such as dabbing, where the availability of evidence-based information is limited, people seek information from others’ experiences.\nSix studies were included in the theme relating to cannabis as a medicine for a specific health issue ( Table 6 ). These studies investigated conversations around the use of cannabis or cannabidiol (CBD) for a specific health issue. The health conditions included glaucoma [ 56 ], PTSD [ 72 ], cancer [ 65 , 101 ], Attention Deficit Hyperactivity Disorder (ADHD) [ 91 ], and pregnancy [ 92 ]. These studies mostly discovered that conversations claimed benefits of cannabis as an alternative treatment for these health conditions, although mentions of harm, and both harm and therapeutic effects, were also present [ 91 ].\nTwelve studies were included in the theme relating to cannabis as a medicine as part of discourse on illness and disease ( Table 7 ). In this theme, the research focus was on social media topics relating to management and treatment options for a range of health conditions rather than on medicinal cannabis per se. Health conditions discussed included inflammatory and irritable bowel disease [ 102 ], opioid use disorder [ 70 , 103 , 104 ], pain [ 71 ], ophthalmic disease [ 44 ], cluster headache and migraine [ 86 ], asthma [ 105 ], cancer [ 93 , 106 ], autism disorder [ 107 ], and brachial plexus injury [ 64 ].\nThere were seven studies in the cannabidiol category [ 61 , 80 , 82 , 89 , 90 , 108 , 109 ] ( Table 8 ). These studies concentrated on conversations related to the benefits of CBD products, product sentiment (positive, negative, or neutral), the factors that impact on a person’s decision to use CBD products, and the trends in therapeutic use of CBD.\nOne paper had a research question that explicitly focused on the detection of adverse events [ 75 ] ( Table 9 ) This study explored the prevalence of internet search engine queries relating to the topic of adverse reactions and cannabis use. Seven other studies contained mentions of adverse effects which were associated with cannabis use [ 47 , 49 , 51 , 53 , 55 , 62 , 77 , 81 ], however these papers were not included under this theme, as their research questions were not centered around the explicit investigation of adverse events.\n\nCurrently, there exist systematic reviews of cannabis and cannabinoids for medical use based on clinical efficacy outcomes from randomised controlled trials [ 18 ] and reviews on the use of social media for illicit drug surveillance [ 110 ]. However, following searches on PROSPERO and the databases listed above, to our knowledge, this paper constitutes the first systematic scoping review examining studies that used user-generated online text to understand the use of cannabis as a medicine in the global community.\nOur scoping review found that the use of social media and internet search queries to investigate cannabis as a medicine is a rapidly emerging area of research. Over half of the studies included in this review were published within the last four years (24, 57.1%), this reflects not only increase community interest in the therapeutic potential of cannabinoids, but also world-wide trends towards cannabis legalisation [ 4 , 7 – 9 , 111 , 112 ]. Regarding social media platforms, Twitter was the data source in eighteen (42.9%) of the forty-two studies, almost three and a half times the number of studies using Reddit (6, 14.3%) and just under three times the number of studies using data from Online forums (5, 11.9%). Three (7.14%) GoFundMe studies and three (7.14%) Google Trends studies were also included in the review. Hence, much of the data in this systematic review comprised posts from the Twitter platform. Several factors may explain this finding, firstly Twitter is real-time in nature, it has a high volume of messages, and it is publicly accessible. These factors makes it a useful data source for public health surveillance [ 113 ].\nRegarding the subjects of the studies, twelve (28.6%) focused on general user-generated content regarding the treatment of health conditions (glaucoma, autism, asthma, cancer, bowel disease, brachial plexus injury, cluster headaches, opioid disorder). These studies were either explicitly designed to investigate cannabis as a medicine or were studies that generated results that incidentally found cannabis mentioned as an alternative or complementary treatment (either formally prescribed or via self-medication).\nQualitative studies featured in the research, but while their contributions are valuable, especially in the context of hypothesis generation, they tend to be limited by their smaller datasets, which frequently comprise manually annotated samples. The recent emergence of powerful machine learning-based natural language processing (NLP) models suggests that it should be possible to automate the continuous processing of far larger datasets using NLP technologies, built upon the insights gained from initial qualitative studies, and even leveraging their annotated data for training purposes. Recent trends in the social science data landscape have shown a convergence between social science and computer science expertise, where the ability to use computational methods has greatly assisted the collection and validation of robust datasets that can form the basis of deeper social science research [ 114 ].\nWe found much heterogeneity in approaches applied to analyse user-related content, and inconsistent quality in the methodologies adopted. While we endeavored to include as many studies as possible, some of the publications initially identified as suitable for inclusion were not suitable based on a minimum quality requirements checklist ( S1 Appendix ). This checklist was designed to ensure that selection of data source, choice of platform, data acquisition and preparation, analysis and evaluation delivers data and conclusions that are appropriate for answering the research questions.\nThe utilisation of user-generated content for health research is subject to several inherent limitations which include; the lack of control that researchers have in relation to the credibility of information, the frequently unknown demographic characteristics and geographical location of individuals generating content, and the fact that social media users are not necessarily representative of the wider community [ 115 ]. Furthermore, the uniqueness, volume, and salience of social media data has implications that need to be considered when used for health information analysis [ 116 ]. Volume is usually inversely related to salience; a platform such as Twitter has a very high volume of information, much of which is not highly pertinent for the analysis of an effect, whereas the information contained in a blog will contain less volume, but will be more salient for analysis. Notwithstanding these limitations, user-generated content comprises large-scale data that provides access to the unprompted organic opinions and attitudes of cannabis users in their own words and is an effective medium through which to gauge public sentiment. To date, insights regarding cannabis as medicine have gained primarily through surveys or focus groups which have their own limitations regarding the format of data collection and potential bias in participant recruitment. A limitation of this scoping review was the lack of inclusion of a computational database such as  IEEE Xplore  in the search strategy, and the exclusion of the search terms ‘infodemiology’ and ‘infoveillance.’ Infodemiology and infoveillance studies explicitly use web-based data for research, and  IEEE Xplore  is a repository that contains technical papers and documents relating to computer science. However, our search was systematic, comprehensive and  IEEE Xplore  is Scopus-indexed, and we expect data loss to be minimal.\n\nOur systematic scoping review reflects a growing interest in the use of user-generated content for public health surveillance. It also demonstrates there is a need for the development of a systematic approach for evaluating the quality of social media studies and highlights the utility of automatic processing and computational methods (machine learning technologies) for large social media datasets. This systematic scoping review has shown that user-generated content as a data source for studying cannabis as a medicine provides another means to understand how cannabis is perceived and used in the community. As such, it is another potential ‘tool’ with which to engage in pharmacovigilance of, not only cannabis as a medicine, but also other novel therapeutics as they enter the market.\n\n(DOCX)\nClick here for additional data file.\n(DOCX)\nClick here for additional data file.\n(DOCX)\nClick here for additional data file.\n(DOCX)\nClick here for additional data file.\n(DOCX)\nClick here for additional data file.\n(XLSX)\nClick here for additional data file.","source_license":"CC-BY-4.0","license_restricted":false}