4-Fluoramphetamine in the Netherlands: Text-mining and sentiment analysis of internet forums

Matthijs Blankers; Daan van der Gouwe; Margriet van Laar

doi:10.1016/j.drugpo.2018.11.016

4-Fluoramphetamine in the Netherlands: Text-mining and sentiment analysis of internet forums

Int J Drug Policy. 2019 Feb:64:34-39. doi: 10.1016/j.drugpo.2018.11.016. Epub 2018 Dec 12.

Authors

Matthijs Blankers¹, Daan van der Gouwe², Margriet van Laar²

Affiliations

¹ Trimbos Institute, The Netherlands Institute of Mental Health and Addiction, Da Costakade 45, 3521VS, Utrecht, the Netherlands; Department of Research, Arkin Mental Health Care, Klaprozenweg 111, 1033NN, Amsterdam, the Netherlands; Amsterdam UMC, Location AMC, Department of Psychiatry, University of Amsterdam, Meibergdreef 9, 1105AZ, Amsterdam, the Netherlands. Electronic address: mblankers@trimbos.nl.
² Trimbos Institute, The Netherlands Institute of Mental Health and Addiction, Da Costakade 45, 3521VS, Utrecht, the Netherlands.

PMID: 30551004
DOI: 10.1016/j.drugpo.2018.11.016

Abstract

Background: Users of new psychoactive substances including 4-fluoroamphetamine (4-FA/4-FMP) frequently share their experiences or opinions in online drug forums. We have tested the potential of computerised analysis of drug users' forum posts for monitoring and early detection of trends. Specifically, we tested whether changes in the volume of 4-FA related posts and sentiments expressed in those posts can be observed around the time 4-FA was increasingly reported by Dutch drug monitoring sources (2012-2017).

Methods: Opening posts from two popular Dutch internet-based drug discussion forums, written between January 1 st, 2012 and January 1 st, 2018 were scraped: Portions of the forum posts about 4-FA were collected. To contrast 4-FA findings against other categories of forum posts, we also collected posts on two other substances (ecstasy and cocaine) and posts not related to a specific substance. Sentiments expressed in these posts were inferred using text recognition software, and analysed for trends using linear mixed modelling.

Results: The number of 4-FA posts increased between 2012 and 2015: 76 posts in 2012, 138 in 2013, 322 in 2014, 323 in 2015, and decreased thereafter: 264 in 2016 and 135 in 2017; X²(5) = 271.8, p < .001. Over time, a decrease in positive sentiment towards 4-FA can be observed starting in 2015, compared to the period before 2015, coinciding with more news searches and reports on adverse events related to 4-FA use. Linear mixed modelling analysis confirmed a significantly higher sentiment score in 2015 compared to 2017 for 4-FA, B = 0.062; SE = 0.023; t(1252) = 2.70; p = 0.007, but not for posts on other substances.

Conclusion: Changes in the volume and sentiments of forum posts coincided with news media exposure related to 4-FA and with trends observed by established drug monitoring sources. Hence, internet forum monitoring facilitates early discovery of trends in the popularity, prevalence and adverse events related to new psychoactive substances.

Keywords: 4-Fluoramphetamine; Internet-based drug forums; New psychoactive substances; Sentiment analysis; Text mining; Trend analysis.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Amphetamines / administration & dosage*
Data Mining
Drug Users
Drug-Related Side Effects and Adverse Reactions
Humans
Internet
Netherlands / epidemiology
Pharmacovigilance
Prevalence
Psychotropic Drugs / administration & dosage*
Social Media

Substances

Amphetamines
Psychotropic Drugs
4-fluoroamphetamine