Globally, the number of people using prescription drugs for reasons other than what they were prescribed, sometimes combining them with other substances such as alcohol to sleep better or stimulants to perform better, has significantly increased. A team of computer scientists and emergency physicians from the American universities of Emory, Oregon and Pennsylvania used AI to analyze drug misuse and the emotions felt by users during consumption. The study titled “Large-Scale Social Media Analysis Reveals Emotions Associated with Nonmedical Prescription Drug Use” was published in the journal Health Data Science.
In 2021, more than 108,000 people died of overdoses in the United States, a figure up 20% from 2020, many of these deaths were caused by the ingestion of prescription drugs, often mixed with other substances. In France, more than 10,000 people die each year from the misuse of drugs.
Non-medical prescription drug use (NMPDU) primarily involves opioids, central nervous system stimulants, and benzodiazepines. Studies investigating its influence on mental health are based on surveys that have shown that NMPDUs involving opioids are strongly associated with psychiatric disorders. For example, an analysis of National Survey of Drug Use and Health (NSDUH) data found associations between opioid abuse and risk factors for suicide.
However, while studies have attempted to characterize the reasons for NMPDU, they do not address the emotional state of consumers. With recent research showing that social media can fill some of the gaps in these traditional survey-based studies, the team analyzed approximately 137 million posts from 87,718 Twitter users in terms of emotions, sentiments, concerns and possible reasons for NMPDU via natural language processing.
Large-scale analysis of twitter posts reveals emotions associated with non-medical prescription drug use
In this study, researchers used natural language processing (NLP) and machine learning approaches to study a large data set from Twitter on three common prescription drug categories and their combinations (opioids, benzodiazepines, stimulants and polysubstances (abuse of two or more different categories of NMPDUs at the same time, generally referred to as coingestion) to study and answer the following main research questions:
- How do the emotional contents expressed in the Twitter profiles of NMPDU groups differ from those expressed in the Twitter profiles of non-NMPDU groups (control group)?
- How do NMPDU tweets sentimentally differ from non-NMPDU tweets?
- How do the personal, social, biological and fundamental concerns expressed in the Twitter profiles of NMPDU groups differ from those expressed in the Twitter profiles of non-NMPDU groups?
They also used thematic modeling on NMPDU tweets to extract potential reasons for non-medical use of each drug category, and compared the distributions (of all the variables mentioned above) between men and women.
The conclusions of the study
These large-scale data analyzes show that there are substantial differences between the message texts of users who self-report an NMPDU on Twitter and the control group, as well as between men and women who report it.
Male users expressed higher anger and lower positivity, joy, anticipation, and sadness. In terms of social and personal content, compared to male users, female users shared more content related to social life (friends and family), health and personal concerns (at home). However, the differences were consistent between the different classes of drugs.
Although social media-based monitoring systems do not replace traditional systems, they can offer complementary insights. The knowledge acquired during this study could thus help to personalize awareness and intervention programs according to the targeted cohorts in order to mitigate the repercussions of the misuse of prescription drugs.
“Large-Scale Social Media Analysis Reveals Emotions Associated With Non-Medical Prescription Drug Use” Health Data Science,
Mohammad Ali Al-Garadi,1 Yuan Chi Yang,1Yuting Guo,2Sangmi Kim,3Jennifer S. Love,4 Jeanmarie Perron,5Abed Sarker.1.6
1Department of Biomedical Informatics, School of Medicine, Emory University, Atlanta, GA, USA
2Department of Computer Science, Emory University, Atlanta, Georgia, USA
3School of Nursing, Emory University, Atlanta, GA, USA
4Department of Emergency Medicine, School of Medicine, Oregon Health & Science University, Portland, OR, USA
5Department of Emergency Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA
6Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, Atlanta, GA, USA