Which social media platforms facilitate monitoring the opioid crisis?

PLOS Digital Health, Journal Year: 2025, Volume and Issue: 4(4), P. e0000842 - e0000842

Published: April 28, 2025

Social media can provide real-time insight into trends in substance use, addiction, and recovery. Prior studies have used platforms such as Reddit X (formerly Twitter), but evolving policies around data access threatened these platforms’ usability research. We evaluate the potential of a broad set to detect emerging opioid use disorder overdose epidemic. From these, we identified 11 high-potential platforms, for which documented regulating drug-related discussion, accessibility, geolocatability, prior opioid-related studies. quantified their volume including informal language by slang generated using large model. Beyond most commonly X/Twitter, with high surveillance are TikTok, YouTube, Facebook. Leveraging variety social instead merely one, yields broader subpopulation representation safeguards against reduced any single platform.

Language: Английский

---

The Use of Natural Language Processing Methods in Reddit to Investigate Opioid Use: Scoping Review

JMIR Infodemiology, Journal Year: 2024, Volume and Issue: 4, P. e51156 - e51156

Published: Sept. 13, 2024

Background The growing availability of big data spontaneously generated by social media platforms allows us to leverage natural language processing (NLP) methods as valuable tools understand the opioid crisis. Objective We aimed how NLP has been applied Reddit (Reddit Inc) study use. Methods systematically searched for peer-reviewed studies and conference abstracts in PubMed, Scopus, PsycINFO, ACL Anthology, IEEE Xplore, Association Computing Machinery repositories up July 19, 2022. Inclusion criteria were investigating use, using techniques analyze textual corpora, source. specifically interested mapping studies’ overarching goals findings, methodologies software used, main limitations. Results In total, 30 included, which classified into 4 nonmutually exclusive goal categories: methodological (n=6, 20% studies), infodemiology (n=22, 73% infoveillance (n=7, 23% pharmacovigilance (n=3, 10% studies). used identify content relevant use among vast quantities data, establish potential relationships between patterns or profiles contextual factors comorbidities, anticipate individuals’ transitions different opioid-related subreddits, likely revealing progression through stages. Most an embedding technique (12/30, 40%), prediction classification approach topic modeling (9/30, 30%), sentiment analysis (6/30, 20%). most frequently programming languages Python (20/30, 67%) R (2/30, 7%). Among that reported limitations 67%), cited was uncertainty regarding whether redditors participating these forums representative people who opioids (8/20, 40%). papers very recent (28/30, 93%), from 2019 2022, with authors a range disciplines. Conclusions This scoping review identified wide variety applications support surveillance interventions addressing Despite clear enable identification opioid-relevant its analysis, there are limits degree interpretive meaning they can provide. Moreover, we need standardized ethical guidelines govern safeguard anonymity privacy forums.

Language: Английский

---

Using Natural Language Processing to Predict Fatal Drug Overdose From Autopsy Narrative Text: Algorithm Development and Validation Study

JMIR Public Health and Surveillance, Journal Year: 2023, Volume and Issue: 9, P. e45246 - e45246

Published: March 7, 2023

Background Fatal drug overdose surveillance informs prevention but is often delayed because of autopsy report processing and death certificate coding. Autopsy reports contain narrative text describing scene evidence medical history (similar to preliminary investigation reports) may serve as early data sources for identifying fatal overdoses. To facilitate timely reporting, natural language was applied texts from autopsies. Objective This study aimed develop a processing–based model that predicts the likelihood an describes accidental or undetermined overdose. Methods all manners (2019-2021) were obtained Tennessee Office State Chief Medical Examiner. The extracted (PDFs) using optical character recognition. Three common sections identified, concatenated, preprocessed (bag-of-words) term frequency–inverse document frequency scoring. Logistic regression, support vector machine (SVM), random forest, gradient boosted tree classifiers developed validated. Models trained calibrated autopsies 2019 2020 tested those 2021. Model discrimination evaluated area under receiver operating characteristic, precision, recall, F1-score, F2-score (prioritizes recall over precision). Calibration performed logistic regression (Platt scaling) Spiegelhalter z test. Shapley additive explanations values generated models compatible with this method. In post hoc subgroup analysis forest classifier, by forensic center, race, age, sex, education level. Results A total 17,342 (n=5934, 34.22% cases) used development validation. training set included 10,215 (n=3342, 32.72% cases), calibration 538 (n=183, 34.01% test 6589 (n=2409, 36.56% cases). vocabulary contained 4002 terms. All showed excellent performance (area characteristic ≥0.95, precision ≥0.94, ≥0.92, F1-score ≥0.92). SVM achieved highest F2-scores (0.948 0.947, respectively). (P=.95 P=.85, respectively), whereas miscalibrated (P=.03 P<.001, “Fentanyl” “accident” had values. Post analyses revealed lower centers D E. Lower observed American Indian, Asian, ≤14 years, ≥65 years subgroups, larger sample sizes are needed validate these findings. Conclusions classifier be suitable potential Further validation studies should conducted ensure detection overdoses across subgroups.

Language: Английский

---

Are treatment services ready for the use of big data analytics and artificial intelligence in managing opioid use disorder?: Viewpoint Paper (Preprint)

Published: March 22, 2024

In this viewpoint paper we explore the use of big data analytics and artificial intelligence (AI) in improving outcomes for people with opioid disorder (OUD) bring to table relevant challenges that must be addressed if new technologies are utilised ethically, effectively equitably. First, conceptualisation AI its relevance OUD treatment services. We then potential as well benefits leveraging enhance patient care keeping international standards OUD. Finally, lay out strategic operational principles which services need address maximize AI. These include greater algorithmic transparency, a framework clinician-technology interfacing, protections vulnerable situations people, adequate capture salient specific environments, resources respond analytical outputs, rebuilding respecting public trust institutions technology, tackling digital exclusion. Ultimately, effective AI-driven change requires full open engagement an system’s complexity, avoiding reductive approaches may discount existing organisational cultures or exaggerate unhelpful attitudes practices. hope, through paper, equip clinician policy maker engage implementation into

Language: Английский

---

Which Social Media Platforms Provide the Most Informative Data for Monitoring the Opioid Crisis?

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: July 7, 2024

Social media can provide real-time insight into trends in substance use, addiction, and recovery. Prior studies have used platforms such as Reddit X (formerly Twitter), but evolving policies around data access threatened these platforms' usability research. We evaluate the potential of a broad set to detect emerging opioid epidemic. From these, we created shortlist 11 platforms, for which documented official regulating drug-related discussion, accessibility, geolocatability, prior use opioid-related studies. quantified their volumes capturing informal language by including slang generated using large model. Beyond most commonly X, with high surveillance are TikTok, YouTube, Facebook. Leveraging many different social instead single platform, safeguards against sudden changes may better capture all populations that opioids than any platform.

Language: Английский

---

Are treatment services ready for the use of big data analytics and artificial intelligence in managing opioid use disorder? (Preprint)

Journal of Medical Internet Research, Journal Year: 2024, Volume and Issue: unknown

Published: March 22, 2024

In this viewpoint, we explore the use of big data analytics and artificial intelligence (AI) discuss important challenges to their ethical, effective, equitable within opioid disorder (OUD) treatment settings. Applying our collective experiences as OUD policy experts, 8 key that services must contend with make most these rapidly evolving technologies: algorithmic transparency, clinical validation, new practitioner-technology interfaces, capturing relevant improving patient care, understanding responding outputs, obtaining informed consent, navigating mistrust, addressing digital exclusion bias. Through paper, hope critically engage clinicians makers on ethical considerations, implications, implementation involved in AI deployment

Language: Английский

---

Exploring profile, effects and toxicity of novel synthetic opioids and classical opioids via Twitter: A qualitative study

Emerging Trends in Drugs Addictions and Health, Journal Year: 2023, Volume and Issue: 4, P. 100139 - 100139

Published: Dec. 13, 2023

Novel synthetic opioids' use has been increasing over the last decade and opioid epidemic attributed to 70% of drug-related deaths worldwide. Lately, Twitter become one key social media platforms where public express their unfiltered honest views opinions anonymously freely. This research comprised a qualitative study that explored motivations, effects toxicity novel opioids from perspectives Tweeters. Tweets were extracted using NVivo 12 Pro by Chrome NCapture thematic content analysis was applied. Extracted data relevant tweets coded into subthemes themes. Five main themes found related uses opioids; knowledge attitude, desired effects, adverse events, harm reduction strategies. For users reported about sources opioids, as well purity, addiction potential lethal effects. The included self-medication for recreational purposes. self-medications, sought against anxiety, depression, pain, overcoming previous addiction. However, events surpassed were: psychosis, addiction, withdrawal, respiratory depression Most linked rather than classical ones. provided valuable source information regarding modalities use, events. These findings benefit practitioners healthcare professionals dealing with users.

Language: Английский

---