# The quality and scope of health information on online drug platforms: a topic modelling and expert evaluation study of a Polish-language forum

**Authors:** Piotr Siuda, Paweł Matuszewski

PMC · DOI: 10.1186/s12954-026-01424-y · 2026-02-16

## TL;DR

This study analyzes health information quality on a Polish drug forum, finding moderate accuracy but significant gaps in safety guidance.

## Contribution

A novel custom evaluation method was developed to assess health information quality on an online drug platform.

## Key findings

- Posts showed moderate quality with scores averaging 35.67 out of 50.
- Information on dosage and consulting specialists was particularly lacking.
- Medical literature associations were mirrored in user discussions, suggesting some evidence-based knowledge.

## Abstract

Determining health information quality on online drug platforms is crucial for revealing and shaping substance use practices. The study aims to assess the scope and quality of health-related information on the largest Polish drug forum, Hyperreal. Ultimately, the goal was to explore how the case of Hyperreal illustrates user-driven knowledge construction around drug use and its potential implications for harm reduction strategies on similar online platforms.

A multimethod approach was employed for the current research. First, the full content of the forum was scraped, and topic modelling was used to pinpoint and analyse posts (N = 159,145) that discussed drugs in a health-related context. From this, 25 posts were selected for expert analysis using specified selection criteria for each topic. As existing standardized tools were found to be unsuitable for the aims of this study, the research team developed a custom questionnaire to evaluate the quality of posts, with psychiatrists serving as expert reviewers.

The topic distribution analysis indicated the most frequent and only sporadically discussed themes. Semantic analysis revealed distinct relationships between specific psychoactive substances and health topics, including associations between ketamine and depression/psychotherapy, marijuana and symptom relief and inflammation, amphetamines and cardiovascular issues, and fentanyl and palliative or oncological care. These patterns closely mirrored associations described in the medical literature, indicating knowledge structures grounded in medical rather than purely anecdotal knowledge. However, although the posts largely adhered to evidence-based medicine, the expert evaluation showed that they contained significant informational gaps. Overall, the mean quality of posts was M = 35.67, SD = 6.70, which indicates moderate health information quality. Experts rated factual accuracy and consistency with medical guidelines as moderate, while safety-related aspects—particularly dosage guidance and encouragement to consult specialists—received the lowest scores.

Online drug platforms are an important source of knowledge on psychoactive substances. However, the health information quality of posts was average, which is especially crucial as these forums shape discourses that replace professional narratives. The study is relevant in the context of harm reduction strategies and emphasizes the need to improve the quality of drug platforms’ content.

The online version contains supplementary material available at 10.1186/s12954-026-01424-y.

## Linked entities

- **Chemicals:** ketamine (PubChem CID 3821), fentanyl (PubChem CID 3345)
- **Diseases:** depression (MONDO:0002050), cancer (MONDO:0004992)

## Full-text entities

- **Diseases:** depression (MESH:D003866), pain (MESH:D010146), psychiatric (MESH:D001523), addiction (MESH:D019966), diabetes (MESH:D003920), vitiligo (MESH:D014820), cancer (MESH:D009369), hypertension (MESH:D006973), inflammation (MESH:D007249)
- **Chemicals:** paracetamol (MESH:D000082), amphetamines (MESH:D000662), amphetamine derivatives (-), heroin (MESH:D003932), amphetamine (MESH:D000661), levothyroxine (MESH:D013974), Xanax (MESH:D000525), fentanyl (MESH:D005283), CBD (MESH:D002185), omeprazole (MESH:D009853), ibuprofen (MESH:D007052), metronidazole (MESH:D008795), morphine (MESH:D009020), THC (MESH:D013759), oxycodone (MESH:D010098), metformin (MESH:D008687)
- **Species:** Cannabis sativa (species) [taxon 3483], Homo sapiens (human, species) [taxon 9606]

## Figures

2 figures with captions in the complete paper: https://tomesphere.com/paper/PMC13015130/full.md

---
Source: https://tomesphere.com/paper/PMC13015130