Social media platforms are a valuable source of signals for public health surveillance, enabling early detection of threats, often before official reports are available. However, dependency on a single platform creates a critical vulnerability: when data access is lost, so is surveillance capacity. This is what happened when Twitter/X restricted its API.

This hands-on workshop presents episomer, a new free, open-source R-based tool developed by ECDC in response to this challenge in 2026. Episomer currently monitors BlueSky and is designed to be extended to other social media platforms, making it resilient to future changes in data accessibility.

The workshop opens with an introduction covering the rationale for social media surveillance in public health surveillance, the epitweetr story, and the systematic review of data accessibility that informed the development of episomer. This is followed by an overview of episomer's architecture, key features and functionalities. The core of the session is a hands-on exercise in which participants explore episomer's dashboard, geolocation, and signal detection functionalities, supported by facilitators. The session closes with a plenary discussion in which participants share observations and potential use cases from their own institutions, followed by a wrap-up on next steps and future developments.

Participants will be able to use episomer after the workshop for their specific use cases since it is a free, open-source tool, and it only requires social media credentials added from the user for data collection.

No technical skills are required. Participants will receive installation instructions in advance of the workshop.

Social media carries large volumes of first-person accounts of vaccine experience alongside circulating misinformation. Both are relevant to public health, and both are difficult to interpret at scale.

This interactive session presents VaxPulse, a social media monitoring platform for vaccine safety and misinformation developed at the Murdoch Children's Research Institute Centre for Health Analytics. Data is collected across multiple platforms and feeds two analysis arms. The first detects vaccine misinformation, including the boundary between misinformation, legitimate safety concern and vaccine hesitancy. The second detects potential adverse events following immunisation using large language models, through identification of personal health mentions, extraction of reported reactions, and normalisation to coded medical terms (SNOMED CT / MedDRA).

Three interactive segments are planned. Participants first triage posts spanning clear falsehood, alarming but legitimate concern, and hesitancy, deciding which warrant action, monitoring or neither. Second, participants classify real posts and extract reported reactions, comparing their classifications with model outputs; disagreement between participants illustrates why inter-rater agreement matters before any model is assessed. Finally, the session closes with a clustered set of reports and the question of whether it constitutes a signal, alongside the governance constraints encountered in this work, including institutional ethics review, privacy impact assessment and platform terms restricting inference of health signals from user content.

This aligns with the conference themes of Novel applications in public health and Privacy & confidentiality, and with Maximising Benefits, Minimising Harms through its focus on both the capabilities and limits of automated monitoring.