JMIR Publications

ABSTRACT

Background:

Cervical cancer is largely preventable, yet it caused about 604,000 new cases and 280,000 deaths worldwide in 2024. Fear can motivate or deter screening, and fear of cancer differs from fear of the exam itself. Online health communities record these concerns in users’ own words, which makes them a practical source of insight for screening programs.

Objective:

This study mapped the topics of cervical cancer screening discussion in peer online health communities, how those topics connect, and where negative emotion and fear concentrate.

Methods:

We analyzed 9851 posts and comments from 7 women’s health subreddits posted between August 1, 2025, and January 31, 2026. Latent Dirichlet allocation and the Analysis of Topic Model Networks procedure identified topics, clusters, and bridging topics. A transformer sentiment model classified each document as negative, neutral, or positive; we validated it, a second transformer model, and the VADER (Valence Aware Dictionary and sEntiment Reasoner) lexicon against 2 coders’ consensus coding of 209 documents. We counted explicit fear terms by topic and their co-occurrence with procedure and disease terms.

Results:

The model yielded 13 substantive topics that formed 3 clusters (modularity 0.30): diagnostic information, the embodied experience of the exam, and personal narrative with peer support. Timing and bleeding, the least prevalent topic (3.5%), had the highest betweenness centrality (0.26) across network thresholds. The transformer model agreed with human coding more closely (macro F1=0.61) than VADER (macro F1=0.41), which labeled 40% of human-coded negative documents as positive. By the transformer model, 36.8% of documents were negative and 19.6% positive, and encouragement and support was the only topic with more positive than negative documents. Fear terms appeared in 3.5% of documents on human papillomavirus (HPV) and cancer risk, the lowest share of any topic, against 21.8% for pain and discomfort and 27.0% for accounts of clinical visits. Sentences expressing fear mentioned the procedure without the disease far more often than the reverse (559 vs 66). HPV and cancer risk connected to the exam-experience cluster only through access to care.

Conclusions:

In this discourse, fear attaches to the screening exam far more than to cervical cancer, and the topics that connect the conversation concern bleeding, pain, and getting an appointment. Screening communication may gain more from building confidence about the exam, through clear information on what to expect, pain management, and self-sampling options, than from further emphasis on cancer risk. Lexicon-based sentiment tools should be used with caution in health contexts, where words such as positive carry clinical meaning.