Abstract
BACKGROUND: Breast cancer affects millions of women and presents not only medical challenges but also emotional, financial, and social burdens. Beyond clinical treatment, patients increasingly turn to online cancer communities (OCCs) for informational support, emotional support, and shared coping strategies. OCCs help patients manage daily life and reduce psychological distress through shared experiences and empathetic engagement. Within these communities, emotional expressions serve as critical cues through which patients communicate their situations and needs with other OCC members. OBJECTIVE: This study explores the relationship between emotions expressed in patients' initial posts and community reactions and engagement in OCCs. Based on the emotions as social information (EASI) model and Plutchik's Wheel of Emotions, this study investigates whether different emotions function as distinct social signals that lead to different patterns of member response, and whether emotions defined as opposites elicit opposing or similar engagement behaviors. METHODS: This study examines how the 8 primary emotions expressed in an initial post-surprise, anticipation, joy, sadness, trust, disgust, fear, and anger-distinctively influence responses of members in OCCs. We collected data from a breast cancer community from 2002 to 2017 and analyzed 23,633 threads from 9137 patients with breast cancer. Using BERT (Bidirectional Encoder Representations From Transformers), we extract emotion scores from patients' initial posts, measure community engagement across 5 response dimensions, and empirically analyze how different emotions are associated with user response behaviors. RESULTS: The analysis shows that certain emotions, such as joy and anticipation, consistently elicit significant effects across all measured response categories (P<.001). Most other emotions, except disgust, also demonstrate significant effects in most categories (3 or 4 of 5 categories; P<.05). Consistent with the EASI perspective, different emotional expressions influence how community members allocate attention, effort, and timing in their responses in distinct ways. Moreover, for most pairs of opposite emotions (eg, surprise vs anticipation, joy vs sadness, and anger vs fear), the impacts on the 5 types of community responses move in parallel rather than opposite directions. CONCLUSIONS: By integrating Plutchik's emotion framework with the EASI model, this study advances understanding of how emotional expressions function as social information and signals in OCCs. The findings show that emotions expressed in patients' initial posts influence community engagement not only through emotional valence, but also through the situational context that community members can infer from the initial posts. In addition, modeling emotions as continuous intensity scales helps reveal how emotional strength amplifies engagement across multiple dimensions, including participation, effort, timing, and topical relevance. These insights extend prior OCC research beyond polarity-based sentiment analysis and offer practical implications for patients, community platforms, and health care practitioners seeking to better support peer interaction and psychosocial care in digital health environments.