Leveraging Pretrained Language Models for Maternal Health Monitoring in Online Communities

Digital maternity support communities are growing in popularity, offering valuable peer support throughout pregnancy and postpartum experiences. These platforms also generate rich textual data that can be leveraged for artificial intelligence (AI) applications. This study applies pretrained language models (PLMs) to classify and analyse 270,195 posts collected from the subreddit “BabyBumps” between 2010 and 2022. Focusing on posts that reflect personal experiences related to pregnancy, postpartum, and related events (85.9%), the analysis reveals that the majority (62.6%) centre on physical health concerns, while nearly half (48.9%) express negative sentiment. Notably, both mental health and negative sentiment–related discussions show a marked resurgence during the COVID-19 pandemic. These findings underscore the evolving emotional and informational needs of expectant and new mothers in online spaces and highlight the potential of AI-driven tools in supporting digital maternal health monitoring.