key: cord-0440643-ev26bd1v authors: Spira, Daphna; Mayat, Noreen; Dreisbach, Caitlin; Poliak, Adam title: Discovering changes in birthing narratives during COVID-19 date: 2022-04-25 journal: nan DOI: nan sha: c59bc980ad4f544c31c638d9ae83da99afb47920 doc_id: 440643 cord_uid: ev26bd1v We investigate whether, and if so how, birthing narratives written by new parents on Reddit changed during COVID-19. Our results indicate that the presence of family members significantly decreased and themes related to induced labor significantly increased in the narratives during COVID-19. Our work builds upon recent research that analyze how new parents use Reddit to describe their birthing experiences. Reddit is a social media platform where users can post anonymous submissions and comments in various subreddits. We use the Pushshift Reddit API (Baumgartner et al., 2020) to collect all submissions posted to nine subreddits related to the birthing experience between April 2009 and June 2021. 1 Following Antoniak et al. (2019) , we remove all posts that do not include any of the terms "birth story," "birth stories," or "graduat," guaranteeing our corpus consists of birthing narratives. We remove all posts that contain less than 500 tokens, 2 resulting in a corpus of 4,484 birthing narratives before and 913 during 3 Method Topic Modeling. To discover distinct topics across our collection of birthing narratives, we apply Latent Dirichlet Allocation (Blei et al., 2003) , as implemented in Mallet (McCallum, 2002) . In initial experiments, k, the number of topics, ranges from 5 to 50. We choose k = 50 based on C v coherence (Röder et al., 2015) . Discovered topics include induced labor, family, breastfeeding, and the first moments between a new parent and child. To determine whether the prevalence of these topics changed during COVID-19, we fit topic-specific Prophet models, an additive regression approach for forecasting time series data (Taylor and Letham, 2018), on the topic's average monthly prevalence in our corpus prior to March 2020. We then compare the topic's actual average monthly prevalence in our corpus during COVID-19 with the corresponding model's forecast. Following Biester et al. (2020) , we quantify how often the actual topic's monthly probabilities fall outside the model's 95% CI and use one-tailed Z-tests to determine statistical significance. Quantifying Personas Presence. Determining the prevalence of types of characters, or personas, in a narrative can illuminate information from an author's perspective, e.g. who is most the relevant, valued, or supportive character. Following Antoniak et al. (2019) , we quantify a persona's prevalence by counting how often they are mentioned, using a dictionary of terms for each persona. 4 We examine the difference in average mentions of each persona before and during 4 Results Figure 1 shows how the forecasted prevalence's for the family and induction topics significantly differ with the topics' prevalence during COVID-19. The increase in the induction topic (Figure 1b ) may reflect the increased recommendation of planned induction, enabling COVID-19 testing of expecting parents in advance of delivery (Goer, 2020) . The decrease in the family topic (Figure 1a ) might correlate with hospitals restricting visitors during the pandemic. Our experiments quantifying the personas' presence demonstrate that the healthcare providers were mentioned at similar rates before and during COVID-19, indicating that from the perspective of birthing parents, providers' roles in birthing narratives remained consistent. We notice a significant drop in the family persona's presence (22.8% , p < 1 −7 ) and significant increase in the partner persona's presence (5%, p < 0.028) during the pandemic. Figure 2 indicates that these differences were usually most apparent from periods 2 to 8 in the stories, which correlates with the time generally spent in the hospital during birthing narratives (Antoniak et al., 2019) . This might suggest that partners supplemented the support missing by families that were unable to visit hospitals before and after delivery. We presented a study demonstrating how parents' self-described birthing experiences significantly changed during COVID-19. Our results indicate that hospital policies may be reflected in birthing narratives. Our work presents a case study in how we can analyze patient experience from their own written narratives and perspectives. Where the system failed: The covid-19 pandemic's impact on pregnancy and birth care Narrative paths and negotiation of power in birth stories 2020. The pushshift reddit dataset Emily Mower Provost, and Rada Mihalcea. 2020. Quantifying the effects of covid-19 on mental health support forums Pregnancy, birthing, and postpartum experiences during covid-19 in the united states Clinical implications and economic effects of the corona virus pandemic on gynaecology, obstetrics and reproductive medicine in germany-learning from italy Covid-19: Should you agree to elective induction of labor? The impacts of covid-19 on us maternity care practices: A followup study Giving birth during the covid-19 pandemic: The impact on birth satisfaction and postpartum depression Mallet: A machine learning for language toolkit Mothers' worries during pregnancy: A content analysis of reddit posts Fathers' worries during pregnancy: A qualitative content analysis of reddit Exploring the space of topic coherence measures Detecting emerging symptoms of COVID-19 using context-based Twitter embeddings Temporal mental health dynamics on social media Forecasting at scale. The American Statistician Identifying pandemic-related stress factors from social-media posts -Effects on students and young-adults The impact of the covid-19 lockdown on the experiences and feeding practices of new mothers in the uk: Preliminary data from the covid-19 new mum study Association for Computational Linguistics. (a) The family members topic fell significantly below the confidence interval in the post-COVID-19 period. (b) The induction topic rose significantly above the confidence The x-axis, y-axis, and red vertical line respectively indicate the date, monthly average topic probability, and beginning of COVID-19. The blue line, shaded region, and black line respectively represent the models' prediction, 95% CI, and actual data Figure 2: How often the persona is mentioned on average (y-axis) during the course of a narrative (x-axis)