An inquiry of “Formula Feeders” sub-reddit using natural language processing approach

• In Event: Using and studying technology in research involving children and families

Abstract

Infancy is a time when nutrition is gained primarily via milk (either breast milk or formula milk). Although exclusive breastfeeding is advocated during the first 6 months of life by the Centers for Disease Control and the World Health Organization (e.g., CDC, 2020; WHO, 2021), infant formula is still found to be an essential resource for nutrition and energy during the first year. It is often used as supplementary food for breastfed infants. Social media, as a large-scale interactive online community, represents one naturalistic setting where caregivers’ formula feeding practices and concerns can be observed via online discussions. The sub-reddit labeled “Formula Feeder” is used by caregivers and parents to post and discuss all topics related to infant formula. This forum provides an unobtrusive way to examine parents’ perspectives about formula feeding. This project summarizes the general topics that caregivers discussed on sub-reddit “Formula Feeders.”

“Formula Feeders” was created in June 2012 for caregivers who are using or have used formula in any capacity to feed their infants. It currently has 13,300 members. All posts, post dates, titles, and content posted between 06/2012 – 09/2022 were collected using python package psaw. Sentiment analysis (i.e., opinion mining that examines whether attitudes are positive, negative, or neutral) of post content was conducted using python package vaderSentiment. Next, post content was pre-processed to remove punctuation and common stop words (e.g., the, a) using python package nltk. Next, topic modeling was performed through Latent Dirichlet Allocation (LDA) using python packages geuism and spaCy on post content. LDA is a generative statistical model that “finds” topics in collections of documents. These topics are then automatically classified in all other posts by calculating how relevant it is to a discovered topic. Topics are defined as a set of terms that collectively suggest a broader shared theme. From this analysis, ten topics emerged. Qualitative content analysis was later conducted to generate most discussed topics.

In total, 8,402 sub-reddit posts from 06/2012 – 09/2022 were collected. Of these, 273 posts were dropped due to empty content, leaving 8,129 posts for analysis. Results from the sentiment analysis indicated that caregivers hold a generally positive attitude about infant formula. Visualization of sentiment score distributions among sub-reddit posts is presented in Figure 1. Specific topics included: 1. formula brand (e.g., coupons, asking for samples, comparing different brands); 2. formula use (e.g., powder dosage, water amount, daily feeding times); 3. child allergy and formula selection (e.g., allergies, soy and dairy-based formula, alimentum and hypoallergenic formula; formula switch); 4. infant reactions to formula (e.g., cry, fussy, scream, spit, hungry, poop); 5. breast milk and breast feeding (e.g., breastfeeding, pumping, supplementary, comparing formula with breastfeeding). Visualization of LDA topics will be presented using an interactive figure. One example of LDA visualization is presented in Figure 2. Findings about sub-reddit posts are informative for researchers and public health officials as they can support the development of effective and targeted communication and public health campaigns.

Authors

• Muqing Liu, Purdue University, West Lafayette
  Presenting Author

• Deborah L. Nichols, Purdue University, West Lafayette