Development and Preliminary Evaluation of a Health Chatbot for Breast Milk Expression Promotion in Mothers of Preterm Infants Admitted to the Neonatal Intensive Care Unit

Abstract

Objective: This developmental research aimed to develop and preliminarily evaluate the usability of a health chatbot designed to promote breast milk expression in mothers of preterm infants admitted to the Neonatal Intensive Care Unit (NICU). Methods: The chatbot development comprised two main phases. Phase 1 involved designing the content and conversational structure, which included reviewing health knowledge on breastfeeding in sick infants, investigating approaches for using chatbots in health behavior promotion, and developing the chatbot’s conversational flow. The content and conversational structure were validated for content validity by a panel of five experts and revised according to their recommendations. Phase 2 focused on constructing the chatbot and conducting preliminary system testing, which included alpha and beta testing. Subsequently, the chatbot was pilot tested with two mothers of preterm infants admitted to the NICU for at least four days. This pilot study aimed to assess the mothers’ engagement with the chatbot, including any usability issues, the frequency of breast milk expression, recording practices, and milk delivery to the NICU. Results: The chatbot, developed on the ManyChat platform, served as a virtual nurse to support mothers’ breast milk expression. It communicated through text messages, infographics, and videos over four consecutive days. System testing revealed that the chatbot consistently delivered automated messages for the entire four-day period. Users were able to read messages, access images and content within infographics, and view videos with clear visuals and audio. Pilot testing with mothers showed that both participants used the chatbot continuously for four days without any usability issues. Both mothers were able to express breast milk, record the time and volume of milk expressed, and deliver the milk to the NICU correctly as advised by the chatbot. Furthermore, the mothers recognized the benefit of the chatbot in reminding them to express milk on schedule. Conclusion: The study findings indicate that this chatbot functioned as intended. Users could consistently access and use it continuously for the specified duration. Furthermore, mothers successfully expressed, recorded, and delivered breast milk correctly according to the chatbot’s guidance. These results suggest the potential feasibility of using a chatbot to promote breast milk expression among mothers of preterm infants admitted to the NICU.