"Rosie" learns Spanish: The AI-powered chatbox bridging disparities in maternal and infant health
In a pioneering effort to combat maternal and infant health disparities, Faculty Associate Quynh Nguyen and collaborator Elizabeth Aparicio created Rosie, an AI-powered chatbot providing crucial information to new mothers, especially in underserved communities. Thanks to a recent $200,000 grant from the National Institutes of Health (NIH), Rosie's capabilities will expand to assist Spanish-speaking mothers, underscoring the project's commitment to inclusivity and accessibility.
This project has the potential to alter the way health information is presented to vulnerable population, promoting a more flexible and tailored approach to reach underserved groups. Racial / ethnic minority women are at increased risk for postpartum depression, and their children are less likely to have had well-child checkups in the past year. Moreover, racial / ethnic disparities are still prevalent for maternal and infant mortality as well as various health behaviors such as safe sleep practices, breastfeeding, and infant nutrition.
Currently, some popular programs involve resource-intensive home visits (limited in scale due to staff and cost constraints) or non-personalized text messages (may not directly address an individual’s questions). This project will develop a chatbot that addresses both of these possible limitations by representing a scalable tool that can have widespread reach across geographies and is personalized and responsive to an individual’s specific informational needs. A prototype of the chatbot, Rosie, capable of engaging in live question-and-answer sessions has already been built. Rosie is able to respond to 334 popular questions that new mothers may have. Pretests with mother groups and Mary’s Center patients have showed a positive reception to the chatbot. During this project the team will leverage recent advances in natural language processing and the emergence of efforts to aggregate massive amounts of health information to assemble a comprehensive health information library. They will further refine Rosie’s dialogue analyzer and response inference engine to robustly recognize and respond to user’s questions in the various and complex ways they can phrase a question. Nguyen and colleagues will test the hypothesis that Rosie may lower risk of postpartum depression, decrease emergency room visits, and increase attendance of well-baby visits.
The investigative team - composed of experts in the field of epidemiology, computer science, biostatistics, and maternal and child health experts - is uniquely suited to implement the study aims. Specific Aims are: 1) Develop technology for a chatbot, Rosie, that will provide health informational support to vulnerable mothers the moment they need it; 2) Evaluate the use of Rosie on maternal and infant outcomes; and 3) Release an open-source packet for the construction of a chatbot. Rosie provides informational support to vulnerable moms the moment they need it and safeguards new moms from misinformation that is common on the web with the ultimate goal of closing the gap in maternal and infant outcomes. Results and tools developed from this proposal can be utilized to inform population-based strategies to reduce health disparities and improve health.
