Mbaza RBC: Deploying and evaluation of an LLM powered Chatbot for Community Health Workers in Rwanda
Keywords: Community Health Workers(CHWs), Large Language Models(LLMs), Low resource settings
Abstract: The emergence of Large Language Models (LLMs) offers an opportunity to support health systems, particularly in low and middle income countries such as Rwanda where there exists limited health infrastructure. By providing information and support to front-line workers, especially community health workers (CHWs), LLMs offer to improve the quality of care by providing quick access to medical guidelines, supporting clinical decision-making, and facilitating health education in local languages. This work deploy and evaluates the performance of Large Language Model (LLM)-based chatbots to assist Community Health Workers (CHWs) in Rwanda, focusing on usability, interaction modalities, and local language processing. A total of 3,000 questions generated by Front-line workers using text and voice input methods were analyzed to determine preferences and error rates. Results indicate a strong preference for text-based queries (66\%), though voice queries showed high satisfaction (97.5\%) with minor transcription errors (2.47\%). The most common focus areas for CHW queries were Maternal and Newborn Health, Integrated Community Case Management, and Nutrition. These findings suggest that, while voice interactions hold some potential, improvements in speech-to-text models are needed for optimal functionality in low-resource settings.
Submission Number: 69
Loading