Maternal health and ChatGPT

SMARThealth ChatGPT : Supporting community health workers to provide guideline-based maternal care in rural India

Background

  • Detecting high-risk conditions in women before complications arise is crucial for averting maternal and newborn mortality and morbidity. In rural India, factors like anaemia, hypertension, and gestational diabetes persist as significant risk factors. 
  • SMARThealth Pregnancy (SHP) improves community-based screening and management of anaemia, diabetes, and hypertension during pregnancy and the postnatal period. Integrating our large language model (LLM) chatbot into the SHP API will provide real-time assistance to community health workers (ASHAs), providing guideline-based information tailored to women.

Aims

  • This project aims to collaboratively develop a large language model to support ASHAs deliver guideline-based care for pregnant and postpartum women in rural India

Research Methodology

  • This pioneering initiative aims to co-create with ASHAs an LLM-based chatbot tailored to their needs to provide contextually precise, clinically accurate, and gender-equitable responses tailored for maternal healthcare. 
  • The approach involves implementing Retrieval-Augmented Generation (RAG) on the pre-trained GPT-4.0 model, followed by developing a repository of gold-standard questions with clinically validated answers. Iterative user testing, gathering ASHAs’ feedback on utility, accuracy, biases, and limitations has been conducted. A user-friendly interface with diverse input/ output options (text to text, text to voice, voice to voice) and user personas tailored to ASHA preference in Hindi and Telugu will be created.

Leads