NLP-based triage of maternal and newborn health questions in Kenya

The Challenge:
One significant driver of maternal and newborn deaths in Kenya is that many women are not informed or empowered to recognize signs of potential complications such as vaginal bleeding during pregnancy or a severe headache and blurred vision. For other women, it’s the lack of ability or opportunity to act on these danger signs in the face of conflicting priorities, including child care and domestic responsibilities, or concerns about income lost during the time seeking care. These converging factors around recognizing and acting on danger signs, contribute to 33% of maternal deaths in Kenya.

Our Solution:
Using data science and machine learning to identify & refer at-risk mothers for care at the right time & place. To address the challenge above, we developed PROMPTS - a low-cost digital health platform that empowers new and expectant mothers to seek care at the right place and time. It improves care-seeking behavior by sending gestation-stage specific “nudge” text messages, alongside a two-way helpdesk service which responds to thousands of messages from mums each day. The PROMPTS platform uses Natural Language Processing (NLP), a machine learning approach, to triage and prioritize messages from mums for clinical danger signs. This, in turn, allows Jacaranda’s clinical helpdesk agents (qualified nurses and medics) to rapidly respond to and refer urgent cases for care - all within an hour.)