This post is to share a mapping of the use of AI/machine learning tools in CHW programs in LMICs. I’m working with a team at Georgetown University on a scoping review on the same topic. As part of this, I carried out this mapping using AI search tools. I’ve identified 38 programs that met the criteria. A few highlights on what this mapping shows:
The large majority (87%) of programs are in sub-Saharan Africa and South Asia. Outside of those regions, there were 4 programs in Latin America and 2 in East Asia.
Types of AI technology employed:
Natural language processing tools and AI chatbots were the dominant types of AI technology used. This suggest keen interest in "voice-first" or "chat-first" interfaces as the most effective way to support low-literacy or busy health workers, allowing them to ask questions naturally rather than navigating complex menus.
Nearly half of tools focused on clinical decision support tools to guide CHWs through patient assessments (e.g., triage or diagnostic algorithms).
8 projects used computer vision for tasks such as interpreting ultrasound video, scanning malaria slides, or screening for oral cancer.
Augmentation not automation: The intended assistance is clearly designed to up-skill the human worker rather than replace them. The main supports to CHWs were clinical decision support and on-demand education and training.
The majority were research or pilot programs. Only 34% showed signs of scaling.
I intend to update this spreadsheet as I learn about any new programs, so if you know of any I missed, please send them my way. But I can’t promise to keep it truly up-to-date. That’s too much work for a volunteer project.
I hope this is useful!