Scope of Online Maternal Health Information in Kinyarwanda and Opportunities for Digital Health Developers
Joselyne Muragijemariya, Valentine Ihogoza, Edith Talina Luhanga
TL;DR
The paper addresses the scarcity and uneven quality of Kinyarwanda online maternal health information. It applies web scraping to collect pregnancy-related pages in Kinyarwanda and uses Non-Negative Matrix Factorization (NMF) topic modeling to uncover five core topics: preventive measures, child care, intimacy, nutrition, and doctor consultations, revealing underexplored areas such as GBV and perinatal mental health. The authors propose three automated tools—topic recommender systems, automated quality scoring, and automated usability/credibility assessment—to guide developers in creating more useful and culturally relevant content. The work highlights practical implications for digital health developers aiming to localize maternal health information in low-resource settings and emphasizes the need for quality and accessibility improvements across language-specific health resources.
Abstract
Maternal health literacy is associated with greater odds of positive pregnancy outcomes. There is an increasing proliferation of websites dedicated to maternal health education, but the scope and quality of their content varies widely. In this study, we analyzed the main topics covered on maternal health websites that offer content in the low-resource Kinyarwanda language (mainly spoken by 12 million Rwandans). We used web scraping to identify maternal health websites. We utilized a topic modeling, using the Non-Negative Matrix Factorization (NMF) algorithm, to identify the topics. We found five main topics: (1) pregnancy danger signs, (2) child care, (3) intimacy (sex), (4) nutrition, and (5) the importance of doctor consultations. However, the articles were short and did not cater to fathers, pregnant adolescents, or those experiencing gender-based violence (GBV) or mental health challenges. This is despite 12.5\% women of reproductive age in Rwanda being victims of GBV and one in five women in low- and middle-income countries experiencing mental illness during the perinatal period. We recommend three automated tools, a topic recommender tool, culturally relevant automated articles, and website quality check tools, to guide software and health content developers.