Development of a medical expert system to improve the quality of antenatal care in Tanzania
Abstract
Maternal mortality remains a global problem, with approximately 830 women dying every day
as a result of childbirth and pregnancy complications. The maternal mortality ratio is as high as
524 deaths per 100 000 live births in Tanzania. The main causes of maternal mortality in
developing countries are all linked to poor prenatal care, which is partially caused by treatment
delays. Studies show that providing women with maternal health information can help achieve
the goal of reducing global maternal mortality to less than 70 maternal deaths per 100 000 live
births by 2030. Through Natural Language Processing (NLP), we leveraged the use of BERT
question and answer model which is a pre-trained model that wasfine-tuned to develop a model
that can diagnose pregnancy complications, explain possible causes in simple language, and
provide recommendations for care and treatment, for Malaria, Pregnancy hypertension (Pre eclampsia), and miscarriage (Threatened abortion) which are the main preventable causes of
maternal mortality in Tanzania. The expert system is embedded in a maternal smartphone app,
MamaApp, that provides weekly information on fetal development, regular and concerning
pregnancy symptoms, and self-care tips. The expert system model was able to diagnose the
three conditions with confidence ranging from 79% to 100%. Validation of MamaApp in
Arusha showed high acceptance from both the expectant mothers and doctors.