Human-In-The-Loop (HITL) application design for early detection of pregnancy danger signs

Melyana Nurul Widyawati (Department of Midwifery, Poltekkes Kemenkes Semarang, Indonesia)
Ery Hadiyani Puji Astuti (Department of Midwifery, Poltekkes Kemenkes Semarang, Indonesia)
Kurnianingsih Kurnianingsih (Department of Electrical Engineering, Politeknik Negeri Semarang, Indonesia)

Belitung Nursing Journal
Vol. 8 No. 2 (2022): March - April

Background: Pregnancy period is a period for mothers to empower themselves to be safe and comfortable. Pregnant women must acquire pregnancy-related information, such as warning signs of pregnancy, to avoid severe complications and even death during pregnancy and childbirth. Therefore, developing an application for pregnant women would be very helpful. Objective: This study aimed to apply Human-In-The-Loop design with an android application to detect pregnancy risk early and avoid maternal morbidity and mortality. Methods: We collected data from the cohort of 5324 pregnant women at the community health centers in the West Lombok District from 2020 to February 2021. The data included age, parity, height, inter-pregnancy interval, hemoglobin levels, upper arm circumference, previous diseases, and bleeding history. We developed a Human-In-The-Loop mobile application and employed the decision tree for identifying pregnancy danger signs. The midwife (human-in-the-loop) reviewed and clarified the data to generate the final detection and made a recommendation. Results: The ordinal regression model revealed that older patients who have more parity, lower height, the distance of children <2 years, hemoglobin <11 g/dl, upper arm circumference (UPC) <23.5 cm, have positive HBsAg, have HIV disease, have a history of diabetes mellitus (DM), have a history of hypertension, positive protein urine, and have other diseases are more likely to have a high maternal risk. The decision tree outperformed and obtained a high accuracy of 92% ± 0.0351 compared to the nine individual classifiers (Nearest Neighbors, Random Forest, Neural Net, AdaBoost, Gaussian Naïve Bayes, Bagging, Extra Tree, Gradient Boosting, and Stacking). Conclusion: The Human-In-The-Loop mobile app developed in this study can be used by healthcare professionals, especially midwives and nurses, to detect danger indications early in pregnancy, accurately diagnose the high risk of pregnancy, and provide treatment and care recommendations during pregnancy and childbirth.