PREDICTION OF THE FINANCIAL STRESS AND HEALTH RISK IN MATERNITY USING MACHINE LEARNING FOR LOW SOCIO-ECONOMIC AREAS OF LAHORE

Abstract

Maternal mortality is the global cause that is affecting the health status of the worldwide population. In low-income countries like Pakistan, financial stress is very common and has been considered a vital cause of maternal mortality. In this research, we have conducted an exploratory study at BHU Kahna Kacha Lahore and general hospital Lahore to understand the financial, health care, and maternity issues in low-socio-economic areas. The sample size used in this study contained 104 expecting mothers, 10 husbands, 3 doctors,1 LHV, and 1 midwife. (The saturation is considered while defining sample size). After thematic analysis and data processing, we extracted the key features that affect maternal mortality. These features are age, Locality, Education Level, Pregnancies Number, Month of pregnancy, Checkups, Last Delivery Type, Income, Monthly Expense, Spent on Last Delivery, Private Tests Cost, Undiagnosed Disease Cost, Transport Cost. We used these key features after data preprocessing to train machine learning algorithms to predict the intensity of maternal health risk and their financial stress level. We trained Support Vector Machine SVM (with different kernels), K-Nearest Neighbors KNN, and Naïve Bayes NB models. We then validated it using cross-fold validation, and our validation measures were Accuracy and F1- Score. According to the results, Naïve Bayes is the best option for the prediction of health risk and financial stress levels. As this research work predicts HR and FS of expecting mothers, it can help the government, insurance companies, social work societies, and banks to devise new health policies, new pregnancy-related insurance schemes, and loan deals while viewing their causes.