According to the World Health Organization, maternal mortality remains a critical public health issue, with 94% of maternal deaths occurring in low- and middle-income countries (LMICs), where rates reached 430 per 100,000 live births in 2020, compared to 13 per 100,000 in high-income countries. Despite this disparity, few studies have explored whether sparse data and features such as vital signs can effectively predict maternal health risks. To address this gap, this study evaluated the predictive capability of vital sign data using machine learning models trained on a dataset of 1,014 pregnant women from rural Bangladesh. Multiple machine learning models were developed using age, blood pressure, temperature, heart rate, and blood glucose, and their performance was assessed using regular, random, and stratified sampling techniques. Additionally, a stacking ensemble machine learning model combining multiple methods was developed to evaluate predictive accuracy. A key contribution of this study is the development of a stacking ensemble model combined with stratified sampling, an approach not previously considered in maternal health risk prediction. The ensemble model with stratified sampling achieved the highest accuracy of 87.2%, outperforming CatBoost (84.7%), XGBoost (84.2%), random forest (81.3%), and decision trees (80.3%) without stratified sampling. These findings demonstrate the feasibility of using sparse data and vital sign features for maternal health risk prediction and suggest that machine learning can provide a practical and accessible solution to improve prenatal care and reduce maternal deaths in resource-constrained LMIC settings.