In this paper, we develop a context-aware, tablet-based learning module for adult education. Specifically, we focus on adult education in healthcare-teaching learners to perform a medical screening procedure. Based upon how learners navigate through the learning module (e.g. swipe-speed and click duration, among others), we use machine learning to predict what comprehensive test questions a user will answer correctly or incorrectly. Compared with other context aware learning applications, this is the first time tablet-based navigation gestures have been used to support learning assessment. We conducted a 21 participant user study, showing the system can predict how users respond to test questions with 87.7% accuracy without user-specific calibration data. This is compared to item response theory methods which achieve about 70% accuracy. We also investigate which attributes are most responsible for making predictions.