Deep neural networks have demonstrated promising prediction performance on many health analytics tasks. However, the interpretability of the deep models is often lacking. In comparison, classical interpretable models such as decision rule learning do not lead to the same level of accuracy as deep neural networks (DNN) and can also be too complex to interpret (e.g., due to large tree depths). In this work, we propose Prototype LeArNing via Rule Learning (PEARL), which iteratively constructs a decision rule list to guide a neural network to learn representative prototypes that can be explained by the associated rules. The resulting prototype neural network inherits both the prediction power of DNNs and interpretability associated with rules, thus can provide accurate and interpretable predictions. Evaluated on real world health datasets, PEARL demonstrates state-of-the-art accuracy to various DNN baselines and interpretable results that are simpler than standard decision trees can provide.