Supplementary Material for “iSurvive: An Interpretable, Event-time Prediction Model for mHealth”

We highlight key differences between the present work and an interpretable, latent state model introduced by (Lian et al., 2014). In it, the model has one sequence of latent (K binary events) states (e.g., progression in a movie); each user experiences the same sequence of latent states but may react differently, resulting in a user-specific intensity function that produces a response to the latent process. In our model, on the other hand, the latent state process evolves independently from user to user. Thus the participants do not experience the same sequence of latent states. This is a key difference for our mobile health application.