Many current context-aware systems only react to the current situation and context changes as the occur. In order to anticipate to future situations and exhibit proactive behavior, these systems should also be aware of their future context. Since predicted context is uncertain and can be wrong, applications need to be able to assess the quality of the predicted context information. This allows applications to make a well-informed decision whether to act on the prediction or not. In this paper, we present prediction quality metrics to evaluate the probability of future situations. These metrics are integrated in a structured prediction component development methodology, which is illustrated by a health care application scenario. The metrics and the methodology address the needs of the developer aiming to build context-aware applications that realize proactive behavior with regard to past, present and future context.