We consider a setup where a sensor is observing a stochastic process of interest which must be communicated to a remote estimator. In addition to the distortion cost, the system incurs costs associated with sensing and communication. The overall cost is to be optimized by jointly designing strategies for controlled sensing, event based communication and remote estimation. Instead of continuously sensing the environment, the sensor is activated only when the remote estimator issues an activation command. Further, when the sensor makes the observation, it can choose not to communicate (and hence reduce communication cost) if the observation is not too informative. We consider the joint optimization of the sensor activation, communication and remote estimation strategies. The resulting optimization problem is a decentralized sequential decision making problem. Unlike prior work, we consider both the controlled sensing and the event based communication aspects of the problem. We extend the analytical approach of [1] to characterize optimal strategies.