State estimation and reconstruction quality of distributed phenomena that are monitored by a network of distributed sensors is strongly affected by communication failures, which is a problem in real-world sensor networks. In this paper, we propose a novel sensor scheduling approach named priority list sensor scheduling (PLSS). This approach facilitates efficient distributed estimation in sensor networks, even in case of unreliable communication, by prioritizing the sensor nodes according to local sensor schedules based on the predicted estimation error. It is shown that PLSS minimizes the expected estimation error for arbitrary packet-loss or transmission probabilities. As prioritizing sensor nodes requires the calculation of several sensor schedules, a novel pruning algorithm that preserves optimal schedules is also derived in order to significantly reduce the computational demand. This is accomplished by exploiting the monotonicity of the Riccati equation and the information contribution of individual sensor nodes in combination with a branch-and-bound technique.