We analyze the operational performance of 202 Turkish rural general hospitals. To help improve performance on both input and output space, we adopt a directional distance approach. We treat a mortality based measure as a "needs indicator". We derive pure technical, scale and output congestion inefficiency measures and show how they vary across size classes. We show that "reducing mortality" involves sacrificing some good outputs. This is a trade off that holds at the potential output level. Second stage regressions of the inefficiency scores against hospital and rural district level variables, pinpoint critical areas for performance improvement. In particular we show the relative scarcity of nurses is linked to output congestion.