This paper investigates the cost efficiency (CE), technical efficiency (TE), allocative efficiency (AE), and scale efficiency (SE) over 34 regional public hospitals in South Korea from 2007 to 2010 using Data Envelopment Analysis (DEA). The CE, AE, and TE of these hospitals during the period are 0.52, 0.71, and 0.74, on average, respectively, indicating that there is a possibility to reduce their inefficiency of 48%, 29%, and 26% by reallocating the input mix or scaling input back. SE of these hospitals during the same period is 0.85, suggesting that most of the regional public hospitals do not operate under the optimal scale which is efficient relative to both constant returns to scale (CRS) and variable returns to scale (VRS) technologies. The empirical result implies that even though half of the regional public hospitals are comparatively efficient allocatively and technically, they have not been good at selecting the cost-minimal input mix. It also indicates that some hospitals have suffered losses from not having the most optimal scale.