Background: Estimation of cost functions is an appropriate tool for optimal resource allocation in hospitals. The present study aimed at estimating cost function for hospitals affiliated to Shiraz University of Medical Sciences. Materials & Methods: In the current cross-sectional study, inputs, costs and resources data for 28 hospitals were gathered from hospital statistics and information system. Required data were collected monthly for two years (2015-2016). Then, The Cobb-Douglas cost function was estimated using panel data analysis. STATA 14 was used for data analysis. Results: Estimates showed that hospitals' cost elasticity concerning the inpatient and outpatient admission were 0.555 and 0.058 for large hospitals, and 0.769 and 0.099 for small hospitals (P<0.01). Among hospitals inputs, the cost of a physician with coefficient 0.175 in large hospitals and 0.481 in small hospitals (P<0.01) was most cost driver. In studied hospitals, average cost (36,406,480 Rial) was higher than marginal cost (26,548,020). Moreover, scale economies index were in small, large, and all hospitals were 0.868, 0.613 and 0.729, respectively. The number of active bed, length of stay and teaching activity had a positive effect on hospitals cost (P<0.01). Conclusion: According to the findings, to control costs, the hospital should manage most cost drivers, such as physicians and pharmaceutical costs. Moreover, According to the average and final cost and scale economies index, hospitals should expand their service provisions with current inputs and resources in order to reach the optimum point of output. [ABSTRACT FROM AUTHOR]