In the recent years most of the studies related to the hospital efficiency focus on the issues of the ownership, size, physician team work, and clinical performance. It is certain that nursing units are the major profit centers for hospitals. However, the relative efficiency of nursing units regarding the resource use and outputs is seldom investigated. Therefore the aim of this study is to investigate the influence of each input and output item on the relative efficiency using Data Envelopment Analysis (DEA). The research samples included 44 units of two government owned hospital accredited as medical centers in 2002. The research samples are divided into 5 groups. The inputs items include full time nurse, continuous education hours, cost expand, specialties, patient care hours and workload. The output items are length of stay, number of admission, occupancy rate and total patient days. Every nursing unit is considered as a Decision Making Unit (DMU). The data were collected from Jan. to Dec. 2002 to evaluate the overall efficiency, technical efficiency and scale efficiency. Efficiency reference sets were found to be reference for inefficiency nursing units. The ways to improve the resources inputs or outputs of those inefficient nursing units were suggested using scale variable analysis. The influence of each input and output variable on the relative efficiency were assessed using sensitivity analysis. The results are summarized as follows:. 1. Overall inefficiency nursing units: A hospital has 4 (20%), B hospital has 5 (20%), both A and B hospitals have 9 (21%), medical nursing units have 5(17%)and surgical nursing units have 7 (47%). 2. Technical inefficiency nursing units: A hospital has 2 (10%), B hospital has 2 (8%), both A and B hospital have 3(6%), medical nursing units has one (3%)and surgical nursing units have 4(27%). 3. Scale inefficiency nursing units: A hospital has 4(20%), B hospital has 5 (20%), both A and B hospital have 9 (21%), medical nursing units have 5 (17%)and surgical nursing units have 7(47%). 4. Relative inefficiency nursing units with overall efficiency value between 0.9-1 belong to the marginal inefficiency units. The overall inefficiency of nursing units due to scale inefficiency. 5. From slack variable analysis the first three inputs needed to reduced are continue education hours, patient care hours and number of full time nurse. The outputs needed to increase are the number of admission and the length of stay. In this study, the input items to evaluate the efficiency of nursing units are mainly based on the data of clinical productivity. However, the quality of nursing care, the index of patient satisfaction, the number of medical equipments and the standard on patient care activities are not considered. It is highly suggested that service index and equipment allocation should be considered while evaluating efficiency.