| Objective: To investigate data envelopment analysis (DEA) and stochastic frontier analysis (SFA) in the evaluation of the efficiency of public hospitals of and above the county level, compare the differences and similarities of the two methods so as to provide reference for the application of the two methods in evaluating hospital efficiency in China. Methods: Empirical study was carried out in our study. We grasped the theoretical principle, operational processes and the application status of DEA and SFA in the evaluation of hospital efficiency through literature analysis. Then we took the evaluation of efficiency of 340 public hospitals of and above the county level in Sichuan province from 2003 to 2007 as an example to carry out case study for the two methods, so as to investigate the application methods and application conditions of the two methods in the evaluation of hospital efficiency. SPSS 15.0 was used to carry out data management and descriptive statistics. In the study by DEA, Deap 2.1 was used to calculate the TE, CE, SE and AE of hospitals. We carried out correlation analysis for the same type of efficiency calculated by different models and index systems. Then we chose the final model and index system according to results from domestic and foreign literature. In the study by SFA, Frontier 4.1 was used to calculate the TE and CE of hospitals and to carry out parametric test and likelihood ratio test. SAS 8.0 was used to carry out principal component analysis so as to reduce the dimension of the indices. Then we chose the final model and index system according to results from domestic and foreign literature.Results: (1) Decision of DEA models in TE process: TE from constant returns to scale (CRS) model was 0.758. The results of total TE, pure TE and SE from variable returns to scale (VRS) model were 0.758, 0.800 and 0.951 respectively. (2) Decision of DEA models in CE process: CE, AE and TE from CRS model were 0.467, 0.900 and 0.522, while those results from VRS model were 0.957, 0.819 and 0.955 respectively. We finally selected VRS model to evaluate TE of hospital by DEA and decided to select the TE results from the TE process rather than the CE process. (3) Decision of DEA index system: We compared the efficiency results with average length of stay (ALOS) in indices and those without ALOS in indices to have finally selected the indices with ALOS. (4) Decision of the explanatory variation of stochastic frontier production function: TE was 0.752 when explanatory variation with price, while that was 0.794 when explanatory variation without price. We finally selected the indices with price. (5) Decision of the variable of stochastic frontier production function: TE was 0.413 when using workload as variable, while that was 0.794 when using the aggregative index number as variable by principal component analysis. We finally selected the latter method to reduce the dimension of variable. (6) Decide the model of stochastic frontier cost function: We selected the first model as result by likelihood ratio test. (7) Decision of the variable of stochastic frontier cost function: CE was 0.540 when using total cost as variable, while that was 0.824 when using total variable cost as variable. We finally selected total variable cost as variable. (8) With the above decided model and index system, we calculated the efficiency of 340 public hospitals of and above the county level. Results of case study by DEA in TE process: The results of total TE, pure TE and SE were 0.758, 0.800 and 0.948 respectively. Results of case study by DEA in CE process: The results of CE, AE and TE were 0.498, 0.904 and 0.554 respectively. TE results by SFA: The results of TE and CE were 0.794 and 0.824 respectively.Conclusion: (1) It is suggested to calculate TE of hospital by DEA and to calculate CE of hospital by SFA. (2) It is suggested to select VRS model when using DEA method. We don't recommend researchers to use the TE result derived from CE process. And we suggest researchers to bring ALOS into the index system. (3) It is suggested to choose total variable cost as explanatory variable when evaluating CE by stochastic frontier cost function. Principal component analysis is a recommendable method for dealing with the output variable of stochastic frontier production function. It is not suggested to use workload as dependent variable when evaluating TE of hospital by stochastic frontier production function. Adding price indices in the explanatory variable is recommended. Researchers may consider adding price indices in explanatory variation. (4) The efficiency of Sichuan hospitals: Most of the hospitals are in middle or high TE. Many hospitals are in high CE. Hospitals in increased returns to scale should increase the utilization of inputs so as to increase output. A few hospitals in decreased returns to scale may consider expanding the scale of operation. Hospitals may properly save some expenditure in the number of beds, number of employees and fixed assets.
|
PDF Full Text Request