Efficiency Evaluation of County-Level Public Hospitals in Hainan, China: A Four-Stage Data Envelope Analysis Model Based on Panel Data
Hainan, located in the southernmost part of China, has been gaining increasing international attention due to its rapid development in various sectors, including healthcare. Despite significant advancements in health services, disparities in efficiency among hospitals remain a critical issue. To address these disparities and formulate effective reform measures, it is essential to evaluate the efficiency of county-level public hospitals in Hainan scientifically. This study aims to assess the efficiency of these hospitals, observe changes in efficiency from 2015 to 2017, and analyze the factors influencing these changes.
The study employed a stratified sampling design to select a representative sample of 88 hospitals from 12 counties in Hainan. The sampling was conducted to balance geographical, economic, and service population characteristics. The sample size was determined using the Data Envelope Analysis (DEA) method, which requires a minimum sample capacity of 2 times the product of the number of input and output indicators. The final sample included 264 observations, consisting of 66 general hospitals (GHs) and 22 traditional Chinese medicine hospitals (TCMHs). Patient information was not included in the study, eliminating the need for an ethics statement or informed consent.
The analysis was performed using Stata v.14.0 software, employing Tobit regression analysis and random effects models based on the results of a Hausman test. The study utilized a four-stage DEA model to evaluate the efficiency of county-level public hospitals in Hainan from 2015 to 2017.
Stage 1: Super-Efficiency DEA Model
The first stage involved the application of a super-efficiency DEA model. This model assesses whether a hospital achieves both technical effectiveness and effective scale. It identifies factors influencing hospital benefits, optimizes resource allocation, and improves efficiency by reducing input without altering output. The super-efficiency DEA model provides a more nuanced understanding of efficiency by allowing for the comparison of hospitals that are already efficient.
Stage 2: Tobit Regression for Slack
In the second stage, Tobit regression was used to analyze the effect of external environmental factors on the slack of the decision-making unit (DMU). Slack variables represent the difference between actual input and the input of the most effective scheme. The dependent variable in the Tobit regression was the total amount of relaxation, which is the sum of ray relaxation and non-radiative relaxation. The independent variables were environmental influence factors. This stage aimed to identify how external factors contribute to inefficiency.
Stage 3: Adjusting the Original Input Factors
The third stage involved adjusting the original input factors based on the results of the Tobit regression model. The goal was to eliminate the impact of external environmental factors by increasing the input of DMUs with a better environment. This adjustment ensures that all hospitals are evaluated under the same external conditions, providing a more accurate assessment of their inherent efficiency.
Stage 4: Adjusted DEA Analysis
In the final stage, the adjusted input factors and original output data were used to perform the super-efficiency DEA analysis again. This step yielded new efficiency values that reflect the true efficiency of the hospitals, free from the influence of external environmental factors.
The descriptive statistical results revealed significant disparities in input and output among the 88 hospitals. The DEA results indicated that 13 hospitals achieved comprehensive technical efficiency, 29 hospitals achieved pure technical efficiency, and 13 hospitals achieved scaling efficiency. The Tobit regression analysis identified disposable income of urban residents, financial subsidy income, and the number of visits per doctor per day as significant factors affecting efficiency. The degree of influence of these variables varied across the 88 hospitals.
The initial input factors were adjusted using the regression results to place all hospitals in the same external environment as much as possible before performing the adjusted DEA. The adjusted DEA results showed that 49 hospitals had increased scale benefits in 2015, 46 hospitals in 2016, and 36 hospitals in 2017. These findings differ from those of Jiang et al., who identified total health expenditure, medical business income, doctors’ daily burden of diagnosis and treatment, the number of hospital beds, and the total number of outpatients and treatments as key factors affecting efficiency.
The inclusion of environmental variables in the analysis helped control and diminish the impact of environmental factors on the evaluation results. After the adjustment, the overall efficiency value decreased, indicating that directly measuring the efficiency value of a hospital without adjusting for external environmental variables leads to an overestimation of the hospital’s efficiency level.
The uniqueness of this study lies in the addition of the super-efficiency DEA model to the traditional DEA framework. This approach allowed for a more intuitive observation of efficiency changes in each hospital. However, due to insufficient data, the study could only observe efficiency changes over three years and could not conduct an in-depth, dynamic evaluation of the overall efficiency of each hospital. Future research should focus on identifying the factors that lead to efficiency differences among different county hospitals and may require a dynamic evaluation assessment.
The study was supported by a grant from the Natural Science Foundation of Hainan Province, China (No. 817139). The findings contribute to a better understanding of the efficiency of county-level public hospitals in Hainan and provide valuable insights for formulating targeted reform measures to reduce disparities and improve overall healthcare efficiency.
doi.org/10.1097/CM9.0000000000001293
Was this helpful?
0 / 0