Measuring Productivity and Eciency of Traditional Chinese Medicine Hospitals From 2009 to 2016: A Bootstrap-Data Envelopment Analysis

Background: Traditional Chinese Medicine (TCM) Hospitals are the dominant agents to provide TCM services in China. How to improve the productivity and eciency of TCM hospitals is quite pivotal for the hospital managers and policy-makers to realize the optimal use of TCM resources. The purpose of this study was to estimate the productivity and eciency of TCM hospitals with the Bootstrap-Malmquist-Data Envelopment Analysis (DEA) from 2009 to 2016 to provide empirical evidence for hospital managers and policy-makers to improve the management and quality of TCM service. Methods: The data of the individual tertiary public TCM hospitals were collected from ocial Yearbooks of Traditional Chinese Medicine of China (2010-2017). Bootstrap-Malmquist-DEA was employed to measure the productivity and eciency of the TCM hospitals from 2009 to 2016. SPSS 23.0 version statistical software was used to conduct the descriptive analysis of the input and output indicators. R 3.2.1 version statistical software was applied to calculate the productivity and eciency of the TCM hospitals with FEAR package. The statistical signicant was set at P < 0.05. Results: The annual average growth rates for each input and output indicator were 6.61% (health professionals), 8.15% (actual open beds), 7.08% (outpatients and inpatients) and 12.50% (discharged patients) respectively from 2009 to 2016. Except the TFPC between 2014 and 2015, more than half of the TCM hospitals had TFPC scores over 1.000, indicating increased productivity. The overall annual geo-means of the scores of the TFPC and its decomposition index were 1.0379 (TFPC), 1.0167 (TEC), 1.0209 (TC), 1.0212 (PTEC), and 0.9955 (SEC) respectively from 2009 to 2016. Conclusions: The overall annual rate of the total factor productivity of the tertiary public TCM hospitals was slightly increased. The technological progress was the main driver to improve the total factor productivity. The decreased TE was more affected by the decreased SE. The TCM hospitals need to pay attention to the development and innovation of the TCM technology, thereby improving the competitiveness. The


Introduction
The Healthy China 2030 Planning Outline approved by China's Central Party Committee and the State Council has proposed that providing e cient and good quality of medical services and taking full advantage of Traditional Chinese Medicine (TCM) are two important goals (1). As an essential part of Chinese health care, the TCM development has been improving since the new medical reform due to a great number of policy documents highlighting TCM released by the Chinese government. As a result, the TCM services have become more accessible and more affordable, thereby decreasing the medical burden (2). For example, the outbreak of the new coronavirus pneumonia, named COVID-19 by the World Health Organization, has become a pandemic. Some traditional Chinese medicines, especially the Lung Cleansing and Detoxifying Decoction (Qing Fei Pai Du Tang), have shown therapeutic effects on mild and ordinary COVID-19 patients (3).
Traditional Chinese Medicine (TCM) hospitals are the dominant agents to provide the TCM services to the public in China (4). The TCM hospitals still face quite a number of challenges, like lower pricing of TCM outpatient services compared to their western medical counterparts (5), serious loss in human resources, and imbalanced structure of health professionals (6). What's worse, most of the TCM programs cannot be reimbursed by public health insurance schemes, which hampers patients from choosing TCM as their rst choice (7). Further, how to provide the good quality of TCM services is also a big challenge for TCM hospitals. One of the key factors to provide the good quality of TCM services is to improve the operational e ciency of the TCM hospitals (8). However, the existing TCM resources are always limited.
Therefore, how to improve the productivity and e ciency of TCM hospitals is quite pivotal for hospital managers and policy-makers to realize the optimal use of TCM resources.
Data Envelopment Analysis (DEA) has been widely applied to calculate the productivity and e ciency of healthcare institutions due to the characteristics of multiple inputs and multiple outputs (9). Furthermore, the Malmquist Productivity Index (MPI) model has been employed to indicate the dynamic productivity and e ciency changes of healthcare institutions or health systems (10,11). There are several studies which adopted the MPI model to measure the productivity and e ciency of TCM hospitals (4,8,(12)(13)(14).
However, there are some gaps on the current literature. None of the studies have focused on the individual tertiary public TCM hospitals at the national level. Besides, the MPI model adopted in current studies is classical. In 2007, Simar and Wilson proposed that Bootstrap could be combined with the classical MPI model to obtain the con dence intervals (CIs) of the productivity and e ciency scores, in order to eliminate the impact of the environmental and random factors on the estimation of productivity and e ciency scores (15,16).
In our previous study, we applied the Bootstrap-DEA model to estimate the e ciency of other kinds of health institutions, like community health service centers (17), public general hospitals (18), and township hospitals (19). The purpose of this study was to measure the productivity and e ciency of the TCM hospitals with the Bootstrap-Malmquist-DEA model from 2009 to 2016 to provide empirical evidence for hospital managers and policy-makers to improve the management and quality of TCM medical services.
It was the rst time that our team conducted the estimation on the productivity and e ciency of TCM hospitals by employing Bootstrap-Malmquist-DEA.

Input and output indicators selection
According to a systematic review on the relative e ciency research of Chinese hospitals with DEA models conducted by Dong & Li (9), they have suggested that the technical e ciency and allocative e ciency were always misunderstood by misuse of combination of monetary and volume indicators. The review and a previous study published by our research team (17) also have mentioned that price information of the indicators is much more di cult than the volume information to be accessed in Chinese hospitals. Therefore, based on the frequency of input and output indicators used most in current studies and the feasibility of data, in this study, the input indicators were the number of health professionals and the number of actual open beds, while the number of outpatients and inpatients and the number of discharged patients were selected as the output indicators.

Research model
The output-oriented Bootstrap-Malmquist-DEA model was employed to measure the productivity and e ciency of the individual tertiary public TCM hospitals. Basically, the total factor productivity changes (TFPC) and its decomposition index-technical e ciency changes (TEC) and technological changes (TC) can be obtained with the classical Malmquist-DEA model on the condition that all decision making units (DMUs)-TCM hospitals are operating in the frontier (20). Moreover, the technological changes can be decomposed into pure technical e ciency changes (PTEC) and scale e ciency changes (SEC). The basic formula is as follows: TFPC = TEC × TC= (PTEC × SEC) × TC Generally, when the productivity and e ciency scores are more than 1.0000, it means the rates are increasing; when the productivity and e ciency scores are equal to 1.0000, it means the rates remain the same; and when the productivity and e ciency scores are less than 1.0000, it means the rates are decreasing.
Considering the productivity and e ciency of all decision making units (DMUs), that is, all TCM hospitals are actually affected by the environmental and random factors (15)(16)(17)(18), the Bootstrap method was introduced to solve this problem (21)(22)(23). Its basic idea is to simulate the data-generating process (DGP) by repeating sample selection, thereby bias-corrected productivity and e ciency scores with CIs being generated. In result, it is much closer to the actual productivity and e ciency of TCM hospitals.
Furthermore, the Bootstrap-Malmquist-DEA model can be able to infer that whether the changes of the productivity and e ciency are statistically signi cant. Statistical analysis SPSS 23.0 version statistical software was used to conduct the descriptive analysis of the input and output indicators. R 3.2.1 version statistical software was applied to measure the productivity and e ciency of the TCM hospitals with FEAR package (24). The number of repeating sample selection was 2000. The statistical signi cant was set at P < 0.05.

Characteristics of the input and output indicators
As shown in Table 1  Summary of the individual changes of the total factor productivity and its decomposition index of the tertiary public TCM hospitals Appendix 1-6 describe the results of the changes of the total factor productivity and its decomposition index of the individual tertiary public TCM hospitals. Except the TFPC between 2014 and 2015, more than half of the TCM hospitals had TFPC scores over 1.000, indicating increased productivity. Summary of the annual geo-means of the changes of the total factor productivity and its decomposition index of the tertiary public TCM hospitals Table 3 shows that the overall annual geo-means of the scores of the TFPC and its decomposition index were 1.0379 (TFPC), 1.0167 (TEC), 1.0209 (TC), 1.0212 (PTEC), and 0.9955 (SEC) respectively (Table 2). Combined with Figure 1, the trend of the TFPC of the TCM hospitals was slightly uctuant. Also, except the TFPC between 2014 and 2015, the rest of the TFPC were slightly increasing. Additionally, from 2011 to 2014, the main factor as the driver to improve the TFP was the technology positive progress. In contrast, both the technology positive progress and the increased technological e ciency improved the TFP between 2009 and 2010 while the technological positive e ciency was mainly the factor to improve the TFP between 2015 and 2016. From 2011 to 2015, the TEC were decreasing due to the decreased SEC.

Discussion
Our study showed that the overall annual rate of the total factor productivity of the tertiary public TCM hospitals was slightly increased because of the increasing invest by the Chinese government. The technological progress was the main driver to improve the total factor productivity, which was associated with the innovation of the TCM technology. The decreased TE was more affected by the decreased SE, indicating the restriction of the scale of the TCM hospitals.
Our study found that the whole annual increasing rate of the TFP was 3.79%, which was a slight growth. This is mainly because after the new medical reform was published, the Chinese government has extensively invested on the development of the public hospitals, including the TCM hospitals. This result was different from that of Yang's study (4). It estimated that the overall rate of the TFP of the TCM hospitals was decreased from 2012 to 2016, which was 1.30%. It may be because of the different DMUs. Though both of the two studies measured the productivity and e ciency of the TCM hospitals at a national level, our study adopted the individual tertiary public TCM hospitals as the DMUs while all levels of TCM hospitals in a province were the DMUs in Yang's study. Compared with the secondary and primary TCM hospitals, the tertiary public TCM hospitals are equipped with more advanced medical instruments, more excellent health professionals and better management systems, thereby leading to Siphon effect on patients (25). This could result in higher productivity and e ciency for the tertiary public TCM hospitals. It may also be because the different input and output indicators were selected in the two studies and Yang's study employed the classical Malmquist-DEA model without introducing the Bootstrap method. These factors could result in different results.
Although the overall TFP rates had increased except the 2014-2015 period, technological progress was found to be the main driver to improve the TFP. The high medical technology is one of the prominent components of the high quality of the medical services and can attract more and more patients to seek medical services. Generally, the TCM hospitals cannot be comparable to the general hospitals, which concentrate on the western medical services (4,5). Therefore, the TCM hospitals especially need to pay attention to the development and innovation of the TCM technology, thereby improving the competitiveness. Also, the development of the TCM technology includes strengthening the training of the current TCM human resources (8). Consequently, the whole level of the TCM technology of a TCM hospital can be improved and then improve the productivity and e ciency of the TCM hospitals.
The TFPC between 2014 and 2015 was 0.9756, which was in uenced more by the decreased TE. Furthermore, the decreased TE was more affected by the decreased SE. This indicated that the scale of the tertiary public TCM hospitals should be restricted in order to be blindly extended. The hospital managers should pursuit the high quality, high e ciency and low cost of the TCM medical services rather than the high volume. This can be achieved by transferring the traditional and single management system to a more re ned management system, which includes strengthening the hospital nancial system, improving the hospital environment and adjusting the price of the TCM services (26).
High productivity and e ciency of the TCM hospitals is highly related to the rational use of health resources (14). The rational use of health resources can be up to the well-structured health professionals (17). Even though this study showed that the number of health professionals was increasing yearly, it could not infer whether the structure of the health professionals was appropriate, especially the ratio between the western clinical doctors/pharmacists and the TCM doctors/pharmacists. Hong analysed the trend of the TCM human resources at a national level from 2010 to 2015 and suggested that the number of the TCM human resources was also continuously increasing (6). However, the increased speed of the TCM pharmacists was slow and the number of the TCM human resources in the TCM hospitals was not enough. What is worse was that the loss of the TCM human resources was also a serious problem. These could have a negative effect on the development of the TCM hospitals and then deteriorate the productivity and e ciency of the TCM hospitals. Although this study showed a slightly increased TFP of the TCM hospitals, it cannot be denied that the TCM hospitals have to strengthen the management level of the TCM human resources. Fang proposed that in order to avoid the loss of the hospital human resources, it was necessary to rebuild the performance and salary management system based on the Diagnosis Related Groups (DRGs) system (25). Speci cally, the DRGs could identify the di culty and scope of the medical services and help assess the medical services. Then, a more advanced performance and salary system can be able to improve the motivation of the TCM human resources to keep them from quitting a job.
In order to keep improving the productivity and e ciency of the tertiary public TCM hospitals, it is imperative for the TCM hospitals to actively participate in the hierarchical diagnosis and treatment system to better paly am important role on the medical alliance (25,27). This can realize the rational use of TCM resources in some way, thereby improving the productivity and e ciency of the tertiary public TCM hospitals based on the abovementioned analysis. However, Wang's study showed that the hierarchical diagnosis and treatment system has not been established within the TCM medical alliance, because it is always the rst choice for the patients to go to the tertiary TCM hospitals to seek the medical services and the ability of a secondary or primary TCM hospital is not enough to meet the patients' medical demands (27). Additionally, to better establish the hierarchical diagnosis and treatment system is associated with the well-established hospital information system (28). It can be able to realize the TCM resource sharing to further achieve the rational use of the TCM resources.

Limitations
This study made a contribution to the current studies by applying a more accurate method (the Bootstrap-Malmquist-DEA model) to measure the productivity and e ciency of the individual tertiary public TCM hospitals at a national level. However, there are still some limitations. First, this study did not include the input and output indicators which represent the characteristics of the TCM medical services. There was no information reported by the o cial Yearbook of Traditional Chinese Medicine of China. It will be better to ask for the information of the relevant indicators from the individual TCM hospitals to further directly analyse the factors which have an impact on the productivity and e ciency of the TCM hospitals. Second, this study only analysed the productivity and e ciency of the TCM hospitals. However, it did not explain the relationship between the quality and the productivity and e ciency, as well as the relationship between the cost and the productivity and e ciency. Quality and cost should be essential part of the further research. Third, due to the duplicated data from 2012 to 2013, it was not clear about the actual productivity and e ciency between 2012 and 2013 and between 2013 and 2014. It should improve the quality of the data collection in the future.

Conclusion
The overall annual rate of the total factor productivity of the tertiary public TCM hospitals was slightly increased. The technological progress was the main driver to improve the total factor productivity. The TCM hospitals need to pay attention to the development and innovation of the TCM technology, thereby improving the competitiveness. The hospital managers should pursuit the high quality, high e ciency and low cost of the TCM medical services. It is necessary to rebuild the performance and salary management system based on the DRGs system. It is imperative for the TCM hospitals to actively participate in the hierarchical diagnosis and treatment system to better paly am important role on the medical alliance and to better establish the hospital information system to realize the TCM resource sharing. Abbreviations