Impact of Health Human Resources on the Output Efficiency of Residents’ Health Status in China


 Background: Along with the development of the Chinese economy, there has been an increasing demand for resident health care services. Human resources for health (HRH) are a cornerstone in the medical system and have a significant influence on the health status of residents. Methods: This study employed the data envelopment analysis (DEA) to explore the output efficiency of residents’ health status. The input variable was medical services and the output variable was the health status of residents. We examined aspects of the provincial and national dimensions in China and used a Tobit regression analysis to explore the impact of HRH on output efficiency scores. Results: We found that in the provincial dimension, all kinds of HRH (except for technicians) had a significantly positive impact on the output efficiency of residents’ health status. In the national dimension, licensed (assistant) doctors, registered nurses, and village doctors had a significantly positive effect on the efficiency in the “with surgical procedures” model. In the “without surgical procedures” model, the effect factors changed to registered nurses and pharmacists. Most importantly, pharmacists played a critical role in both the provincial and national output efficiency scores. For pharmacists, the influence coefficient was three times higher than licensed (assistant) doctors and 11 times higher than registered nurses.Conclusions: There is asymmetry between the demand for drug therapy and the lack of clinical pharmacists. Optimizing the allocation of HRH should be a top priority for healthcare systems and the government should adopt a long-term perspective.


Introduction
Along with the development of the Chinese economy, there has been an increasing demand for resident health care services. In 2016, China proposed the strategic theme of "healthy China" and issued a notice on key tasks that could enhance reform in the medical system by 2019. A guiding ideology in this movement is to "transform the focus from curing diseases to people's health". Human resources for health (HRH) are capable of providing basic medical services and are essential for continuous improvement of residents' health status. Moreover, HRH ensure social stability and economic development.
It is important to analyze the impact of HRH on the output efficiency of residents' health status so that the government can modify health policy to improve the overall efficiency of medical services and residents' health status. This study employed the data envelopment analysis (DEA) to explore the output efficiency of residents' health status. The input variable was medical services and the output variable was the health status of residents. We examined aspects of the provincial and national dimensions in China and used a Tobit regression analysis to explore the impact of the HRH on output efficiency scores.
The remainder of this paper is organized as follows. Section 2 is the literature review. Section 3 details the materials and methods. Section 4 presents the empirical study. Section 5 is the discussion. Section 6 offers conclusions and policy recommendations.

Literature Review
HRH are a cornerstone of the medical system. Although many researchers have indicated that HRH availability has improved in some countries, out-of-balance, insufficient, and limited HRH availability persists in many other countries. Therefore, many countries still need to develop a long-term HRH plan [1][2][3][4][5][6][7][8] . Narasimhan (2004) [9] claimed that the quantity and distribution of HRH directly influence the quality of medical services and the long-term development of the medical system. Furthermore, HRH are the main force involved in dealing with public health emergencies. Fraher (2020) [10] argued that HRH are responsible for how well a country responds to and handles the COVID-19 crisis. Pharmacists have been deemed the health professional nearest to residents and therefore particularly qualified to provide convenient and efficient medical care for everyday diseases [ 11 -13 ] . Ung (2020) [ 14 ] stated that pharmacists have made a great contribution to community health during the COVID-19 outbreak.
Health economists have put forward a theory to welfare economics that argues health maximization is a resource allocation goal [15][16] . Medical service equity is a key factor for social fairness, but service delivery and health resource allocation in China is still inefficient [17][18][19] . Meng (2019) [20] proposed that health resources in China are inefficient because healthcare is overused and health resource distribution is uneven. Fu (2018) [ 21 ] suggested that the Chinese government focus on optimizing and allocating need-based health resources. Chowdhury (2014) [22] noted that countries with inadequate health resources need to focus on health service efficiency and hospital resource productivity. Tomblin (2013) [23] posited that health care planning becomes 第 4 页 共 22 页 difficult and unpredictable when resources are constrained and new health conditions surge (i.e., the COVID-19 pandemic).
Many studies have employed the DEA and used the Tobit regression analysis to evaluate health care efficiency and analyze influential factors. Domestic scholars have studied the regional efficiency differences in China's medical system [24 -27 ] . Other scholars have compared efficiency scores before and after Chinese healthcare reform [28 -30 ] . Foreign scholars have also widely researched efficiency and its contributing factors in the medical system [ 31 -36 ] . A number of studies have shown that DEA approaches are widely used to assess healthcare system efficiency. Table 1 briefly summarizes the input and output variables developed in previous empirical studies that monitored efficiency. Table 1. Input and output variables used in previous studies.

Authors
Inputs Outputs Cheng [28] Labor and capital Number of outpatients, emergency visits, and inpatients Hamidi [37] Number of beds, doctors, nurses, and non-medical staff Number of treated inpatients and outpatients Lee et al. [38] Number of beds, doctors, and nurses Number of inpatient and outpatient visits Afonso [39] Physicians, nurses, acute care beds, and MRI Life expectancy, infant mortality, and potential years of life lost Chen [40] Number of physicians, nurses, and beds Number of outpatient visits and inpatient cases Hu [27]

Network DEA Methodology
In this study, the DEA was used to explore the output efficiency of residents' health status. The DEA was first proposed in 1978 by Charnes and Cooper [45] , two famous American operational research experts, and has great advantages in dealing with multiple input and output problems. Yang [42] Population into three groups by year Physicians, hospital beds, and medical expenditures Ng [43] Number of doctors, nurses, pharmacists, other staff, and beds Number of outpatient and inpatient cases Kontodimopoulos [44] Number of doctors, nurses, and beds Outpatient visits, admissions, and preventive medical services According to the basic ratio-based DEA model, we can always pick the right is not optimal in these n decision units, we find the right u and v to maximize o  . This particular DMU performance can be measured by the classic CCR (Charnes, Cooper, and Rhodes,1978) programming model: It is difficult to calculate an optimal solution in Model (2), so conversion to a linear form using the Charnes-Cooper [46] transformation is necessary. Let The CCR model is based on the assumption that the return to scale is constant.
However, in some circumstances, the return to scale changes. Banker et al. (1984) [47] extended the CCR model to account for variable returns to scale, which became known

Tobit regression
Once we calculated the efficiency scores, a censored variable was formed and we used the Tobit regression to further examine determinant efficiency factors. The Tobit regression was first proposed by Tobit [48] in 1958 and the standard form is as follows: where * y is the potential dependent variable,  is the constant term, i x is the impact factor,  is the coefficient vector, and i  is the error term. The Tobit regression follows a normal distribution. In this study, we generally used the following form (Carlucci, 2018 andChaouk, 2020) [49][50] : For this study's purpose, the Tobit regression equations were established as follows:

Theil index
In 1967, Thiel proposed the Theil index [51] , which is a measure of inequality based on the information theory. It can be written as where T is the Thiel index, i y represents the actual value of the observed indicator in region i , and y represents the average value of all regional observations. The smaller the Thiel index, the smaller the distribution difference will be.
The results of the network DEA were obtained using DEA-SOLVER Pro5.0 software. The results of the Tobit regression analysis were obtained using EViews 10 software. The results of the Theil index were obtained using R x64 3.6.1 software.

Data Sources and Description
According this paper's purpose, the variables were divided into three aspects: input variables (medical service), output variables (residents' health status), and independent variables (HRH).

Input and Output Variables
The variable selection was guided by previous research and data availability. We focused on analyzing the efficiency of the provincial and national dimensions, respectively. Three variables were taken into account: "outpatient visits", "inpatient visits", and "surgical procedures", which were used as medical services for residents' health status. "Surgical procedures" had an apparent correlation with physicians and registered nurses and was used as a control variable. According to health indicators proposed in the "healthy China 2030" plan, which included "life expectancy", "infant 第 9 页 共 22 页 mortality", "maternal mortality", and "mortality in children younger than 5 years", we selected the above four variables to apply to resident health status.
However, because we still lack provincial data regarding "infant mortality" and "mortality in children younger than 5 years", these two variables were replaced in the provincial dimension with "perinatal mortality" and "prevalence in low weight children younger than 5 years", respectively. Moreover, variables in the DEA models were non-negative [52] except for "life expectancy". This paper applied a monotone decreasing transformation by taking the reciprocal of the other three undesirable variables and multiplying the reciprocal by 100 or 10, which allowed the output variables to be positive.

Independent Variables
In the Tobit regression analysis, HRH were used as independent variables and impacted the output efficiency of residents' health status. There are six major categories of HRH in Chinese medical institutions: licensed (assistant) doctors, registered nurses, pharmacists, technicians, village doctors, and others (novitiate HRH).
Considering the data dimension of the regression model, this paper instead used the proportion of each HRH category. Table 2 presents variable selection.  Table 3. Correlation analysis of input and output variables in the national dimension.
Note: ***, **, and * denote significance at the 1, 5, and 10% statistical level, respectively. Table 4 shows a correlation analysis of input and output variables in the provincial dimension in China (31 provinces) from 2017. Only "perinatal mortality" was associated with "outpatient visits" and "surgical procedures". There was almost no significant correlation between the other variables.

The Efficiency Scores of Each Province
In the provincial dimension, we used data from 31 provinces in China (2017).
Each province was treated as a DMU. Figure 1 shows the visual representation of the efficiency for each province, revealing that the efficiency scores with and without the variable of "surgical procedures" presented no significant difference.

The Efficiency Scores of the Whole Country
In the national dimension, we used time series data from 2007 to 2017. Each year was treated as a DMU. There is a rule of thumb that the number of units should be at least twice the number of inputs and outputs in order to preserve discriminatory power [53 -56 ] . In this article, the factor analysis reduced the dimension of residents' health status variables.
where y represents the residents' health status and is the only output variable, 1 o is the positive change of infant mortality, 2 o is the positive change of mortality in children younger than 5 years, 3 o is the positive change of maternal mortality, and with "surgical procedures" without "surgical procedures" o is life expectancy. Input variables were the same for the provincial analysis. We also separated samples with and without the "surgical procedures" variable.

Results of Tobit Regression Analysis
We considered HRH allocation as a possible influential factor on efficiency.
Efficiency scores were used as dependent variables and the proportions of each type of HRH (x1、x2、x3、x4、x5、x6 ) were used as independent variables. Table 5 shows the estimation results of the Tobit regression analysis from the provincial dimension. The impact factors in both models (with and without "surgical procedures") were almost the same. Except for technicians (x4), every other HRH category in the provincial dimension was statistically significant and positive (p < 0.05).

Determinants of Provincial Efficiency in China
When analyzing the "with surgical procedures" model, the largest correlation with "surgical procedures" without "surgical procedures" Therefore, we concluded that almost all kinds of HRH have a generally significant impact on efficiency. More importantly, pharmacists (x3) have the greatest influence; they were found to be three times more influential than licensed (assistant) doctors.
This highlights the importance of improving the provincial output efficiency of residents' health status.

Results of Theil Index
For further discussion, we used the Theil index to measure inequality trends of HRH distribution among provinces from 2007 to 2017. Based on our above conclusions, this paper analyzed practicing (assistant) physicians, registered nurses, pharmacists, and technicians. Figure 3 shows a daunting equity challenge for HRH distribution in China. Although the Theil index decreased, the most severe inequality was seen for pharmacists. The Theil index for technicians revealed a similar decreasing trend, but it was the minimum compared to the other three HRH categories since 2011.
It is also worth noting that the inequality distribution of practicing (assistant) physicians and registered nurses remarkably increased. In conclusion, if inequalities were a problem, our data could support the above research conclusions.

Discussion
Considering the above analysis, it is clear that the role of Chinese pharmacists merits further discussion. With the improvement of health care awareness, there is asymmetry between the demand for drug therapy and the lack of clinical pharmacists.
In addition, Chinese clinical pharmacists do not have prescription rights and physicians are typically valued more highly than pharmacists, which often leads to unreasonable drug use. Over the past decade, the Chinese government has increased their investment of medical resources and the number of HRH have increased. However, after an increase in the total number of personnel, the proportion of pharmacists has declined.
According to the China health statistics yearbook, in 2017 China had 45,298 pharmacists in medical institutions, which accounts for 5% of the total number of HRH.
Since the abolition of drug price additions, hospital pharmacy windows are full of people and pharmacists spend a lot of time dispensing and supplying medications instead of conducting clinical pharmacy work in hospitals. This paper considered the impact of HRH on the output efficiency of residents' health status. In the provincial dimension, there was no significant difference between the two models, proving that almost all kinds of HRH (except for technicians) had a generally significant impact on efficiency. In the national dimension, licensed (assistant) doctors, registered nurses, and village doctors had significantly positive effects on efficiency in the "with surgical procedures" mode. Meanwhile, the effect factors changed to registered nurses and pharmacists in the "without surgical procedures" model. More importantly, pharmacists showed the greatest influence on both the provincial and national output efficiency of residents' health status.

Policy Recommendations
Optimizing the allocation of HRH should be a top priority for healthcare systems and the government should adopt a long-term perspective. First, there should be an increased focus on professional training clinical pharmacists and promoting the implementation of prescription rights for pharmacists so that they can provide convenient medical services, reduce irrational drug use, and monitor patients who need long-term medication. Secondly, the health sector should pay more attention to increasing the number and quality of HRH in order to narrow the gap across provinces.
Moreover, additional policies should be established to attract more health workers to work in primary hospitals or rural areas. Finally, more research should be carried out on the layout of HRH, such as discerning the most effective proportion between different types of HRH or allocating HRH among different medical institutions. The ultimate goal is to improve the overall efficiency of medical services and residents' health status.