The CCR model is designed to measure technical efficiency (TE) and can extend single-output efficiency to multi-output efficiency. While similar to the BCC model in DEA, the CCR model differs in its approach. Technical efficiency in the CCR model is the overall efficiency, which can be decomposed into scale efficiency (SE) and pure technical efficiency (PTE). PTE indicates the level of resource utilization, while SE reflects changes in scale returns due to increased inputs (14). Let there be n DMUs, for m inputs and s outputs. For the k-th DMU:

Input vector:

Output vector:

The CCR model solves for output weights u, and input weights v to maximize the efficiency θ_k:

Function:

Constraints:

The BCC model is a DEA model first developed by Banker (39) as a measure of pure technical efficiency by introducing the concept of the distance function (40). The BCC model overcomes the assumption of constant returns to scale in the CCR model by assuming variable returns to scale and reflects both pure technical efficiency and efficiency of scale (15).

The BCC model, which assumes variable returns to scale and is a model that only considers pure technical efficiency (PTE) that excludes scale-efficiency, is a CCR model with the addition of the constraint ∑j=1nλj=1. The optimization constraint equation is as follows:

where s⁺ and s⁻ represent input and output slack variables, respectively. The CCR model calculates the technical efficiency (TE) of a DMU, while the BCC model isolates the impact of scale, yielding pure technical efficiency (PTE). Scale efficiency is then derived by dividing TE by PTE, with a standardized threshold of 1 under the BCC model.

This study evaluates economies of scale by analyzing increasing, decreasing, and constant returns to scale. When all inputs grow proportionally, output expands at the same rate under constant returns to scale. If output grows at a slower rate, returns to scale decrease; if it grows faster, returns to scale increase. Increasing returns to scale indicate economies of scale, while decreasing returns signal dis-economies of scale. In hospital efficiency analysis, institutions with decreasing returns should downsize, whereas those with increasing returns should enhance efficiency by expanding key assets (41).

Fried et al. (37) argues that DMU efficiency is influenced by operational inefficiencies, environmental factors, and statistical noise, necessitating SFA regression analysis to separate these effects. Originally introduced by Aigner et al. (42), SFA has been widely applied in empirical research. Kumbhakar and Lovell (43) further developed the methodology by demonstrating how SFA can effectively separate managerial inefficiency from environmental influences in healthcare settings.

In practical terms, the SFA model serves as a “filter” that helps us understand why some hospitals appear inefficient. When we observe that a hospital is not performing optimally, this could be due to three reasons: poor management decisions (true inefficiency), unfavorable external conditions such as geographic location or patient demographics (environmental factors), or simply random events like equipment breakdowns or staff shortages (statistical noise). The SFA model mathematically separates these three components, allowing us to identify which hospitals are genuinely underperforming due to management issues vs. those facing external challenges beyond their control.

In the SFA model, the error term comprises two components: random error (v) and technical inefficiency (u). Typically, v follows a normal distribution, while u is a non-negative random variable. This decomposition enables the precise estimation of DMU technical efficiency.

In the first stage, the slack variable (s_ni) represents the deviation from optimal efficiency. SFA regression employs input factors as dependent variables and environmental factors as independent variables, modeled as:

where s_ni is the slack variable, f(z_i, β_n) denotes environmental impact, v_ni represents random error, and u_ni indicates technical inefficiency. With k environmental variables, z_i = z_1i + z_2i + ⋯ + z_ki, and β_n as their corresponding parameters. Here, vni~N(0,σv2), and u_ni follows a truncated normal distribution. These two terms are assumed to be independent.

If the variance of management inefficiency is σv2, and the variance of the random disturbance term is σun2, satisfying γ=σv2σv2+σun2, then as γ approaches 1, the management factor S_in is the main influencing factor. As γ approaches 0, the random factor is the main influencing factor. The formula for separating management inefficiency is as follows:

where σσ=σuσvσ, σ=σu2+σv2, and λ=σuσv. Thus, the random error term serves as a mechanism for relaxing variables, reducing environmental impact to decrease the value of management inefficiency.

The practical significance of this adjustment process can be understood through a simple analogy: imagine comparing the performance of two hospitals, one located in an affluent urban area with abundant medical resources and another in a remote rural region with limited infrastructure. Without SFA adjustment, the rural hospital might appear less efficient simply due to its challenging operating environment rather than poor management. The SFA correction essentially “levels the playing field” by adjusting for these environmental differences, allowing us to make fair comparisons of managerial efficiency across different operating contexts.

Coefficient significance in the SFA model is crucial for adjusting slack variables. A negative coefficient reduces slack and enhances efficiency, while a positive coefficient increases slack, lowering efficiency. Even if insignificant, coefficient values influence input variable trends. Based on SFA results, DMU input values are corrected to account for environmental and noise effects, ensuring a more accurate efficiency assessment.

The adjusted input value (x_ni) is computed as:

where maxi(zip)-zip standardizes external environmental conditions, and m(x_i(v_ni) − v_ni) ensures uniform random error adjustments. This adjustment process transforms the original input data into “environmentally-neutral” values, enabling more accurate efficiency comparisons by removing the influence of factors beyond management control. The resulting adjusted inputs provide a cleaner foundation for subsequent efficiency analysis using the Super-SBM model.

While traditional DEA models provide valuable insights into hospital efficiency, they face significant limitations when multiple hospitals achieve perfect efficiency scores of 1.0, making it impossible to rank or differentiate among the best performers. Additionally, standard DEA models assume proportional changes in all inputs and outputs, which may not reflect the complex reality of hospital operations where different resources can be adjusted independently. To address these limitations and build upon the environmentally-adjusted inputs from the SFA stage, this study employs the Super-SBM model, which offers enhanced discriminatory power and more realistic efficiency measurements.

In the CCR and BCC models, all outputs are proportionally expanded, or inputs are contracted by a common factor along the input direction. Anderson's super-efficiency model addresses the issue of all DMUs being on the production frontier (44). Tone and Tsutsui (45) extended this approach by developing the Super-SBM model specifically to handle slack variables and provide more nuanced efficiency measurements. The principle is explained in Figure 2, where points B, C, and D are on the optimal frontier, and it is impossible to measure the efficiencies of these three DMUs. By constructing a super-efficiency model, the efficiencies of the DMUs on the optimal frontier can be excluded from the model while the efficiencies of the DMUs on the optimal frontier can be measured, for example, by excluding point C from the optimal point, we can see that a new frontier BC'D is formed, and the efficiency value of point C, which has been excluded, is the new frontier BC'D. The efficiency value of point C, which has been excluded, is the new frontier BC'D, which has been excluded. The efficiency value of the excluded point C is the ratio of the distance from OC' to the distance from OC on the new frontier, which is greater than one, so point C was originally on the original efficient frontier, and can still be made more efficient by increasing the input values of X₁ and X₂ compared to the present. The traditional DEA approach treats all inputs and outputs as proportionally adjustable, but this assumption fails to capture the reality that hospitals can independently adjust different resources (such as staff, beds, or equipment) without maintaining fixed ratios. The Super-SBM model overcomes this limitation by explicitly accounting for input excesses and output shortfalls, providing a more realistic assessment of hospital efficiency.

Based on the super-efficiency model (22), the Super-Slack Measurement (Super-SBM) compensates for the shortcomings of the BCC model. The Super-SBM model deals with input excesses and output deficiencies and gives a scalar measure of all inefficient units by using an additive model (46). Sun et al. (32) demonstrated the superior performance of Super-SBM in healthcare applications, particularly for ranking efficient hospitals and identifying specific areas for improvement. The Super-SBM model can be expressed as:

The integration of SFA-adjusted inputs with the Super-SBM model creates a powerful analytical framework that first removes environmental bias through SFA corrections and then applies sophisticated efficiency measurement through Super-SBM. This two-stage approach ensures that efficiency scores reflect genuine managerial performance rather than external circumstances, while providing the discriminatory power needed to rank and compare hospital performance effectively.

We bring the input and output indicators into the DEAP software, and after some processing, we can get the following results, which are shown in Table 6 and Figure 3. Through the DEA-CCR and BCC methods, we can derive the stages of development of TE, PTE, SE, and the corresponding DMU size gains of the 21 public TCM hospitals, which are further analyzed in the following section.

Comprehensive technical efficiency (TE) is a key indicator for measuring the overall operation of Chinese medicine hospitals in Hebei Province. It reflects the effectiveness of resource allocation and is used to evaluate resource utilization and hospital performance (54). As shown in Table 6, the average TE of TCM hospitals is 0.778. Five hospitals, accounting for 23%, exhibit effective TE, indicating good resource utilization and efficient operation. These hospitals benefit from strong supervisory mechanisms and smooth communication between management levels. However, 16 hospitals, or 76%, have ineffective TE, highlighting the need for improvement in their operations. Most hospitals have compensable operational inefficiencies, with TE values ranging from 0.6 to 0.8, indicating low overall efficiency. Further analysis of pure technical efficiency (PTE) and scale efficiency (SE) is needed. The lowest TE values belong to DMU11 and DMU10, both county-level TCM hospitals, suggesting a correlation between operational efficiency and geographic factors. These factors will be removed through stochastic frontier analysis in future steps.

Pure technical efficiency (PTE) measures input productivity at the optimal scale, reflecting the effectiveness of resource management (55). Hospitals with low PTE are far from achieving optimal efficiency. According to Table 6, the average PTE of TCM hospitals is 0.813. 5 hospitals (23%) have effective PTE, indicating maturity in management and technology, while 16 hospitals (76%) have ineffective PTE, showing significant room for improvement. The majority of these hospitals have PTEs between 0.6 and 0.8, suggesting weaknesses in management and technology. Enhancing the training of TCM technicians and improving hospital management practices can lead to quicker improvements. By optimizing resource use, strengthening process management, and implementing scientific methods, the technical efficiency of TCM hospitals can be enhanced (56, 57).

Scale efficiency (SE) reflects the development stage of Hebei Provincial Hospital of Traditional Chinese Medicine. While continuous investment in fixed assets may negatively impact hospital operations, the theory of returns to scale suggests that efficiency rises initially and then declines. To optimize investment, it is necessary to reduce input for declining returns (DRS) and increase input for increasing returns (IRS) (58). Table 6 shows that 5 out of 21 hospitals are efficient and at the fixed SE stage (23.8%), one hospital has decreasing SE (4.8%), and 15 hospitals (71.4%) have increasing SE. The data indicate most hospitals are improving their SE, suggesting the need for further investment by the government to enhance operational efficiency. However, some hospitals are experiencing diminishing returns, signaling issues such as resource wastage, internal management problems, and miscommunication (59).

According to Appendix Table A4, we can see that the average values of TE, PTE, and SE of TCM hospitals are high, but the minimum values are unstable, and the smallest comprehensive efficiency is already below 0.4, which indicates that the development of TCM hospitals is very unbalanced. This is partly due to environmental and location factors, which need to be determined and eliminated by parametric methods.

According to the development of the efficiency of TCM hospitals, those with efficiency scores greater than 0.9 are defined as the high band, 0.6 to 0.9 as the medium efficiency band and below 0.6 as the low efficiency band. From the analysis of the efficiency bands in Table 7, it can be seen that the majority of TCM hospitals are in the high and middle bands of the efficiency scores. There are 6 TCM hospitals in the high band of TE, 12 in the middle band, and 3 in the low band; 8 TCM hospitals in the high band of PTE, 12 in the middle band, and 1 in the low band; and 17 TCM hospitals in the high band of SE, and 4 in the middle band. In general, there are relatively few TCM hospitals in the low-efficiency band. The efficiency levels provide a practical reference for hospital administrators and policymakers to focus on performance improvement efforts. Hospitals in the low-efficiency band may require capacity building programs or external support, while high-efficiency hospitals could serve as benchmarks for best practices. Moreover, identifying hospitals in the middle band offers opportunities for transitional support strategies to elevate them to higher efficiency levels.

According to the National Health Commission's classification, Chinese medicine hospitals in Hebei Province are categorized based on bed numbers: small (below 900 beds), medium (900–1,200 beds), and large (above 1,200 beds). Appendix Table A5 shows that 3 medium and small-sized hospitals (14.3%) have average efficiencies of 0.802, 0.749, and 0.845; 6 medium-sized hospitals (28.6%) have average efficiencies of 0.855, 0.891, and 0.939; and 12 large-sized hospitals (57.14%) have average efficiencies of 0.837, 0.897, and 0.856. The TE for hospitals of these sizes averages around 0.85.

The data reveal that medium-sized hospitals (900–1,200 beds) have the highest efficiencies, followed by large-sized hospitals (over 1,200 beds) and small-sized hospitals (under 900 beds). This suggests that maintaining around 1,000 beds optimizes hospital efficiency, aligning with National Health Commission guidelines. Public hospitals should avoid the uncontrolled expansion of fixed assets like beds, as it can waste resources and decrease efficiency.

SFA, or stochastic frontier analysis, introduces influencing factors through parametric methods and is able to link the correlation between multiple inputs and outputs, which is applicable to the efficiency evaluation of various fields. Through the SFA model, we can get the specific impact of the external environment and statistical noise on the efficiency of hospitals, and in the subsequent study, the random factors are eliminated, and only the relevant factors of the hospital's own management and inputs are retained, to provide a more objective and realistic evaluation of the TE of each type of hospital (60). According to the previous section on indicator selection, we choose the logarithmic values of these two variables for comparative regression analysis because the government's financial input and science and education-related inputs are externally determined, and the data are accurate and easy to obtain, and have nothing to do with the hospital's own management and configuration factors, and for a number of other reasons.

The regression results are shown in Table 8a. Based on the λ-values, we can see that the equations for slack variables 1, 2, and 3 are significant, which suggests that the number of healthcare workers, fixed assets, and number of beds in operation in Chinese medicine hospitals are affected by both external environmental factors and management factors. The λ values of the equations for slack variables 1 and 3 are at a certain distance from 1, indicating that random noise also affects the input variables and that management and random factors need to be separated. In the SFA equation for slack variable 3, the logarithm of financial and technological inputs is “+” and is significant at the 1% level, which means that increasing financial and technological inputs has a positive effect on the investment in fixed assets of TCM hospitals and that an increase in the number of resources needed to produce the same output does not help to increase TE. In the regressions of input variables 1 and 2, the logarithm of scientific research inputs is (–), and the coefficient in the regression for healthcare technicians is significant at the 1% level. This indicates that investment in science and education can increase the utilization of human resources; in the regression of input variable 2, the logarithm of financial input is (–), and the coefficient of the regression of the number of beds at the end of the year is significant at 1%, which indicates that the higher financial input can increase the utilization of beds, further proving the effect of the environmental variables on the efficiency of the hospitals.

Table 8b indicates that a “+” sign means the input indicator positively impacts resource utilization, thereby improving technical efficiency. Further analysis shows that in Hebei Province's Chinese medicine hospitals, fixed asset input and bed numbers are significantly influenced by regional factors, affecting the accuracy of internal efficiency measurements. The input of health and technical personnel is influenced by random factors. Local financial conditions impact the allocation of fixed assets and bed numbers, thus affecting hospital efficiency. This aligns with our analysis, as central cities, with more financial and educational resources, experience higher efficiency in their local hospitals.

Except for bed input, financial inputs negatively affect the other efficiency intermediaries, suggesting that continued subsidies may expand inputs without improving hospital efficiency. Furthermore, medical and nursing personnel are still scarce in TCM hospitals. Investing in scientific education can enhance the technical efficiency of TCM hospitals. Cultivating more TCM practitioners and improving bed utilization will help meet the growing demand for TCM services.

One limitation of the DEA model is that it cannot rank multiple effective DMUs (efficiency = 1). To address this, we used SFA regression analysis to eliminate external environmental factors and noise, then conducted Super-SBM analysis on the SFA-adjusted input data. The technical efficiency results for TCM hospitals are shown in Table 9. From 2017, DMU4, DMU2, DMU8, DMU3, and DMU7 hospitals were deemed efficient, with DMU4 being the most efficient, followed by DMU2 and DMU8. In 2018, DMU2 hospitals were rated the most efficient, followed by DMU4 and DMU5. This approach allows for year-to-year analysis of efficiency changes, facilitating recommendations and evaluations for each hospital.

Compared to the traditional DEA method (see TE values in Table 6), DMU10 and DMU11 county hospitals ranked last in technical efficiency. After adjusting for external and random factors, DMU10 moved up three places, indicating that local financial and educational conditions contributed to better efficiency. Although there is still room for improvement, DMU10 is no longer last. Conversely, DMU11 remained at the bottom in 2018, highlighting that its poor technical efficiency is due to internal management issues rather than external factors.

Through Figure 4, we analyze the efficiency of 21 TCM hospitals in Hebei Province in a 2-year comparison. We can see that the technical efficiency of 14 TCM hospitals has been slightly improved, the efficiency of 2 TCM hospitals has a huge drop, and the other hospitals remain in their reasonable range of stable fluctuations, which indicates that the technical efficiency of TCM hospitals as a whole did not get big progress in 2018 and that the government's relevant input policies still need to be improved in 2018, especially because some hospitals have a huge drop in TE, which is a great waste of resources and management deficiencies that must be brought to the attention of regulators. In particular, In particular, DMU6 and DMU7 experienced significant declines in technical efficiency, dropping from rank 12 to 13 and from rank 5 to 12, respectively. These shifts may reflect local mismanagement, resource misallocation, or disruptions associated with the ongoing implementation of the hierarchical medical reform, such as patient diversion or cost-sharing mechanisms. Conversely, DMU10 rose three positions in ranking after adjustment, suggesting that improved local conditions—such as financial input, administrative oversight, or staff retention—may have begun to alleviate prior inefficiencies. Such rank changes, when viewed alongside environmental adjustments, provide indirect evidence on the differential institutional responses to provincial policies and internal constraints. This also highlights the need for localized management strategies beyond uniform policy incentives.

The modified three-stage method accurately measures the technical efficiency of hospital units, eliminating the impact of environmental and random factors. This allows us to assess efficiency based solely on internal decision-making, offering a more reliable measure. However, this study is static and does not capture dynamic changes in efficiency for Chinese medicine hospitals in Hebei. To address this, we applied the Malmquist model, re-analyzing the Malmquist index with input data from the corrected three-stage model. This allows us to evaluate the dynamic technical efficiency of Hebei's public Chinese medicine hospitals from 2014 to 2018. The Malmquist model measures the growth rate of total factor productivity (TFP). When TFP > 1, efficiency improves, and when TFP < 1, efficiency declines:

where Malmquist Productivity Index (MPI) represents the total factor productivity change index, TECI is the technical efficiency change index, TCI is the technical progress change index, PTECI is the pure technical efficiency change index, and SECI is the scale efficiency change index. This formula reveals not only year-to-year changes in efficiency but also the technological development of hospitals (61).

We used the MPI model to measure and decompose the total factor productivity (TFP) change in TCM hospitals across regions over time, as shown in Table 10. The MPI index for Hebei's TCM hospitals decreased slightly over five years, with an average value of 0.983, representing a 1.7% decline. This change results from a 5.8% decrease in technological change and a 1.5% increase in technical efficiency. The increase in technical efficiency is driven by a 0.2% rise in pure technical efficiency (PTECI) and a 0.7% rise in scale efficiency (SECI).

The decline in productivity is mainly due to technological regression, with minimal improvement from technological efficiency. Despite internal efforts to optimize operations, technological regression driven by external factors is the primary cause. Specifically, technical change increased by 11.6% in the first period but decreased by 11.3% in the second, and by 3.8% and 4.3% in the third and fourth periods, respectively. This is likely related to the decline in TCM hospitals' technical levels, the migration of TCM professionals, and imbalanced incentives. The sustained drop in TCI coincides with the implementation of the Development Plan for TCM Health Services (2015–2020), which prioritized infrastructure construction, integration of TCM into elderly care, and prevention-focused services. These priorities temporarily redirected investment and institutional focus toward primary-level development rather than hospital-based technological advancement. The rate of technical efficiency change increased by 17.0% in the second period after a 13.2% drop in the first, and then increased by 2.2% and remained unchanged in the third and fourth periods. This fluctuation in TECI aligns with the phased rollout of Hebei's Hierarchical Medical System Implementation Guidelines (2015), which began in 2015 and achieved province-wide coverage by 2017. By reshaping referral patterns and redistributing patient flow, the reform indirectly affected the case mix and staffing needs of public TCM hospitals, contributing to shifts in technical efficiency alongside other operational adjustments such as depreciation accounting and ICDS expansion. The spike in technical efficiency in the second period may be due to changes in depreciation expense accounting in 2016 and the expansion of the Integrated Care Delivery System (ICDS) in 2015, which required more nursing staff. Lower bed utilization and higher labor costs further reduced operational efficiency (62). From 2016–2018, despite integrated care reforms, technical efficiency showed minimal real improvement beyond accounting effects.

According to Appendix Table A6, the technical efficiency change index (TECL) shows that from 2014 to 2018, DMU4, DMU6, DMU8, and DMU9 experienced a decline in technical efficiency. For DMU4 and DMU6, the decrease was mainly due to a drop in pure technical efficiency, with resource waste from mismanagement being the key factor. For DMU8 and DMU9, the decline in efficiency resulted from both reduced pure technical efficiency and scale efficiency. By decomposing efficiency changes, we can identify the factors affecting hospital technical efficiency and service production over time, allowing for targeted interventions to improve the overall operational efficiency of TCM hospitals (63).

In the previous section, we used the modified three-stage model to measure the technical efficiency of Hebei Provincial Hospital of Traditional Chinese Medicine (HBHM) and analyzed the results. However, as this is a static analysis, we applied the Malmquist index model to examine dynamic efficiency, breaking down total factor productivity change into technology change, technical efficiency change, pure technical efficiency change, and scale efficiency change. Still, we cannot quantitatively determine which specific input and output factors affect the technical efficiency of Hebei TCM hospitals. Therefore, in this section, we apply the Tobit model to regressively analyze these factors (64).

The Tobit model, a restricted dependent variable model, is suitable as the technical efficiency values derived from the modified three-stage model are continuous and greater than 0. Traditional OLS regression would yield biased results, so we use maximum likelihood estimation with the Tobit model.

We selected variables based on operational and financial factors affecting hospital efficiency. After removing environmental influences in the three-stage model, we considered data availability, representativeness, and comprehensiveness, including operational indicators like hospital scale, medical services, and financial factors such as income structure, development capacity, asset operation, and cost management (65).

Operational factors include the number of health and technical personnel, beds, and depreciation expense, reflecting hospital scale. Bed utilization, outpatient numbers, and average length of stay represent service quality, with higher outpatient numbers indicating greater patient acceptance and a shorter length of stay reflecting better medical technology.

Financial factors include drug and sanitary material expenditure ratios, management expense ratios, personnel expense ratios, net fixed asset ratios, total asset turnover ratios, and outpatient income cost ratios. Higher ratios suggest poor internal management, while lower personnel expense ratios reflect better resource allocation. Net fixed asset ratios show asset condition, while total asset turnover ratios indicate operational strength. The outpatient income cost ratio reflects cost management, with higher ratios signaling better resource utilization and internal control. See Figure 5 for specific indicators.

In this paper, the value of the technical efficiency of TCM hospitals in Hebei Province obtained by the third stage SBM model is used as the explanatory variable, and the above indicators are used as the dependent variables. To avoid coefficient bias, the Tobit regression model is used to analyze the factors influencing the technical efficiency of TCM hospitals in Hebei Province in 2014–2018, and the specific regression equations are as follows:

The relevant data from 2014–2018 was brought into the software for Tobit regression, and based on the modeling assumptions above, we obtained the regression results as shown in Appendix Table A7.

The panel data regression results indicate that five variables significantly affect the technical efficiency of Hebei Provincial Hospital of Traditional Chinese Medicine at the 5% level: depreciation expense, bed utilization rate, management expense ratio, personnel expense ratio, and total asset turnover ratio. Depreciation expense and management expense ratio negatively affect technical efficiency, while bed utilization rate, personnel expense ratio, and total asset turnover ratio positively impact it. Additionally, the number of beds negatively influences technical efficiency at the 10% significance level.

Regarding operational factors, only depreciation expense and bed utilization rate were statistically significant at the 5% level. Depreciation expense, which reflects the size of fixed assets, is negatively correlated with technical efficiency. Larger fixed assets lead to higher depreciation costs, which in turn reduce technical efficiency. This suggests that Hebei hospitals have older fixed assets, and future investment in new equipment could enhance service and efficiency (66). Bed utilization rate has a positive effect, as improving this rate enhances hospital service efficiency, reduces average hospitalization time, and improves technical efficiency. The negative coefficient for the number of beds at the 10% significance level suggests that further expansion of bed numbers could harm overall efficiency. This may be due to Hebei's proximity to Beijing and Tianjin, where people tend to seek medical care, causing a bed oversupply in Hebei.

For financial factors, the management expense ratio, personnel expense ratio, and total asset turnover ratio were significant at the 5% level. The management expense ratio negatively impacts technical efficiency, indicating poor internal control and excess management personnel. The personnel expense ratio has a positive correlation with technical efficiency, suggesting that increasing salaries for health technicians can enhance efficiency. The total asset turnover ratio, however, negatively affects technical efficiency, implying that improving asset utilization and focusing on income generation could boost efficiency. This negative association may appear counterintuitive, but it reflects the fact that some hospitals with higher turnover ratios tend to prioritize revenue generation over optimizing internal processes or investing in long-term improvements. In such cases, high asset turnover could be driven by excessive reliance on outpatient volume or service expansion, which may not correspond to improved technical efficiency under the DEA framework. Moreover, these hospitals may be under pressure to meet fiscal performance targets, potentially at the cost of sustainable operational efficiency. Therefore, hospitals should optimize asset use, maintain current investment levels, improve services, and promote Chinese medicine to attract local patients.