This study utilizes official provincial-level annual data from the National Bureau of Statistics. The dataset covers the period from 2014 to 2023 and has been collated and validated for consistency. Following the classification standards of the National Bureau of Statistics, the 31 provinces of mainland China (excluding Hong Kong, Macao, and Taiwan) are categorized into three regions: Eastern, Central, and Western. The eastern region comprises 11 provinces: Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Hainan, Liaoning; The Central Region encompasses 8 provinces: Shanxi, Anhui, Jiangxi, Henan, Hubei, Hunan, Jilin, Heilongjiang; The Western Region includes 12 provinces: Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang.

This study applies the BCC model in Data Envelopment Analysis (DEA) to conduct a static efficiency measurement, and adopts the Malmquist index to evaluate the dynamics of total factor productivity (TFP). For comparative analysis, the 31 provinces in China are categorized into three broad geographical regions—eastern, central, and western—to further elucidate the spatial distribution patterns and interregional disparities in the efficiency of healthcare resource allocation.

Data Envelopment Analysis (DEA) is a non-parametric method used to evaluate the relative efficiency of decision-making units (DMU) with multiple inputs and outputs, and it has been widely applied in healthcare efficiency measurement (10). Among various DEA models, the Charnes-Cooper-Rhodes (CCR) and Banker-Charnes-Cooper (BCC) models are the most commonly applied (11). This study utilizes the DEA-BCC model, which assumes variable returns to scale, to assess three key efficiency metrics: Comprehensive Technical Efficiency, Pure Technical Efficiency, and Scale Efficiency. Comprehensive Technical Efficiency is the product of PTE and SE, with scores ranging from 0 to 1. A score of 1 indicates relative efficiency, whereas a score below 1 indicates relative inefficiency.

The Malmquist index model is a dynamic analysis framework used to examine changes in total factor productivity across different periods for decision-making units (12). An index value greater than 1 indicates an improvement in efficiency, while a value less than 1 signifies a decline. This index can be decomposed into two components: technical efficiency change and technological progress change. Through this decomposition structure, it is possible not only to identify trends in efficiency changes but also to pinpoint the primary sources of efficiency variation.

While the DEA-BCC model provides a static evaluation of resource allocation efficiency within a single year, the Malmquist index captures efficiency changes over time and distinguishes whether they stem from improvements in technical efficiency or advancements in technology. Since static analysis alone provides a limited view of dynamic changes, combining DEA with TFP analysis enables a more holistic assessment of both the current state and long-term trends in medical resource allocation. Therefore, TFP analysis is not separate from efficiency research but rather extends it by uncovering the underlying drivers of efficiency change.

This study not only follows the basic requirements of the data envelopment analysis model for the number of decision making units when selecting indicators, that is, the number of decision making units is at least twice the total number of input and output indicators, but also draws on the relevant research results of Xu Pingping (13), Duan Guimin (14) and Cui Chengsen (15). Finally, the input and output indicators is determined: the input indicators selects health technicians, institutional beds and medical facilities from the three levels of manpower, material resources and institutional scale, which correspond to the core manpower, material resources and infrastructure resources of the medical and health system, and can directly reflect the accessibility of medical and health service capabilities. The output indicators focus on the core achievements of medical services, and select the number of patient visits and hospital discharges. It not only captures the intensity of service utilization, but also is a classic indicator in the field of medical and health efficiency evaluation. Together, it comprehensively reflects the supply capacity and actual service output level of the provincial medical and health system.

The period 2014–2023 witnessed commensurate expansion across all three input indicators within China’s healthcare system. Health technicians demonstrated the most pronounced growth trajectory at an average annual rate of 5.69%, followed by institutional beds (4.92%) and medical facilities (4.48%). In contrast, the growth of output indicators remains modest and exhibits considerable volatility. The average annual growth rates for outpatient visits and hospital discharges are 2.57 and 4.45%, respectively. Notably, service output experienced significant fluctuations in 2020 due to the COVID-19 pandemic, with the number of patient visits and discharges declining by 11.23 and 13.29%, respectively. This sharp contraction can be primarily attributed to constrained access to healthcare facilities, the deferral of non-urgent medical care, and a reduced propensity of patients to seek treatment amidst pandemic control measures. A gradual recovery commenced in 2021–2022, culminating in a rebound to record-high levels by 2023. Over the ten-year span, the cumulative growth in resource inputs (health technicians: +64.54%; hospital beds: +54.12%; medical facilities: +48.32%) generally surpassed the growth in service volume (patient visits: +25.64%; hospital discharges: +47.93%). This suggests a suboptimal translation of resources into effective service delivery, highlighting issues of low utilization efficiency (Table 1).

To assess the robustness and reliability of the findings, this section performs robustness checks along two dimensions: model specification and variable selection. Given that Data Envelopment Analysis is sensitive to its underlying assumptions and input–output variable sets, we re-estimate efficiency scores by altering key model configurations. This approach tests whether the core conclusions remain unchanged under alternative methodological setups.

First, in the context of static efficiency analysis, we recognize that medical service output is multidimensional and that the original indicator, “number of discharged patients”primarily captures service volume. To better reflect the technical complexity and resource intensity of care, we substitute it with “Inpatient Surgical Procedures” sourced from the National Bureau of Statistics. Efficiency scores are then re-estimated using the DEA-BCC model and compared against the baseline. The Spearman rank correlation test indicates a statistically significant positive correlation between the two sets of efficiency rankings (ρ = 0.66, p < 0.001), demonstrating that the static efficiency evaluation maintains moderate stability and robustness under this alternative output measure.

Second, for the dynamic efficiency analysis, we examine the potential influence of returns-to-scale assumptions on total factor productivity measurement. Specifically, the Malmquist index is recalculated using the DEA-CCR model and compared with the results from the DEA-BCC model. A significant positive correlation is found between the indices derived from the two models (ρ = 0.56, p < 0.001). Although there are some differences in the specific values of the productivity index due to different assumptions of returns to scale, the overall trend is consistent, indicating that the conclusion of this paper on the dynamic evolution of medical resource allocation efficiency has good robustness.

Between 2014 and 2023, China’s healthcare system underwent substantial expansion in both infrastructure and human resources, evidenced by cumulative growth of 64.54% in health technicians, 54.12% in hospital beds, and 48.32% in healthcare institutions. However, the growth rates for outpatient visits and hospital discharges were only 25.64 and 47.93%, respectively—significantly lower than the pace of resource investment. The asymmetry between input and output growth rates indicates a structural imbalance in the allocation of medical resources. Further analysis of slack variables in DEA-inefficient provinces reveals that this imbalance is manifested as the coexistence of input excess and output shortfall. This shows that the efficiency of resource utilization has not been improved simultaneously with the input, which is reflected in the obvious resource mismatch at the provincial level. In 2023, the average comprehensive technical efficiency across 31 provinces stood at 0.869, with average pure technical efficiency and scale efficiency at 0.913 and 0.954, respectively. Over half of the regions failed to achieve DEA effectiveness, these findings collectively indicate suboptimal resource utilization efficiency nationwide, highlighting substantial potential for optimization.

Therefore, the government should comprehensively account for local healthcare demands and demographic distribution patterns, enhance regional health planning, and optimize existing medical resources. It is essential to avoid haphazard expansion of tertiary hospitals to achieve precise alignment between resource allocation and public health needs. Concurrently, policy should encourage patients with common and chronic conditions to utilize primary care facilities, thereby fostering complementary development and synergistic use of healthcare resources. Furthermore, establishing multidimensional performance metrics centered on service quality, patient satisfaction, and cost containment is essential to transition from a volume-oriented to an efficiency-oriented healthcare delivery (16). Finally, according to local conditions, redundant human resources should be transformed into talents that meet the needs of different levels of institutions through standardized training, specialist training and other ways.

From a regional perspective, the eastern region demonstrates the highest efficiency in healthcare resource allocation, with its comprehensive technical efficiency significantly exceeding the national average. In contrast, the central region exhibits the lowest efficiency, while the western region slightly outperforms the central region yet remains below the national benchmark. Three primary factors explain these disparities. First, the eastern region benefits from robust economic foundations within the Yangtze River Delta, Pearl River Delta, and Bohai Rim economic circles, which consistently support technological upgrades and talent development. Since the 18th National Congress of the Communist Party of China, the Party Central Committee has introduced a series of major regional strategies, including Beijing-Tianjin-Hebei coordinated development and Yangtze River Delta integration have enabled the eastern region to pioneer tightly-knit medical alliances and telemedicine networks, thereby enhancing technical efficiency and resource coordination. Second, although the central region rise strategy emphasizes equalization of public services, specific policies have largely focused on infrastructure construction, with insufficient attention to talent cultivation and technological innovation. Third, provinces including Xinjiang, Tibet, and Guizhou have utilized national counterpart assistance mechanisms to establish preliminary regional collaboration and telemedicine platforms, enhancing healthcare service capabilities in select remote areas. Additionally, regional strategies like the Chengdu-Chongqing Economic Circle have brought new opportunities to the western regions, partially alleviating healthcare workforce shortages. This dual approach of “external support and endogenous capacity building” has progressively optimized healthcare resource allocation efficiency in western China.

The Guiding Opinions call for optimized spatial distribution of healthcare resources to bridge urban–rural and interregional disparities, and narrow the gap in healthcare resource distribution between central-western and eastern regions. Therefore, Implementation should prioritize telemedicine infrastructure, medical consortium development, and targeted technical assistance to foster balanced regional development. According to the “Opinions of the Central Committee of the Communist Party of China and the State Council on Promoting High-Quality Development in the central region in the New Era” synergies should be strengthened with major regional strategies such as Beijing-Tianjin-Hebei coordination and Yangtze River Economic Belt integration. By absorbing the relocation of high-quality medical resources from the Beijing-Tianjin-Hebei region and collaborating with the Yangtze River Economic Belt to establish a cross-regional medical technology coordination system, the efficiency gap in regional healthcare resource allocation can be narrowed. Concurrently, refined incentive policies for healthcare professionals, promoting their migration to underdeveloped central and western regions through enhanced compensation and welfare measures (17).

Between 2014 and 2023, China experienced an average annual decline of 1.6% in total factor productivity, with a mean value of 0.984. Technological efficiency remained stable during this period, while technological regression—also decreasing at 1.6% per year—emerged as the principal determinant of TFP deterioration. One contributing factor is the significant lag in the iterative upgrading of medical equipment, particularly within primary healthcare institutions at or below the county level, the allocation rate of high-end medical equipment remains low, and advanced technological innovations struggle to permeate effectively to these grassroots levels (18). Secondly, a structural shortage exists in the supply of intermediate and senior healthcare technical personnel. In particular, the central and western regions face an insufficient total number of such personnel, imperfect talent echelon development, and a lag in the professional competency of staff to operate new technologies and equipment. This makes it difficult to translate hardware investments into tangible improvements in service efficiency (19).

From a spatial evolution perspective, the eastern and western regions demonstrated gradual TFP improvement, whereas central China exhibited lower levels and slower growth. Moreover, all provinces recorded technical progress values below 1, reflecting a dual challenge of systemic inefficiency and significant regional fragmentation in healthcare technology advancement. Although the national implementation plan advocates for “expanding high-quality medical resources and achieving balanced regional distribution” (2), policy execution displays marked regional disparities. The eastern region leverages economic advantages to attract skilled professionals and introduce advanced equipment, creating a virtuous cycle; the western region has established telemedicine platforms through policy support, while in the central region, high-quality technical resources are concentrated in provincial capitals. This centralization, coupled with slow equipment renewal and staffing shortages in county-level institutions, has created a hierarchical fragmentation in technology diffusion that fundamentally constrains TFP growth.

Consequently, the Healthy China 2030 Blueprint should be operationalized by institutionalizing healthcare technology R&D investment as a performance metric in local government health assessments, while curricular expansion in medical technology innovation should be implemented at central China’s universities to cultivate an indigenous technical talent pool. Furthermore, a structured collaboration framework for technology translation should be established between eastern and central regions, supported by creating specialized intermediary hubs for medical technology transfer and regional healthcare centers. In alignment with the Implementation Plan’s primary-care competency standards, sustainable training mechanisms for grassroots medical personnel must be established. Telemedicine operations and smart device utilization should be incorporated into competency evaluations, ensuring technologies become both accessible and applicable in routine practice.

The replicable common characteristics observed in the eight DEA-efficient provinces identified in this study offer a clear optimization pathway for inefficient provinces. First, they exhibit a coordinated resource input structure that effectively avoids single-factor redundancy. These provinces maintain a more balanced ratio among health technicians, beds, and institutions (20). In contrast, inefficient provinces frequently exhibit problems of over-investment in a single resource, exemplified by the excessive number of hospitals in Hebei and the surplus of beds in Heilongjiang. This observed redundancy underscores that an optimal allocation structure is more critical than merely increasing investment. Second, technology and management innovation drive intensive, quality-oriented development. Most of these efficient provinces are pioneers in healthcare reform. Regions such as the Yangtze River Delta and the Pearl River Delta lead in promoting smart hospitals and telemedicine to enhance service efficiency (21). Meanwhile, Guangxi and Chongqing have facilitated resource allocation to grassroots levels through the establishment of tight-knit medical alliances (22). Inefficient provinces, conversely, commonly lag in technology adoption and management practices. Third, they possess sound regional coordination mechanisms. Leveraging national-level strategies, they have constructed cross-regional cooperation networks, breaking down the localization of resources through a dual approach of “external linkage and internal integration” (23). The Beijing-Tianjin-Hebei coordinated development initiative and the Yangtze River Delta integration strategy serve as typical examples.

To enhance the resource allocation efficiency of inefficient provinces, it is imperative to move beyond the mindset of merely increasing investment. In accordance with “Guidance on Optimizing the Layout and Construction of Primary Healthcare Institutions” (24), a dynamic resource monitoring and adjustment mechanism should be established. This involves regularly evaluating and rebalancing key factors such as beds and institutions, and implementing a two-way flow mechanism between tertiary and primary healthcare—combining downward rotations with dispatched training. This approach not only revitalizes redundant resources but also strengthens primary-level service capabilities.

It is equally crucial to strengthen the dual drivers of technology and management. On one hand, it is essential to promote state-advocated digital service models, such as the “telemedicine service network” and the model of “distributed inspection with centralized diagnosis.” On the other hand, the requirements of the DRG/DIP payment reform must be fully implemented to promote the integration of digital tools with performance management and regulatory mechanisms.

Furthermore, based on “the Opinions on Further Ensuring and Improving People’s Livelihood and Striving to Resolve the People’s Urgent Concerns and Expectations,” provinces should integrate into national regional strategies. By leveraging tightly knit county-level medical consortia, they can compensate for shortages of high-quality resources through mechanisms for co-construction, resource sharing, and talent support.

This study has several limitations that should be acknowledged. First, the output indicators selected in this study are primarily measures of service volume. While such indicators capture aspects of process and quantity, they do not encompass outcome-oriented measures related to health status or service quality. This may compromise the comprehensiveness of the efficiency assessment. Consequently, this limitation affects the interpretation of the DEA results. For instance, a high DEA efficiency value merely indicates that a region efficiently produces service volume given its inputs; it cannot be extrapolated to infer superior service quality or health improvement outcomes. Second, the data envelopment analysis model employed in this study does not incorporate external environmental factors as control variables. While this approach facilitates the measurement of resource transformation capability from a purely technical perspective, it also implies that the estimated efficiency scores may be substantially influenced by prevailing environmental characteristics. Consequently, this may limit the explanatory power of the findings regarding causal relationships. Third, DEA results are sensitive to indicator selection and may not reflect stochastic variations. Finally, although we provide an optimization path by combining efficiency decomposition and redundancy analysis, due to the limitation of macro data and research methods, we fail to conduct detailed cost–benefit analysis, local adaptability assessment or priority ranking for each recommendation. Future research should incorporate quality-oriented indicators to complement existing measures of service volume. Furthermore, external environmental factors should be explicitly integrated into the analysis. This would provide a more robust explanation for regional efficiency disparities and strengthen the causal interpretation of the findings.