Preamble

Research on the Mechanism of Artificial Intelligence Empowering Primary Health Care Service Performance

Abstract

As a core component of the "Healthy China" strategy, the performance of primary health care services directly impacts the accessibility and equity of national health services. With the rapid development of digital technology, artificial intelligence (AI) has become a critical driver for the transformation and upgrading of primary health care. This study explores the mechanism by which AI empowers primary health care service performance, aiming to provide a theoretical basis and practical guidance for improving the quality and efficiency of grassroots health services. By analyzing the multi-dimensional impact of AI on clinical decision support, chronic disease management, and administrative efficiency, this paper constructs a theoretical framework for AI-driven performance enhancement.

1. Introduction

Primary health care institutions serve as the "first line of defense" for public health. However, these institutions currently face challenges such as a shortage of high-quality medical resources, insufficient technical capabilities of general practitioners, and heavy administrative burdens. The integration of machine learning, deep learning, and natural language processing into the primary care setting offers a transformative opportunity to address these bottlenecks. AI empowerment is not merely a technological upgrade but a systemic reconstruction of service delivery models.

2. Theoretical Framework and Mechanism Analysis

The impact of AI on primary health care performance can be analyzed through three primary dimensions: clinical precision, operational efficiency, and patient engagement.

2.1 Enhancing Clinical Decision Support

AI-powered clinical decision support systems (CDSS) utilize machine learning algorithms to analyze patient data against vast medical databases. For general practitioners at the grassroots level, these tools provide evidence-based recommendations for diagnosis and treatment, effectively reducing the rate of misdiagnosis and ensuring that clinical practice aligns with the latest medical guidelines. This technological support bridges the "knowledge gap" between primary care providers and specialists in tertiary hospitals.

2.2 Optimizing Chronic Disease Management

Chronic disease management is a cornerstone of primary health care. AI facilitates a shift from reactive care to proactive, personalized health management. Through predictive modeling, AI can identify high-risk populations and provide early warnings for complications. Wearable devices integrated with AI platforms allow for continuous monitoring of physiological parameters, enabling timely interventions and improving the long-term health outcomes of residents.

2.3 Improving Administrative and Operational Efficiency

Administrative burdens often detract from the time general practitioners can spend with patients. AI applications, such as automated medical record generation and intelligent scheduling systems, significantly reduce the time spent on repetitive tasks.

School of Public Affairs, Zhejiang University, Hangzhou 310058, China

School of Mathematical Sciences, Zhejiang University, Hangzhou, Zhejiang Province; Jinhua Municipal Hospital of Traditional Chinese Medicine, Affiliated Hospital of Zhejiang Chinese Medical University, Jinhua, Zhejiang Province. Doctoral Supervisor: Kong Dexing; Chief Physician: Kong Jiangming.

Background

China is currently leveraging artificial intelligence technology to enhance the standardization and homogenization of primary healthcare services. This initiative is driving universal health coverage and demonstrating the nation's global leadership within the digital health sector.

This study empirically reveals the underlying mechanisms through which artificial intelligence empowers the performance of primary healthcare services and proposes corresponding optimization pathways.

Methods

This study focuses on a large-scale, multi-center policy pilot case involving the deployment of an "Ultrasound Artificial Intelligence-Assisted Diagnostic System" across 109 public medical institutions in Puyang City, Henan Province, from July 2022 to May 2024.

The research utilizes the organizational change dynamics model as its primary theoretical framework. Data collection was primarily conducted through questionnaire surveys. The methodological approach for data analysis includes descriptive statistical analysis, exploratory factor analysis (EFA), confirmatory factor analysis (CFA), analysis of variance (ANOVA), and structural equation modeling (SEM).

Results

A total of 429 valid questionnaires were collected for this study. The performance evaluation index system designed herein for AI-empowered primary health services comprises two dimensions: internal optimization performance and social adaptation performance. The results indicate that the social adaptation performance of system applications is higher than the internal optimization performance. Furthermore, the application of these systems not only yields direct performance outcomes, such as improved medical quality and increased operational efficiency, but also leads to more significant improvements in social adaptation performance, particularly regarding sustainability and satisfaction.

The primary contextual triggers for performance enhancement are identified as "policy environment," "industrial support," and "transformation of achievements." Meanwhile, the core enabling factors consist of "medical insurance support," "technical proficiency," and "purchasing power." Based on these findings, three critical optimization paths are proposed to enhance the performance of AI-empowered primary health services:

1. "Policy Environment / Industrial Support $\rightarrow$ Technical Proficiency $\rightarrow$ Social Adaptation Performance / Internal Optimization Performance"
2. "Policy Environment / Industrial Support $\rightarrow$ Purchasing Power $\rightarrow$ Social Adaptation Performance"
3. "Policy Environment / Transformation of Achievements $\rightarrow$ Medical Insurance Support $\rightarrow$ Social Adaptation Performance / Internal Optimization Performance"

Conclusion

Analysis of the Influence Mechanism of Artificial Intelligence Empowering the Performance of Primary Health Services

Based on a comprehensive understanding of the dual value of deploying medical artificial intelligence (AI) devices—namely, their profitability and their public utility—the public sector should employ a variety of policy instruments. Efforts should focus on creating a leading and encouraging policy environment, enhancing the completeness of industrial supporting facilities, and accelerating the cultivation of mechanisms for the transformation of scientific achievements. Continuous progress must be made in improving the technical standards of medical AI equipment, strengthening the purchasing power of primary healthcare institutions, and accelerating the inclusion of AI diagnostic applications into medical insurance payment systems. These actions are essential to further enhance the performance of AI in empowering primary health services.

Keywords: Artificial Intelligence; Primary Health Services; Empowerment; Performance; Influence Mechanism; Policy Pilot; Structural Equation Model

1. Introduction

The integration of artificial intelligence into the healthcare sector represents a transformative shift in how medical services are delivered, particularly at the primary care level. As medical AI devices possess both commercial potential and significant public value, understanding the mechanisms through which they improve service performance is critical for policymakers and healthcare administrators alike.

Primary health services serve as the frontline of the healthcare system. However, these institutions often face challenges such as limited diagnostic capabilities, a shortage of skilled medical personnel, and insufficient funding. Artificial intelligence offers a promising solution to bridge these gaps by providing high-accuracy diagnostic support, optimizing administrative workflows, and enabling personalized patient care.

Despite the potential benefits, the adoption of AI in primary healthcare is not without obstacles. Issues ranging from high procurement costs and technical complexity to the lack of standardized reimbursement frameworks hinder widespread implementation. Therefore, it is necessary to analyze the specific pathways through which AI empowers primary health service performance and to identify the policy levers that can facilitate this transition.

This study utilizes structural equation modeling (SEM) to explore the influence mechanisms of AI empowerment. By examining the interplay between technological advancement, institutional capacity, and policy support, we aim to provide a theoretical and empirical basis for optimizing the deployment of AI in the public health sector. Through targeted policy interventions—such as fostering a supportive regulatory environment and integrating AI services into medical insurance—the performance of primary health services can be significantly enhanced, ultimately leading to better health outcomes for the general population.

Zhejiang University, Hangzhou 310058, China.
Zhejiang Chinese Medical University, Jinhua 321017, China.
Kong Dexing: Professor/Doctoral supervisor.
Kong Jiangming: Chief physician.

Background

China is using artificial intelligence technology to enhance the standardization and homogeneity of primary health services and drive universal health coverage.

Objective: To empirically reveal the mechanism by which artificial intelligence (AI) empowers the performance of primary health services and to propose corresponding policy recommendations.

Cite this article:
KONG Zihe, BIAN Qingyang, KONG Dexing, et al. Analysis of the influence mechanism of artificial intelligence empowering the performance of primary health services [J]. Chinese General Practice, 2025. DOI: 10.12114/j.issn.1007-9572.2024.0536. [Epub ahead of print].

Editorial Office of Chinese General Practice. This is an open access article under the CC BY-NC-ND 4.0 license.

Methods

A large-scale and multi-center policy pilot case on the ultrasound AI-assisted diagnostic system (the AI system) deployed in 109 public medical institutions in Puyang City, Henan Province, from July 2022 to May 2024, was selected for research. The Organizational Change Dynamics Model was used as the main theoretical framework, and a questionnaire survey was used as the main data collection method. Descriptive statistical analysis, exploratory factor analysis, confirmatory factor analysis, variance analysis, and structural equation model analysis are the main data analysis methods.

Results

A total of 429 valid questionnaires were obtained. The performance evaluation index system of artificial intelligence empowering primary health services designed in this study included two dimensions: internal optimization performance and social adaptation performance. The social adaptation performance of system applications was higher than that of internal optimization. Applications not only brought about direct performance results such as improvements in medical quality and operational efficiency, but also led to more prominent enhancements in social adaptation performance such as sustainable development and satisfaction. The main situational triggers for performance improvement were "policy environment", "industrial support" and "technology transfer", while the main enabling factors were "medical insurance support", "technological level" and "purchasing power". The three key optimization paths for enhancing the performance of primary health services empowered by artificial intelligence are "policy environment/industrial support $\rightarrow$ technological level $\rightarrow$ social adaptation performance/internal optimization performance", "policy environment/industrial support $\rightarrow$ purchasing power $\rightarrow$ social adaptation performance", and "policy environment/technology transfer $\rightarrow$ medical insurance support $\rightarrow$ social adaptation performance/internal optimization performance".

Conclusion

On the basis of fully understanding that the deployment of the AI system has the dual values of profitability and publicity, the public sector should adopt a variety of policy tools, starting from creating a leading and encouraging policy environment, strengthening the integrity of industrial supporting, and accelerating the cultivation of results transformation mechanism. It is essential to make efforts to improve the technical level of medical AI equipment, to strengthen the equipment purchasing ability of primary health institutions, to accelerate the inclusion of AI diagnostic applications in medical insurance payments, and to further improve the performance of AI enabling primary health services.

Key words: Artificial intelligence; Primary health services; Empowerment; Performance; Influence mechanism; Policy pilot; Structural equation model

China is currently seeking to utilize next-generation artificial intelligence (AI) technology to enhance the standardization and homogenization of primary health services. This initiative aims to drive universal health coverage and demonstrate global leadership in the field of digital health. General Secretary Xi Jinping has emphasized the need to establish a "Grand Health" concept, promoting the downward shift of health resources, advancing the construction of county-level medical communities, and improving primary infrastructure. These efforts are intended to enhance the quality and efficiency of medical services and ensure the effective operation of the hierarchical medical system. In recent years, AI technology has achieved revolutionary breakthroughs and has been effectively applied across various health sectors, including disease screening and prediction, health management, auxiliary diagnosis, chronic disease management, rehabilitation, and the integration of medical and elderly care. These advancements have expanded the scope of the health industry and created favorable conditions for improving the performance of primary health services in China. Theoretically, the empowering role of AI in primary health services includes at least the following aspects: first, improving the accuracy of disease diagnosis to enhance the operational efficiency of the medical system and save overall medical costs \cite{8-11}; second, using diagnostic standardization and intelligence as a foundation to enhance the feasibility of mutual recognition for laboratory and examination results, thereby avoiding repetitive testing and promoting the rebalancing of high-quality medical resources across regions \cite{12-13}; and third, improving the technical proficiency of general practitioners in primary health institutions \cite{14-15}, which enhances the equity and accessibility of medical services. However, due to practical limitations, few empirical studies have revealed how China's public sector can effectively enhance the impact of AI empowerment on primary health services.

This study selects a policy pilot case involving the deployment of an "Ultrasound AI-Assisted Diagnostic System" across 109 public medical institutions in Puyang City, Henan Province. Using the organizational change motivation model as a theoretical framework, the research employs questionnaire surveys as the primary data collection method and structural equation modeling (SEM) as the main data analysis method. The study empirically reveals the influence mechanism of AI empowerment on the performance of primary health services and proposes corresponding optimization paths.

1.1 Research Object

This study selects the large-scale, multi-center policy pilot case of the "Ultrasound Artificial Intelligence-Assisted Diagnostic System" deployed across 109 public medical institutions in Puyang City, Henan Province, from July 2022 to May 2024 as the research object. On July 11, 2022, the General Office of the People's Government of Puyang City issued the "Notice on the Promotion and Application of Artificial Intelligence-Assisted Diagnosis and Treatment Systems and Palm Vein Intelligent Identification Systems," initiating the deployment of medical artificial intelligence equipment in healthcare institutions throughout the region.

As a critical component of this initiative, the AI-SONIC ultrasound artificial intelligence-assisted diagnostic system relies on the "DE-Light" deep learning technology platform independently developed by Chinese experts. It integrates core frontier technologies from fields such as artificial intelligence, medical imaging, and information security to establish an intelligent diagnosis and treatment framework that combines image recognition, feature extraction, diagnostic judgment, and result generation. The system possesses several core functions, including ultrasound image standardization, real-time analysis, and high-precision recognition. It supports automatic detection, benign-malignant determination, and quantitative analysis for areas such as the thyroid, breast, carotid artery, and pelvic floor. Furthermore, it can automatically generate high-precision examination reports, making it widely applicable for diagnostic support, clinical training, and scientific research data collection.

As of May 31, 2024, the AI-SONIC system has been deployed across 109 public medical institutions at or above the primary level in Puyang City. A total of 291 modules have been installed and utilized, including 30 units in tertiary medical institutions, 66 units in secondary medical institutions, and 195 units in medical institutions below the secondary level. To date, the system has screened a total of 281,663 patient visits and detected 3,129 cases of thyroid nodules.

Following the deployment of the system, the diagnostic capabilities and service efficiency of primary healthcare institutions have been effectively enhanced. The system can reduce the time required to issue a traditional ultrasound report from approximately 15 minutes to just 0.2 seconds. Consequently, the average daily patient capacity of physicians has increased from 20–25 individuals to approximately 40 individuals, an efficiency increase of 37.5% to 50.0%.

The system achieves an impressive identification rate of 97% and a diagnostic accuracy for benign and malignant tumors exceeding 95%. These figures are significantly higher than the average performance of attending physicians in the ultrasound departments of China's top-tier (Grade 3A) hospitals, which typically hovers around 75%. Furthermore, the system provides more detailed examination data; for instance, when detecting carotid intima-media thickening, it generates precise measurements to two decimal places (e.g., 1.38 mm or 1.69 mm). This level of precision far surpasses the conventional ranges (such as 1–2 mm or 2–3 mm) typically provided by hospitals. Additionally, the system can automatically calculate stenosis ratios and other critical metrics.

The implementation of this technology has also led to a significant shift in patient flow, with nearly 75% of ultrasound patients returning from county-level hospitals to primary-care township and community health institutions. This shift has strengthened the capacity for initial diagnosis at the grassroots level and reduced the incidence of redundant examinations. From the perspective of medical insurance economics, if the system is integrated into insurance coverage at a rate of 5 to 10 RMB per person-visit, the incremental expenditure would represent only 1/14 to 1/7 of the costs associated with repeated examinations. This translates to a cost-saving ratio of 85.7% to 92.9%.

These data points fully demonstrate that the system plays a significant role in expanding screening coverage and enhancing the efficiency and accuracy of ultrasound diagnostics. This progress marks a demonstrative milestone in Puyang City’s development of an artificial intelligence-driven medical service system.

1.2.1 Concepts and Models

Perceived performance can accurately characterize the empowerment of behaviors and outcomes in primary health services by artificial intelligence \cite{16-17}. Perceived performance reflects the public's overall perception of service quality; compared to efficiency evaluations based on input-output ratios, it possesses more comprehensive explanatory power and better reflects public value. Based on the "Task Performance-Contextual Performance" model \cite{19-20}, this study enriches the performance evaluation dimensions by including internal optimization performance (such as medical quality and operational efficiency) and social adaptation performance (such as sustainable development and social satisfaction).

Furthermore, this study introduces the Organizational Change Dynamics Model as the theoretical framework for analyzing the influence mechanism. This framework effectively explains how the power transmission mechanism for improving performance is formed when primary healthcare institutions are equipped with medical Artificial Intelligence (AI) devices. The model emphasizes that improvements in organizational performance stem from the interaction between external driving forces and the internal capability responses of the organization.

Based on this logic, this paper defines achievement transformation, the policy environment, and industrial configuration as externally driven independent variables. These represent the mature foundation for AI technology entering medical practice, the degree of institutional support and protection, and the industrial ecological support for technology implementation, respectively. They serve as the key external situational triggers for the deployment of AI medical systems at the grassroots level.

Simultaneously, integrating the theoretical perspective of "capability adaptation and internal transformation" within organizational change, we further introduce technical level, purchasing power, and medical insurance support as mediating variables. These reflect the absorption and response capabilities of primary medical institutions toward technological change and constitute the core enabling factors within the organization. Specifically:
- Technical Level reflects the advancement of the AI system itself, serving as the technical prerequisite influencing the adoption intention and usage effectiveness of primary medical institutions.
- Purchasing Power reflects the financial capacity of primary institutions regarding equipment procurement and maintenance.
- Medical Insurance Support is a key factor in measuring whether the application of AI systems possesses the conditions for institutionalization and sustainability.

The logic behind these variable settings aligns with the "External Drive - Internal Response - Performance Outcome" theoretical chain of the Organizational Change Dynamics Model, providing a clear theoretical basis and structural support for the subsequent empirical analysis.

1.2.2 Data Collection Methods

Based on the theoretical concepts and models previously discussed, and in conjunction with the policy pilot implementation in Puyang City, the research team conducted multiple rounds of group discussions and expert consultations prior to questionnaire design to ensure the scientific rigor and practical utility of the measurement tools. The group consisted of six members with strong interdisciplinary backgrounds, including professors, senior medical professionals, and doctoral students from fields such as artificial intelligence (AI) application management, hospital management, and public administration. The expert consultants included technical specialists in AI medical system development, heads of medical informatics departments, and professors specializing in medical AI from higher education institutions. During the discussion process, the research team employed a hybrid "brainstorming-structured interview-iterative revision" approach. Initially, the questionnaire dimensions and item concepts were established through brainstorming sessions. Subsequently, semi-structured interviews were conducted to gather expert advice, and the content was repeatedly optimized based on their feedback. This process culminated in a formal measurement questionnaire based on a 7-point Likert scale (where 1 to 7 represent strongly disagree, very low agreement, low agreement, moderate agreement, high agreement, very high agreement, and strongly agree, respectively). The sampling survey was organized by the Puyang Institute of Big Data and Artificial Intelligence, with questionnaires distributed extensively to ultrasound practitioners in primary regions where ultrasound AI-assisted diagnostic systems have been deployed. Data collection was completed in May 2024, with a total of 493 questionnaires recovered, of which 429 were valid.

Statistical Methods

In this study, SPSS 27.0 software was utilized to perform descriptive statistical analysis, exploratory factor analysis (EFA), confirmatory factor analysis (CFA), and analysis of variance (ANOVA) on the collected questionnaire data. Quantitative data are expressed as ($\bar{x} \pm s$). Comparisons between two groups were conducted using independent samples $t$-tests, while comparisons among three or more groups utilized one-way ANOVA. If the assumption of homogeneity of variance was met, LSD post-hoc tests were referenced; otherwise, Tamhane’s T2 test results were used. AMOS software was employed to conduct structural equation modeling (SEM) path testing. The application performance of the ultrasound AI-assisted diagnostic system deployed at the primary level served as the dependent variable. Independent variables included achievement transformation, policy environment, and industrial configuration, while technical level, purchasing power, and medical insurance support served as mediating variables. The reliability of the evaluation system was assessed using Cronbach’s $\alpha$ coefficient, with a threshold of $>0.8$ indicating good reliability. Validity was assessed based on whether the KMO value was $>0.8$ and whether the significance level of Bartlett’s test of sphericity was $<0.05$; a KMO value $>0.8$ combined with a significance level $<0.05$ indicates good validity. The system application performance was calculated by summing the weighted indicators equally; a performance score $>5$ indicates high perceived system performance. A $P$-value of $<0.05$ was considered statistically significant.

2.1 Empirical Measurement of Performance

2.1.1 Sample Demographics, Indicator Design, and Reliability and Validity Testing

Among the 429 samples, there were 61 males (14.2%) and 368 females (85.5%). The most frequent age group was 21–30 years old (40.3%), and educational attainment was primarily concentrated at the junior college (51.0%) and undergraduate (46.6%) levels. In terms of work experience, the 1–5 years category was the most prevalent (29.4%), as detailed in [TABLE:1].

The performance evaluation indicator system for AI-enabled primary health services designed in this study comprises two dimensions: internal optimization performance and social adaptation performance. These dimensions encompass four domains—medical quality, operational efficiency, sustainable development, and satisfaction—and 17 specific indicators, as shown in [TABLE:2]. Empirical measurements yielded a Cronbach’s $\alpha$ coefficient of 0.987 for the performance evaluation indicator system, indicating high reliability. The KMO test value was 0.967, which is not only $>0.9$ but also very close to 1. Furthermore, the significance level of the Bartlett’s test of sphericity was $0 < 0.05$, suggesting that the theoretical indicators have good validity and are suitable for factor analysis. Using the principal component method to extract two fixed factors and applying varimax rotation, the cumulative explained variance of the two factors reached 86.229%. The factor loading coefficients for all indicators were $>0.5$, and with the exception of indicators 4 and 5, the classification of indicators was consistent with the original conceptual framework.

Cross-loadings were observed for indicators 4, 5, 8, and 11. Since other items measured similar content and these indicators had a negligible impact on the overall construct, they were removed. Indicators 9, 10, and 13 also exhibited cross-loadings; however, they were retained because their content is relatively important and no other items measured the same content.

To further ensure the appropriateness of the performance system structure, a confirmatory factor analysis (CFA) was conducted. Initial results showed that the standardized loading coefficients for each measurement indicator were $>0.8$ and $<0.95$, and the reliability coefficients were all $>0.7$ and $<0.9$, meeting the standard requirements. However, the model fit parameters did not initially reach reasonable standards. After adjustments based on the modification indices provided by AMOS, the ratio of chi-square to degrees of freedom ($\chi^2/df$) for the revised model was $2.482 < 3$, the Goodness of Fit Index (GFI) was $0.961 > 0.9$, the Adjusted Goodness of Fit Index (AGFI) was $0.924 > 0.9$, and the Root Mean Square Error of Approximation (RMSEA) was $0.059 < 0.08$, all of which reached ideal states. The Composite Reliability (CR) for the two factors of the performance system were 0.973 and 0.968, respectively (both $>0.7$), and the Average Variance Extracted (AVE) values were 0.879 and 0.793 (both $>0.5$), meeting the standard requirements. Consequently, the reliability and convergent validity of the two factors are considered acceptable.

2.1.2 Indicator Performance Analysis of System Application

The perceived performance of the AI-assisted diagnostic system following its implementation in primary healthcare settings (hereinafter referred to as "system application") yielded mean scores ranging from 5.00 to 5.34. The differences between the mean values of various indicators were minimal, and the standard deviations remained relatively low, indicating that respondents generally provided a high performance evaluation of the system's application. Analyzing the two primary dimensions, the social adaptation performance of the system application was rated higher than its internal optimization performance. This suggests that the application not only produced direct performance outcomes, such as improved medical quality and increased operational efficiency, but also led to more pronounced improvements in social adaptation metrics, including sustainability and overall satisfaction. The item "increased direct revenue for medical institutions" received the lowest performance rating ($5.00 \pm 1.76$), suggesting that concerns remain regarding the procurement of medical artificial intelligence equipment when considering short-term economic benefits, particularly in the absence of coverage by medical insurance funds, as shown in [TABLE:3].

2.1.3 Analysis of Performance Differences Across Demographic Characteristics

When comparing optimization performance, social adaptation performance, and total performance within groups of different educational levels, no statistically significant differences were observed ($P > 0.05$). However, statistically significant differences were found in optimization performance, social adaptation performance, and total performance when comparing groups across different age ranges and years of work experience ($P < 0.05$), as shown in [TABLE:4].

2.2.1 Design and Analysis of Performance Influencing Factors

Guided by the organizational change dynamics model proposed by Greenwood and Hinings, this study identifies six factors that influence the social behavior of artificial intelligence empowerment in primary healthcare services. These factors are categorized into contextual triggers (achievement transformation, industrial supporting facilities, and policy environment) and enabling factors (technological level, purchasing power, and health insurance support), as detailed in [TABLE:5].

The analysis of performance-influencing factors reveals that the scores for all indicators exceeded 5 points (on a 7-point scale). This indicates that the various stakeholders actively involved in the application projects perceive these factors as having a significant impact on the performance of system implementation.

2.2.2 Construction and Analysis of the Structural Equation Model

Based on the theoretical foundations of factor mechanism analysis and the specific characteristics of artificial intelligence (AI) empowerment in primary healthcare services, we constructed a structural model of the relationships between variables [FIGURE:1]. The results indicate that both contextual triggers and enabling factors influence overall performance. Specifically, contextual triggers primarily impact social adaptation performance, while enabling factors simultaneously affect both internal and external performance. Furthermore, contextual triggers can influence performance indirectly by empowering enabling factors.

This study utilized AMOS software to conduct path testing for the structural equation model (SEM). The application performance of the ultrasound AI-assisted diagnostic system deployed at the primary level served as the dependent variable, while scientific achievement transformation, policy environment, industrial configuration, technical level, purchasing power, and health insurance support served as independent and mediating variables. The initial SEM fitting results did not reach ideal values. Consequently, the model was adjusted based on the modification indices provided by the software. The resulting modified model addressed internal optimization performance and social adaptation performance. In the analysis of variance (ANOVA) based on age, since there was only one participant in the $<21$ and $>61$ categories and fewer than 30 participants over the age of 51, age was reintegrated into three groups ($<31$ years, 31–40 years, and $\ge 41$ years). Similarly, for the ANOVA based on educational level, as the number of postgraduate students was only 3, the data were reintegrated into three groups (high school and below, junior college, and bachelor's degree and above).

The model fitting results showed a chi-square to degrees of freedom ratio ($\chi^2/df$) of $2.947 < 3$, $GFI = 0.929 > 0.9$, $AGFI = 0.879 < 0.9$, $RMSEA = 0.067 < 0.08$, $TLI = 0.974 > 0.9$, and $CFI = 0.983 > 0.9$. Although the AGFI did not reach the standard threshold, all other critical diagnostic indices met the ideal values, indicating that the model fit is satisfactory.

2.2.3 Path Analysis of the Revised Structural Model

Except for the three paths "Achievement Transformation $\rightarrow$ Technical Level," "Achievement Transformation $\rightarrow$ Purchasing Power," and "Industrial Support $\rightarrow$ Social Adaptation Performance," all other path factors had a positive impact on the performance of system applications, demonstrating a promoting effect ($P < 0.05$), as shown in [TABLE:6]. Overall, the influence of context-triggering factors and enabling factors on performance, as well as their respective paths, has been verified. Specifically, there are three key optimization paths for improving the performance of AI-enabled primary health services:

(1) Key Path 1: "Policy Environment / Industrial Support $\rightarrow$ Technical Level $\rightarrow$ Social Adaptation Performance / Internal Optimization Performance." This path suggests that a robust policy framework and industrial infrastructure directly enhance technical capabilities, which in turn improves both the social integration and internal operational efficiency of AI systems.

(2) Key Path 2: "Policy Environment / Industrial Support $\rightarrow$ Purchasing Power $\rightarrow$ Social Adaptation Performance." This trajectory indicates that policy incentives and industrial maturity strengthen the financial capacity of primary healthcare institutions to acquire technology, thereby facilitating better social adaptation outcomes.

(3) Key Path 3: "Policy Environment / Achievement Transformation $\rightarrow$ Medical Insurance Support $\rightarrow$ Social Adaptation Performance / Internal Optimization Performance." This path highlights that effective policy environments and the successful commercialization of research findings can secure greater support from medical insurance systems, ultimately driving improvements in both social adaptation and internal service optimization.

3 Discussion

The deployment of an ultrasound artificial intelligence (AI) auxiliary diagnostic system at the primary level in Puyang City, Henan Province, represents a public health service innovation project closely integrated with cutting-edge technology. Faced with the current shortage of primary healthcare resources and the uneven distribution of high-quality medical assets, deploying AI-assisted diagnostic systems at the grassroots level is of great significance. This initiative enhances the service capabilities of primary medical institutions, promotes hierarchical diagnosis and treatment, facilitates the mutual recognition of medical examinations, and reduces government medical insurance expenditures. Furthermore, it provides the international community with an excellent model for "AI technology empowering primary health services" and serves as a practical example of creating "barefoot doctors for the new era" in rural areas. Based on an empirical analysis of the mechanism by which AI empowers primary health service performance in Puyang City, the following three recommendations are proposed.

3.1 Public Sectors Should Fully Understand the Dual Value of Medical AI Equipment

Medical AI devices possess the dual values of profitability and public interest. Regardless of the specific situational triggers involved, the social adaptation performance and internal optimization performance of AI-enabled primary health services have been effectively enhanced through the mediation of enabling factors such as technical proficiency, purchasing power, and medical insurance support. Specifically, the two situational triggers of policy environment and achievement transformation directly contribute to the improvement of social adaptation performance. Consequently, although some health institutions have expressed concerns regarding the short-term financial returns of procuring medical AI equipment, the empowerment of primary health services by artificial intelligence generates positive externalities. The public value created by these systemic applications is even more pronounced than the direct operational value generated for primary medical institutions.

3.2 Medical AI Industrial Support Should Be Highlighted and Valued

The importance of transforming research findings into practical applications cannot be overlooked. Research results indicate that optimizing industrial support serves as a critical lever for enhancing the technical competitiveness of medical artificial intelligence products, while simultaneously strengthening the purchasing intent and capacity of primary healthcare institutions.

3.2.1 Actively Develop the Medical AI Industry

(1) Medical artificial intelligence products, represented by intelligent auxiliary diagnostic systems, should be listed as key projects for rural revitalization and health-based poverty alleviation. Efforts should be made to promote these technologies across national and global markets. (2) Attention must be directed toward the standardization of the medical AI industry and its products to establish advanced experiences that are both replicable and scalable. (3) Industry-university-research cooperation projects should be actively promoted. Through policy guidance and financial support, deep collaboration between medical institutions, universities, research institutes, and AI enterprises should be fostered to accelerate the clinical application and translation of medical artificial intelligence achievements.

3.2.2 Increase Government R&D Investment in Medical AI

(1) Establish specialized scientific research funds that offer varying funding amounts and cycles tailored to different R&D stages—such as basic theoretical research, technical application development, and clinical product validation—to provide precise support for innovative exploration at every link. (2) Construct R&D innovation platforms by investing in the creation of professional medical artificial intelligence research centers, laboratories, and industrial parks. These facilities should provide research teams with advanced hardware, massive medical data resources, and convenient spaces for technical exchange and collaboration. (3) Cultivate and recruit professional talent by adding medical AI-related majors and courses in universities and vocational colleges, establishing specialized scholarships to encourage students to pursue studies in this field, and supporting joint training programs between universities and enterprises for targeted talent development. Furthermore, highly attractive talent recruitment policies should be formulated—including competitive compensation, research start-up funds, and housing security—to recruit top AI scientists, medical experts, and interdisciplinary talents globally to strengthen domestic R&D capabilities.

3.3 Strengthen Policy Support and Reform

The policy environment is an unavoidable and critical starting point for improving the performance of artificial intelligence in empowering primary healthcare services. Only on the foundation of robust policy support can the technical proficiency of AI systems, the purchasing power of primary medical institutions, and the extent of health insurance support be effectively enhanced. Therefore, macro-level medical administrative policy support should be strengthened, and relevant reforms should be advanced in a coordinated manner within the framework of deepening healthcare reform.

3.3.1 The Government Should Act as a "Guide"

To optimize institutional supply and fulfill the government's responsibility for healthcare provision, it is essential to increase investment in the intelligent infrastructure of primary healthcare institutions. Drawing on the successful experience of Puyang City in Henan Province regarding the deployment of AI-assisted diagnostic systems, local authorities should establish dedicated task forces, define key priorities and task allocations, and create standardized work ledgers. By implementing robust support measures, the rapid promotion and application of AI-assisted diagnostic systems can be facilitated. Accelerating the integration of high-tech AI products into primary healthcare facilities will ensure that the achievements of technological development benefit the public more quickly, effectively, and extensively.

3.3.2 Establish Financial Subsidy Mechanisms Adapted to Local Economic Levels

Growth mechanisms should be established to enhance the capacity of primary healthcare institutions to introduce and operate artificial intelligence (AI) systems, thereby increasing their willingness to adopt these technologies. First, it is recommended that a special financial subsidy fund be established to provide targeted subsidies for primary healthcare institutions purchasing medical AI equipment and software. This would reduce procurement costs and improve the ability of these institutions to introduce advanced technologies—particularly in resource-scarce areas—enabling them to rapidly integrate medical AI products and improve healthcare quality. Second, the government should provide tax incentives, such as value-added tax (VAT) reductions and corporate income tax preferences, for enterprises producing medical AI products and engaging in related technological research and development.

3.3.3 Balance Public Welfare with Incentive Mechanisms

To ensure the sustainable development of digital health, it is essential to establish a service-oriented medical billing mechanism. This involves the timely integration of clinically effective and professionally recognized new technologies and projects into the medical insurance payment and pricing systems. Simultaneously, the compensation structure must be optimized by implementing the "Two Allows" policy (allowing medical institutions to adjust their total payroll and allowing them to determine internal distribution), thereby incentivizing primary healthcare workers to actively participate in the intelligent transformation of the sector. Finally, the construction of an integrated health and wellness system should be advanced. This requires strengthening the synergy between primary healthcare institutions and resources in public health and elderly care to build a "health-centered" service framework. Such efforts will facilitate the deep integration and collaborative application of artificial intelligence systems across all stages of the healthcare continuum, including disease prevention, diagnosis, and rehabilitation.

3.3.4 Strengthen Medical Insurance Support

In Puyang City, artificial intelligence has already been applied within primary healthcare institutions; however, it has not yet been incorporated into the scope of medical insurance fund payments, and official charging standards for these medical items remain unavailable. Establishing appropriate charging standards for AI-aided diagnostic services and including them in medical insurance coverage would result in additional expenditures far lower than the "waste" of insurance funds caused by redundant examinations. It is estimated that this could reduce medical insurance expenditure costs by 85.7% to 92.9%. Taking thyroid ultrasound AI as an example, five provinces—Hebei, Shandong, Shanxi, Jiangsu, and Guizhou—have already integrated it into their provincial medical service price catalogs. It is recommended to accelerate the promotion of AI-related diagnostic and treatment services as items eligible for medical insurance fund payments.

In summary, the sustainable development of AI-empowered primary healthcare services depends not only on the advancement of the technology itself but also on the synergistic support of policies across finance, medical insurance, personnel, and service delivery systems. Only through such coordination can the healthcare reform goals of "improving efficiency, reducing costs, and increasing equity" be truly realized. Due to the rarity of "large-scale, multi-center" experimental cases and the scarcity and low accessibility of data, this study primarily utilizes subjective perception data supplemented by objective data to analyze the mechanisms by which AI influences the performance of primary healthcare services. As medical AI equipment is deployed more extensively and deeply across various fields, more high-quality objective data will become available. Future research could further incorporate indicators such as frequency and depth of use to achieve higher precision and multi-dimensional verification of AI-empowered primary healthcare service performance.

Author Contributions: Kong Zihe: conceptualized the research, designed the research plan, proposed and designed the research propositions, performed data collection and analysis, and drafted and revised the manuscript; Bian Qingyang: performed data collection and statistical analysis, and drafted and revised the manuscript; Kong Dexing and Kong Jiangming: conceptualized the research, designed the research plan, implemented the research process, and revised the manuscript.

The authors declare no conflicts of interest.

References

[1] Health News. China demonstrates global leadership in the field of digital health [N/OL].
[2] Xi Jinping. The Governance of China, Volume II [M]. Beijing: Foreign Languages Press.
[3] You Lulu, Du Shilin. Research on the current status of service capabilities in primary medical institutions [J]. Chinese General Practice.
[4] Li Xiaoli, Zhang Bo, Wang Kang, et al. Development and application of artificial intelligence [J]. Journal of Beijing University of Technology, 2020, 46(6): 583-590.
[5] Yuan Bo, Dai Hua, Wu Jia, et al. Application of artificial intelligence in the field of general practice [J]. Chinese General Practice, 2021, 19(9): 1433-1436, 1572.
[6] Ren Fuji. Can artificial intelligence gallop across the vast fields of "Grand Education" and "Grand Health"? [J]. CAAI Transactions on Intelligent Systems, 2020, 15(2): 405-408.
[7] Zhao Jiaying, Gao Peng, Zhu Yongjun, et al. The application of artificial intelligence will improve the efficiency of primary healthcare services in China [J]. Chinese General Practice, 2017, 20(34).
[8] WALLIS C. How artificial intelligence will change medicine [J]. Nature, 2019, 576(7787): S48. DOI: 10.1038/d41586-019-
[9] Song Xuanke, Zhang Wenqiang, Zhang Dan, et al. Application and Effectiveness of Scenario-Based Simulation Teaching in Standardized Residency Training for Thoracic Surgery [J]. Henan Journal of Surgery, 2020, 26(5).
[10] MINTZ Y, BRODIE R. Introduction to artificial intelligence in medicine [J]. Minim Invasive Ther Allied Technol, 2019, 28(2): 73-81. DOI: 10.1080/13645706.2019.1575882.
[11] Hu Zhibo. Current status and prospects of artificial intelligence applications in the medical field [J]. Electronic Components and Information Technology, 2020, 4(12): 61-62.
[12] TANG S L, MENG Q Y, CHEN L, et al. Tackling the challenges to health equity in China [J]. Lancet, 2008, 372(9648): 1493-1501. DOI: 10.1016/S0140-6736(08)61364-1.
[13] Xu Guoping, Wang Jiaji. Primary Health Care: Basic medical and health services should be correctly understood and fully implemented in China's new health care reform [J]. Chinese General Practice, 2015, 18(32): 3893-3900.
[14] Zhou Lingli, Wang Heping, Xie Long, et al. Analysis of the current status and equity of general practitioner resource allocation in China [J]. Chinese General Practice, 2017, 20(19): 2311-2315.
[15] Yu Xiaosong. Development and prospects of general practice in China since the founding of the People's Republic of China 70 years ago [J]. Chinese Journal of General Practitioners, 2019, 18(11): 1017-1021. DOI: 10.3760/cma.j.issn.1674-4152.001058.
[16] CAMPBELL J P, MCCLOY R A, OPPLER S H, et al. A theory of performance [A]//SCHMITT M J, BORMAN W C. Personnel selection in organizations. San Francisco: Jossey-Bass, 1992.
[17] YUCHTMAN E, SEASHORE S E. A system resource approach to organizational effectiveness [J]. Am Sociol Rev, 1967, 32(6): 891. DOI: 10.2307/2092843.
[18] Fan, B., & Jin, J. (2016). The mechanism of the impact of public service provision on perceived public service performance: The mediating role of government image and the moderating effect of public participation. Management World, 32(10), 50-61, 187-188. DOI: 10.19744/j.cnki.11-1235/f.2016.10.006.
[19] Zhang, D., Zhang, H., & Fan, B. (2018). Performance of local government functional transformation: Conceptual definition, dimensional design, and empirical evaluation. Chinese Public Administration, (5), 33-39.
[20] BORMAN W C, MOTOWIDLO S J. Expanding the criterion domain to include elements of contextual performance [A]//SCHMITT M J, BORMAN W C. Personnel selection in organizations. San Francisco: Jossey-Bass, 1993: 71-98.
[21] GREENWOOD R, HININGS C R. Understanding radical organizational change: bringing together the old and the new institutionalism [J]. Acad Manag Rev, 1996, 21(4): 1022.
[22] Kong Zihe, Kong Dexing. What Kind of Artificial Intelligence is Needed at the Primary Healthcare Level? [N]. Health News, 2025-06-09 (05).

What Kind of Artificial Intelligence is Needed at the Primary Healthcare Level?

In recent years, the rapid development of artificial intelligence (AI) has brought transformative opportunities to the medical field. However, as these technologies transition from high-level research institutions to primary healthcare facilities, a critical question emerges: what kind of AI do grassroots medical institutions actually need? To bridge the gap between advanced technology and practical clinical application, AI development must move beyond mere technical complexity and focus on accessibility, reliability, and integration into the existing primary care ecosystem.

Enhancing Diagnostic Accuracy and Efficiency

The primary challenge at the grassroots level is often the shortage of specialized medical personnel and the resulting pressure on general practitioners. Therefore, AI systems designed for these settings must prioritize enhancing diagnostic accuracy for common diseases while streamlining clinical workflows. Rather than replacing doctors, AI should serve as a "force multiplier" that assists in screening, preliminary diagnosis, and chronic disease management. For instance, AI-powered imaging tools for chest X-rays or fundus photography can help primary care physicians identify abnormalities that might otherwise require a specialist's review, ensuring timely referrals and reducing the rate of misdiagnosis.

Adaptability and Ease of Use

For AI to be truly effective at the primary level, it must be characterized by high adaptability and low barriers to entry. Many grassroots facilities operate with limited hardware resources and varying levels of digital infrastructure. Consequently, AI solutions should be lightweight, compatible with existing Hospital Information Systems (HIS), and require minimal specialized training to operate. The user interface must be intuitive, providing clear, actionable insights rather than opaque technical data. If a tool is too complex or disrupts the established workflow, it is unlikely to be adopted, regardless of its theoretical performance.

Reliability and Interpretability

Trust is the cornerstone of medical practice, especially in community settings where long-term doctor-patient relationships are vital. AI models must demonstrate high levels of robustness and interpretability. "Black box" algorithms are often met with skepticism; therefore, AI systems should provide "explainable" results that allow physicians to understand the reasoning behind a suggestion. Furthermore, these systems must be reliable across diverse patient populations to ensure equity in healthcare delivery.