Glaser and Strauss founded grounded theory methodology in 1967 (26). It was a well-known methodology for generating conceptual frameworks or theories from the collected data (26) when little is known about a phenomenon. It has been applied to various regions, including management (27), sociology (28), and healthcare (29) largely. As little research on the policy feedback mechanism of health improvement has been explored, our research aims to construct a theoretical model to guide health improvement based on the Chinese healthcare improvement situation. Therefore, the grounded theory method is appropriate for our research purpose. The grounded theory methods guided the whole research. Selected cases and other text materials were coded using three-level codings based on the grounded theory method. Through the research process, we constantly wrote memos, conducted a theoretical saturation test, and finally constructed a theoretical model.

As Glaser espoused the dictum, “all is data” (26). GT data could include various potential sources, such as political speeches, newspapers, government reports, policy documents, newspaper editorials, and so on (30). Therefore, our research collected text data through multiple channels. We conducted inclusion and exclusion criteria based on the grounded theory method. Based on this criterion, we collected our data using the purposive, theoretical, and snowball sampling methods (18). Figure 1 presents an overview of the data collected.

Based on the inclusion and exclusion criteria, first, we complied with a grounded theory academic research dataset for the collected cases, policy documents, and literature materials (Figure 1). NVivo software is a strong qualitative analysis software used for the grounded theory approach (34). It assists in coding and querying documents, thus helping the researchers analyze large qualitative materials more effectively (34). Our study used NVivo 12 Plus software to do a three-level coding process, write memos, conduct a theoretical saturation test, and do other research works. Then, we compared the components of our grounded theory model with the original components of classic theories to explore whether they could fit the Chinese healthcare situation. Finally, we elaborated and analyzed our grounded theory model.

The detailed steps were based on the grounded theory method: (1) Opening coding. We collected the raw data and generated initial concepts and categories in open coding in NVivo. (2) Axial coding. We transformed primary data into more abstract concepts built during the open coding phase (26). (3) Selective coding. We discovered the core categories from the main categories and confirmed a storyline to clarify the theory (26). Based on the coding results, our storyline is “Policy Feedback Mechanism for Healthcare Improvement in China.” We used this storyline to construct a grounded theory. The detailed coding process is presented in Table 1, and the partial original examples are listed in Table 2. (4) Theoretical Saturation validation. Saturation is defined as no new codes occur while adding instances of the same codes (31). We combined policies, survey reports, and other texts, repeatedly analyzed the research materials, conducted theoretical sampling, and selected other cases for the same coding process. The model was theoretically saturated until no new important categories and relations appeared (31). (5) Rigor. Based on the grounded methodology, three steps were taken to ensure the result’s reliability. First, writing memos and following the qualitative guidance improves theoretical sensitivity and coding credibility (18, 35). Second, the coding was validated by experts (co-authors with many years of hospital management experience). Third, the author has verified the coding results after some time to ensure the rigor of the findings. (6) We compared the components of our grounded model with components of policy feedback theory, policy tool theory, and Andersen’s model. Then, we further explained our newly grounded theory and explored its generality and particularity.

Through the three coding processes (Table 1 for partial coding details), we inductively generated 49 mainly initial concepts, 26 initial categories, 11 main categories, and 5 core categories. Then, we constructed the model of policy feedback mechanism for healthcare improvement (PFMHI) (Figure 2).

Our model and case data showed that the healthcare improvement policy involved multiple objects of government, medical organizations, and patients. At the government level, the government needed to consider the existing problems. Then, a series of policy tools could be used to change the beliefs and behaviors of medical organizations. With patients’ feedback, the government could continue improving existing policies. Although medical organizations suffered from limited supply at the hospital level, they were impacted by the resource and interpretive effects of policy tools. They optimized hospital behavior and beliefs to enhance the patient experience and satisfaction, obtain medical organization achievements, and implement healthcare improvement policies. At the patient level, patients with different needs were directly or indirectly affected by government policy or various healthcare improvement behaviors in medical organizations. With patients’ positive feedback on trust, experience, and satisfaction increasing, their feelings (including their sense of gain, happiness, and safety) could help medical organizations and the government further optimize services and policy design. The above confirmed a complete policy feedback mechanism process.

Our model could answer the policy feedback mechanism for healthcare improvement in China. Based on the existing policy context, from the pathway direction of government to patients, at the policy time of t1, the government used various policy tools to produce policy feedback effects (resource or interpretive effects) on hospitals. Hospital beliefs and behaviors influenced various aspects of patient feedback. The policy feedback results are from patients to the government. The patient’s feelings, medical organization outcomes, and policy improvement will create new policy tools at the policy time 2.

As our research question involved health policy and services, Andersen’s model is a classical conceptual framework for analyzing healthcare utilization (36, 37). It could explore healthcare access and utilization mechanisms and offer guidance in designing interventions to improve health services (37). This model focuses on individuals’ health behaviors and patients’ service satisfaction (37). It contains contextual characteristics, individual characteristics, health behaviors, and outcomes (21, 37). Although it stresses that improving healthcare focuses on contextual and individual factors and has some consideration of health policies, most research focuses on individual characteristics and lacks a deep exploration of policy feedback factors (21). The policy feedback theory and policy tool theory are important research regions in policy research (38, 39). As our model not only attaches importance to the policy feedback factors but also fully considers the individual factors, we compared our PFMHI model with the above classic policy feedback theory, the policy tool theory, and Andersen’s model, as shown in Table 2, to analyze the unique contributions of our model. We found that the PFMHI model reinforced some elements of the policy feedback theory, Andersen’s model, and the policy tool. This finding indicated that the three classical theories had some applicability and rationality in the context of Chinese health policy. Besides, we have added Chinese-specific elements of “Patient Sense of Gain,” “Patient Happiness,” and “Patient Sense of Safety” in the module of “Patient Feelings”. Therefore, our model further enriched the original three classical theories and emphasized patient-centered care, focusing more on the public (patients) perspective on healthcare improvement policy. However, our cases include various types of patients, like patients with chronic diseases (A22), patients with diabetes (A3), patients with cancer (A33), and patients with stroke (A4). Based on the grounded theory method and related literature, we defined the concept of “patient” from a universal perspective to clarify our grounded theory framework. Therefore, the term “patient” in our research represents the potential citizens of health care services (18, 40). Our model is not specified for certain disease patients, which it is similar to Andersen’s model (21).

The main research on policy feedback theory was how policies impact citizens or political organizations (41). However, research on the interaction among policy, organization, and citizens was rare. Our PFMHI model explored the research scenarios of the interaction among the three objects, which addressed the current gap in this research field. As most typical cases were closely related to the interaction between the government, hospitals, and patients, it provided the conditions for us to construct the multi-level interaction mechanism of policy feedback theory. Our PFMHI model showed that the healthcare improvement policy feedback mechanism had multi-level objects of the government, medical organizations, and patients (see Table 2, A19,A14,A3).

In some cases, the government played an essential leading role in the policy of improving healthcare services. They formulated health policies, investigated and grasped the public’s attitudes toward healthcare services, promoted medical alliances, and granted Internet hospital licenses (A11, A50, A25). On the macro level, the government analyzed questions about public access to healthcare, studied health policy, and guided medical organizations to improve healthcare (B16, B1) (12, 42). Hospitals were also the primary actors for those policies. All cases involved medical organizations, which had taken a series of actions to improve healthcare service and patient experience (A10, A11, A14). We also found that patients were the vital objects of those policies. Compared with previous literature, patients were generally passive in the doctor-patient relationship (18). The policy goal of improving healthcare services emphasized enhancing patient experience (B16) (12). It showed that patients were important in evaluating the outcome of those policies (A3, A4, A22, A33). Since the introduction of Engel’s biopsychosocial model of health care in 1977 (18), the medical culture has changed. Patients and healthcare professionals had a more balanced relationship, with more self-determination and autonomy rights recognized than before (18). Therefore, the “patient” module was considered a specific policy performance. Based on the grounded theory method and related literature, we defined the concept of “patient” from a universal perspective to clarify our grounded theory framework. Therefore, the term “patient” in our research represents the potential citizens obtaining healthcare services (18, 40).

The interpretation of the theory depended on the specific research situation (26). We constructed the policy context module of our policy feedback mechanism, which included “problems,” “supplier conditions,” and “demander needs” (Table 2). Next, we analyzed the three main components in the policy context module, combining the model diagram and case material.

Firstly, the module on “problems” referred to the government’s analysis of improving healthcare. The following three levels were explored. On the macro level, the policy design was not clear and specific. The high-quality healthcare resources were insufficient. The hospitals and disciplines only played limited roles in healthcare improvement due to a lack of systematic coordination mechanisms (A12). On the medium level, hospitals lacked overall planning with vague development ideas (A12). On the micro level, the problem of difficult and costly access to healthcare persisted due to the vast demand of patients and limited medical services capacity (A17).

The next module, “supplier conditions,” mainly considered hospital-level issues. As the leading supplier of healthcare services, some hospitals with limited medical service levels and uneven health resource allocation restricted the improvement of the medical service supply capability (13). As mentioned in the cases, hospitals lacked many hardware and software facilities. The patient had a longer waiting time, and the treatment process was complex (A32). Although the patient was eager for excellent healthcare, the whole healthcare system layout was imperfect. High-quality resources were mainly in big cities and large-medium hospitals (A17).

Finally, the module on “demander needs” mainly considered patient-level issues. The literature reported that patients were healthcare service consumers (18) and feedback providers on health policies (43). In our cases, we found that people had an increasing demand for healthcare services, and different age groups with diseases had different needs for healthcare services. Besides, patients with diseases may face various psychological conditions such as fear and anxiety (A58, A17, A61). Therefore, the demander’s needs included the factors of patient needs (A9), patient expectation (A43), patient preferences (Investigation Report 16) (44), and patient health literacy (Investigation Report 23, Investigation Report 56) (45, 46). The demander’s need, as promoted by the physical-psycho-society medical model and the patient-oriented idea, should be considered in the policy context of our model.

Policies and goals are made using policy tools (39). The literature documents that policy tools could better explain the dynamic nature of policy and help policymakers make effective decisions (39). There was little research that combined policy tools with policy feedback theory. Furthermore, research into the impact of policy tools and mixes on policy feedback theory was rare. In our selected cases and policy texts, we constructed the “policy tools driven” module to represent the specific policy measures to improve healthcare service. For example, “carry out a pilot ambulatory surgery” (A58), “strengthen the continuity of care” (A50), and “mutual recognition of medical examination result” (A1). By analyzing these policy tools and mixes, we could clearly explain how the government promoted the policy feedback progress of improving healthcare service by policy tools and reached the policy outcome to improve the patient experience (Table 2).

In this module, we first analyzed the levels of government and medical organizations. The model reinforced the research content of policy feedback theory, that was, how policy impacted organizations (e.g., their resources, priorities, political opportunities, or incentive structures) (41) (Figure 2).

The government sometimes uses policy tools to impact public hospitals’ beliefs and behaviors. On the aspect of belief, policy tools like “goal planning” and “systems and mechanisms” produced interpretive effects to guide public hospitals to strengthen the idea of “patient-centered” medical services (A18, A7). Besides, hospitals established the policy goal orientation of implementing the health reform and the National Healthcare Improvement Initiative (A9, A18). On the aspect of behavior, policy tools, such as the utilization of medical insurance funds and the encouragement and guidance of hospitals, exerted resource and interpretive effects that significantly influenced the healthcare behaviors of numerous public hospitals. Table 3 lists the initial categories induced based on the grounded theory data analysis. We summarized and listed the following five behaviors in Table 3: hospital operations and management, medical service supply, hospital supportive services, political participation behavior, and medical organizations interaction. The above five types of behaviors represent the behaviors of hospitals in the policy feedback process. They not only theoretically explain the impact of policy feedback on organizational behavior, but also, at the practice level, these listed behaviors could be a practical guide for hospitals to enhance healthcare quality and patient experiences.

For the patient feedback module, the previous research on policy feedback theory has mainly regarded patient feedback (public feedback) as a result of policy feedback (47). However, our model took patient feedback as a process variable, reinforcing the policy feedback notion that “new policies create new politics” (41). The case showed that patients interacted with governments and medical organizations while improving the healthcare service process (A15, A45, A5). Patient experience, satisfaction, and trust indeed increased when public hospitals implemented a series of policy tools by public hospitals (A13, A14, A59). Hence, patients were more willing to participate in improving healthcare service. For example, patients participated in the Health Commission’s survey on patient satisfaction (A14, A11). The survey found that public hospitals enhanced patient satisfaction by implementing various policy tools (A14, A11). Besides, the hospital also used policy tools like “education and promotion” to encourage more patients to try new healthcare service models (A23) and participate in hospital management (A45).

We compared classic policy feedback theory (48, 49) and explained the policy feedback process driven by our policy tools (Figure 2). At public policy time 1, the government used policy tools to improve healthcare service—the policy tools generated both resource and interpretive effects that influence medical organizations and patients. The detailed analysis is as follows: on the medical organization level, the policy tool generated an interpretive effect shaping hospitals’ beliefs. Additionally, it produced both interpretive and resource effects that shaped hospital behaviors. Through these influences, medical organizations affect patient experience, satisfaction, trust, and engagement through various behaviors. On the patient level, the “patient feedback” module demonstrated that government and medical organizations influence patients both directly and indirectly. This influence encompasses improvements in patient experience, patient satisfaction, trust, and engagement behaviors. This process involves complex relationships and interactions between the government, hospitals, and patients. Based on the grounded theory method and related literature, we defined the concept of “patient” from a universal perspective to clarify our grounded theory framework. Therefore, the term “patient” in our research represents the potential citizens of health care services (18, 40). The “patient feedback” module contributes to understanding how patients interact with other actors like governments or hospitals (18). It is not focused on patients categorized according to their illness or conditions (18). Therefore, this research did not consider the specific health settings, health professionals, or health status. However, demographic factors will be emphasized in the subsequent quantitative empirical tests based on our theoretical model. Consequently, our PFMHI model established a theoretical pathway for the policy feedback process driven by the policy tools, which verified the multi-level effects logic of policy feedback. We further emphasized a “patient-oriented” and actively listened to the patient’s (public) voice (41), placing significant importance on patient experience. We further put the “patient” perspective into the policy feedback process to improve healthcare.

The rightmost column of our PFMHI model lists the policy feedback results of improving healthcare (Figure 2). It mainly concluded with “patient feelings,” “medical organization outcomes,” and “policy improvement.” Next, we combined the PFMHI model diagram and case data to explain the three main policy feedback results.

Firstly, the “patient feelings” module focused on the patient level and reinforced the policy objective of enhancing healthcare in China by improving patient perceptions and experiences. Terms such as “sense of gain,” “happiness,” and “sense of safety” are prevalent in Chinese health policy. Sense of gain refers to people’s subjective feeling of acquisition of actual benefits compared to their expected benefit (50, 51). Happiness refers to people’s subjective feeling of a better life aligned with their current situation (50, 51). Sense of safety refers to people’s subjective affirmation of a stable and peaceful life (50, 51). However, empirical literature to measure the three conceptions is scarce. From these cases, we found that the patient’s feelings mainly showed in the aspect of the patient’s sense of gain (A14), the patient’s happiness (A13), and the patient’s sense of safety [B16 (12), A15, A9]. Those terms reflect political concepts with Chinese characteristics. Moreover, combined with the health situation and the coding results, we can operationalize the patient’s sense of gain on the patient experience, the patient’s happiness on the patient satisfaction, and the patient’s sense of safety on patient engagement and trust and other factors in the future empirical studies. The Chinese government and hospitals increasingly value those patients’ feelings, as they are important indicators for evaluating policy effectiveness from the patient’s perspective.

Besides, the “medical organization outcomes” module mainly focused on the hospital level. The case showed that implementing healthcare improvement policies has resulted in many achievements for the medical organization. Internally, these policies enhanced the quality of healthcare services, elevated scientific research levels, improved medical staff satisfaction, and strengthened talent teams. Externally, they boosted the hospital’s image, increased economic and social efficiency, and received recognition from the government departments (A13, A44). Through this module, our case showed that while policies impacted patients through medical organizations, the organizations themselves also obtained positive outcomes. These research results align with the policy feedback theory, which focuses on the impact of policy on groups (49).

Finally, the “policy improvement” module was mainly on the policy level. Our PFMHI model showed that the government obtained a crucial policy feedback outcome upon completing the policy feedback process: improving policy design. At public policy time t2 (shown in Figure 2), the government received feedback from patients and hospitals. When the feedback was positive, the government was likely to promote the policy further by establishing specialist medical alliances (A37) and advancing the national medical reform policy (A25). This research primarily focused on typical cases that received more positive feedback. Therefore, our model mainly emphasized positive feedback. It was consistent with the classical policy feedback theory, which focuses more on studying positive feedback (49).

Meanwhile, through theoretical sampling, we added a series of national surveys to improve healthcare in China. Those surveys revealed the shortcomings of policy implementation through third-party evaluations, namely the existence of negative feedback. All this feedback could motivate the government to improve policies at public policy time 2 and reshape the new policies [Surveys Report 1 (52), Survey Report 12 (53)]. All those analyses showed that the theoretical mechanism of the PFMHI model we constructed corresponded to the core idea of policy feedback theory: the result of policy feedback leads to reshaping politics (41).