Enhancing health system efficiency in China: considering the interaction between use of primary care and the demand for secondary and tertiary care

Abstract

This study examines the substitution effects and complementary dynamics between outpatient and inpatient services across different levels of hospital care in China's tiered healthcare system. The data of this study originated from official administrative medical insurance reimbursement records from 2013 to 2019, with a final sample size of 1 520 263 patients. Using individual-level data and controlling for regional variations through fixed-effects models, we identify significant patterns in healthcare utilization that provide actionable insights for enhancing system efficiency. We have found a notable substitution effect: increased utilization of primary care services was negatively associated with the demand for secondary and tertiary care, thereby supporting ongoing health policy reforms. Additionally, outpatient services at primary care facilities could reduce the demand for both outpatient and inpatient services at higher-level hospitals. The homogeneity of outpatient services further facilitated substitution across care levels, allowing primary and secondary care to increasingly manage clinical cases previously handled by tertiary hospitals. Finally, we explored the complementary relationship between outpatient and inpatient services within the same care level, emphasizing highlighting how financial incentives contribute to induced hospitalization in China's healthcare system. These findings suggest that healthcare policies must be adjusted to address systemic inefficiencies and realign financial incentives in order to improve resource allocation and patient care.

Key messages.

• Substitution effects between primary and tertiary care have the potential to reduce patient load on higher-level hospitals, but require well-structured primary care systems.

• Outpatient services at primary care facilities are associated with lower utilization for both outpatient and inpatient services at higher-level hospitals.

• The homogeneity of outpatient services across levels of care facilitates a smoother substitution process, allowing lower-level facilities to absorb more demand.

• Induced hospitalization complicates the complementarity between outpatient and inpatient services in China, suggesting that policy measures could help reduce unnecessary admissions and improve the efficiency of the healthcare system.

Introduction

In the evolving landscape of global healthcare, primary care is universally acknowledged as the foundation of effective and sustainable health systems (Kruk et al. 2018). From the 1978 Alma-Ata Declaration to the more recent Astana Declaration, there is a clear consensus on the critical role of robust primary care systems centered on the person (Hanson et al. 2022). The general practitioner system in the United Kingdom and the family doctor system in Germany have demonstrated significant benefits in improving health outcomes and optimizing healthcare costs (Busse et al. 2017, Anderson et al. 2022). Notwithstanding these developments, numerous health systems, particularly those in low- and middle-income countries (LMICs), continue to struggle with challenges such as elevated hospitalization rates, overuse of specialty medical services, and mounting healthcare costs, with primary care frequently failing to effectively fulfill its gatekeeping function (Bitton et al. 2017), which is often due to primary care’s insufficient gatekeeping capacity.

Primary care serves as the first point of contact, managing a broad spectrum of health issues and coordinating patient care holistically (Li et al. 2020). In contrast, secondary and tertiary care, as typically provided by hospitals, address more complex conditions requiring specialized knowledge and technology (Eberly et al. 2020). The separation and integration of these care levels vary globally, reflecting diverse policy designs. In China, this tiered framework aims to meet the needs of a large, heterogeneous population by promoting a clear demarcation and interaction among care levels, with policies encouraging primary care to alleviate the service burden on higher-level hospitals (Yip et al. 2019).

Studies from various healthcare systems indicate that the role of primary care in influencing hospital service utilization varies. In the United States, increased access to primary care has reduced specialty service use, suggesting a substitution effect (Reiss-Brennan et al. 2016). However, studies from England and Norway indicate a complementary relationship, where primary care may enhance referrals and utilization of hospital services (Ringberg et al. 2013, Lau et al. 2021). These disparate outcomes illustrate the contextual dependency of the effects of primary care enhancements, emphasizing the importance of tailored approaches in healthcare policy and system design. Generally, in the context of healthcare delivery, substitution occurs when an increase in the utilization of one type of service (e.g. primary care) leads to a corresponding decrease in the use of another (e.g. tertiary hospital services). Conversely, complementarity describes a dynamic where the use of one service increases the demand for another, typically by improving diagnosis or facilitating referrals. Understanding the balance between these two forces is crucial for optimizing the balance between primary and hospital care.

The extensive utilization of secondary and tertiary healthcare services in China has led to notable inefficiencies and a mounting financial burden across the healthcare system. Notwithstanding the government’s endeavors to enhance primary care through reforms aimed at establishing a tiered diagnosis and treatment system, the relationship between these reforms and healthcare efficiency remains under-explored. The available evidence indicates that although primary care facilities in China are designed to address a wider range of health concerns, patients frequently choose to bypass these in favor of higher-tier hospital services due to concerns about the quality of care and the public’s perception of the latter (Yip et al. 2019). This perception undermines the potential of primary care to serve as an effective gatekeeper within the healthcare system.

China’s healthcare system is an exemplification of these complexities. Although reforms aim to establish a tiered diagnosis and treatment system, many patients bypass primary care facilities due to perceived quality concerns, thereby limiting primary care’s gatekeeping potential (Yip et al. 2019). Pilot programs introducing family doctor contract services have demonstrated efficacy in reducing outpatient visits to tertiary hospitals, indicating potential substitution effects (Li et al. 2020). However, challenges persist, including resource disparities and fragmented referral pathways (Yip et al. 2019). The absence of exhaustive empirical investigations into these dynamics signifies a pivotal lacuna in our comprehension of the efficacy of China’s healthcare reforms.

This study makes several key contributions to the growing body of literature on healthcare system optimization by empirically examining the substitution and complementary effects between primary care and hospital services within the context of China’s tiered healthcare system. The primary contribution of this research is its focus on the role of primary care in a rapidly developing healthcare system, specifically China, where primary care is comparatively underdeveloped in relation to higher-tier hospital services. Whilst the significance of primary care in enhancing healthcare efficiency has been extensively acknowledged in global research, particularly within Western healthcare models, this study uniquely shifts the focus to the LMICs context, thereby addressing a critical gap in the existing literature. By leveraging a large-scale, representative dataset that spans multiple years, the study provides valuable empirical insights into the dynamics between primary, secondary, and tertiary care in China, offering a comprehensive view of healthcare resource allocation in a non-Western setting.

The second key contribution of this study is its exploration of how primary care can alleviate the burden on secondary and tertiary hospitals through the substitution of lower-acuity cases, leading to enhanced resource allocation and cost savings. Whilst the majority of preceding studies have concentrated on the impact of primary care in high-income contexts, this research provides compelling evidence of its potential to reduce hospital service demand and associated healthcare costs in a LMIC.

Finally, this study contributes to the international discourse on healthcare system integration by presenting empirical evidence in support of the complementary role of primary care in improving overall healthcare outcomes. This understanding of primary care’s dual role in both substituting and complementing hospital services has been underexplored in existing studies, especially within the context of developing countries. By providing new evidence that may inform healthcare reforms, not only in China but also in countries with similar healthcare structures, the study helps shape a broader global conversation about the integration of primary and hospital care, with implications for more sustainable, cost-effective, and equitable healthcare systems worldwide.

Context in China

China's healthcare system operates on a tiered model aimed at optimizing resource allocation and addressing the medical needs of a vast population. The system is structured across three levels of care. Primary care facilities, such as urban community health centers and rural township hospitals, are responsible for the provision of preventive care, basic medical services, and chronic disease management; secondary hospitals serve as intermediaries, offering routine inpatient and outpatient care; while tertiary hospitals deliver specialized and complex treatments, including advanced surgeries and critical inpatient care. This framework is supported by a hierarchical referral system, which is designed to direct patients to the most appropriate level of care, thus ensuring efficiency and equity (Zhou et al. 2021b).

Notwithstanding its meticulous design, the system encounters considerable operational challenges. Over the past two decades, there has been a notable increase in the hospitalization rate, from 4.7% in 2003 to 21.4% in 2023. This reflects an increase in demand for hospital services but also highlights potential systemic inefficiencies (Li et al. 2020). Patients frequently circumvent primary and secondary care facilities, opting instead for tertiary hospitals, even for minor ailments, due to perceptions of superior quality and better outcomes (Li et al. 2021). This trend has led to severe overcrowding at tertiary hospitals, where resource depletion compromises their ability to effectively manage complex cases.

The underutilization of primary care is one of the system’s most critical weaknesses. Studies consistently highlight deficiencies in primary care facilities, including limited diagnostic accuracy, inadequate infrastructure, and a lack of skilled personnel (Li et al. 2020). These deficiencies create a vicious cycle: as primary care institutions fail to attract patients, they struggle to secure resources and talent, further eroding their credibility (Li et al. 2017b). Furthermore, economic policies and insurance incentives exacerbate this imbalance by favoring hospital-based care, discouraging patients from utilizing primary facilities (Liu et al. 2018). As a result, tertiary hospitals face unsustainable workloads, while primary care remains stagnant, undermining the system’s long-term sustainability. Weak integration among the tiers of care further compounds these challenges. The referral system is often ineffective, with inadequate coordination between primary and higher-level facilities disrupting continuity of care (Lian et al. 2019). This fragmentation hampers the efficiency of patient transitions and increases cost.

While these challenges are significant, the tiered healthcare system in China also offers a clear path toward improvement. The inauguration of tiered care reform initiatives, particularly in the domain of chronic disease management, proffers an opportunity to mitigate the pressure borne by tertiary hospitals while concomitantly augmenting the quality of care dispensed within primary healthcare facilities (Hu et al. 2021). One of the earliest pilot programs of tiered healthcare reforms showed promising results in managing chronic diseases more effectively, suggesting that with adequate support and resources, the primary care sector can play a pivotal role in enhancing the overall healthcare system (Xu et al. 2020). Should the reforms successfully surmount the issue of underutilization plaguing primary care and fortify the referral mechanisms, the tiered system stands to attain enhanced efficiency and equity, culminating ultimately in improved patient prognoses and curtailed costs throughout the entire healthcare system (Zhou et al. 2021b).

Looking ahead, the full potential of the tiered healthcare model will depend on effectively addressing current challenges while seizing emerging opportunities. Pivotal challenges, including the underutilization of primary care facilities, the inefficiency of the referral system, and the dearth of patient awareness regarding primary care, continue to be the main bottlenecks hindering the development of the tiered healthcare system (Li et al. 2021). Nevertheless, with the further refinement of policies and more efficient allocation of resources, the tiered healthcare system harbors substantial potential for advancement. Future policies ought to focus on enhancing the appeal of primary care through incentive mechanisms and fostering patient trust, thereby enabling primary care to effectively substitute for or complement tertiary services. The central aim of this study is to explore the optimal implementation of tiered healthcare policies, with a view to improving the overall efficiency and equity of healthcare services and advancing the further development of China’s healthcare system (Lu et al. 2019).

Literature review

The interplay between different tiers of healthcare services, particularly the roles of primary care and hospital services, as either substitutes or complements, remains a central question in health systems research. A comprehensive understanding of this dynamic is essential for the development of efficient and equitable healthcare systems. A number of studies have examined this interaction in a variety of global contexts, demonstrating a range of outcomes contingent on the specific structural, policy, and cultural characteristics of health systems.

Complementary or substitutive relationships are not confined to the domain of primary care and hospital services, rather, they extend across a range of healthcare delivery modalities, including specialist outpatient services, community-based care, telemedicine, and alternative medicine practices (Alsharif 2021, Schmitz-Grosz 2021, Dehghan et al. 2023). For example, in some systems, community-based services are designed to provide preventive and rehabilitative care, thereby substituting for certain hospital-based interventions. The integration of telemedicine into care pathways provides a further illustration of this interplay. In many cases, it serves to complement traditional care by facilitating follow-ups and chronic disease management, while in certain instances it may substitute for face-to-face consultations (Yu et al. 2023). Similarly, alternative and complementary medicine (e.g. acupuncture, homeopathy) has been investigated as a substitute or adjunct to conventional care, particularly in the context of chronic pain management (Kim et al. 2017).

The role of primary care in health systems varies significantly, reflecting both substitution and complementarity in different contexts. In the United States, the Veterans Affairs healthcare system demonstrated that increasing access to primary care through Community-Based Outpatient Clinics resulted in a reduction in the usage of specialty care and hospitalization rates. This underscores the potential of primary care to substitute for specialized services in well-integrated systems (Fortney et al. 2005). In contrast, Lau et al. (2021) identified only a weak substitution effect from England, and their findings suggest that enhancing primary care may result in a redistribution of patient loads away from hospitals, although this does not necessarily lead to significant cost savings. This implies that primary care often serves to complement hospital services by enhancing the overall responsiveness of the health system. The complexity of these dynamics is further highlighted by studies on extended primary care hours. In Italy, extended primary care availability reduced non-urgent emergency department visits, effectively complementing hospital services by improving patient triage and care delivery for less severe cases (Lippi Bruni et al. 2016).

China’s healthcare system provides a distinctive perspective on the interplay between primary care and hospital services, given the country’s dual challenges of high demand for tertiary care and underutilized primary care services. Some studies in China indicate that primary care plays a limited role in substituting for hospital services, partly due to public perceptions of quality differences between primary and hospital care. For example, Yip et al. (2019) observed that patients frequently bypass primary care facilities in favor of tertiary hospitals, driven by trust and perceived expertise, which further exacerbates inefficiencies in the health system. Pilot programs designed to enhance the gatekeeping role of primary care have yielded promising results. In certain regions, there has been a notable increase in the utilization of primary care, with patient visits to community health centers rising by >50% (Xu et al. 2020). Consequently, there has been a reduction in hospital outpatient visits and expenditure, which suggests that the initial objective of redistributing patient flows has been achieved (Li et al. 2017a, Xu et al. 2020). Additionally, improvements in quality have been observed, particularly in areas such as care coordination and accessibility (Liang et al. 2019). Furthermore, satisfaction with factors such as convenience and waiting times has exhibited a modest increase (Wu et al. 2016). Nevertheless, these endeavors are still in their infancy, and significant challenges remain, including an unequal distribution of resources, a scarcity of primary care capacity, and a lack of cohesion in referral pathways (Qian and Ramesh 2024).

While substantial evidence exists from high-income countries, research on the substitutive and complementary roles of primary care in developing health systems, such as China’s, remains limited. The specific challenges of underfunded primary care, uneven service distribution, and cultural perceptions of care quality in developing countries necessitate context-specific investigations. Moreover, the urgency of addressing the inefficiencies in China’s healthcare system highlights the necessity for research to inform policy and system design, particularly in strengthening the gatekeeping role of primary care and integrating care pathways across tiers.

Methods

Data and design

The data used in this study are derived from official administrative medical insurance reimbursement records under basic medical insurance, covering a period of 7 years from 2013 to 2019. The dataset includes repeated cross-sectional annual samples of insured individuals from 66 cities across China, including four municipalities, 27 provincial capitals, and 34 prefecture-level cities from different provinces. While the data contain unique patient identifiers within each year, it is not possible to longitudinally track individuals across years. The database encompasses both outpatient and inpatient medical information, including data on urban and rural residents, as well as urban employees insured in these cities. The sampling samples are drawn from the medical insurance production databases of each coordination area, with theoretical sampling of ≥2% for provincial capitals and municipalities, and ≥5% for prefecture-level cities. All inpatient records within the ‘calendar year’ are included for all sampled patients who receive benefits. It is evident that the records under scrutiny contain a plethora of information, including personal information (e.g. the individual’s unique patient identifier, gender, date of birth, insurance type, and insured region), medical institution-related information (e.g. the institution’s name, its level, and the specific department responsible for treatment), treatment details [e.g. location, serial number, admission and discharge dates, disease diagnosis, and the International Classification of Diseases, Tenth Revision (ICD-10) codes], cost information (e.g. total cost, insurance fund payments, patient out-of-pocket payments), and medical insurance settlement methods. With regard to the authenticity of the data and the extent of the sample coverage, the dataset is considered to be highly representative.

The initial subset used for analysis encompassed 1 697 551 patient records, comprising annual aggregated data such as total medical visits, total cost, insurance-fund payments, patient out-of-pocket payments, and breakdown by hospital level and visit type. The demographic variables included age, gender, insurance type, and the most frequent diagnosis code per patient annually. The models employed incorporate fixed effects for city, year, and disease type to control for confounding factors at these levels. This approach is advantageous in that it accounts for temporal trends, regional heterogeneity, and variations in case mix, thereby improving estimation accuracy.

In our main analyses, we control for case-mix at the most granular level available by using fixed effects based on the specific primary discharge diagnosis codes from ICD-10. ICD-10 is a globally recognized diagnostic tool maintained by the World Health Organization, which provides a standardized system for classifying diseases and health problems. Controlling for individual diagnosis codes enables our models to account more precisely for the wide variation in patient health conditions and disease severity, thus minimizing potential confounding from case-mix and providing more reliable estimates of the substitution and complementarity effects.

To ensure the robustness of our findings, we also conducted two sets of sensitivity analyses, which are presented in the online supplementary material. The first set of analyses (supplementary Tables S1–S3, see online supplementary material) uses a broader disease categorization based on the first letter of the ICD-10 code. This approach groups diseases into major chapters as defined by the World Health Organization (e.g. ‘J’ for respiratory diseases) and tests whether our findings hold with a more aggregated case-mix control. The second set (supplementary Tables S4–S6, see online supplementary material) employs fixed effects for the patient’s discharge department (e.g. cardiology, pediatrics) as an alternative proxy, capturing clinical specialty groupings rather than diagnostic classifications. The consistency of results across these different specifications, as detailed in the results section, strengthens the validity of our conclusions.

In accordance with the established approaches in the literature (Woodger et al. 2018), a series of data-cleansing procedures was implemented, drawing upon numerical length-of-stay thresholds and percentile cutoffs that typically exceed the 90th to 95th percentiles. This approach was adopted with the objective of identifying potential outliers. Patients with an annual total number of medical visits >14 were excluded to remove cases likely to represent chronic or complex conditions that are not representative of the general population. This step resulted in the exclusion of 177 288 records, thereby yielding a final analytic sample of 1 520 263 patients. The detailed sample selection process is illustrated in supplementary Figure S1 in the online supplementary materials.

Due to the administrative nature of the database, missing data were minimal. The majority of missing data was observed in the field of diagnosis coding, as not all hospitals consistently provided detailed ICD diagnosis codes. In order to address this limitation, all records with missing or unknown diagnosis codes were uniformly assigned to an ‘unspecified’ category. Other key variables such as demographic information and utilization measures had near-complete data coverage, and no imputation methods were applied.

For transparency and robustness, we present both unadjusted models without additional covariates or fixed effects, and fully adjusted models including patient demographics and fixed effects. The main text reports fully adjusted models to reduce confounding and improve estimation accuracy, while results of unadjusted regression models are presented in supplementary Tables S7–S9 in the online supplementary material.

Variables

In our model, we consider primary care ($Primar{y}_{it}$), which captures the healthcare utilization at primary care facilities, including both the number of visits and associated costs. Similarly, healthcare utilization at secondary ($Secondar{y}_{it}$) and tertiary ($Tertiar{y}_{it}$) levels is captured, including details of outpatient and inpatient visits and costs. Following the approach of existing literature, all variables related to cost have been logarithmically processed. Furthermore, the study incorporates key covariates to adjust for potential confounding factors. Age is treated as a continuous variable, reflecting its continuous impact on healthcare needs. Gender is coded as a binary variable (0 for female, 1 for male), and insurance type is categorized into two groups: Urban Resident Basic Medical Insurance (0) and Urban Employee Basic Medical Insurance (1). This classification system allows for the exploration of variations in service usage based on the type of insurance coverage (Han et al. 2020). Detailed variable definitions are presented in supplementary Table S10 in the online supplementary material.

Statistical analysis

In order to estimate the substitution and complementary effects between outpatient and inpatient services, fixed-effects regression models were employed. In our analytical framework, these effects are operationalized as follows: a statistically significant negative coefficient (β < 0) for a lower-tier service utilization variable (e.g. primary care visits) when predicting a higher-tier service outcome (e.g. tertiary hospital visits) is interpreted as evidence of a substitution effect. Conversely, a statistically significant positive coefficient (β > 0) suggests a complementary effect. These models are well-suited for controlling time-invariant individual characteristics that could potentially bias our results, such as inherent health conditions, regional healthcare infrastructure, and policy variations. The fixed-effects approach allows for a focused analysis of within-individual and within-region variations over time, thereby enhancing the precision of our estimations regarding healthcare utilization patterns.

It is important to note that this study primarily investigates the relationships between primary, secondary, and tertiary healthcare institutions in China under the backdrop of the broadly implemented tiered healthcare system, rather than focusing on a specific policy reform. The tiered healthcare system itself serves as a contextual framework for the analysis, the aim of which is to examine how healthcare utilization behaviors shift between different levels of hospitals. In this context, the study does not directly assess the impact of specific tiered healthcare policies; rather, it explores the substitution effects within hospital services across levels, providing insights into how the broader policy environment influences the dynamics between primary, secondary, and tertiary care.

In order to address potential concerns regarding the impact of tiered healthcare policies, additional year-by-year sub-sample analyses were conducted. The findings indicate that the coefficients across the years remain largely consistent, suggesting that the direct effect of specific policy changes related to tiered healthcare on healthcare utilization patterns has been limited. This finding aligns with the primary outcomes of this study, which emphasized substitution dynamics between hospital levels rather than a substantial shift driven by policy reforms. These results are further illustrated in supplementary Tables S11–S15 in the online supplementary material.

The model specifications included fixed effects for year, city, and disease type to control for temporal policy shifts, geographical disparities in healthcare infrastructure, and inherent treatment variations across medical conditions. The model is formally expressed as follows:

$$Secondar{y}_{it}={\alpha }_0+{\beta }_{1}Primar{y}_{it}+{\beta }_2{X}_{it}+{\lambda }_t+{\delta }_j+v_i+{\epsilon }_{itj}$$ (1)

$$Tertiar{y}_{it}={\alpha }_0+{\beta }_{1}Primar{y}_{it}+{\beta }_2{X}_{it}+{\lambda }_t+{\delta }_j+v_i+{\epsilon }_{itj}$$ (2)

$$Tertiar{y}_{it}={\alpha }_0+{\beta }_{1}Secondar{y}_{it}+{\beta }_2{X}_{it}+{\lambda }_t+{\delta }_j+v_i+{\epsilon }_{itj}$$ (3)

In equations (1) to (3), the subscripts i, t, and j represent the individual, the year in which the individual is observed, and the prefecture-level city administrative unit to which the individual belongs, respectively. ${\lambda }_t$, ${\delta }_j$ and $v_i$ are year fixed effects, city fixed effects, and disease fixed effects, respectively. $X_{it}$ are a series of control variables regarding the status of individual i in year t, while ${\epsilon }_{itj}$ denotes a random disturbance term. Our particular interest is the coefficient ${\beta }_1$.

Results

Descriptive statistics

The descriptive statistics summarized in Table 1 illustrate the patterns of healthcare utilization and cost across primary, secondary, and tertiary levels of care. On average, patients made ∼2.98 visits to all levels of health care facilities, with the majority of visits occurring at primary care facilities (mean = 1.15 visits). Secondary and tertiary hospitals saw fewer visits, with means of 0.76 and 1.07 visits, respectively.

Table 1.

Descriptive characteristics of study population.

Variable	n	Mean	Std	Min	Max
Total visits	1 520 263	2.98	2.84	1	13
Total visits at primary	1 520 263	1.15	2.02	0	13
Total visits at secondary	1 520 263	0.76	1.62	0	13
Total visits at tertiary	1 520 263	1.07	2.06	0	13
Total cost, USD	1 520 263	374.80	1600.25	0.01	217451.52
Total cost at primary, USD	1 520 263	43.09	315.94	0	104432.13
Total cost at secondary, USD	1 520 263	100.96	645.76	0	149584.49
Total cost at tertiary, USD	1 520 263	230.77	1404.69	0	217451.52
Gender	1 520 263	0.490	0.500	0	1
Age	1 520 263	44.382	20.031	0	119
Health Insurance type	1 520 263	0.646	0.478	0	1

In terms of healthcare costs, the average total cost per patient was ∼USD 374.80. Costs at primary care facilities averaged ∼USD 43.09 per patient, while costs at secondary and tertiary hospitals were higher, averaging USD 100.96 and USD 230.77 respectively. The maximum cost recorded reached up to ∼USD 217 445.52, reflecting the high cost of treatments in certain cases.

The demographic profile showed a nearly equal gender distribution with a mean of 0.490, indicating a nearly balanced mix of male and female patients. The average age of the cohort was 44.38 years, with ages ranging from 0 to 119 years, demonstrating a board demographic span. Regarding insurance coverage, 64.6% of the observed population had Urban Employee Basic Medical Insurance.

Relationship between primary care services and secondary or tertiary hospital services

Table 2 reveals that each additional visit to primary care facilities is associated with a 38.8% reduction in the probability of visiting a secondary hospital (β= −0.388, P < 0.01). Specifically, each additional visit to primary care facilities is associated with a reduction of 0.127 visits (β = −0.127, P < 0.01) and a 27.7% decrease in costs at secondary hospitals (β = −0.277, P < 0.01). Moreover, a 1% increase in costs at primary care facilities is associated with 0.134 reductions in visits (β = −0.134, P < 0.01) and a 34.8% reduction in costs at secondary hospitals (β = −0.348, P < 0.01).

Table 2.

Relationship between primary care utilization and secondary hospital utilization.

	(1)	(2)	(3)	(4)	(5)
	Secondary visited	Total visits at secondary	Total cost at secondary	Total visits at secondary	Total cost at secondary
Primary visited	−0.388***
	(0.001)a
Total visits at primary		−0.127***	−0.277***
		(0.001)	(0.001)
Total cost at primary				−0.134***	−0.348***
				(0.001)	(0.001)
Gender	−0.003***	−0.042***	−0.035***	−0.040***	−0.027***
	(0.001)	(0.003)	(0.005)	(0.003)	(0.004)
Age	0.001***	0.002***	0.006***	0.003***	0.009***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Health insurance type	0.037***	0.239***	0.513***	0.166***	0.293***
	(0.001)	(0.003)	(0.006)	(0.003)	(0.006)
City FEsb	Yes	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes	Yes
Disease FEs	Yes	Yes	Yes	Yes	Yes
n	1 449 692	1 449 692	1 449 692	1 449 692	1 449 692
R²	0.265	0.098	0.163	0.120	0.220

^aRobust standard errors clustered at individual level are in parentheses. ***, **, * are significant at the 1%, 5%, and 10% levels. ^bCity FEs, Year FEs, and Disease FEs represent fixed effects at the city, year, and disease type (ICD-10), respectively. All variance inflation factors < 10.

Table 3 demonstrates that each additional visit to primary care facilities is associated with a 32.0% reduction in the likelihood of visiting a tertiary level hospital (β = −0.320, P < 0.01). Each additional primary hospital visit is associated with 0.147 fewer visits (β = −0.147, P < 0.01) and a 29.0% reduction in costs at tertiary hospitals (β = −0.290, P < 0.01). Furthermore, a 1% increase in costs at primary care facilities may be associated with 0.137 decrease in visits (β = −0.137, P < 0.01) and a 33.8% reduction in costs at tertiary hospitals (β = −0.338, P < 0.01).

Table 3.

Relationship between primary care utilization and tertiary hospital utilization.

	(1)	(2)	(3)	(4)	(5)
	Tertiary visited	Total visits at tertiary	Total cost at tertiary	Total visits at tertiary	Total cost at tertiary
Primary visited	−0.320***
	(0.001)a
Total visits at primary		−0.147***	−0.290***
		(0.001)	(0.001)
Total cost at primary				−0.137***	−0.338***
				(0.001)	(0.001)
Gender	−0.029***	−0.119***	−0.170***	−0.117***	−0.163***
	(0.001)	(0.003)	(0.005)	(0.003)	(0.005)
Age	−0.001***	−0.002***	−0.001***	−0.001***	0.002***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Health insurance type	0.089***	0.432***	0.932***	0.367***	0.730***
	(0.001)	(0.004)	(0.006)	(0.004)	(0.006)
City FEsb	Yes	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes	Yes
Disease FEs	Yes	Yes	Yes	Yes	Yes
n	1 449 913	1 449 913	1 449 913	1 449 913	1 449 913
R²	0.336	0.183	0.260	0.194	0.297

^aRobust standard errors clustered at individual level are in parentheses; ***, **, * are significant at the 1%, 5%, and 10% levels. ^bCity FEs, Year FEs, and Disease FEs represent fixed effects at the city, year, and disease type (ICD-10) respectively. All variance inflation factors < 10.

To further explore the dynamics over time, we performed year-by-year sub-sample analyses (supplementary Tables S11–S15 in the online supplementary material). The results confirm that the substitution effects are robust and persistent. For instance, increased primary care utilization was significantly associated with reduced inpatient visits at tertiary hospitals in every year from 2013 to 2019 (supplementary Table S11). Although the effect size varied annually—ranging from a coefficient of −0.0322 in 2017 to −0.121 in 2019—the relationship remained consistently negative and statistically significant. This pattern suggests that while the intensity of substitution may be influenced by short-term factors, the underlying dynamic is a stable characteristic of the healthcare system throughout the observation period, strengthening the main conclusions of our study.

Supplementary Table S16 (see online  supplementary material) provides further insights into the interactions between hospital levels, offering supplementary statistical evidence to support the substitution effects between secondary and tertiary hospital services. Each increase in visits and costs at secondary hospitals is significantly associated with reduced visits and expenditure at tertiary hospitals.

Further analysis on the relationship in terms of different types of healthcare

As demonstrated in Table 4, for each additional outpatient visit at primary care facilities, the probability of an inpatient visit to a tertiary hospital may be reduced by 5.6% (β = −0.056, P < 0.01). In addition, for each additional outpatient visit at primary care facilities, there may be a reduction of 0.012 visits (β = −0.012, P < 0.01) and a 7.8% decrease in costs for inpatient services at tertiary hospitals (β = −0.078, P < 0.01). Moreover, an augmentation of 1% in outpatient costs at the primary level is associated with a 0.014 decrease in inpatient visits (β = −0.014, P < 0.01) and a 9.4% reduction in inpatient costs at tertiary hospitals (β = −0.094, P < 0.01).

Table 4.

Relationship between outpatient services at primary care facilities and inpatient services at tertiary hospitals.

	(1)	(2)	(3)	(4)	(5)
	Inpatient tertiary visited	Total visits at inpatient tertiary	Total cost at inpatient tertiary	Total visits at inpatient tertiary	Total cost at inpatient tertiary
Outpatient primary visited	−0.056***
	(0.001)a
Total visits at outpatient primary		−0.012***	−0.078***
		(0.000)	(0.001)
Total cost at outpatient primary				−0.014***	−0.094***
				(0.000)	(0.001)
Gender	−0.033***	0.002***	−0.004	0.002***	−0.001
	(0.001)	(0.001)	(0.004)	(0.001)	(0.004)
Age	−0.001***	0.001***	0.005***	0.001***	0.005***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Health insurance type	0.162***	0.017***	0.117***	0.012***	0.076***
	(0.001)	(0.001)	(0.005)	(0.001)	(0.005)
City FEsb	Yes	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes	Yes
Disease FEs	Yes	Yes	Yes	Yes	Yes
n	1 449 692	1 449 692	1 449 692	1449 692	1 449 692
R²	0.270	0.211	0.255	0.213	0.259

^aRobust standard errors clustered at individual level are in parentheses; ***, **, * are significant at the 1%, 5%, and 10% levels. ^bCity FEs, Year FEs, and Disease FEs represent fixed effects at the city, year, and disease type (ICD-10) respectively. All variance inflation factors < 10.

Table 5 indicates that for each additional outpatient visit at primary care facilities, the likelihood of an outpatient visit at a tertiary hospital may be reduced by 14.5% (β = −0.145, P < 0.01). Furthermore, for each additional outpatient visit at primary care facilities, there may be a 0.134 decrease in visits (β = −0.134, P < 0.01) and a 21.5% reduction in costs at tertiary outpatient services (β = −0.215, P < 0.01). Moreover, an increase of 1% in outpatient costs at primary care facilities is associated with a 0.136 decrease in visits (β = −0.136, P < 0.01) and a 25.2% decrease in costs at tertiary outpatient services (β = −0.252, P < 0.01).

Table 5.

Relationship between outpatient services at primary care facilities and outpatient services at tertiary hospitals.

	(1)	(2)	(3)	(4)	(5)
	Outpatient tertiary visited	Total visits at outpatient tertiary	Total cost at outpatient tertiary	Total visits at outpatient tertiary	Total cost at outpatient tertiary
Outpatient primary visited	−0.145***
	(0.001)a
Total visits at outpatient primary		−0.134***	−0.215***
		(0.001)	(0.001)
Total cost at outpatient primary				−0.136***	−0.252***
				(0.001)	(0.001)
Gender	−0.002***	−0.121***	−0.175***	−0.116***	−0.166***
	(0.000)	(0.003)	(0.004)	(0.003)	(0.004)
Age	0.000***	−0.003***	−0.003***	−0.002***	−0.002***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Health insurance type	0.011***	0.423***	0.910***	0.374***	0.802***
	(0.001)	(0.003)	(0.005)	(0.003)	(0.004)
City FEsb	Yes	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes	Yes
Disease FEs	Yes	Yes	Yes	Yes	Yes
n	1 449 692	1 449 692	1 449 692	1 449 692	1 449 692
R²	0.256	0.187	0.276	0.197	0.301

^aRobust standard errors clustered at individual level are in parentheses; ***, **, * are significant at the 1%, 5%, and 10% levels. ^bCity FEs, Year FEs, and Disease FEs represent fixed effects at the city, year, and disease type (ICD-10) respectively. All variance inflation factors < 10.

Supplementary Tables S14–S17 (see online supplementary material) expand upon these analyses by exploring the substitution effects within hospital services at a more granular level. Supplementary Table S17 illustrates the association of primary care outpatient visits with secondary hospital inpatient services, indicating significant reductions in both visits and costs. In a similar manner, supplementary Table S18 (see online supplementary material) examines the reduction in outpatient visits and costs at secondary hospitals due to enhanced primary care outpatient services. Supplementary Table S19 (see online supplementary material) assesses the influence of outpatient services at secondary hospitals on inpatient services at tertiary hospitals, showing a decrease in both inpatient visits and costs. Supplementary Table S20 (see online supplementary material) focuses on the relationship between outpatient services across secondary and tertiary hospitals, revealing substantial decreases in both visits and associated costs at tertiary hospitals.

Relationship within the same level of hospital in terms of outpatient and inpatient services

Table 6 indicates that each additional outpatient visit at a tertiary hospital is associated with a marginal increase in the probability of an inpatient visit at the same level of hospital, with a 3.8% rise observed (β = 0.038, P < 0.01). This consequently results in an increase of 0.009 in inpatient visits (β = 0.009, P < 0.01) and a 5.7% increase in inpatient costs (β = 0.057, P < 0.01). Additionally, an increase of 1% in outpatient costs at tertiary hospitals is associated with a 0.004 increase in inpatient visits (β = 0.004, P < 0.01) and a 2.0% increase in inpatient costs (β = 0.020, P < 0.01). Supplementary Table S21 (see online supplementary material) shows that each additional outpatient visit at a secondary hospital is also associated with an increase in the likelihood of an inpatient visit at the same level of hospital, as well as an increase in inpatient visits and costs. An increase in outpatient costs at secondary hospitals is also associated with an increase in inpatient visits and costs at the same level of hospital.

Table 6.

Relationship between outpatient services and inpatient services at tertiary hospitals

	(1)	(2)	(3)	(4)	(5)
	Inpatient tertiary visited	Total visits at inpatient tertiary	Total cost at inpatient tertiary	Total visits at inpatient tertiary	Total cost at inpatient tertiary
Outpatient tertiary visited	0.038***
	(0.001)a
Total visits at outpatient tertiary		0.009***	0.057***
		(0.000)	(0.001)
Total cost at outpatient tertiary				0.004***	0.020***
				(0.000)	(0.001)
Gender	−0.033***	0.003***	0.001	0.002***	−0.003
	(0.001)	(0.001)	(0.004)	(0.001)	(0.004)
Age	−0.001***	0.001***	0.004***	0.001***	0.004***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Health insurance type	0.164***	0.020***	0.130***	0.020***	0.137***
	(0.001)	(0.001)	(0.005)	(0.001)	(0.005)
City FEsb	Yes	Yes	Yes	Yes	Yes
Year FEs	Yes	Yes	Yes	Yes	Yes
Disease FEs	Yes	Yes	Yes	Yes	Yes
n	1 449 692	1 449 692	1 449 692	1 449 692	1 449 692
R²	0.270	0.209	0.253	0.209	0.252

^aRobust standard errors clustered at individual level are in parentheses; ***, **, * are significant at the 1%, 5%, and 10% levels. ^bCity FEs, Year FEs and Disease FEs represent fixed effects at the city, year, and disease type,s (ICD-10) respectively. All variance inflation factors < 10.

Discussion

This study provides novel perspectives on the substitution effects among different levels of hospital care, utilizing a nationally representative sample. It represents the first comprehensive examination of the relationships between primary, secondary, and tertiary institutions in China. The findings robustly demonstrate that, despite current limitations in primary healthcare capacity, primary care can significantly alleviate challenges within the healthcare system. The study emphasizes the transformative capacity of effective primary care in optimizing healthcare delivery, particularly in LMICs.

Primarily, our results reveal a distinct substitution effect, wherein increased utilization of primary care services leads to a significant reduction in the use of secondary and tertiary hospital services. This effect represents an effective reallocation of resources, driven by patient incentives to seek less intensive care when appropriate and by healthcare providers’ efforts to manage more conditions at the primary level. Such shifts optimize resource allocation by decreasing the unnecessary reliance on higher-level, more costly care facilities. For instance, Danish research highlights that robust general practitioner networks with gatekeeping functions are pivotal in reducing superfluous specialist consultations and hospitalizations (Krämer and Schreyögg 2019). Cost-effective primary care can attenuate the reliance on resource-intensive tertiary facilities, as has been witnessed in LMICs. Other Chinese studies suggest that enhanced utilization of primary care contributes to a reduction in hospital admissions and outpatient visits for preventable conditions (Yip et al. 2019, Li et al. 2020). In contrast, studies from the United States illustrate a complementary relationship, where increased primary care utilization enhances referrals and diagnostic precision, ultimately leading to higher demand for specialty care (Starfield et al. 2005). This is because greater use of primary care often results in more referrals to specialty services due to improved screening and case detection (Aliberti et al. 2024, Islam et al. 2024). While this dynamic differs from the substitution effects observed in China, it reveals a mechanism through which robust primary care can enhance overall healthcare quality rather than simply reducing resource use. In China, where primary care remains relatively underdeveloped in terms of diagnostic and therapeutic capacities, this observed substitution effect points to a promising pathway for healthcare reform.

Thus, this study highlights the substitutability of outpatient services across various levels of hospital care, a concept increasingly explored in health systems worldwide as a strategy for optimizing resource utilization and improving patient outcomes. Our findings indicate a substitution effect, where the use of outpatient services at primary care facilities reduces the demand for both outpatient and inpatient services at higher-level hospitals. This underscores the critical role of outpatient services in lower-tier institutions, particularly in reducing the utilization pressure on tertiary hospitals. This finding aligns with the international literature, which notes a trend toward service standardization across different healthcare levels. In Canada, regional care networks that integrate primary and secondary outpatient services have successfully diverted significant patient demand from tertiary care (Kearon and Risdon 2020). Similarly, the United Kingdom’s focus on expanding primary care capacity has resulted in fewer unnecessary outpatient visits to specialists, demonstrating the benefits of strengthening service delivery at lower levels (Winpenny et al. 2016). Longitudinal studies in Europe indicate that improved access to outpatient care in underserved areas is linked to reduced avoidable hospitalizations and inpatient costs, underscoring the potential of well-implemented outpatient services (Elek et al. 2019).

The overlap in service provision between tertiary hospitals and primary care facilities plays a critical role in the observed substitution effects. Primary care facilities provide a wide array of services, including outpatient care, family doctor contract services, chronic disease management, rehabilitation, and minor surgical procedures. This overlap enables primary care facilities to manage a substantial portion of patient care typically handled by higher-level hospitals, diverting cases that do not require specialized interventions. Such service overlaps are evident in many healthcare systems around the world. For instance, in Germany, primary care physicians are able to manage a significant proportion of the healthcare burden by providing timely treatment for common ailments and chronic diseases, reducing the pressure on tertiary hospitals (Busse et al. 2017, Krämer and Schreyögg 2019). Furthermore, the provision of more accessible and convenient primary care services has the effect of intercepting patients who would otherwise have sought care at a tertiary hospital. This encompasses the management of common illnesses, the conducting of regular follow-ups for chronic conditions, and the provision of preventive services such as immunizations and health screenings. These interventions effectively prevent many patients from utilizing unnecessary hospital resources, thereby reducing patient load in both outpatient and inpatient departments at tertiary hospitals (Yip et al. 2019). The core function of primary care extends beyond treatment to disease prevention and health management. As a result, fewer patients require specialized care at higher-level hospitals, leading to a more efficient use of healthcare resources (Li et al. 2020).

The indirect effects of primary care in reducing inpatient hospital admissions are closely linked to its role in substituting for outpatient services at tertiary hospitals. Specifically, primary care outpatient services first act as a substitute for outpatient visits to tertiary hospitals, which are often the initial point of contact for patients requiring specialized care. Concurrently, the correlation between outpatient services and inpatient admissions at tertiary hospitals manifests a complementary feature. In essence, primary care outpatient visits have been shown to contribute to two distinct pathways of substitution. Firstly, they have been observed to directly reduce the number of tertiary hospital outpatient visit. Secondly, they have been shown to indirectly reduce the need for inpatient services at tertiary hospitals, which are often initially triggered by outpatient consultations. International studies support this dual pathway of substitution. The reduction in outpatient visits is directly linked to fewer hospital admissions. Outpatient services are the primary source of hospital admissions, particularly in tertiary hospitals, where a significant portion of inpatient admissions is initially triaged through outpatient consultations, often in specialized departments. This ‘induced demand’ mechanism, whereby hospitals deliberately promote or generate heightened demand for inpatient services through the facilitation of excessive outpatient consultations or unnecessary diagnostic procedures, has been extensively documented in China. For instance, Zhou et al. (2021a) demonstrated that, under price regulation, tertiary hospitals often engage in such demand-induced behaviors by converting cases that could be handled in outpatient settings into inpatient admissions. Consequently, a decline in outpatient visit to tertiary hospitals, especially in areas characterized by high levels of demand-induced behavior, is directly associated with a decrease in inpatient admissions. When primary care effectively manages patients who do not require tertiary hospital resources, those patients who remain at such tertiary hospitals are more likely to need specialized treatment or hospitalization, thus enhancing the efficiency of tertiary hospital utilization. This approach is predicated on the notion that it can prevent the progression of minor ailments into more severe diseases. The timely and accessible nature of primary care services can play a crucial role in averting the exacerbation of conditions, thereby obviating the need for more costly and intensive hospital treatments (Elek et al. 2019).

From an incentive standpoint, tertiary hospitals face a misalignment of economic incentives that often promote unnecessary admissions. In the prevailing fee-for-service or Diagnosis-Related Group payment systems, inpatient services constitute a pivotal revenue stream for hospitals. This economic structure incentivizes hospitals to maximize bed occupancy, even in cases where clinical criteria for hospitalization are not strictly met. Furthermore, it is important to note that tertiary hospitals may engage in demand-induced behavior, such as excessive testing, splitting hospital stays, or upgrading treatments to maintain revenue streams. This behavior is exacerbated by hospital performance incentives linked to the volume of complex surgeries, bed numbers, and research outputs, rather than the efficiency of patient management and referral to primary care (Liu et al. 2020). Conversely, primary care facilities frequently possess inadequate financial incentives to manage chronic diseases effectively. The financial constraints experienced by primary care providers, particularly in rural areas, and their limited ability to receive adequate reimbursement for chronic disease management, contribute to a vicious cycle of underperformance. This has been demonstrated to result in a reduced number of patients receiving appropriate care, which consequently leads to lower income for primary care providers (He 2022).

Consequently, the misaligned incentives within primary care facilities and tertiary hospitals engender inefficiencies and hinder the optimization of resource use. While tertiary hospitals benefit from high-volume inpatient services, they also face financial pressures to maintain high occupancy rates, which can result in unnecessary hospitalizations. Conversely, primary care providers struggle with inadequate resources and financial rewards, leading to underutilization and inefficiencies in the management of chronic diseases. Addressing these misalignments through policy reform is crucial to ensuring that healthcare resources are allocated more effectively and that primary care is integrated as an active coordinator of care, ultimately reducing unnecessary hospital visits and admissions.

This study has yielded several clear policy recommendations that are essential for improving the efficiency and equity of China’s healthcare system. It is imperative to acknowledge the necessity of implementing reforms to the payment system as a matter of utmost priority. The introduction of a ‘regional insurance prepayment with residual retention’ model should be prioritized. In this system, regional insurance funds are allocated to health alliances in advance, with any remaining funds distributed between tertiary hospitals and primary care providers according to a predetermined ratio. This approach would incentivize tertiary hospitals to proactively refer patients to primary care settings, thus promoting the effective use of primary care resources. Concurrently, primary care should adopt a ‘per capita weighted payment’ system, where payment is based on the number of residents under contract, with additional weight given to populations with chronic conditions such as hypertension and diabetes (1.5 to 2 times the standard payment). This payment model has the potential to promote more precise and effective management of chronic diseases at the community level.

A redesign of incentives for higher-level hospitals is necessary. This includes the re-evaluation of performance-assessment criteria with a view to integrating referral rates and the effectiveness of primary care support into the performance evaluation of tertiary hospitals. Special allowances should be provided to incentivize hospital physicians to provide care at the primary care level, thereby enhancing the coordination between higher-level hospitals and primary care providers. Additionally, policies should focus on improving primary care capabilities, such as reinforcing the family doctor contract services, which play a pivotal role in improving primary care delivery and reducing the service demands placed on higher-level hospitals.

In addition, regulatory frameworks must be enhanced to ensure transparent and evidence-based hospitalization criteria. The establishment of robust accountability mechanisms would serve to prevent the unwarranted admission of patients to tertiary hospitals, a practice that is often driven by financial incentives or demand-induced behavior. This is of crucial importance in mitigating the overuse of hospital services, particularly in systems where financial incentives for inpatient care may encourage unnecessary admissions (Zhou et al. 2021a). By implementing stringent standards for admissions and ensuring oversight, China can reduce the over-reliance on tertiary care, improving both efficiency and equity within the healthcare system.

Finally, as emphasized in this study, the transformation of China’s healthcare system necessitates enhanced integration of digital tools. Investment in digital infrastructure, including shared electronic health records and population health management systems, should be accelerated. These tools would improve coordination between primary and tertiary care by enabling seamless referral tracking and providing data-driven accountability. International experience has shown that integrating these systems can increase the effectiveness of substitution by 22%–31%, and so improve the efficiency of the whole healthcare system (Orton et al., 2018). These digital systems would also help to realign provider incentives with patient-centered outcomes, ensuring that primary care is not just a passive substitute, but an active coordinator of care that manages a broad spectrum of conditions. Collaborative family doctor teams, including secondary-level specialists, could play a key role in this shift by addressing patient needs at the local level without excessive reliance on tertiary hospitals. This comprehensive approach will provide China with a roadmap to move away from its current hospital- and inpatient-centric system toward a more sustainable, efficient, and equitable healthcare system based on strengthened primary and community-based care.

It is crucial to position these findings within the broader policy context. While this study does not aim to causally evaluate any single policy intervention—a task complicated by the heterogeneous and concurrent nature of Chin’s healthcare reforms—it provides critical foundational evidence for the ongoing policy debate. Our results demonstrate a robust, system-level substitution dynamic, suggesting that policies designed to strengthen the capacity and quality of primary care are a promising pathway for enhancing overall health-system efficiency. These findings, therefore, serve not as a direct policy evaluation, but as an essential characterization of the systemic environment in which policies operate, offering a strong empirical rationale for prioritizing investments in primary care infrastructure, workforce, and service integration.

There are several limitations that should be acknowledged. First, while fixed-effects models control for regional variations in healthcare access and quality, the analysis does not account for temporal variations or changes in healthcare policy that may affect substitution effects over time. Such time-varying, unobservable factors—including evolving healthcare policies, hospital management practices, patient preferences, or improvements in healthcare quality—could influence healthcare utilization patterns and thus may confound causal inference regarding substitution effects. Second, our analysis lacks detailed patient severity information and explicit indicators of referral behaviors. Without severity-level adjustments, the observed substitution effects may primarily reflect shifts in utilization among patients with less severe conditions, rather than the genuine replacement of higher-tier services by primary care for complex cases. Incorporating referral indicators would allow a more precise evaluation of how primary care facilities relieve pressure from higher-level hospitals by reducing unnecessary referrals, thus providing a more appropriate measure of substitution. Third, the absence of longitudinal tracking restricts our ability to assess long-term health outcomes or persistent changes associated with the substitution between outpatient and inpatient services. Lastly, our sample is limited to insured urban populations, limiting the generalizability of our findings to rural or uninsured groups, who experience distinct barriers to healthcare access and utilization patterns. Future research incorporating broader populations, longitudinal tracking, and severity-specific referral information would help validate and extend the insights provided by our findings.

Conclusion

This study provides critical insights into the substitution effects and complementarity dynamics between different levels of hospital care in China’s tiered healthcare system, with important implications for health policy and resource allocation. First, our analysis reveals a clear substitution effect where increased use of primary care services effectively reduces reliance on secondary and tertiary care, supporting ongoing policy reforms aimed at optimizing resource allocation. Second, we show that the homogeneity of outpatient services across hospital levels facilitates substitution within outpatient care, with patients increasingly choosing primary and secondary facilities for services traditionally sought in tertiary hospitals. Finally, the study highlights the complementary relationship between outpatient and inpatient services within the same level of care, which is complicated by China’s issue of induced hospitalization, where financial incentives may encourage unnecessary admissions. These findings underscore the need for targeted policy interventions to improve healthcare system efficiency, ensure appropriate resource use, and address systemic challenges such as induced demand. To improve the efficiency and sustainability of healthcare, policymakers need to address these systemic distortions by aligning financial incentives for different levels of hospital care, enhancing the regulatory framework, and promoting integrated care models that foster better coordination between levels of care.

Finally, while our study provides a comprehensive overview of the substitution and complementarity dynamics, its observational design limits our ability to attribute these patterns to specific policy reforms. To build upon our findings and provide more definitive guidance for policymakers, future research should employ quasi-experimental designs to isolate the causal effects of tiered-care interventions. Methodologies such as difference-in-differences, regression discontinuity designs, or interrupted time series analyses, paired with granular data on the timing and intensity of local policy implementation, would be invaluable. Such studies could more precisely quantify the impact of specific reforms—for instance, changes in reimbursement schemes or the expansion of family doctor contracts—on patient behavior, healthcare costs, and system-wide efficiency, thereby advancing evidence-based policymaking in China and other similar contexts.

Supplementary data

Supplementary data is available at Health Policy and Planning  online.

Author contributions

R.J., J.T., and Y.S. contributed to conception or design of the work. R.J., J.T., and C.N. contributed to data analysis and interpretation. R.J. and J.T. contributed to drafting of the article. All authors contributed to critical revision of the article and approved the manuscript prior to submission.

Reflexivity statement

The authors of the study include two females and two males and span multiple levels of seniority and diversity of disciplinary expertise. While two of the authors specialize in health policy and health systems in China and Asian regions, the third is a senior demographer who focuses primarily on population health and the impact of health systems on it; the last is an economist who specializes in policy evaluation and causal inference. All four authors have extensive experience conducting quantitative studies in China.

Ethical approval

Ethical approval for this type of study is not required by our university.

Funding

This work was supported by grants from the National Natural Science Foundation of China (grant number: 72404274) the Fundamental Research Funds for the Central Universities, and the Research Funds of Renmin University of China (grant number: 25XNK J07). The funders had no role in the study design, data collection and analysis, the writing of the manuscript, or in the decision to submit this article for publication.

Data availability

The data that support the findings of this study are not publicly available due to their confidential nature, which contains sensitive patient information. Data sharing is restricted to protect patient privacy in accordance with ethical and legal requirements.