Network characteristics of referrals for chronic disease patients: results from hierarchical medical reform in Shenzhen, China

Abstract

Background

The Chinese government has introduced a series of hierarchical medical policies to ensure continuity of care, but referrals remain difficult to implement effectively. This study aimed to evaluate the chronic disease referral network and explore the problems associated with the specific implementation of referrals.

Methods

This study was a repeated cross-sectional study of monthly data collected between August 2017 and December 2023 in Luohu district, Shenzhen, China. Social network analysis was used to construct a referral network for chronic disease patients. Density, degree centrality, and betweenness centrality were calculated to assess the relationships and status among different levels of care and their trends.

Results

Over seven years, 104,682 chronic disease patients were referred, with a predominance of downward referrals. The number of upward referrals (Z = 2.5776, P_trend < 0.01) and downward referrals (Z = 4.7723, P_trend < 0.001) increased significantly. Primary care facilities (PCFs) were strongly associated with the tertiary hospital (0.51–0.98). The out-degree of all levels of medical institutions showed a significant increasing trend (P_trend < 0.05). The coronavirus disease 2019 (COVID-19) pandemic did not cause significant level changes in network metrics but accelerated the upward trend in the out-degree of secondary hospitals (P < 0.05). The in-degree of secondary hospitals and PCFs showed a significant increasing trend (P_trend < 0.01). Public PCFs had significantly higher network metrics compared to private PCFs (P < 0.001).

Conclusions

The referral network has a vertical flow pattern conducive to the division of labour, cooperation, and resource integration of medical institutions in the region, and a hierarchical medical order is taking shape. However, poor communication between secondary hospitals and other institutions, high demand for data informatisation, and the gap between private and public PCFs may hinder further progress.

Introduction

Annually, 17 million people worldwide die from chronic noncommunicable diseases (NCDs) before the age of 70.86% of these premature deaths occur in low- and middle-income countries. According to the World Health Organization, deaths from chronic diseases are projected to reach 52 million by 2030, with 77% of these deaths occurring in low- and middle-income countries [1]. The control of chronic NCDs typically requires continuous treatment and management [2, 3]. However, health systems in low- and middle-income countries are primarily designed to address acute diseases [4]. The inverted triangle of health resource allocation and fragmentation of the healthcare service system in China is not conducive to the need for sustained treatment of chronic NCDs [5].

To address these challenges, the Chinese government has introduced a series of hierarchical medical reform policies. The Guiding Opinions on Promoting the Construction of a Hierarchical Diagnosis and Treatment System, issued by the General Office of the State Council in 2015, propose gradually developing a hierarchical diagnosis and treatment model that includes grassroots first diagnosis, two-way referrals, the separation of emergencies and chronic diseases, and up-and-down linkages, and establishing a hierarchical diagnosis and treatment system consistent with national conditions [6]. This system aims to optimise resource allocation and enhance service efficiency through integration and coordination within healthcare networks. To further strengthen the hierarchical medical system, initiatives such as the Outline of the National Health and Medical Service System Plan (2015–2020) [7] and the Management Measures for Medical Consortia (Trial) [8] have been introduced to promote medical consortia, enhance telemedicine, and establish unified two-way referral platforms, thereby facilitating the vertical flow of high-quality medical resources. In 2023, the State Council of China issued the ‘Opinions on Further Improving the Healthcare Service System,’ advocating the exploration of various forms of integrated management between district-level hospitals and community health service institutions, as well as enhancing continuous and smooth two-way referral service pathways.

Despite the increasing urgency of hierarchical medical reform in China, the implementation of two-way referral systems continues to encounter significant challenges. Challenges such as an incomplete two-way referral mechanism, underdeveloped benefit-sharing systems, fragmented medical consortia, and inadequate communication of medical information hinder the effective implementation of hierarchical diagnosis and treatment [9, 10]. Existing research on referral systems has predominantly focused on theoretical analyses of system structures and policy mechanisms. Alternatively, studies examine residents’ attitudes towards referrals and their willingness to seek care at primary care facilities to facilitate the implementation of two-way referrals [11, 12]. However, there is limited research using real data to specifically analyse the problems with two-way referrals.

Effective referral networks have been demonstrated to improve healthcare services for chronic disease patients in some developed countries [13, 14]. Social network analysis is based on a relational perspective, examining the connections between individuals or organisations as networks [15]. Quantitatively analysing the characteristics of such networks can provide better explanations of the behaviour of individuals or organisations. This approach has been gradually used in medicine in recent years to help understand, explain, and modify behavioural patterns and disease transmission [16–18]. The analysis of referral networks is essential for assessing the effectiveness of hierarchical medical systems. Referral networks reflect the intensity of interactions and role allocation among medical institutions, providing a dynamic perspective on policy functionality in practice. Referral network analysis identifies key institutions driving policy implementation, uncovers systemic bottlenecks, and evaluates patient flow across medical tiers, providing scientific evidence for policy optimisation. This study analyses the chronic disease referral network in Luohu district to address the following questions: Has the referral network of different levels of medical institutions in Luohu district, Shenzhen, been coordinated up and down following the hierarchical medical reform? What issues exist in the referral network? How can the implementation of the hierarchical medical policy be further promoted?

Methods

Study setting

At the end of 2014, the Shenzhen Municipal Health Commission decided to initiate primary care reform. Luohu district was selected to implement the pilot reform of hierarchical diagnosis and treatment. The Luohu medical reform was launched in August 2015 with the official establishment of the Luohu Hospital Group. The Luohu Hospital Group represents a successful exploration of the urban medical group model. At present, Shenzhen has initially established an integrated, high-quality, and efficient medical service system. The system is based on district medical institutions, promotes group operations, and is responsible for promoting the health of the population. The health-centred, grassroots-focused healthcare reform in Shenzhen is advancing rapidly [19].

Luohu district has a resident population of 1.14 million, with chronic diseases such as diabetes and hypertension being significant health issues among residents [20]. The district’s healthcare system consists of three levels of medical institutions, forming a multi-level medical service network offering comprehensive services, from basic community healthcare to specialist treatment. Luohu district has implemented a series of policies to promote hierarchical diagnosis and treatment. (a) A region-wide two-way referral platform has been established [21]. Standardised data entry and updating processes have been implemented, and automatic verification and consistency checking mechanisms have been introduced to ensure the information is complete and reliable. A dedicated team is responsible for regular maintenance, employing strict encryption measures and role-based access control, adhering to relevant laws and regulations, and safeguarding data privacy. Adopts common medical data exchange standards to ensure consistent data formats and coding across various information systems, enabling seamless data sharing. Ensures that the information systems of primary care facilities, hospitals, and the Health Luohu Application (APP), are interconnected with residents’ health records. Staff at different levels of healthcare facilities have clear responsibilities to ensure communication and coordination among all institutional levels. Authorised general practitioners and hospital specialists can handle patient referrals using primary care institutions, hospital information systems, and the Health Luohu APP. (b) Utilises the regulatory role of economic levers [22]. Adjusts the payment ratio for general outpatient services, as well as the deductible and payment ratio for inpatient services, based on the level of the medical institution. (c) Promotes contracted services for family doctors [23]. Integrates basic medical services and public health services, and uses residents’ health records to provide health management services for different populations. Provides priority referrals to specialists and other services for patients who have signed a contract with a family doctor. (Fig. 1)

Fig. 1.

Framework for referrals in Luohu district, Shenzhen

Data sources

Patients with chronic diseases require prolonged rehabilitation, treatment, and care, often involving referrals between different medical institutions. Efficient hierarchical referrals enable patients with chronic diseases to receive continuous healthcare services and to improve treatment outcomes. The two-way referral electronic information system was established in August 2017. Therefore, this study extracted monthly referral data from the two-way referral electronic information system for patients with chronic diseases such as hypertension, diabetes mellitus, chronic obstructive pulmonary disease, and malignant tumours in the Luohu district from 1 August 2017 to 31 December 2023. Records were excluded if they involved referrals to institutions outside Luohu district, lacked critical information (e.g., missing referral dates or institutions), or pertained to patients who declined the referral. Although additional information fields were added to the referral system over time, the key data required for this study, including referral dates, referring and receiving institutions, and disease categories, have been consistently recorded since 2017. This ensured data completeness throughout the study period. Variations in the naming of medical institutions (e.g., abbreviations or alternative spellings) were standardized and assigned unique identifiers to ensure consistency. All information used was de-identified, and no ethical review was involved.

Study design

Medical institutions are categorised into three grades based on their qualifications: tertiary hospitals, secondary hospitals, and primary care facilities (PCFs). PCFs are divided into two groups: public PCFs, representing those within the Luohu Hospital Group, and private PCFs, representing those provided by non-government providers. Changes in referrals were analysed based on the number of upward and downward referrals between medical institutions in Luohu district, Shenzhen, China, over seven years. The referral network consists of institutions and the links between them, where nodes represent medical institutions and links represent the relationships between these institutions. The referral network was divided into two levels of analysis: the group network and the individual medical institutions. The group network analysis included the density of referrals among the three levels of medical institutions. Medical institution analysis involved calculating the centrality of all individual members of the network, including degree centrality and betweenness centrality. Public PCFs and private PCFs were compared in terms of the number of referrals and the differences in density and centrality between these and higher-level medical institutions. Finally, to assess the impact of the coronavirus disease 2019 (COVID-19) pandemic on referral patterns.

Statistical analysis

The calculation of the network metrics was conducted in Ucinet (V. 6.791), and the graphical visualisation of the network was generated in Gephi (V. 0.10.1). Mann-Kendall trend analysis was used to assess changes in referral and network metrics over seven years. The method is a non-parametric test that does not require the data to follow a normal distribution, is robust to outliers, and is particularly sensitive to monotonic trends in time series. This makes it well-suited for capturing long-term changes in chronic disease referral patterns over the study period. Interrupted time series (ITS) analysis was employed to evaluate the impact of the COVID-19 pandemic on referral characteristics and to provide insight into the effectiveness of hierarchical medical policies. ITS is particularly suitable for assessing the effects of interventions or external events on trends over time, as it does not require a control group, and instead uses pre- and post-intervention data collected at multiple time points to estimate changes. Mann-Kendall trend and ITS analyses were performed in R (V. 4.3.2), and Newey-West standard errors were applied in the ITS model to adjust for autocorrelation by weighting the lagged covariance matrix of the residuals. Kruskal-Wallis test was used in SPSS 25 to assess the differences in network metrics across different levels of medical institutions. Mann-Whitney U test was used to evaluate the differences in the number of referrals and network metrics between public and private PCFs. P < 0.05 was considered statistically significant.

Results

Overview of referral networks

Chronic disease referral records were collected from August 2017 to December 2023. A total of 42,825 records were excluded due to incomplete information, declined referrals, or referrals outside the Luohu district. After processing, the final dataset contained 104,682 referral records. Referrals of chronic disease patients in Luohu district are mainly downward referrals, except for 2020 and 2021, when the proportion of upward referrals (59.6%, 62.0%) was greater than that of downward referrals (40.4%, 38.0%). For all other years, downward referrals ranged from 64.2 to 89.0%, consistently outpacing upward referrals (11.0–35.8%) (Fig. 2). The total number of referrals in Luohu district showed a significant upward trend from 2017 to 2023 (Z = 6.0961, P_trend < 0.001). The number of upward referrals (Z = 2.5776, P_trend < 0.01) and downward referrals (Z = 4.7723, P_trend < 0.001) showed significant upward trends (Table 1). Figure 3 shows the visualisation of the chronic disease referral networks in Luohu district. The referral network shows an increasing number of nodes, which are becoming more connected.

Fig. 2.

The number and proportion of upward and downward referrals

Table 1.

The number (median (min, max)) of upward and downward referrals in Luohu district from 2017 to 2023

	n	Upward referrals	Downward referrals	Total
Years
2017	55	170 [123, 193]	434 [383, 521]	594 [526, 713]
2018	59	235 [103, 309]	506 [302, 609]	731 [405, 895]
2019	63	346 [152, 431]	579 [359, 900]	886 [511, 1306]
2020	61	341 [77, 477]	264 [26, 493]	531 [247, 938]
2021	62	396 [234, 548]	215 [1, 9039]	675 [258, 9473]
2022	59	286 [165, 412]	1892 [959, 2371]	2198 [1140, 2725]
2023	61	272 [120, 470]	2298 [1374, 2684]	2515 [1490, 3088]
Z	-	2.5776	4.7723	6.0961
P trend	-	< 0.01	< 0.001	< 0.001

Fig. 3.

Referral networks in 2017 (A), 2020 (B), and 2023 (C). (Each node in the network corresponds to a healthcare institution, with its size indicating the degree of centrality. The thickness of the lines between nodes indicates how frequently referrals are made between institutions.)

Tight linkages between the tertiary hospital and PCFs

The results of the group density analysis across different levels of medical institutions are shown in Fig. 4. The density of the referral network from the tertiary hospital to PCFs was stable, except in 2017, fluctuating mainly between 0.93 and 0.98. The lowest density in 2017 may have resulted from delays in the operation of the information technology platform. The density of the referral network from PCFs to the tertiary hospital remained between 0.51 and 0.68. Overall, the referral network between the tertiary hospital and PCFs was tight, especially the link from the tertiary hospital to PCFs. The tertiary hospital exhibited the highest out-degree, in-degree, and betweenness centrality in the network (P < 0.001) (Table 2), confirming its central role in coordinating chronic disease care. The out-degree of tertiary hospitals was significantly higher than the in-degree (P < 0.001), reflecting their role in transferring patient care responsibilities to lower-level institutions. In contrast, PCFs exhibited a greater in-degree than out-degree centrality (P < 0.001), highlighting their role as recipients of patients referred downward. This dynamic reflects the successful implementation of downward referrals, where tertiary hospitals alleviate care responsibilities by redistributing chronic disease patients to PCFs.

Fig. 4.

Density between different levels of the medical institutions

Table 2.

The average degree and betweenness centrality at different levels of medical institutions

		n	2017	n	2018	n	2019	n	2020	n	2021	n	2022	n	2023	Z	P trend
Out-degree
	Tertiary hospital***	1	8.9530	1	7.8105	1	9.1935	1	3.9194	1	15.6025	1	18.5905	1	20.0847	3.7957	< 0.001
	Secondary hospitals	1	0.0041	3	0.0488	3	0.0923	3	0.1162	3	1.7181	4	4.2007	4	4.1399	8.1591	< 0.001
	Primary care facilities	53	0.0700	55	0.0631	59	0.0902	57	0.0972	58	0.1133	54	0.0898	56	0.0827	2.4323	< 0.05
	Public PCFs	26	0.1238	28	0.1136	30	0.1486	30	0.1593	32	0.1864	31	0.1449	33	0.1403	-2.4906	< 0.05
	Private PCFs	27	0.0154	27	0.0109	29	0.0298	27	0.0282	26	0.0233	23	0.0131	23	0.0000	5.1768	< 0.001
In-degree
	Tertiary hospital***	1	2.9091	1	3.3669	1	4.4879	1	4.2042	1	5.1339	1	3.5805	1	3.3361	0.8577	0.391
	Secondary hospitals	1	0.4804	3	0.1192	3	0.2787	3	0.4458	3	0.4795	4	0.4526	4	0.3236	3.2243	< 0.01
	Primary care facilities	53	0.1851	55	0.1349	59	0.1605	57	0.0749	58	0.3579	54	0.5671	56	0.6544	3.9233	< 0.001
	Public PCFs	26	0.2345	28	0.1843	30	0.2147	30	0.0982	32	0.6061	31	0.8689	33	0.9945	3.1933	< 0.01
	Private PCFs	27	0.1348	27	0.0838	29	0.1044	27	0.0490	26	0.0524	23	0.1472	23	0.1664	5.1856	< 0.001
Betweenness
	Tertiary hospital***	1	0.4563	1	0.3061	1	0.3930	1	0.2765	1	0.2003	1	0.3586	1	0.2829	-2.6522	< 0.01
	Secondary hospitals	1	0.0122	3	0.0005	3	0.0146	3	0.0250	3	0.0227	4	0.0402	4	0.0202	4.1038	< 0.001
	Primary care facilities	53	0.0005	55	0.0002	59	0.0011	57	0.0015	58	0.0007	54	0.0010	56	0.0006	1.8341	0.066
	Public PCFs	26	0.0007	28	0.0002	30	0.0018	30	0.0019	32	0.0009	31	0.0013	33	0.0010	-2.6236	< 0.01
	Private PCFs	27	0.0003	27	0.0002	29	0.0004	27	0.0009	26	0.0004	23	0.0005	23	0.0000	0.1803	0.8569

*** Kruskal-Wallis test comparing centrality at different levels of hospitals. P < 0.001

Challenges facing secondary hospitals

The density of the referral network from secondary hospitals to PCFs ranged from 0.01 to 0.44 (Fig. 4), and the density of the referral network from PCFs to secondary hospitals ranged from 0.08 to 0.16. The referral network between PCFs and secondary hospitals was not dense. The density of the referral network between the tertiary hospital and secondary hospitals was 0, indicating that no referrals were made between tertiary and secondary hospitals. However, there were signs of improvement. The out-degree centrality (Z = 8.1591, P_trend < 0.001) and in-degree centrality (Z = 3.2243, P_trend < 0.01) of secondary hospitals showed significant upward trends (Table 2), indicating a growing role in the network. Additionally, the betweenness centrality of secondary hospitals increased significantly (Z = 4.1038, P_trend < 0.001), indicating their emerging role as a bridge. Nevertheless, secondary hospitals continue to be underutilised compared to tertiary hospitals.

Impact of the COVID-19 pandemic

Before the COVID-19 pandemic, the baseline trends for total referrals, downward referrals, the out-degree of secondary hospitals, and the in-degree of PCFs showed a significant rise (P < 0.05) (Table 3). After the pandemic, no significant changes in levels were observed (P > 0.05), indicating no immediate effect. However, total referrals, downward referrals, the out-degree of secondary hospitals, and the in-degree of PCFs exhibited accelerated trends over time, with the pandemic having a significant impact on the long-term trends of these indicators.

Table 3.

The interrupted time series analysis of referral patterns before and after the COVID-19 pandemic

Variable	Baseline Level	Baseline Trend	Post-pandemic Level Change	Post-pandemic Trend Change
Referrals
Upward	183.6074 ***	1.6666	67.1317	-1.9183
Downward	57.0935	13.5430 *	-567.6169	39.9901 *
Total	240.7009 **	15.2096 ***	-500.4853	38.0718 **
Tertiary hospital
Out-degree	6.2436 ***	0.1007	-6.7244	0.3841
In-degree	3.3012 ***	0.0152	0.8547	-0.0396
Betweenness	0.4013 ***	-0.0018	-0.0869	0.0031
Secondary hospitals
Out-degree	-0.6030 **	0.0313 ***	-1.0867	0.1015 **
In-degree	0.2726	-0.0011	0.2603	-0.0010
Betweenness	-0.0014	0.0005	0.0106	-0.0004
Primary care facilities
Out-degree	0.1198 ***	0.0003	0.0391	-0.0007
In-degree	0.0751	0.0060 *	-0.2894	0.0212 *
Betweenness	0.0003	0	0.0005	0

***P < 0.001, **P < 0.01, *P < 0.05

Comparison of public and private PCFs

In terms of the number of referrals, upward and downward referrals were significantly greater in public PCFs than in private PCFs (P < 0.001) (Table 4). Figure 4 shows the density of referral networks from public and private PCFs to higher-level providers. The density of the referral network from the tertiary hospital to public PCFs was stable, ranging from 0.93 to 0.97, while the density of the referral network to private PCFs spanned a wider range, from 0 to 1.00. The densities of referral networks from secondary hospitals to public and private PCFs ranged from 0.04 to 0.45 and from 0 to 0.42, respectively. The densities of referral networks from public PCFs to the tertiary hospital and secondary hospitals ranged from 0.87 to 0.91 and from 0.03 to 0.21, respectively. The densities of referral networks from private PCFs to the tertiary hospital and secondary hospitals ranged from 0 to 0.48 and from 0 to 0.22, respectively. This shows that the linkage between public PCFs and the tertiary hospital is stronger than that between private PCFs and the tertiary hospital. Neither public PCFs nor private PCFs are strongly connected to secondary hospitals. In terms of referral network centrality, public PCFs had significantly greater out-degree centrality, in-degree centrality, and betweenness centrality than private PCFs (P < 0.001).

Table 4.

Comparison of public PFCs and private PFCs

	Publica	Privatea	P
n	36	31	-
Upward referrals	272	35	P < 0.001
Downward referrals	904	155	P < 0.001
Total referrals	1176	183	P < 0.001
Outdegree	0.1478	0.0180	P < 0.001
Indegree	0.4936	0.1000	P < 0.001
Betweenness	0.0012	0.0004	P < 0.001

^a Mean

Discussions

This study reviewed the referrals of chronic disease patients in Luohu district, Shenzhen, China, over seven years. The characteristics of the referral network among different levels of medical institutions were analysed, highlighting differences in referrals between public and private PCFs. The results confirm the effective implementation of referral measures in Luohu district, especially in public PCFs within the medical consortium. This indicates that chronic disease management tasks are being shared at the grassroots level and that a hierarchical medical order is taking shape.

The status and influence of all levels of care in two-way referrals have increased following the implementation of the Luohu district’s internal referral programme. The tertiary hospital, in a leadership position, has a high level of influence in the two-way referral network and can largely control the referral network. This highlights the coordinating role of the leading hospital in integrating the healthcare system. This study observes the phenomenon of vertical patient flow in Luohu district. China differs from the United States and the United Kingdom, where primary care diagnosis is mandatory, and patients are required to visit a PCF before being referred to a higher-level hospital [24, 25]. PCFs in China generally operate in isolation from higher-level hospitals [26, 27]. The links between PCFs and higher-level hospitals have long been weak [28, 29], even after the introduction of the hierarchical medical policy [30]. The implementation of the hierarchical medical policy in Luohu district has effectively facilitated the upward and downward flow of patients. However, there is close cooperation between the upper and lower levels of care, mainly between tertiary hospitals and PCFs. There was weak communication between secondary hospitals and PCFs, and no referrals were made between secondary and tertiary hospitals. Francetic et al. [31] found that, due to geographic advantages, urban residents were more likely to choose tertiary care hospitals than rural residents in Tanzania. This study was conducted in Shenzhen, a large city, and the high accessibility of quality healthcare resources for city residents may have caused patients to bypass intermediate levels of hospitals and choose the highest-level hospitals when referring upwards. Additionally, favourable policies and the fact that patients in the stabilisation period do not need complex healthcare services may lead patients referred from tertiary hospitals to choose PCFs instead of secondary hospitals. This trend highlights the challenges faced by secondary hospitals. Tertiary hospitals continue to exhibit a siphoning effect, and PCFs have been strongly supported by Chinese policy, with the gap between PCFs and secondary hospitals gradually narrowing. Patients are treated in PCFs for mild diseases and visit tertiary hospitals for more severe diseases. In the first half of 2023, the number of visits to tertiary hospitals and PCFs in China increased, while the number of visits to secondary hospitals declined by 7.6% [32]. Due to a general preference for higher-tier hospitals, perceived to offer better quality of care [33, 34], and a lack of policy incentives favouring the choice of secondary hospitals [35], patients often bypass these facilities, leaving them neglected in the referral network. Similarly, in South Africa, the healthcare system is centred around primary care, where women with low-risk pregnancies are encouraged to deliver at PCFs, and women with high-risk pregnancies are referred to secondary or tertiary hospitals. However, many patients with low-risk pregnancies still choose tertiary hospitals [36]. Furthermore, a study has shown that approximately one-third of patients receiving tertiary surgical services in South Africa could have been treated at secondary care facilities [37]. In Pakistan, patients often bypass secondary hospitals due to the perceived superior quality of tertiary hospitals [38]. Such phenomena not only affect the rational allocation of medical resources but also highlight the weaknesses in the hierarchical diagnosis and treatment system in practice. These challenges reflect the common difficulties faced by hierarchical healthcare systems in many countries.

Our findings further reflect this issue, with the high betweenness centrality of tertiary hospitals suggesting their key role in the patient referral process. This centrality can affect the continuity of patient care and may strain resources in cases of poor management. This centrality plays a crucial role in coordinating two-way patient referrals, ensuring continuity of care and timely access to care, and contributing to improved health outcomes for patients. However, its efficiency largely determines the overall performance of the referral network. When overloaded, hospitals with a high degree of centrality may delay care, triggering an imbalance in resources and leading to a less efficient system. It is essential to enhance the capacity of secondary hospitals to ensure the stability of the referral network. Moreover, our analysis revealed that the COVID-19 pandemic did not adversely impact the referral network. This stability can be attributed to targeted policies and adaptations that maintained the continuity of care throughout the crisis. For example, at the early stages of the epidemic, non-COVID-19 patients in the study area were primarily admitted to regular hospitals, while COVID-19 patients were redirected to sentinel hospitals for specialised care. This separation ensured that routine healthcare services for patients with chronic diseases were largely unaffected. Meanwhile, the integrated healthcare system played a particularly crucial role during this period. For example, the sharing of medical testing centres enabled patients to have tests conducted at any PCF within the system, and the decentralisation of specialists to PCFs allowed patients to be seen by specialists from higher-level hospitals and to receive faster referrals when necessary. These policies ensured continuity of care and optimised referral efficiency. Additionally, the pandemic increased patients’ focus on reducing the risk of COVID-19 exposure [39]. Patients avoided high-traffic tertiary hospitals and increasingly sought care at lower-level providers, such as secondary hospitals and PCFs, to minimise the risk of COVID-19 exposure. This shift is reflected in the elevated role of these institutions within the referral network, which contributed to the strengthening of the hierarchical care system. These findings suggest that the role of secondary hospitals in the referral network should be further reinforced. As key bridge nodes in the referral network, secondary hospitals can not only effectively triage patients during routine care, thereby alleviating the pressure on tertiary hospitals, but also enhance the resilience and responsiveness of the entire network in responding to public health emergencies.

In July 2024, the Decision of the Central Committee of the Communist Party of China on Further Deepening Reform and Advancing Chinese-Style Modernisation emphasised the need to ‘accelerate the construction of a hierarchical diagnosis and treatment system’ [40], underscoring its significance in healthcare reform. However, the current underutilisation of secondary hospitals contradicts this goal, revealing a critical gap in the system. We propose strengthening policy guidance by implementing the hierarchical diagnosis and treatment system for major diseases. This includes clarifying the functions of medical institutions at all levels and ensuring strict adherence to their designated responsibilities. Financial incentives should be fully utilised. Diagnostic and treatment services within the defined scope should form the basis for health insurance settlement, receiving full reimbursement rates, while services outside this scope would be reimbursed at a reduced rate. Differentiated reimbursement policies could be introduced for patients using secondary hospitals. For example, higher reimbursement rates could be offered to those following the hierarchical diagnosis and treatment pathway. The government should also increase financial support to improve specialist facilities and enhance professional training in secondary hospitals. Secondary hospitals could specialise in rehabilitation or geriatric care, tailored to regional and population needs. Tertiary hospitals should enhance their support by dispatching specialists to secondary hospitals and providing advanced training opportunities for secondary hospital doctors. Specialist alliances should be formed to foster partnerships with specialised medical institutions outside the medical consortium, improving specialist capabilities and promoting resource sharing. Referral pathways can be optimised with information technology and artificial intelligence. These tools can analyse patient conditions and regional resource distribution, recommend referral organisations, and reduce biases against secondary hospitals.

Many developed countries have well-established general practitioner systems and reasonable downward transfer rates, with their studies focusing more on upward referrals [41, 42]. In China, downward referrals are far fewer than upward referrals, and it is more difficult to refer patients downward than upward [43, 44]. In this study, chronic disease patients were successfully transferred to lower-level institutions, effectively sharing the burden of care in higher-level hospitals. Several studies have found that healthcare costs influence patients’ willingness to select downward referrals [45]. More favourable healthcare policies in Luohu district for attending PCFs may have contributed to this phenomenon. In addition, the Luohu Hospital Group has established a two-way referral electronic information system applied to the Luohu district, which facilitates the integration of patients’ healthcare information and is important for advancing two-way referrals. The Luohu two-way referral system has made significant efforts in data security, interoperability, and communication and coordination.

Inadequate referrals are a barrier to effective two-way referrals. Thakkar et al. [46] indicated that weak provider-provider linkages and the lack of an organised downward referral system in the hypertension referral network in western Kenya were important determinants of the strength of the referral system. In an interview, patients within the medical consortium complained of frustration with repeated questions from doctors about their condition after referral. Many studies have promoted the use of information technology in referrals [47]. This is considered an effective way to promote the integration of medical resources and information sharing [43]. This is also reflected in the Luohu two-way referral system. However, it must be mentioned that the Luohu two-way referral electronic information system is only a regional referral information system. Doctors cannot access their past medical records if patients seek referrals beyond this region. England [48] and Australia [49] have established national healthcare data exchange platforms to facilitate data sharing among medical institutions. While these developed countries have better penetration of electronic health information systems, developing countries encounter numerous challenges in establishing national electronic health information systems [50, 51]. Rural areas in India struggle with inadequate infrastructure, including unreliable internet connectivity, insufficient power supply, and lack of necessary hardware. These problems hinder the effective deployment and use of electronic information systems in these regions [52]. The experience of some developing countries highlights the lack of information technology skills among users, political and administrative support for projects, interoperability frameworks, insufficient funding, logistical support, and system maintenance [53]. By the end of 2023, the national health information platform for all people in China will have been essentially completed. On this basis, the Chinese government has put forward new requirements, including the universal promotion of electronic health cards, the promotion of the interoperability and sharing of examination and test results, and the promotion of cross-province access to electronic health records [54]. Given the vastness of China’s landmass and the differences in infrastructure and economic bases between different regions, there are many challenges in promoting health information management on a nationwide scale. Adequate allocation of resources, effective communication, proper planning and project management, local support from stakeholders, implementation based on a holistic approach, and adequate technical support are all factors that can contribute to the successful implementation of an electronic health information system [55, 56]. To improve the interconnectivity of medical information and support the formation of a hierarchical diagnosis and treatment model, the government should provide policy and financial support, implement comprehensive planning, and establish unified health information standards, including data content, coding rules, and security specifications, to ensure seamless communication across regions and sectors. In addition, data security should be enhanced by adopting advanced encryption technologies and strict access controls to protect patient privacy. Interoperability should be promoted through the adoption of common data exchange standards to ensure compatibility across information systems. Communication and coordination should be strengthened, cross-level collaboration mechanisms established, and the efficiency of information sharing among medical institutions enhanced.

The number of referrals, centrality, and closeness of ties to higher-level hospitals were greater in public PCFs than in private PCFs within the medical consortium. The referral rate in public healthcare facilities in Indonesia is nearly twice that of private healthcare facilities [57], aligning with the study. First, the differences between public and private PCFs in this study may mainly be attributed to the structure of the medical consortia. Public PCFs are part of an integrated healthcare system, which enables the coordination of human, financial, and material resources [58, 59]. This structure not only promotes collaboration between different levels of care and ensures continuity of care for chronic patients, but also strengthens the ability of public PCFs to manage complex cases and identify patient referrals promptly and effectively. In contrast, private PCFs operate independently, lacking the institutional linkages and resource support provided by medical consortia, limiting their integration within the referral network. Second, differences exist in the financial models between public and private PCFs. Private PCFs often operate on a self-financing model, leading to a greater focus on financial performance. Physicians need to actively retain patients to maintain their revenue streams, which may limit their willingness to refer patients to higher levels of care [60–62]. However, public PCFs receive financial support from the government, with referrals serving as an important indicator of performance evaluation. As a result, public PCFs prioritise patient health outcomes. Third, the policy favours public PCFs in resource allocation. Basic public health services are primarily provided by public PCFs, which are government-led and funded through financial allocations to support essential programmes, such as health record management, chronic disease prevention and treatment, and immunisation planning [62, 63]. These resources and responsibilities reinforce the role of public PCFs within the hierarchical medical system. The Chinese government places high value on non-government participation in the provision of healthcare services [64]. We propose enhancing the integration of the healthcare system through policies that encourage private PCFs to join medical consortia. A financial subsidy mechanism could be established based on referral compliance and volume, providing referral incentives to private PCFs. Additionally, collaboration between public and private PCFs should be promoted, while strengthening the capacity of private PCF doctors in patient management and referral through joint training.

Limitations

Although this study provides a long-term, comprehensive assessment of referrals in Luohu district, it is limited to referrals at the district level. Considering the impact on hospitals of different levels and jurisdictions of geographic government units, the medical institutions we included were all under the jurisdiction of Luohu district. Some patients with chronic diseases may be referred to other regions in Shenzhen or Guangdong province, which were not assessed. Only one tertiary hospital can be included, even though it shows prominent centrality. This may result from the strong influence of tertiary hospitals in the district healthcare system. In addition, we did not categorise chronic diseases because this study focuses on the population to which continuing healthcare services apply. This may result in certain chronic diseases not being covered by this study.

Conclusions

Referrals between medical institutions in Luohu district of Shenzhen increased significantly after implementing the hierarchical medical policy, and the role of medical institutions at all levels in the referral network expanded. The referral network exhibits a vertical flow pattern led by the highest-level hospitals, closely linked to PCFs. This pattern facilitates the division of labour and resource integration among medical institutions, and a hierarchical medical order is taking shape. However, there are some problems. Poor communication between secondary hospitals and upper- and lower-level medical institutions, greater needs for health data informatisation, and a large gap between private and public PCFs may hinder the further advancement of the hierarchical medical system. The future should focus on increasing policy support and financial incentives while enhancing the professional capabilities of secondary hospitals and private PCFs through collaborative efforts. Furthermore, information technology should be utilised to optimise referral processes, further enhancing the tiered healthcare system.

Abbreviations

PCFs

Primary care facilities

NCDs

Noncommunicable diseases

APP

Application

COVID-19

Coronavirus disease 2019

ITS

Interrupted time series

Author contributions

Conceptualization: FG, PT, YZ and HL. Formal analysis: FG and HL. Data curation: FG, JL, ZW and YZ. Software: FG and YZ. Visualization: YZ. Validation: FG, YZ and GH. Supervision: HL. Writing – original draft: FG, JL and YZ. Writing – review & editing: FG, JL, PT, YZ and HL. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China, grant number 72374214.

Data availability

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The data used in this study is administrative management data from referrals between medical institutions. All information has been de-identified, and patient identities cannot be restored from the existing information. According to Article 32 of the ‘Ethical Review Measures for Life Science and Medical Research Involving Humans’ issued by the National Health Commission of China (https://www.gov.cn/zhengce/zhengceku/2023-02/28/content_5743658.htm), this study is exempt from ethical approval, accordance and consent to participate by the Shenzhen Luohu People’s Hospital Research Ethics Committee.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.