STRENGTHS AND LIMITATIONS OF THIS STUDY

We evaluated the impact of this reform using a difference-in-differences analysis with multiple periods.

The progressive inclusion of members of the Hubei Pediatric Medical Alliance (HPMA) provides us with a ‘natural experiment’ that allows us to compare ‘treated’ (joined the HPMA) and ‘untreated’ (not joined the HPMA) local hospitals within the same region of the country.

From a data accessibility perspective, as of 2019, the data for this study were primarily derived from research conducted by members of the HPMA and the paediatric quality control system, and the data were limited in scope.

Considering the actual policy implementation stage, most members joined the HPMA in 2016, and the corresponding support measures had not yet been completely implemented. The policy effect is challenging to present in the short run.

Introduction

Health inequality caused by the scarcity and uneven distribution of paediatric service resources is common worldwide.1 Hospital strategic alliance is one of the widely adopted responses,2 3 aiming to improve service efficiency and health equity by strengthening interhospital professional and service cooperation.4 5 For example, the Catalan Hospital Alliance in Spain and the Hospital Strategic Alliance in south-western Ontario of the USA have improved the accessibility and quality of medical services for children in remote areas through technical support.6–8 However, controversies exist regarding the practical and long-term effects of hospital strategic alliances. Research has suggested that despite the short-term service quality improvement coming along with the alignment, there is a tendency to further concentration of resources and power to the superior hospitals, thus strengthening the monopoly position of these hospitals in the long run.9

In China, high-quality paediatric services are unevenly distributed in tertiary or children’s hospitals in large cities.10 11 The paediatric medical services accessible to children in small towns or rural areas are minimal.12 Therefore, to improve the paediatric service supply in these areas, the Chinese government encourages the establishment of strategic alliances among hospitals to strengthen service technology dissemination and service cooperation. The Hubei Pediatric Medical Alliance (HPMA) is a provincial paediatric strategic alliance initiated by Wuhan Children’s Hospital. Since its establishment in 2016, the alliance members have gradually increased from 157 to 186 in 2022, covering all the prefecture-level administrative areas in Hubei Province. The alliance aims to enhance the paediatric service capabilities of local hospitals through measures including the assignment of medical experts, paediatrician training and optimisation of the referral process. However, it is unclear whether these measures have genuinely improved the paediatric service capabilities of local hospitals.

This study aims to evaluate whether the HPMA strategies impact the paediatric service capacity of local hospitals and explore its intrinsic mechanism to provide a reference for the practice of paediatric strategic alliances in other countries.

MethodResearch setting

Hubei Province is located in central China, with 8870 700 resident children in 2018.13 There were 2.4 paediatric beds and 0.4 paediatricians per 1000 children in Hubei Province.13 The number of paediatricians per 1000 children was lower than the average in China (0.61).13 14 Furthermore, the paediatric service resources in Hubei are highly unevenly distributed, as evidenced by the fact that the region covering 90% of the total geographical area of Hubei Province had only 50% of the number of paediatricians and beds.13 Data collected by the HPMA in 2023 show that Wuhan, the capital city of Hubei Province, had 0.96 paediatricians per 1000 children, while other areas of the province had only 0.54 paediatricians per 1000 children. This disparity indicates a significant concentration of paediatric resources in the capital city, exacerbating the health equity problem for children in remote areas of Hubei.

Cooperating strategy

HPMA strengthened interhospital cooperation mainly through long-term medical expert assistance, medical team on-site guidance, special continuing education programmes, and the creation of green referral channels. To improve the quality of paediatric services in local hospitals, HPMA sent medical experts to deliver long-term services in local hospitals. Paediatric medical teams are also organised regularly to provide on-site guidance in their wards. The clinicians of the local hospitals were sent to the HPMA’s leading hospitals for continuing education programmes based on the urgent health needs of the local children. Green referral channels were also built to facilitate the accurate and timely referral of local paediatric patients. However, due to various factors such as geographical location, initial service capacity and regional demand, the degree of participation of alliance members in these strategies varies among different regions. Table 1 shows the detailed elaboration of HPMA’s support measures for local hospitals.

Table 1

Support measures provided by the HPMA to local hospitals

Research assumptions

In this study, we focus on the impact of the strategies of HPMA on the paediatric service capacity of local hospitals. According to the resource dependence theory,15 organisations must interact with environmental factors they rely on to obtain the resources for survival and growth. Superior hospitals typically possess advanced medical service technologies and management expertise that are often lacking in local hospitals. Thus, we contend that the cooperation strategies can significantly influence how resources are shared and managed, thereby affecting the overall quality and availability of paediatric services provided by the local hospitals. Hospital strategic alliance can better meet the demand of local hospitals for resources such as the stable and adequate supply of patients with common and frequently occurring diseases, stable patients with chronic conditions and patients in recovery, as well as technology and knowledge,16 which can enhance the health service capacity of local hospitals.

We propose the following hypotheses based on the study above.

H1: The interhospital cooperation strategies of the HPMA improves the local hospitals’ paediatric service capacity.

H2: The effect of the interhospital cooperation strategies of the HPMA on the paediatric service capacity of local hospitals is stronger under a tightly integrated cooperation model compared with a loosely integrated model.

Research subjects and groups

For a more comprehensive and accurate assessment of policies and related outcomes, our inclusion criteria are: (1) Local hospitals that joined the HPMA between 2016 and 2019 and (2) Local hospitals that did not join a hospital strategic alliance and did not receive paediatric support between 2016 and 2019. Exclusion criteria: (1) Local hospitals that are members of the HPMA but also belong to other hospital strategic alliances and receive paediatric support; (2) Local hospitals that are not members of the HPMA and belong to other hospital strategic alliances, receiving paediatric support. At last, 119 local hospitals were included in the study, including 58 HPMA members and 61 hospitals without joining the HPMA. In 2016, 48 institutions joined the HPMA, with an additional 4 joining in 2017 and 6 more in 2018. We used these hospitals’ 2015–2019 panel data from the Hubei Pediatric Quality Control Center and the Hubei Provincial Statistical Yearbook.

The degree of participation in HPMA of local hospitals varies in practice. We counted the frequency of the significant collaborative initiatives of the 58 local hospitals participating in the HPMA, including the frequency of aid experts, special programme trainees, dual referrals and on-site guidance. We further performed hierarchical clustering of the implementation of core initiatives. When using the hierarchical clustering method, we did not preset the number of clusters, but imported the frequency of aid experts, special programme trainees, dual referrals and on-site guidance. Set the distance type to squared Euclidean and select the nearest neighbour as the clustering method. We found that two types of cooperation patterns can be divided. The hierarchical clustering analysis of local hospitals are provided in online supplemental figure 1. In addition, we plotted 3D scatter plots to visualise the differences in initiatives across models, as shown in figure 1. Thus, we named the group with more connections with HPMA as the close cooperation group (n=40); the frequency of the significant collaborative initiatives in this group is consistently high and significantly greater than that of the other group. Meanwhile, another group was named the loose cooperation group (n=18).

Figure 1

3D scatter plot of local hospitals incorporating cooperation models.

Variables

We used the policy dummy variable as the independent variable. If a hospital joined the strategic alliance, this variable takes on a value of 1. If a hospital did not join the strategic alliance, this variable takes on a value of 0. Since the cooperation activities of the strategic alliances take time to implement, we delayed the assignment of this variable by 1 year. For example, if a local hospital joined the HPMA in November 2016, its policy variable was 0 before 2016 and 1 after 2017, respectively. It should be noted that no hospital has left the HPMA since joining, thus there were no instances where the variable needed to be recorded to 0 again.

We adopt to the indicators of ‘outpatient visits’ and ‘inpatient visits’ used by Aksan et al in their study on changes in the capacity of public and private hospitals in Turkey17 and use ‘number of outpatient visits’ and ‘number of inpatient visits’ as our dependent variables to provide a visual representation of the capacity of care providers,12 which reflects to some extent the paediatric service capacity of local hospitals to develop or improve.

In addition to the above independent and dependent variables, we also set control variables, mainly referring to the control variables used by Alexander and Lewis in evaluating the operational performance of contract management in US public hospitals.18 These variables, representing hospital and environmental characteristics, have significantly impacted hospital service capacity and performance. Given the availability and consistency of the data, we have chosen the indicator ‘number of beds available’ as the primary hospital characteristic variable. The indicator ‘number of health personnel per thousand permanent resident population’ was selected as the primary environmental distinct variable.

See table 2 for details of all variable definitions.

Table 2

Study variables and definitions

Empirical strategyData processing

In this study, we used Stata/SE V.16.0 for data processing. The cross-section of the n=119 panel data, time T=5, is a short panel.

The overall impact of joining the HPMA

To estimate an average overall effect of joining the HPMA on the paediatric service capacity of local hospitals, we used the multistage difference-in-differences (DID) estimation method by contrasting the evolution of the number of outpatient and inpatient visits between those who joined and those who did not join the HPMA, as follows:

(1)

where：

p is for local hospitals, r stands for region and t for the period (year). Yprt is each of our outcome variables (number of outpatient or inpatient visits) in each local hospital p of each region r in year t; Treatp takes the value 1 if a local hospital joined the HPMA during the 2015–2018 period (ie, adopting local hospitals, belonging to the treatment group), and 0 otherwise (ie, non-adopting local hospitals, serving as the control group). Postt is a time dummy that equals 1 if the observation is within the experimental period (2015–2018) and 0 otherwise. The interaction term Treatp×Postt is the policy variable of interest. When Treatp×Postt=1, it indicates that the local hospital belongs to the treatment group and is in the experimental period. Conversely, Treatp×Postt=0 represents the control group, which consists of local hospitals that did not implement the policy during the specified experimental period. Yeart is year fixed effects; institutionsp are local hospitals selected effects; Xprt are covariates representing the characteristics of each region and local hospitals that vary over the year: health technicians per 1000 resident population, number of actual beds; and ∈prt is the random error term. β1 aims to measure the overall effect of joining a paediatric alliance on the paediatric service capacity of local hospitals.

There are two main concerns in our analysis. First, we developed four models to analyse further the impact of different cooperation models on the paediatric service capacity of local hospitals. Model 1 takes as the treatment group those observations that join the HPMA, that is, when the subject enters the HPMA, treat takes the value of 1, while, on the flip side, observations that do not enter the HPMA are included in the control group, treat takes the value of 0. Model 2 uses the treatment group of subjects with a loose-knit cooperation model, that is, treat takes 1 when the subject joins the HPMA and adopts a loose-knit cooperation model. Conversely, subjects who do not enter the HPMA are included in the control group, and treat takes the value of 0. Model 3 uses the treatment group of subjects with a close-knit cooperation model, that is, when the subject joins the HPMA and adopts a close-knit cooperation model, the treat takes the value of 1. On the other hand, subjects who do not join the HPMA are included in the control group, and the treat takes the value of 0. Model 4 also uses the treatment group of subjects with a close-knit cooperation model, with the treat taking the value of 1, and uses the control group of subjects with a loose-knit cooperation model, with the treat taking the value of 0.

The second primary concern is the parallel trends assumption. The central DID assumption implies that the pre-existing trends in the number of outpatient or inpatient visits in both groups of local hospitals should be parallel before joining the HPMA, conditional on the set of local hospitals’ characteristics that we control for. We performed an event study to test for this assumption by including leads and lags in our model. Furthermore, this model allowed us to explore the dynamic effect of paediatric alliance support on the paediatric service capacity of local hospitals over time and is represented as follows:

(2)

Paediatric Alliance Support for k periods is a dummy variable that equals 1 if the local hospitals in year t joined the HPMA in year k. These are dummy variables for the adopting local hospitals for each preimplementation period, up to 3 years; and for each postimplementation period, up to 4 years, leaving as base category the year when the hospital first joined the HPMA (year 0).

Patient and public involvement

Patients were not involved in this study.

ResultsDescriptive analysis

Each variable’s descriptive statistical analysis is provided in online supplemental table 1. In online supplemental table S1, we report the descriptive statistics for the four key variables—the number of outpatient and emergency visits (NOVS), the number of inpatient visits (NIVS), the number of available beds (NAB), the number of health personnel per thousand permanent resident population (HTPP) —used in our analysis. The means, maximums and minimums are presented according to the sample sizes included in each respective model. This approach ensures that the descriptive statistics directly correspond to the models discussed in the paper.

Parallel trend test and dynamic effect analysis

The test results are shown in table 3. None of the factors passed the significance test before implementing the mechanism, and they started to be significant in the year in which the tool was implemented or one year after the implementation. This indicates that there was no significant difference in outpatient visits and discharges between the experimental and control groups prior to local hospitals joining the HPMA. Therefore, the results of all four models satisfy the requirements of the parallel trend hypothesis, and the year-to-year coefficient variation after implementation reflects the policy effect’s time-dynamic effect.

Table 3

Dynamic parallel trend test

DIDs results

The baseline regression was conducted in two steps. In the first step, no control variables were added to the four models. The regression results are detailed in online supplemental table 2. Then in the second step, all control variables are added to the four models, and the regression results are described in table 4. The results of Models 1 and 3 showed that the cross-over Treat×Post coefficient was significantly positive at a 1% level with or without the control variable increase, except for the coefficient of the number of outpatients in the Model 2, which is significantly positive at the 5% level, suggesting that membership in the HPMA did increase the paediatric service capacity of local hospitals, validating hypothesis H1.

Table 4

Baseline regression results of the model based on different cooperation models

Effect of different cooperation models on the paediatric service capacity of local hospitals. Model 4 showed that after controlling for the influence of other variables, the cross-over Treat×Post coefficient was significantly positive at a 1% level. Meanwhile, the cross-over Treat×Post coefficient of the Model 3 is larger than the Model 2, which shows that the net benefits of the loose-knit cooperation model are more significant than those of the close-knit model. Furthermore, the cross-over Treat×Post coefficient of the number of inpatients in the Model (2) was insignificant, indicating that the loose-knit cooperation model did not significantly increase the number of inpatients. Based on Models 2, 3 and 4, we can conclude that under both indicators of the number of outpatients and the number of inpatients, local hospitals joining the HPMA have a significant improvement in paediatric service capacity than local hospitals not joining the HPMA. There are also differences in the advancement of paediatric services under different models. The close-knit cooperation model significantly improves the efficiency of the paediatric service capacity of local hospitals, thereby validating hypothesis H2 is tested.

Sensitivity analysis

To address potential endogeneity issues and ensure the robustness of our findings, we performed a sensitivity analysis. The initial test for parallel trends provided a basis for ruling out endogeneity to some extent and validating our approach. However, to further rule out the influence of other unknown factors on facility selection and to confirm that the observed effects are sorely due to membership in the HPMA, we conducted a placebo test using a multiple-period DID design.

Specifically, we randomly generated a list of healthcare organisations that would hypothetically join the HPMA or a similar cooperative arrangement. We then estimated coefficients for these randomly selected (hypothetical) groups, expecting them to be spurious (ie, not significant). This process was repeated 500 times, and the distribution of the 500 estimated coefficients was plotted. As shown in figure 2, a kernel density plot of the estimated coefficients reveals that they are centred around a value of zero and follow a normal distribution. The accurately estimated coefficients fall outside the main distribution, indicating that they are outliers. This finding is consistent with the expected outcome of a placebo test. In other words, the positive effect of hospital strategic alliances on paediatric service capacity is not attributable to random factors, and the results of our benchmark regression are robust.

Figure 2

Placebo test for Models 1–4 virtual disposal experimental group. (a) The kernel density plot for Model 1 with NOVS as the explanatory variable. (b) The kernel density plot for Model 1 with NIVS as the explanatory variable. (c) The kernel density plot for Model 2 with NOVS as the explanatory variable. (d) The kernel density plot for Model 2 with NIVS as the explanatory variable. (e) The kernel density plot for Model 3 with NOVS as the explanatory variable. (f) The kernel density plot for Model 3 with NIVS as the explanatory variable. (g) The kernel density plot for Model 4 with NOVS as the explanatory variable. (h) The kernel density plot for Model 4 with NIVS as the explanatory variable. NIVS, number of inpatient visits; NOVS, number of outpatient and emergency visits.

Discussion

This analysis provides preliminary results about the impact of hospital strategic alliances on the paediatric service capacity of local hospitals. We found that HPMA can significantly improve the health service capacity of local hospitals, and this conclusion still holds by parallel trend test with placebo test. In our study, the number of outpatients and inpatients in local hospitals increased significantly after joining the HPMA. Moreover, this effect is positively associated with the intensity of the cooperation measures.

These significant results may be related to some of the intrinsic characteristics in this case. First, it is the highly complementary cooperative relationships that we have evaluated in this study. Some superior hospitals have joined the HPMA and play the role of technology providers rather than receivers. However, complementary skills and resources organisations that are complementary as opposed to competitive are one of the essential prerequisites for the effective operation of strategic alliances, just as Lewis has suggested. Therefore, we only evaluated the impact of HPMA on the paediatric service capacity of local hospitals.2 19 Second, the superior hospitals’ commitment to support plays a vital role in this case.19 The alliance claims to unite the foremost hospitals to support the local hospitals and puts these promises into practice. These substantial assistants seem to be making short-term sacrifices for long-term benefits. Third, serving a particular group of children is more likely to stimulate resonance and trust among alliance members. To solve the dilemma of paediatric development in the province, the alliance has adopted the core concept of ‘solidarity for warmth’ and the purpose of ‘equality, openness, cooperation and win-win’ for value integration. It can reduce tensions between primary care, secondary care and tertiary care providers and promote the formation of common values pursued by member institutions.20 Several authors have also reported that a shared sense of values, purpose and trust among network members is required to implement clinical networks successfully.20–22

Superior hospitals are also motivated to perform these substantial assistance activities if the macro policy is considered. The Chinese government uses a series of medical insurance payments and government tools to guide superior hospitals to focus on diagnosing and treating complex and complicated diseases rather than common ones. Medical insurance has unique policies for superior hospitals in payment to cover more difficult medical activities. At the same time, the national performance evaluation of public tertiary hospitals also includes the difficulty of diagnosis and treatment activities in the review. To shunt general patients and receive more stubborn patients, superior hospitals must establish close cooperative relationships with local hospitals. In this context, strategic alliances have become another competition between foremost hospitals. Distinguished hospitals can enhance their core competitiveness and radiation influence in the cooperation, which makes them more potent in competition with other top paediatric hospitals in the region. As Garrick argues in his study, hospitals must ‘get beyond their walls to win the market competition’. Hospitals must develop alliances with smaller community and rural hospitals and providers to ensure a flow of patients through their doors and, simultaneously, respect the individuality of their new partners.16

Thus, it is not difficult to understand why closer cooperative relationships can bring more service growth in local hospitals.23–25 The alliance provides additional clinical and management support for close partners. On one hand, the coalition offers more targeted subspecialised training programmes for intimate partners. These programmes include standardised training, written exams and certifications. On the other hand, the alliance dispatches long-term stationed management teams and one-on-one clinical experts to provide support based on the actual needs of close partners. This improves local hospitals’ diagnostic and treatment capabilities and brings them the latest management ideas and techniques.26–28

Strengths and limitations

The significant advantage of our study is that the progressive inclusion of members of the HPMA provides us with a ‘natural experiment’ that enables a comparison between ‘treated’ (joined the HPMA) and ‘untreated’ (not joined the HPMA) local hospitals within the same region of the country. Furthermore, we examined the impact of different cooperation intensities on the service capacity of local hospitals and we analysed the potential reasons for these differences, including policy, resource relationships and values.

In terms of the limitations of our study, first, from a data accessibility perspective, as of 2019, the data for this study were primarily derived from research conducted by members of the HPMA and the paediatric quality control system, and limiting the scope of our analysis in scope. Second, considering the actual policy implementation stage, most members joined the HPMA in 2016, and the corresponding support measures had not yet been fully implemented. The policy effect is challenging to demonstrate in the short run. Another potential limitation is the possibility of anticipation effects, where hospitals may begin to change their practices in anticipation of joining the HPMA. Such effects could violate the parallel trends assumption underlying our analysis. To address this limitation, future research could incorporate an event study design to examine any deviations in outcomes prior to the official implementation of the HPMA. This would help to determine whether hospitals began to change their behaviour in anticipation of joining the HPMA and provide a more robust assessment of the policy’s impact.

Conclusion

The HPMA enhanced the paediatric service capacity of local hospitals. Furthermore, the HPMA’s close cooperation model is a more effective solution for addressing health inequality caused by the scarcity and uneven distribution of paediatric service resources. These results might be related to particular policy environments, such as hospital performance assessment and medical insurance payment reform.

Data availability statement

Data are available upon reasonable request. The data that support the findings of this study were collected by HPMA and are not publicly available due to restrictions set by the data provider. Requests to access these data sets should be directed to the corresponding author. The data may be made available on a case-by-case basis, subject to approval from HPMA and any necessary confidentiality agreements.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.