Integrating interventions supported by development assistance for health into local health system: evidence from a China–World Bank–UK rural health system strengthening project (1998–2007)

Study design

We used the case study approach that allows in-depth exploration of complex systems.45 From December 2021 to May 2022, we reviewed 64 published papers and policy or project documents, and conducted semistructured interviews with 22 key informants. We used a policy triangle framework built on Walt and Gilson’s model for health policy analysis46 and the WHO’s health system building blocks47 for data collection and initial coding. From February to March 2023, the final analysis was conducted under the analytical framework developed by Atun et al on integrating targeted health interventions into health systems.36

Data collection

This study has adopted a policy triangle framework, which was originally developed for a multicountry research project titled ‘Sustaining Effective Coverage in the Context of Transition from External Assistance—Lessons from Countries’ (The project was sponsored by the Alliance for Health Policy and Systems Research, the WHO, the WHO Department of Health Systems Governance and Financing and UHC 2030).48 The framework is consistent with Walt and Gilson’s health policy analysis model,46 emphasising four essential components: context, content, actor and process. The WHO’s health system building blocks serve as the foundation for the ‘content’ component.47 Specifically, the framework considers the complex interplay of various elements, including the political, economic, institutional and sociocultural context, the content areas of governance, financing, inputs, and service delivery, the array of actors such as the government, development partners, and civil society, as well as the policy-making and implementation processes before, during and after the DAH interventions. Collectively, this approach aims to streamline the collection of comprehensive data pertinent to the DAH initiative. Based on this framework, a data collection tool was customised to fit the unique circumstances of China.

To better understand the BHSP and its context, we searched and selected 64 relevant published papers and project or policy documents (see online supplemental material 1 for the scoping review process and online supplemental material 2 for these documents). The documents include credit agreements, project appraisal documents, progress and completion reports, slides, videos and national policies or strategies on rural health. Most project documents can be downloaded from the World Bank website.49 We also obtained completion reports accomplished by subnational governments and slides through personal correspondence. We then identified additional project documents by reviewing the reference lists of related published literature and secured them through online searches. We further identified the relevant policy documents from the already-obtained project documents, related literature and key informants throughout the data collection process.

Drawing on a preliminary literature and document review, we purposively selected key informants who have been engaged in various stages of the project—namely design, implementation and evaluation—for a minimum of 4 years. These individuals were chosen from diverse backgrounds, including donor agencies, governmental bodies and academic institutions, ensuring a multifaceted perspective. We then encouraged them to nominate additional key informants, thus employing a snowball sampling technique. The outreach efforts were meticulously crafted to achieve an equilibrium between national and subnational stakeholders, predominantly involved in the project implementation, and national and international stakeholders, who primarily contributed to high-level decision-making. Finally, out of the 37 contacted key informants, 22 participated in the interviews—4 were national government stakeholders, 11 were subnational government stakeholders (covering 4 provinces, 1 prefecture, 2 counties and 2 townships), 4 were donor stakeholders, 1 was a stakeholder from an external development partner and 2 were the country’s technical experts involved in the project (see online supplemental material 3 for respondents’ characteristics).

Before each interview, we designed a separate set of guides based on the policy triangle framework and the document and literature review. We then iteratively revised prompts based on previous interviews and background research of the respondents (see online supplemental material 4 for an example of the interview guide). The interviews lasted 1.5–2.5 hours; they were recorded with the respondents’ consent and transcribed promptly in their original languages. Three interviews were conducted in English, and the rest were in Chinese. An additional team member validated the transcripts to ensure their accuracy. We discussed the key findings after each interview.

Data analysis

To keep the consistency of the abovementioned multicountry research project, we used a framework approach to code the documents and interview transcripts50 51 from June 2022 to December 2022, according to the policy triangle framework. Two coders familiarised themselves with the data and conducted a pilot coding for five purposively selected respondents’ transcripts to cover different perspectives. The pilot coding used a deductive approach to test out the major themes in the policy triangle framework and develop a coding scheme for formal coding. In the formal coding, recurrent themes and subthemes were identified by inductively examining the data to develop more specific concepts. We triangulated the results from different data sources to identify patterns and linkages between themes and concepts and explore relationships and explanations. MAXQDA 2022 (Release V.22.3.0) was used for coding.

The research question of this study was framed iteratively throughout the data coding process. We then used the analytical framework for the integration of targeted health interventions into health systems, as developed by Atun et al36 (see table 2). This framework, previously employed in the extant literature on intervention integration within global health,52 encompasses various components that may impact integration: the problem, the broad context, the intervention, health system characteristics and the adoption system. These components interact in a dynamic, non-linear way. As this analytical framework is compatible with the policy triangle framework, it proved instrumental in analysing the interactions and interconnections of the data collected surrounding the research question.

Apart from the data supposedly to be coded based on the policy triangle framework, additional valuable data needed for this analytical framework were further identified through a re-examination of interview transcripts and related documents. These included nuanced details regarding interventions. Finally, framed by this analytical framework, the previously coded themes and subthemes were systematically reorganised, and new themes were identified from the additional valuable data.

Favourable context and strategic timing for addressing systemic health issues

The BHSP was set in a favourable context (figure 1), where addressing systemic issues in China’s health sector was crucial due to the country’s transition to a market economy. As the government adopted a pragmatic, experimental approach in its policy formulation, a shift in political priorities towards the health sector and growing economic capacities allowed feasible experiments that could align with the country’s health system needs and be integrated on a larger scale.33 53

Request permission

BHSP is one of the major pilot schemes.54 During the project’s inception (1994–1998), the State Council issued a decision paper with general principles for public health reform in 1997.55 These principles guided the project design,30 while the project was perceived as part of the government’s approach to exploring a practical way forward.32 Throughout the project years (1998–2007), the country’s central leadership transition resulted in greater political commitment towards improving people’s livelihood and well-being over economic growth. The government launched NCMS and MA in 2003, which had been piloted by the project CMS and MFA. Coincidentally, the SARS outbreak put strengthening the public health system on the political agenda.56 Towards the end of the BHSP project, China was already considered an emerging and resilient economy, particularly evident during the 2008 financial crisis. In 2009, the Chinese government launched a healthcare system reform under this promising economic climate. The primary objective of this reform was to enhance basic health services in the country, which resonated with the BHSP.33 The reform also integrated some project interventions as routine health system functions. Respondents emphasised the importance of the broad context, with one of them noting that the BHSP integration could have been challenging without this context:

One of the reasons that the BHSP worked is that the timing was right… If someone had tried to do that project three or four years earlier, it would not have worked. And if someone tried to do it in other countries, I’m not sure it worked… And only as the overall economic and political system was able to learn and adapt could this project take place… It’s what we call in donor parlance, particularly in World Bank terms, a “Christmas tree project”—it means that it has all kinds of bells, whistles, and trinkets on everything. Normally those projects are a disaster.—Donor R4

The intervention characteristics and the corresponding integration outcomes

The BHSP’s interventions were intricate and interrelated. The respondents drew connections between the intervention characteristics and various outcomes of interventions’ integration.

The ideas and values instilled by the project required a supportive context and system for long-term learning and transformative changes in the health system. Respondents reported that principles including ‘promoting government accountability’ (national government R1 and R2; subnational government R11), balancing hard and soft infrastructure investments, and prioritising demand-side versus supply-side investments were endorsed by the 1997 public health reform’s policy text.55 Project practices and the 2009 healthcare reform integrated these ideas and values across the project areas and the country. However, integrating the concept of ‘county health resource planning’ was limited in the project areas, resulting from this concept’s incompatibility with China’s top-down institutional structure of policy planning that extends beyond the health sector:

For county health resource planning, people felt that it was relatively vague or unclear—it was effective only in project areas but could not be sustained in the long run… Because pushing for reforms is not easy, and it is closely related to our institutional system.—National government R1

CMS and MFA prioritised institutional building, which required beneficiary mobilisation and cross-sector collaboration. Respondents indicated that their integration entailed significant adaptation. Unlike other project priority interventions covering all project areas and populations, CMS and MFA were piloted smaller before the NCMS and MA were launched in 2003. Respondents, project documents57 58 and published literature32 revealed that before 2003, CMS was initially tested in selected townships of project counties and was once considered a failure. MFA targeted either 20% or 5% of the population in poverty (The project counties were divided into two parts considering different extent of poverty. MFA targeted 5% families in part A counties and 20% in part B counties. UK DFID’s grant supported 8%–10% families in selected pilot counties among part A counties. These criteria were adapted as the project MFA was transformed to national MA since 2004 (see Page 15 in Liu et al58)), but these targets were not reached at the provincial level.57 58 It was only after 2003 that CMS and MFA were transformed into NCMS and MA, respectively, with the project counties serving as pioneers of national pilot schemes of NCMS and MA.33 57 58 NCMS avoided most of the errors of its predecessor, CMS, by ensuring significant government investment in the institutionalisation of insurance management and achieving more extensive and effective coverage (see online supplemental material 5 for NCMS coverage in the project and non-project counties). Additionally, project elements surrounding MFA were adapted to the national social security system primarily managed by the Ministry of Civil Affairs (MoCA).33 58 These adaptations enabled the project to leverage its experiences and resources, making them valuable for national implementation.

With their effects perceived in routine health system activities, interventions like upgrading health facilities (A2), improving health service delivery (B1) and priority health interventions (B2) were viewed as ‘shadows of BHSP’ (donor R2 and subnational government R11) in the 2009 healthcare system reform. According to subnational respondents, even between 2007 and 2009, some project provinces continued to support priority health interventions, and most project areas adhered to guidelines and protocols developed by the BHSP in their routine health system activities. Similarly, the tangible outcome of health facilities upgrading by the project was sustained in the project areas following the project’s completion, making the township health centres serve as major service delivery organisations for the rural population.

Project embeddedness: adapting to health system characteristics

According to project documents and respondents, the BHSP’s service delivery and management systems, including those for finance, procurement and information, shared human resources and infrastructure with China’s existing health system (see online supplemental material 6 for the governance structure), aligning with the government’s strong administrative capacity. This embeddedness largely allowed for the smooth integration of project interventions into routine health system functions with few disruptions.

The project management offices (PMOs) were part of the government. One arrangement was that the existing office performing routine health system functions served as the PMO. Project-designated staff received the same salary as the staff from any other office. This ‘one office, two functions’ approach (subnational government R2 and R9) allowed for combining project activities with routine operations and gave the previous PMO managers and staff significant policy influence after the project’s completion. Subnational respondents reported the success of this approach, with most subnational managers we interviewed obtaining high-level positions at health bureaux in the postproject period. This leadership continuity facilitated the further integration of project legacies into the local health systems of the project areas. In two of the four project provinces covered by the interviews, the health bureaux’s Department of Planning and Finance served as the PMO, enabling the BHSP’s experiences to shape further provincial decision-making and practices. Health bureaux doubled as PMOs at the county level, promoting mutual progress between the project and routine operations.

In another arrangement, the PMO operated as a standalone entity. The national PMO, the Ministry of Health Foreign Loan Office, possessed significant expertise in managing projects supported by external development partners, especially since the World Bank began supporting China’s health sector in 1983.58 However, respondents reported that few of its managers or staff continued to engage in rural health reforms. Two other provincial PMOs studied also adopted similar standalone arrangements. Some key PMO managers of these two provinces were not involved in policy-making in postproject interventions, probably hindering project interventions’ integration:

Our province’s PMO is standalone—it did not share office space with other departments. We established this PMO and assigned some people to work specifically on the project, so sustaining project interventions was difficult. Currently, all the personnel assigned have been dispersed to various organisations or departments, and there is no one left in the provincial health bureau now.—Subnational government R10

In addition, the absence of the prefectural government was a gap in the alignment of the health system, limiting the potential spread of project experiences at the prefectural level and in non-project counties. According to subnational respondents, project finance was adjusted for effectiveness during implementation, transferring funding directly from project provinces to counties and thus bypassing the prefectures. In addition, as project activities diverted resources for the routine functions of prefectural governments, their involvement gradually faded. Furthermore, although the spread of project experience to non-project counties was part of project activities, it was merely marginal.57 59 Thus, the BHSP had rather indirect policy influence related to the 2009 national healthcare system reform at prefectural levels and in non-project counties.

Another important aspect of project embeddedness was the cofinancing arrangement between the government and donors, which positively and negatively impacted project implementation and intervention integration. On the negative side, the arrangement financially burdened subnational governments. They were required to provide all the counterpart funds.32 60 However, these subnational governments could only meet 84.9% of the initial estimate due to their impoverished status.57 As the resource-constrained subnational governments preferred investing in health facility upgrading, investment in ‘soft infrastructure’ (donor R1 and R2; national government R2; subnational government R2, R3, R8 and R9)—capacity and institutional building—was compromised,57 61 together with the compromised integration of these interventions. On the positive side, project managers dedicated considerable effort to advocacy for project interventions with government leaders to gain political support and sufficient counterpart funds. This advocacy helped strengthen local ownership and commitment conducive to integrating project interventions. Subnational respondents and a project report57 attested to this positive impact.

Through providing counterpart funds, we reported [the project progress] to the [government] leaders, and they spontaneously knew that someone was carrying out the project interventions and thus gave us money. They had to pay attention to these interventions.—Subnational government R4

Adoption system: role of project leading groups and technical support

As part of the adoption system, the project leading groups (see online supplemental material 6) at the national, provincial and county levels were critical in strengthening political commitment and cross-sectoral collaboration. According to respondents and project documents,30 57 the leading groups mobilised related ministries and departments (including those for health, finance, civil affairs, etc), raising the project’s visibility among the sectors involved. At the subnational levels, the leading groups were chaired by executive leaders of the subnational governments,57 58 who were responsible for supervising the project implementation at the subnational level. Especially in the semiannual project supervisions, these subnational government leaders worked with project managers and experts at the subnational, national and international levels for project problem-solving.59 Respondents reported that the project leading groups and the according supervisions elevated the subnational government’s attention to the health sector and the project interventions’ sustainability:

The technical support system (see online supplemental material 6) played a vital role in these supervisions as well as the adaption of international norms and practices through evidence-based research and consultancy.32 57 ‘Mutual learning’ characterised the relationship between international and local experts.32 59 Through the interplay of international and local knowledge and frequent fieldwork in the project, international and local experts jointly helped to adjust the project activities to suit the ever-changing local needs. They also developed innovative approaches that aligned with China’s context.32 62 Under this technical support system, donors insisted on a need-based approach for systemic improvement in the negotiation, ensuring local adaptation and recipient leadership in the project.

Respondents highlighted the example of MFA adaptation. Given the subnational governments’ preference for facility upgrading over institutional building, MFA preparedness became a precondition for receiving credits for facility upgrading, ensuring MFA implementation. A national expert hammered out this ‘setting precondition’ approach, while the international experts primarily introduced the concept of medical assistance.44 Moreover, when the national MA was in its initial stage around 2003, there was a major change in MoCA personnel. To introduce project lessons and models to the new MoCA personnel, local project experts and DFID experts jointly worked to design and evaluate MoCA’s MA pilots in BHSP project counties.32 This technical support system penetrated at every level and highlighted the empowerment of local experts, ensuring contextualised adaptation and facilitating the integration of project interventions into the local health system.

It Is important to involve local experts to achieve sustainable development. After our project ended and international experts left, knowledge, theories, and concepts were maintained in the country. These were applied into practices and promoted further by local experts.—Donor R1