Original research

Using a policy learning lens to understand health financing policy outcomes: the case of translating strategic health purchasing into policy and practice in Burkina Faso

Abstract

Introduction Achieving universal health coverage (UHC) through an effective health financing system is a challenge for many low-income countries. Learning is key to success due to many uncertainties and unknowns. Using the case of translating strategic health purchasing into policy and practice in Burkina Faso, our study seeks to understand how policy learning can shape policy processes and outcomes.

Methods We used a qualitative case study design and Dunlop and Radaelli’s conceptualisation of policy learning to identify which modes of learning did or did not occur, what helped or hindered them and the resulting policy outcomes. Dunlop and Radaelli frame policy learning as epistemic, reflexive, negotiative or hierarchical. We collected data through documentary review and in-depth individual interviews with 21 key informants. We analysed the data manually using pattern-matching techniques.

Results The introduction of strategic health purchasing in Burkina Faso was initially seen as an opportunity to reduce the fragmentation of the health financing system by coupling a performance-based financing scheme and a user fee exemption policy. However, this has faltered, and our findings suggest that an inability to harness all modes of learning has led to blockages. Indeed, while reflective learning was present, epistemic, hierarchical and learning through bargaining were absent, preventing national policy actors from defending their own policy or scheme from reaching compromises. But thanks to facilitating processes led by a well-resourced organisation and contextual elements that encouraged the emergence of more pluralistic modes of learning, some progress was achieved in operationalising strategic health purchasing.

Conclusions Some modes of learning seem to be overlooked in countries’ efforts to achieve UHC. Facilitation techniques and initiatives that encourage the use of all modes of learning, while supporting countries to take full ownership and responsibility for consolidating their own learning health systems, should be promoted.

What is already known on this topic

  • Learning is increasingly recognised as an essential element in strengthening health systems and ensuring the effectiveness of health policies.

  • However, the concept of learning remains elusive and difficult to operationalise. There are still several dimensions of learning that require further research.

What this study adds

  • Given the paucity of actionable learning frameworks in health, we used a policy learning analytical framework from political science to assess its empirical relevance in health.

  • We apply this analytical framework to the concept of strategic health purchasing. Translating this concept into policy and practice is challenging. Different modes of learning need to be used.

How this study might affect research, practice or policy

  • Our study confirms that learning matters for successful health financing policy processes.

  • The appropriate harnessing of different learning modes, either concurrently or sequentially, is particularly important in this regard.

Introduction

Universal health coverage (UHC) is high on the political agenda of many low- and middle-income countries (LMICs), and developing an effective health financing system is particularly important for advancing this agenda.1–3 Health purchasing is one of the three main functions of any health financing system, along with revenue collection and pooling.4 The purchasing function consists of ‘the transfer of pooled resources to service providers on behalf of the population for which the funds were pooled’4 (p. 180). How this function is enacted determines the extent to which health financing is efficient, equitable and supportive of UHC goals.2 5 In this regard, there is a consensus among experts that health purchasing should be strategic rather than passive.6 7 Passive purchasing, for example, involves the automatic allocation of funds to providers on a historical basis, with no supporting documents or prior assurance that appropriate and quality services have been provided.8 In contrast, strategic health purchasing involves, at a minimum, monitoring providers and linking their payment to their performance and/or the health needs of the population they serve.7 However, the classification of purchasing into passive and strategic is more for heuristic purposes, as the reality is far from such a binary approach—it is rather a continuum between these two extremes.6

While strategic health purchasing is now well conceptualised7 9 10 and is more and more researched,11–13 it is still not yet always interpreted in the same way.14 For instance, strategic health purchasing has sometimes been confused with14 or assumed to have many similarities with specific schemes such as results-based financing.15–17 Others argue that it should not be seen as a specific policy or scheme, but rather as a set of policy tools that policy/decision-makers can use effectively to improve the impact of UHC-related policies—the so-called ‘functional approach to strategic health purchasing’.18 This functional approach focuses on strategic thinking about what services to buy, for whom, from whom to buy them, how to pay providers and at what rate and frequency, what reliable information system to put in place that would accurately report on provider performance, what governance mechanisms to put in place to oversee and regulate the whole system.

These different approaches to strategic health purchasing indicate that there are still open questions about how to operationalise it in a specific setting. There is no single blueprint, and each country must find its own way forward. Indeed, achieving such operationalisation is complex, for at least three reasons. First, the complexity lies in the way strategic health purchasing is framed, communicated, understood and applied in practice by policy actors—it is a relatively new concept in LMICs and not a very intuitive one. Second, in most countries, strategic health purchasing could suggest a broad, substantial and challenging agenda for health system reform.13 Third, strategic health purchasing is not a one-size-fits-all solution that can be implemented in a standard way across countries. These elements suggest that policy learning is a crucial factor for countries aiming to advance towards strategic health purchasing and, consequently, towards UHC. This perspective has been previously articulated by numerous scholars in the field.19–22

Policy learning has been approached and conceptualised in different ways by political scientists,23–26 and it is difficult to find common ground on how to make it actionable. This challenge also results in a dearth and fragmentation of empirical studies on policy learning.27 Dunlop and Radaelli define policy learning as ‘the updating of beliefs based on lived or witnessed experiences, analysis or social interaction’ (p. 599).28 Policy learning gained prominence when Heclo drew attention to the importance of knowledge and ideas in politics and policy processes: ‘Politics finds its sources not only in power but also in uncertainty—men collectively wondering what to do. […]. Governments not only ‘power’ (or whatever the verb form of that approach might be); they also puzzle. Policy-making is a form of collective puzzlement on society’s behalf; it entails both deciding and knowing’ (p. 305).29 Several conceptions, theories and analytical frameworks related to learning have subsequently emerged, such as Heclo’s ‘political learning’,29 Etheredge’s ‘governmental learning’,30 Sabatier’s ‘policy-oriented learning’,31 Rose’s ‘lesson-building’32 and Hall’s ‘social learning’.33 A bibliometric analysis of learning by Goyal and Howlett in 2018 confirms this fragmentation, showing that political scientists work in silos and do not learn enough from each other’s work,23 which hinders progress in operationalising the concept.

Learning is often valued for its result, that is, as an outcome, but it is also a process. For example, Bennett and Howlett25 suggest looking at learning through three key elements: Who is learning? What is being learnt? To what effect(s)? Regarding learners, Moyson et al34 found that they can be at a microlevel, involving individuals or groups/teams; at a mesolevel, involving organisations; or at a macrolevel, pertaining to learning across organisations or government units, for example, through policy transfer,35 policy diffusion36–38 or lesson-drawing.32 As for learning contents, Vagionaki and Trein39 assert that policy actors can learn about new information and ideas, or about specific policy instruments (eg, laws and regulations, market-based incentives, taxes, subsidies, information control). The most common outcomes expected from learning are positive changes, but these are far from being automatic and clear-cut. In fact, several other factors also influence policy change, such as political economy or path dependency.40 Moreover, learning itself and the changes it is intended to bring about are difficult to materialise, capture and measure.24 34 In addition, learning can be inefficient or go wrong, if some scope conditions are not met,41 42 or if, for example, the wrong teachers are listened to or the wrong lessons are implemented.41 43 44

Despite the extensive literature on ‘evidence to policy’,45 and the growing interest of global health actors and researchers in learning in the health sector,19–22 46–49 some aspects of ‘learning health systems’ and ‘learning for UHC’ in low-income countries remain under-researched. For example, while it is recognised that learning is an important component of the policy process,21 43 50 few studies in the health sector have taken a policy learning perspective to investigate it empirically and to examine and infer the modes of learning that shape or result from policy-making. Dunlop and Radaelli have shown that policy learning can be either epistemic, reflexive, bargaining-oriented or hierarchical,28 and that these modes of learning reflect characteristics of the policy process.43 In a previous paper, we have already shown that such a conceptualisation of learning is relevant and potentially applicable in the health sector, particularly in UHC processes.22

This paper aims to go further and illustrate policy learning empirically by examining the introduction of strategic health purchasing in Burkina Faso and efforts to translate it into policy and practice. Specifically, our work seeks to provide insights into how learning, as an important component of the policy process, can shape it and contribute to understanding some of the challenges associated with building a coherent health financing system in a resource-constrained country like Burkina Faso—a country that has undertaken major health financing reforms in recent years.51–53 To achieve these objectives, our research questions (RQ) are as follows:

RQ 1. What modes of learning have or have not been used in the process aiming at translating strategic health purchasing into policy and practice?

RQ 2. What helped or hindered learning to occur?

RQ 3. How have these learning or missed learning opportunities affected the outcomes of the processes of translating strategic health purchasing into policy and practice?

Study context

To improve access to and quality of healthcare for its population, the government of Burkina Faso has introduced several health financing policies. These included (1) the national policy of subsidising 80% of the direct medical costs of emergency obstetric and neonatal care,54 55 known as ‘SONU’, which was implemented between 2006 and 2015 and was fully financed by the national budget; (2) the pilot phase of a performance-based financing (PBF) scheme between 2011 and 201856–58 or (3) the national user fee exemption policy for women and children under 5 years of age since 2016, known as ‘gratuité’,59–61 which has replaced ‘SONU’ and is still ongoing.

The process of adopting and formulating health policies in Burkina Faso can be described as following different pathways. It may be that the government initiates the policy development process and commissions technical departments or experts to propose and evaluate different alternatives. This is exemplified by ‘SONU’.54 Alternatively, ideas or initiatives may originate from civil society, non-governmental organisations (NGOs), associations or pressure groups, or from technicians lobbying politicians or decision-makers to adopt new or revised policies. This second scenario may be achieved by taking advantage of windows of opportunity, such as elections or social movements, as was the case with ‘gratuité’.62 These health policies are more likely to be endorsed and funded by the national budget, thereby ensuring a certain degree of sustainability. Another scenario is that of donor-funded projects and programmes, which often present sustainability challenges when funding ends while there is a lack of government or national ownership, or when the policy has not been institutionalised. This was evidenced by the experience of the PBF scheme.63

PBF was initially introduced in Burkina Faso in 2011 by the government as a pilot intervention in three health districts (out of 63 districts) with the financial support of the World Bank (WB), following a study tour in Rwanda. PBF links funds, payments or bonuses to health service providers based on results achieved against predefined quality and quantity criteria. It affords these providers a degree of autonomy in the utilisation of these resources. The rationale for introducing PBF was to improve the use and quality of maternal and child health services. After initial promising results, the government decided in 2014 to expand the pilot project to 12 health districts (out of 63) in six regions (out of 13), with continued financial support from the WB. Following a reorganisation of the health district map in 2015, which resulted in the splitting of some health districts, the total number of districts covered by PBF increased to 19 (out of 70). To implement PBF, a National Technical Unit was set up in 2014, comprising national technical assistants and policy-makers. The technical assistants had previous experience in implementing PBF, either acquired in Burkina Faso between 2011 and 2013 or in other countries. They received training in PBF from Sina Health,64 with some of them subsequently assuming the role of PBF trainers. They were tasked with transferring their expertise to policy-makers. Both the technical assistants and the policy-makers had vested interests and technical authority, but they had limited decision-making power on whether to continue PBF after the pilot phase.

‘Gratuité’ is almost 100% funded by the state budget and was previously piloted by international NGOs in five health districts since 2007.62 65–68 The policy was formulated and implemented by policy-makers involved in the pilot projects. These policy-makers had both the technical expertise, the necessary political support and the decision-making power to devise the policy and draft the guidelines essential for its successful implementation. Following a 2-month pilot phase in three regions, ‘gratuité’ was rolled out nationally in all public health facilities. For children under the age of 5, all direct medical costs related to curative care are covered, with a focus on the conditions and diseases included in the country’s Integrated Management of Childhood Illnesses strategy. For women, ‘gratuité’ provides comprehensive coverage for women’s healthcare needs related to pregnancy, encompassing antenatal care, childbirth and postpartum care up to 42 days after childbirth. The benefit package also includes the screening and treatment of precancerous lesions of the cervix, the physical examination of the breasts (for the early detection of breast cancer) and the treatment of obstetric fistulas. Funds are disbursed to health facilities on a quarterly basis under a third-party payment system. In this system, the state acts on behalf of the target population to pay for healthcare on a fee-for-service basis.

In introducing ‘gratuité’, the government of Burkina Faso was concerned about the increasing fragmentation of health financing and considered how this policy could be complementary and synergistic with PBF to avoid duplication of payments and allow each scheme to build on the strengths and assets of the other. The Secretary General of the Ministry of Health (MOH) therefore asked PBF and ‘gratuité’ policy-makers, that is, senior technicians, to reflect on how the two schemes can be coupled to make their purchasing function more efficient and performing. The concept of ‘strategic health purchasing’ thus emerged to brand this coupling and to achieve the latter four meetings were held in August 2016, March 2017, April 2017 and October 2017. The aim of the first meeting was to identify where the two schemes converged and diverged and what needed to be harmonised or changed to facilitate their coupling. The other three meetings were set up to develop a ground document for the implementation of a coupling strategy, including the definition of key principles for the coupling, the development of technical and financial management tools, and the conditions for successful implementation. These four meetings received financial support from the WB which sought to use this coupling as the basis for component 1 of the Health Services Reinforcement Project (HSRP), its new operation planned for 2018–2023, a successor to PBF. Each of these meetings ended with recommendations, but they were never implemented until the next meeting, which contributed to the coupling that was expected to lead to a new health financing scheme not being achieved by 2020.

Given this situation and the stalemate in the process, the idea of coupling PBF and ‘gratuité’ was abandoned. Instead, the MOH, in agreement with the WB, which was concerned about the low disbursement of funds for component 1 of the HSRP, decided to implement strategic health purchasing mechanisms only through ‘gratuité’. The HSRP was subsequently restructured to reflect these changes. However, like the coupling, the integration of strategic health purchasing approaches into ‘gratuité’ proved to be elusive. Indeed, disputes arose over how to proceed, prompting the Minister of Health to request technical assistance from the Strategic Purchasing Africa Resource Centre (SPARC) in February 2020 to help operationalise strategic health purchasing.69 SPARC is a resource hub launched in 2017 by the international NGO Results for Development, with funding from the Bill and Melinda Gates Foundation.70 From 2022, SPARC is hosted by the international NGO Amref Health Africa and aims to build and support capacity in strategic health purchasing across sub-Saharan Africa.

SPARC’s support to the MOH began in the last quarter of 2020 and was still ongoing at the time of writing. In September 2020, an initial consultative meeting was held with key MOH stakeholders to clarify needs and priority interventions, including (1) building a common understanding of what strategic health purchasing is and what it entails; (2) getting component 1 of the HSRP back on track, as its implementation had been severely delayed and (3) developing and implementing a roadmap for the overall strengthening of strategic health purchasing.

Methods

Conceptual framework

Our study adopts Dunlop and Radaelli’s conceptualisation of policy learning.28 50 Their framework, developed from a political science perspective, shows that policy learning can be either epistemic, reflexive, negotiative or hierarchical—these modes of learning reflect the characteristics of the policy process.43 Indeed, policy-makers/decision-makers may be faced with (1) intractable problems or complex policy issues with many uncertainties or, on the contrary, (2) ‘simpler’ policy problems for which solutions are available or easy to find. In the first case, epistemic learning is sought when specialised/expert knowledge is needed to find an answer; reflective learning arises or is useful when ‘expert’ knowledge is unavailable or insufficient (no evidence, no expert)—everyone’s contribution is encouraged through open and frank exchange and debate to address the situation, and thus all forms of knowledge (eg, lay, professional, expert) are equally valued and used.28 50 In the second case, learning through bargaining occurs when policy actors with often divergent ideas or interests discuss or negotiate, among peers or without subordination, to reach compromises and agreements on specific issues, while hierarchical learning occurs when policy actors follow formal and/or informal rules, instructions, orders and other more or less specific provisions.28 50 Each mode of learning has characteristics, as shown in table 1.

Table 1
|
Key features of the different modes of policy learning

Study design

Our study used an exploratory qualitative case study design. Our work was exploratory because it focused on an under-researched field: health policy learning in sub-Saharan Africa, and in particular, the empirical application of Dunlop and Radaelli’s framework to UHC policies, especially health financing. A qualitative case study approach is appropriate for our study because it allows complex events or phenomena to be examined in their real-world context,71 and it has proved relevant in other studies of policy learning.72–74 The case in our study is defined as the actions taken by the MOH of Burkina Faso to translate strategic health purchasing into policy and practice. We focus on two specific actions, which are our units of analysis or ‘cases within the case’: (1) the attempt to couple PBF and ‘gratuité’ between 2016 and 2019 and (2) SPARC’s technical support to the MOH since 2020 and ongoing.71

Study participant sampling

Our key informants were purposively selected based on their involvement in discussions related to strategic health purchasing and the possibility of contacting them, as some were not in Burkina Faso at the time of data collection. A total of 21 people, including 18 national and three international experts, participated in the study, and their function, role and the case for which they were interviewed are listed in table 2.

Table 2
|
Key informants’ function and role and case(s) in which they were interviewed (policy-maker here means senior technician)

Data collection

We collected data through document review and in-depth interviews with key informants.

The documents we reviewed included scientific (n=3) and grey literature (n=23), the latter including decrees and laws (n=2); policy documents (n=3); terms of reference for meetings (n=4); PowerPoint presentations (n=5); minutes of meetings (n=4); technical and concept notes (n=3) and activity/mission reports (n=2). We selected these documents based on their relevance to the research topic. The objective was to collate as much documentary evidence as possible from key informants to facilitate an accurate historical account of the formulation and implementation of PBF and ‘gratuité’, including their key developments and milestones. We used a grid, more specifically a coding tree based on the elements of our analytical framework, to extract data from these documents.

We conducted the interviews between August 2021 and July 2022, using a semistructured guide that covered the elements of our conceptual framework. We formulated the questions broadly—some examples are given in box 1—so that respondents could express their opinions and impressions at length. These questions were inspired by the elements that operationalise our analytical framework, as presented in table 1. The first author conducted the in-depth interviews in French, both face to face in the respondents’ offices and remotely via Zoom or Microsoft Teams platforms, depending on the key informants’ preferences and their residence in Burkina Faso or abroad. Several interviews were sometimes conducted with the same key informant, either because he/she was involved in both the attempt of coupling PBF and ‘gratuité’ and SPARC’s support to the MOH, or because there was a need to explore some issues further from previous interviews. We recorded most of the interviews and notes were made of data that were not recorded. The latter were subsequently incorporated into the coding and analysis of the data. The recorded data were transcribed verbatim into French in a Word file. The verbatims presented in the results section were translated into English by the first author just for the purpose of this manuscript.

Box 1

Some questions from the interview guide

  • What does strategic health purchasing mean to you?

  • How did you first hear about it?

  • What do you think is at stake for Burkina Faso and for the policy actors?

  • What was the role of each policy actor?

  • What topics or issues were raised during meetings or discussions?

  • What was the atmosphere at these meetings or discussions?

  • What was the nature of the discussions?

  • What was the attitude, the state of mind of the policy actors?

  • What were the main objectives of the meetings or discussions?

  • Were these objectives achieved? If so, what contributed to this? If not, why not?

Data processing and analysis

We began data analysis in the field, through the reflective and analytical notes we took on memo cards during or immediately after the interviews. We developed a coding tree in Excel, based on the elements of our analytical framework (table 1). To guarantee the dependability, strength and thoroughness of the findings, we employed a triangulation strategy whereby the data from the literature review, the interview transcripts, the notes, and our understanding of the context and health financing landscape in Burkina Faso, were analysed separately and then compared, using the coding tree we established. This approach enabled us to identify similarities and differences between the data sets. Points of divergence were explored in greater depth through document review or formal and informal discussions with the same or new key informants. The first author (JAK) coded the data manually and this coding subsequently underwent critical review by the other coauthors.

We used content analysis to identify the patterns that emerge from the data.71 We sought to characterise policy-making processes during the attempt of coupling PBF and ‘gratuité’ as well as SPARC’s technical support to the MOH. We then inferred the modes of learning (reflexive, epistemic, bargaining and hierarchical) that emerged or were constrained, the factors that explained these results, and the effect of these learnings or missed learning opportunities on policy outcomes.

Positionality and reflexivity

Our statement on reflexivity, as required by the journal guidelines, is included in online supplemental file 1.

The first author of this paper (JAK) has worked as a consultant for the WB and SPARC, even though he was not directly involved in either the attempt of coupling PBF and ‘gratuité’ or in SPARC’s technical support to the MOH. He is familiar with most of the actors who were/are involved in these processes, has built trusting relationships with them and has even collaborated or is collaborating with some of them in professional and research activities on health financing in Burkina Faso. This familiarity could limit the necessary distance to be maintained when analysing the data to minimise bias, especially as these actors are sometimes at odds. Aware of this risk, the first author tried to maintain a balanced position between all the actors. Rather, he used his knowledge of the actors and the context to access grey literature, to ask and probe meaningful questions during interviews, to have informal discussions and to understand the underlying issues. Moreover, although some of the coauthors (MDA and BM) have prior research experience in Burkina Faso, they occupy a more peripheral position, which affords greater neutrality towards the interviewees and the local context, and brings more objectivity to the analysis and interpretation of the findings. Ultimately, this mix of ‘insider’ and ‘outsider’ positions in the study team promises more interesting and comprehensive insights into policy processes.75

Results

Our findings are presented for each learning mode in relation to the coupling of PBF and ‘gratuité’ on the one hand, and to SPARC’s support to the MOH on the other one. Results are presented to reflect our three research questions and in line with our conceptual framework. First, we examine what learning modes have emerged and have been used or were stymied in these two cases (RQ 1). Second, to identify the factors that influence learning (RQ 2), we look at how did key policy actors related to each other; where knowledge and insights came from, were created, transferred and used for; and how interactions took place. Third and last, we present any policy outcomes (RQ 3). To make our results easier to understand, we often refer to the RQs and features associated with each learning mode in our analytical framework (see table 1).

Epistemic learning

The attempt to couple PBF and ‘gratuité’

Little evidence of epistemic learning emerged in the coupling discussions,76 with basic contextual scientific knowledge or technical expertise in strategic health purchasing being relatively weak as this concept was still in its infancy in Burkina Faso (RQ 1).

We did not know what to call the coupling. Should we say ‘PBF-gratuité’ or ‘gratuité-PBF’? And in the meantime, when we met with donors, a WHO expert told us that it was not worth looking for a complex word, that it was part of strategic purchasing; [KII-5].

From our interviews and document review (minutes of meetings and activity reports), key actors involved in the coupling discussions included (feature No 1): (1) policy-makers from various MOH departments, some in charge of implementing ‘gratuité’, others monitoring PBF implementation to ensure transfer of skills to the MOH, and others overseeing health financing issues in general, some of whom acted as facilitators at some meetings; (2) national technical assistants in charge of PBF implementation; (3) administrative staff and healthcare providers from health regions, districts and facilities where PBF was implemented; (4) members of PBF contract development and verification agencies and (5) and donors. The health financing specialists from the MOH who acted as facilitators during the meetings had a long and good experience of the Burkinabe health system but were not well acquainted with the concept of strategic health purchasing (RQ 2 and feature No 7). They were appointed by the health authorities on an ad hoc basis, more for their facilitation skills and their experience of the health system than for their expertise.

When I came back to the MOH in 2017 [after working for several years in an international organization], I found that it [strategic purchasing] was a concept that was being developed. I became part of the group that was working on developing the national health financing strategy, and that is when I heard more about strategic purchasing and started to learn about it; it is a concept that I had not really mastered before; [KII-12].

I was co-opted to attend the meeting and facilitate the sometimes-heated debates, because I am an old hand in the system that many people respect. But basically, I was curious, I wanted to see people’s perspectives on the coupling, how they dealt with the concept of strategic purchasing, and give my own perspective; [KII-12].

As a result, there was no experienced expert to guide participants, and it was difficult to reach consensus among policy actors on what strategic purchasing was (RQ 3 and feature No 7). For example, some actors equated strategic health purchasing with PBF, or simply rebranded PBF under a different name with slight modifications, while others already saw its functional value.

People talk about strategic purchasing all the time, but it is nothing new. It is all there [referring to the book ‘Performance-Based Financing Toolkit’]; [KII-11].

Strategic purchasing considers the needs of the population and the performance of providers. Much of this is already present in ‘gratuité’, so we no longer need to implement PBF tools; [KII-2].

SPARC’s support to the MOH

In contrast to the attempt to couple PBF and ‘gratuité’, our interviews and documents review (scientific papers, PowerPoint presentations, technical and concept notes) revealed that SPARC’s support to the MOH drew on a body of scientific knowledge and expertise on strategic health purchasing produced or supported by many expert organisations (RQ 1 and feature No 2), such as the RESYST’s consortium,76 the WHO,77 the Health Finance and Governance (HFG) project78 or SPARC itself, which worked at this time on a special series on the topic in nine African countries, which has now been published.79 SPARC’s legitimacy and expertise were recognised, its facilitators were aided by experts from WHO and SPARC headquarters, and their contribution was highly valued (RQ 2 and features No 1 and No 6). Indeed, many policy actors relied on them to help them better understand what strategic health purchasing is and what it entails (feature No 5).

When SPARC came in, their facilitators brought some interesting perspectives. We understood that it is a whole set of behaviors that you need when you are talking about strategic purchasing… personally, I did not have that full picture of things; [KII-5].

There were a lot of questions during the meeting and almost everyone had a question. There was a thirst to learn, to understand what strategic purchasing is. For some, it was a continuation of PBF; for others, it was something different. Many wanted to know how strategic purchasing differs from PBF; [KII-20].

Our interview with SPARC’s facilitators revealed that they developed and applied an assessment tool derived and adapted from a framework proposed by the RESYST consortium,76 to measure strategic health purchasing practices within ‘gratuité’ (features No 3 and No 6). This tool was deemed instrumental in demonstrating how strategic health purchasing can be incorporated into a given policy (RQ 2 and feature 4). This tool presents 23 desirable attributes for the purchasing function from a strategic purchasing perspective (see box 2 for some examples of these desirable attributes). The RESYST framework was used because of its relative simplicity, intuitiveness and ease of use (RQ 2 and feature 6). For each of the attributes, a scoring system was defined according to whether the practices associated with it were counterproductive, neutral or satisfactory in terms of strategic health purchasing. The assessment of ‘gratuité’ was carried out in several successive meetings with the MOH’s key stakeholders using the discovery method: each desirable attribute is presented, explained and discussed with the participants and the latter, based on their knowledge and experience, review the current ‘gratuité’ practices against these attributes and draw their own conclusion by assigning a score, usually after discussion among themselves (RQ 2 and features No 3 and No 6). The next step, which was not discussed in the meetings, but which emerged from our interviews, would be to identify changes or reforms that could be introduced to improve practices that were deemed counterproductive or neutral (feature No 5). These changes or reforms had not yet been implemented at the time of writing.

Box 2

Some desirable attributes from a strategic purchasing perspective drawn from the tool

  • Select providers on the basis of their location, availability and the quality of services they offer.

  • Establish (performance) contracts with providers.

  • Audit the invoices of health facilities.

  • Pay health facilities regularly and appropriately.

  • Develop, manage and use effective information systems to measure provider performance.

  • Allocate resources equitably across the country.

  • Inform citizens of their rights and responsibilities.

  • Ensure that citizens can exercise their rights.

  • Establish effective mechanisms for responding to public complaints.

  • Report publicly on resource use and service provider performance.

  • Fill the gaps in infrastructure and equipment for service delivery.

Reflexive learning

The attempt to couple PBF and ‘gratuité’

Reflexive learning was apparent during the attempt to couple PBF and ‘gratuité’, even if it was sometimes hindered (RQ 1). Indeed, in the four meetings convened to discuss the coupling, our interviews and document review (terms of reference and minutes of meetings) highlighted that several working methods were used to gain insights from policy actors, mainly based on their experiences in being involved in the implementation of either PBF or ‘gratuité’ (RQ 2 and features No 2 and No 3). These working methods included: using PowerPoint presentations to highlight the key features of each scheme; sharing Burundi’s experience in coupling PBF with a user fee exemption policy; presenting the results of previous meetings from the second meeting onwards; having group discussions to share viewpoints, experiences and perspectives; and reporting group discussions findings in plenary.

The minutes revealed that participants in the first meeting brainstormed to identify where PBF and ‘gratuité’ converged and diverged conceptually, and what needed to be harmonised or changed to achieve the coupling (features No 3, No 4 and No 5). From these interactions, it emerged that PBF and ‘gratuité’ could be coupled under certain conditions, with regard to (1) purchased services—if they are aligned because ‘PBF purchases health indicators while ‘gratuité’ purchases health services as they should be charged to health service users’ (KII-4); (2) management tools—if services purchased under ‘gratuité’ are integrated into PBF tools, in particular, the business plans and the indices tool; (3) verification—if the same verification agencies and procedures are used for the two schemes and (4) payment methods and mechanisms—if they are harmonised, as ‘gratuité’ uses a fee-for-service payment (with prepositioning of funds), while PBF uses a lump sum with a retrospective payment system.

Based on the results of this first meeting, the other meetings were focused on the design and development of the basic document for the implementation of the coupling (features No 3, No 4 and No 5), as outlined in their terms of reference and minutes. However, these meetings were plagued by strong opinions, clear-cut positions and preconceived ideas, which were not helped by the lack of effective facilitation (RQ 2 and feature No 7).

To be honest, we had the will to couple PBF and ‘gratuité’, but at the same time we were not ready to give up our PBF practices so easily. If you look at the characteristics of strategic purchasing, you also have them in PBF. In fact, the real problem is that PBF is tagged as World Bank, and strategic purchasing as Word Health Organization; [KII-7].

We had the impression that some people were trying to destroy PBF, thinking that for ‘gratuité’ to succeed, PBF had to disappear. We did everything to prove them wrong, we held several meetings but at a certain point we realized that we were preaching in a desert; [KII-6].

PBF has never worked and will never work; it is not a sustainable policy. We cannot meet its generally unrealistic demands. […] A study tour to Burundi was planned because they have experience in coupling PBF and ‘gratuité’. This trip did not take place because of a political unrest in Burundi. But it was an unnecessary trip because PBF as implemented cannot work in our context; [KII-1].

A dialogue of the deaf then ensued because no one was willing to listen to the other, to learn and change their perspectives or preferences (RQ 2 and feature No 7). As a result, our inquiries showed that proposed changes to each scheme to facilitate their coupling did not happen because no one was willing to let go (RQ 3).

SPARC’s support to the MOH

Like the attempt to couple PBF and ‘gratuité’, reflective learning was present in SPARC’s support to the MOH (RQ 1). In fact, although the SPARC label gave them some legitimacy as they were supported by experts from WHO and SPARC headquarters, the facilitators stated during the interviews that they did not claim to have extensive expertise in strategic health purchasing in their interactions with participants at the meetings (feature No 2). Instead, they positioned themselves more as coaches and facilitators to reconcile positions, as evidenced by a brief describing this process.69 Furthermore, while meetings participants were receptive to the many insights into strategic health purchasing provided by the facilitators, their interactions with them were not always in a teacher-learner mode, with the former listening ‘religiously’ to the latter.

People did not come just to listen to someone spouting words, and I did not come as an expert on anything… SPARC itself defines its in-country approach as a coaching approach, a facilitation approach; [KII-20].

The SPARC facilitators recalled that the floor at these meetings was free and conducive to open exchange, without no one participant monopolising it. Participants were free to express their views, experiences, doubts or disagreements (RQ 2 and features No 3 and No 6). During one meeting, a facilitator used a metaphor based on everyone’s everyday life to stimulate participants’ interest and allow them to reflect and discuss among themselves what strategic health purchasing could/would mean in practice. He referred to the example of a person who regularly goes to the market to buy tomatoes. This consumer will not buy poor quality tomatoes from the same vendor every day without questioning. On the contrary, each time the consumer realises that he/she has bought tomatoes that do not meet his/her expectations, he/she will consider changing his/her behaviour to get good quality tomatoes—this can be considered as strategic purchasing (RQ 2 and features No 4, No 5 and No 6).

This work yield positive results, reconciling views on what strategic health purchasing is and what it entails (RQ 3 and feature No 5).

What I can see is that the understanding has changed. Compared to the first meetings, where people did not have the same perspectives, in the subsequent meetings, when I talk to people, it is more or less the same vision of strategic purchasing that they have now; [KII-21].

Learning through bargaining

The attempt to couple PBF and ‘gratuité’

The scope conditions for learning through bargaining to emerge in the discussions regarding the attempt to couple PBF and ‘gratuité’ were not met (RQ 1). Indeed, key actors had vested interests and different perspectives that they could not reconcile (RQ 2 and features No 1 and No 7). For example, while some wanted PBF, which was due to end in June 2018, to continue and become more sustainable and saw the coupling as an opportunity to fulfil their wishes, others felt that it might not be worth continuing PBF.

We wanted PBF to become a government-funded program. So, we thought that coupling PBF with ‘gratuité’ would be an opportunity to make this shift; [KII-10].

‘Gratuité’ was a national policy and therefore could not be stopped. We saw that Burundi had successfully combined PBF and free health care, so we thought: well, why not try to replicate Burundi’s experience in Burkina Faso?; [KII-9].

Relevant assets of PBF were already considered in the design of ‘gratuité’, so we did not see the need to change anything. For example, quality issues are considered in the certification process that we planned to implement; we also already have the separation of functions in ‘gratuité’, namely the provider-payer-verifier split; [KII-1].

In addition, some policy actors interviewed felt that they were in a weaker position than others who, in their view, would not let go (RQ 2 and feature No 7). Indeed, for them, the relationship between PBF policy-makers and ‘gratuité’ policy-makers was unequal, as ‘gratuité’ was a recent national policy adopted by the Council of Ministers, while PBF was a pilot project in 19 of the 70 health districts, funded only by the WB, which was nearing completion and had not been institutionalised, that is, integrated into the routine of the health system. Thus, ‘gratuité’ had more political and public budgetary support than PBF, leaving the former with little to lose and the latter with much to lose if the coupling failed—in such a situation of inequality, real negotiation and learning opportunities are limited (RQ 1).

We realized that sometimes the other side would not concede. Do you know why? Because for them it is a policy adopted by the Council of Ministers, and they know that we cannot do anything against ‘gratuité’, no matter what we say or do, they know that ‘gratuité’ will continue; [KII-4].

It was also reported during the interviews that some of the key players were sometimes disinterested or casual, leaving ongoing meetings unaccounted for and unpunished, or being represented by people with little decision-making power (RQ 2 and feature No 7).

Some people do not feel compelled to come to the meetings, or they come to the opening and then leave and never come back. They know that whatever decision you make would not apply to them; [KII-10].

As a result, although the points of convergence and divergence between the two schemes were identified as highlighted in the ‘Reflexive learning’ section, our inquiries revealed that a consensus was never reached to meet these conditions and to adopt a basic implementation document. Therefore, a new policy combining PBF and ‘gratuité’ has never emerged (RQ 3).

SPARC’s support to the MOH

Similarly to what described in relation to the attempt to couple PBF and ‘gratuité’, learning through bargaining was not of much use when SPARC supported the MOH (RQ 1). In fact, the experience of the failed attempt to couple PBF and ‘gratuité’ meant that policy actors were already aware of each other’s preferences, some being intractable (feature No 5).

We were told that 1+1=1 meaning that PBF + ‘gratuité’ = ‘gratuité’; we were told this at a meeting. And that meeting was not an easy one…, so we understood that the coupling was not worth dreaming about; [KII-9].

After the failure of coupling PBF and ‘gratuité’, the MOH decided to implement strategic health purchasing only through ‘gratuité’, leading to a restructuring of the HSRP, as described in the ‘Study setting’ section.

A decision had to be made to move forward, especially with the HSRP deadline looming. The MOH therefore opted for ‘gratuité’, which was already being implemented nationwide and was largely funded from the national budget. The World Bank funds would complement this budget and address challenges faced by ‘gratuité’ such as insufficient and late payments to health facilities and poor quality of care; [KII-11].

However, this decision did not revive the process as quickly as expected (RQ 3). In fact, some actors suspected that others were still trying to replicate key features of PBF, as implemented between 2014 and 2018, within ‘gratuité’ (feature No 7).

If you look at [the first draft of] the proposed strategic purchasing implementation document, it is basically a PBF implementation document that has just been repackaged with some new words; [KII-1].

SPARC’s facilitators elucidated that their mission statement was twofold (feature No 5): first, to cultivate a unified conceptualisation of strategic health purchasing among national policy actors, and second, to facilitate implementation and get component 1 of the HSRP back on track; otherwise, the WB money could be lost.

There was an emergency agenda and a long-term agenda. The emergency agenda was to rapidly operationalize strategic health purchasing through ‘gratuité’ with funds from the HSRP. However, the HSRP was not able to disburse funds that were at risk of being returned to the World Bank, and we should help prevent this from happening. The long-term agenda was to identify the systemic changes needed to make strategic health purchasing through ‘gratuité’ sustainable and replicable in other health financing schemes; [KII-20].

Our interviews revealed a consensus among policy actors to act without delay to prevent the loss of financial resources. The HSRP was, therefore, implemented in a piecemeal fashion, following a technical note that set out a roadmap for the holistic strengthening of strategic health purchasing in the Burkinabe healthcare system (RQ 3).

Hierarchical learning

The attempt to couple PBF and ‘gratuité’

During the meetings, the policy actors discussed on an equal footing in terms of hierarchy, so this mode of learning was rarely present (RQ 1). In fact, hierarchical learning was largely absent, mainly because the MOH authorities were unable to give clear guidance to meeting participants on what to do. There was no prior national knowledge or experience of the coupling to draw on to give instructions on how to proceed or what milestones to achieve—almost everything had to be devised by the technicians, who were left to work out for themselves how to deal with the challenges (RQ 2).

There were no clear instructions from the hierarchy to help the actors in the discussions. The authorities did not get involved, they left us to our own devices, which was a pity. This probably explains why we have never reached a consensus; [KII-8].

Furthermore, PBF was perceived by many policy actors as having little or no sustainability because it was poorly institutionalised and owned by the MOH (RQ 2)—there was no transfer of powers. Indeed, our interviews highlighted that unlike ‘gratuité’, PBF had little political support, was developed in isolation, was driven by a few actors working in remote offices outside the main MOH buildings and was not sufficiently reported to higher authorities (RQ 2).

PBF was perceived as the affair of a limited number of technicians operating in isolated buildings, with minimal engagement with other directorates of the MOH. Therefore, PBF was not widely known, and we failed to engage the top officials of the MOH. For many, PBF represented an approach that merely provided financial resources to health workers when they already had salaries from the national budget. This suggests that if you had to live with ‘gratuité’ at that time, you either accepted what ‘gratuité’ [policymakers] wanted, or you disappeared; [KII-9]

There was significant turnover at the highest political and technical levels of the MOH—there have been five ministers of health and five secretaries-general between 2011 and 2018; so, few had enough time to really understand what PBF is about and the issues at stake to make informed decisions or give clear direction (RQ 2). As a result, the involvement of the highest health authorities in the discussions around the coupling, including the enforcement of recommendations coming out of the meetings, has been minimal or non-existent—the tendency has always been to favour ‘gratuité’.

Between 2011 and 2018, the period of PBF implementation, we have had several ministers of health. The time it takes to explain to a minister what PBF is, that minister is removed. So institutional instability and changes in leadership, whether at the top of the MOH or at the level of the general and technical directorates, hospitals or decentralized units, are not conducive to the sustainability of policies, but rather hinder them. For example, if we could have a minister who stayed for 5, 6 or 7 years, that would be ideal; [KII-8].

The highest authorities of the MOH did not intervene in our discussions. This is not surprising, by the way, because you cannot really get involved in something you do not know enough about, with the recurrent changes in the head of the ministry; [KII-5].

SPARC’s support to the MOH

Unlike the attempt to couple PBF and ‘gratuité’, the restructured HSRP received more political and hierarchical attention and support from the highest authorities in the MOH (RQ 1 and RQ 2), on which SPARC facilitators relied. The interviews revealed that the concern about losing WB funding was a significant factor in this outcome, and SPARC itself was expected to play a role in preventing such an eventuality, as mentioned above (see ‘Learning through bargaining’ section). Also, political authorities sometimes intervened to direct which regions or interventions should be funded.

With the restructuring of the HSRP, the government wanted to redirect funds to regions with a lot of internally displaced people [Burkina Faso has been facing terrorist attacks with massive population displacements since 2015]. The new minister [of health] also wanted the HSRP to fund hospitals to improve the quality of care, with his slogan ‘changing the face of hospitals’; [KII-1].

This political and hierarchical attention and support helped to drive action, including the development of the roadmap for the holistic strengthening of strategic health purchasing (see ‘Learning through bargaining’ section), although this was tempered by a high turnover at the top of the MOH which led to frequent restarts (RQ 3).

Discussion

Our study takes an original approach to investigating the potential of policy learning to inform the policy process—it gives greater insight into the role of learning in health policy-making. Our findings indicate that the capacity of policy actors to engage with diverse modes of learning shapes policy outcomes, particularly regarding the process of translating strategic health purchasing into policy and practice through the coupling of PBF and ‘gratuité’ or SPARC’s support to the MOH. Some contextual and political economy factors have played a role in determining whether these modes of learning have occurred or not, in relation to the position of the actors involved. Considering the empirical evidence presented, we discuss these results and their implications for the global health financing community. We also discuss the relevance and limitations of employing Dunlop and Radaelli’s framework in the context of UHC with a focus on our case studies.

Modes of learning, their determinants, their impact on policy outcomes and policy implications

As for the attempt to couple PBF and ‘gratuité’, we have seen that reflexive learning was predominantly employed in contrast to the other three learning modes (ie, epistemic learning, hierarchical learning and learning through bargaining). In 2016, the concept of strategic health purchasing was still in its infancy in sub-Saharan Africa and few people knew what it entailed, and Burkina Faso was in virtually uncharted waters trying to operationalise it. The attempt to couple PBF and ‘gratuité’, which was initially used to flesh out the concept, was a new venture—there were few similar experiences in the world, and little expertise or scientific literature available. The Burundian experience was presented, but it remained abstract and not contextualised as the study tour did not take place. National policy actors had different interpretations of strategic health purchasing and its operationalisation. Our findings suggest that implementing strategic health purchasing requires high-level expertise and an advanced understanding of how to optimise a health financing system, and expert organisations and international bodies may need to set up learning communities, processes, agendas,19 80 to effectively assist countries. This need is true for reasonably well-defined policies or schemes readily understood such as PBF or user fee exemptions, and is certainly even more true for approaches, concepts or ideas that are more subject to various interpretations as it is the case for strategic health purchasing.14 Beyond the difficulty to mobilise epistemic learning, whose role remains crucial as explained above, and although reflective learning was present, the poor bargaining and hierarchical learnings in the process probably explain the failure to couple PBF and ‘gratuité’. In fact, without neglecting the political economy of the situation and the vested interests at stake, with everyone defending their own position, the scope of the negotiations was too broad and ill defined, and the technocrats had to reach agreement among themselves in the absence of clear guidelines from their superiors and from the political level.

The gap in learning processes prevented national policy actors from achieving meaningful policy outcomes. Nevertheless, some progress was made, although piecemeal, in operationalising strategic health purchasing when the use of additional learning modes (ie, epistemic and hierarchical learning) or the reinforcement of reflexive learning was facilitated by the involvement of SPARC, with the influence of compelling contextual factors. It is reasonable to assume that SPARC’s relative success stems from (1) a more mature theoretical and operational understanding of strategic health purchasing—for example, the functional approach was instrumental in this respect11; (2) a greater involvement of the hierarchy, which issued some instructions and guidelines; (3) a right timing, linked to the fact that policy-makers knew the end result they wanted to achieve and had to find a solution and compromise if they did not want to lose WB funding and (4) the ability of SPARC to trigger reflective learning through a broker/coach/facilitator approach rather than simply wearing the mantle of expert69—this appears to be a step in the right direction and highlights the importance of learning facilitation and coaching.81 82

Berta et al define facilitation as ‘a goal-oriented, context-dependent social process for implementing new knowledge into practice or organisational routines. It typically involves individuals learning together in the context of a recognised need for improvement and supportive relationships. Effective communication and interactive problem solving are key process components’ (p. 7).83 As for Harvey et al, they assert that ‘the facilitator’s role is concerned with enabling the development of reflective learning by helping to identify learner needs, guide group processes, encourage critical thinking, and assess the achievement of learning goals’ (p. 305).84 Furthermore, it is essential that facilitators are equipped with the necessary skills to facilitate the activation of diverse learning modes. To illustrate, in instances of contention between stakeholders, as evidenced in our study, facilitators must possess the capacity to discern which disagreements may require to be resolved through decisions by higher authorities or by monitoring (ie, by activating hierarchical learning), which ones can be resolved through scientific evidence or expert opinion (ie, by activating epistemic learning), and which ones can be left open for discussion (ie, by activating reflexive learning) or negotiation, including with the right approach (ie, by activating learning through bargaining).

Our findings confirm that the learning modes are not mutually exclusive and that they can coexist and sometimes occur simultaneously22—a mix of learning modes is even desirable and essential. The absence of learning through bargaining in our case study may be a contributing factor to the observed mixed results, even in the context of SPARC’s intervention. In fact, health financing systems in many countries are fragmented and require multiple actors to be considered.85 86 This suggests that the negotiation phase between technocrats and the establishment of facilitation mechanisms should be given more attention by political authorities and, more broadly, by global health actors. Our study shows that even when technocrats agree on many points, the process of reforming health financing can run into tensions at a very early stage. This is likely to be a problem related to a ‘scheme’ mindset: technocrats are defensive, with strong underlying interests in maintaining the leadership on individual schemes and/or with entrenched positions or dogmatic views on how the scheme they are promoting should be implemented, making it difficult for them to be flexible and to learn.

Applicability, relevance and limitations of Dunlop and Radaelli’s framework

There are few actionable frameworks for empirically assessing policy learning,28 50 and to our knowledge, this is the first time that the Dunlop and Radaelli’s framework has been applied in a low-resource setting. Our findings support the relevance and applicability of Dunlop and Radaelli’s framework, which we have adapted, to the health sector and to health financing.

From a policy research perspective, the contribution of Dunlop and Radaelli’s framework (and other policy learning frameworks31) is to recognise that policy is not just the result of political power relationships—knowledge and ideas also influences policy.29 87–90 In fact, policy processes for UHC involve managing complexity and uncertainty, so a constant focus on policy learning is key, and different modes and mixes of learning may be needed. Epistemic learning is probably the most visible,22 supported by the strong push in recent years for ‘evidence-based policy’ or ‘translating research into policy and practice’.91–93 However, Dunlop and Radaelli’s framework and our findings show that other modes of learning and knowledge are also useful, if not inevitable, for health policy to be successful, as ourselves and others have already pointed out.22 94 Dunlop and Radaelli identified several factors that facilitate or impede learning, listed in table 1, which were largely corroborated by our empirical data, although with minor discrepancies. This was particularly the case for the barriers to hierarchical learning, where the ones we identified were different and even contradictory to those described in the analytical framework. Indeed, the latter mentioned elements such as veto power, insubordination, hierarchical arrogance and excessive control that inhibits personal initiative. In our case study, hierarchical learning was hampered by the inaction of the hierarchy, which was unable to provide clear instructions or guidelines due to a lack of technical expertise or interest in the pertinent issues. In addition, the frequent change in leadership at the helm of the MOH likely diminished the significance of instructions and, consequently, the efficacy of hierarchical learning, as the directives were likely to evolve swiftly with the advent of each new leader. In essence, the efficacy of hierarchical learning is contingent on the clarity and efficacy of the directives issued by line managers. Consequently, the durability of this learning process may be compromised in instances where there is a considerable turnover of these line managers.

According to Dunlop and Radaelli’s framework, the pivotal elements that shape learning modes are the ‘tractability of the problems’ and the ‘level of certification’ of the actors.28 In the context of our study, two forms of power that stem from policy actors’ specific roles and identities (see table 2) were the most significant factors influencing actors’ certification. The first one is the power conferred by the specific policies they were supporting. The second one is power derived from the knowledge and ideas they possess. However, the impact of these two forms of power on learning processes and outcomes was found to vary. Indeed, the use of knowledge/ideas has facilitated learning processes and policy outcomes, as evidenced by reflexive learning in the cases of the coupling of PBF and ‘gratuité’ and SPARC’s support to the MOH, or by epistemic learning in the case of SPARC’s support to the MOH. Conversely, the use of power as conferred by the specific policies has impeded learning processes and policy outcomes, as evidenced by the instances of the stalling of learning through bargaining in both cases.

While Dunlop and Radaelli’s framework allows for a pluralistic approach to learning, it is worth noting that it does not encompass every potential learning scenario. Indeed, this framework is primarily interactionist in its approach to learning,28 which explains, for example, why our study focused on learning during meetings. However, alternative learning mechanisms exist, such as those that facilitate direct field-based learning, whether in a collective or individual capacity. These include ‘learning by experimenting’ during pilot phases or ‘learning by doing’ during large-scale implementation phases, which were evident in both the PBF and ‘gratuité’ policy-making processes. One feature of learning consists in removing the unknown which could have at least two potentially related sources: an imperfect flow of information and uncertainty about the future. The framework proposed by Dunlop and Radaelli, with its interactionist approach, appears to be effective in addressing the first source, while the second source may require a broader approach. Direct action in the field through experimentation, learning by doing, past or lived experience (either personally or from third parties), or self-taught capacity building may prove a more suitable approach for addressing the latter.

Study limitations

While our study yielded pertinent and meaningful findings, it is not without some limitations. First, data were collected retrospectively, 5 years after the process of translating strategic health purchasing into policy and practice had begun, although it was still ongoing at the time of writing. It is, therefore, likely that the reconstruction of certain events may be subject to memory bias. However, we have used several sources to triangulate the data collected, thereby minimising this bias. In addition, our observations were primarily focused on the interactions between policy actors during meetings; other potential interactions may have occurred in other contexts, and thus have not been captured in our study.28 However, it is common practice in Burkina Faso (and in analogous contexts in Africa) to devise and assess policies through meetings gathering diverse actors as in our case studies, thereby minimising the risk of missing key interactions and therefore potential bias.

Avenues for further empirical work

In terms of the different stages of the heuristic framework (ie, the different stages of the policy process), the process of translating strategic health purchasing into policy and practice described in this study was mainly at the formulation stage. Had we had a more complete process, including the agenda setting, adoption, implementation and evaluation stages, we would have seen a different pattern in the emergence (or non-emergence) of the four learning modes. We hypothesise that epistemic learning and reflective learning are useful for policy formulation—as evidenced in our study; learning through bargaining for policy adoption; and hierarchical learning for policy implementation, monitoring and evaluation. Epistemic learning can help identify causalities and develop the policy’s theory of change—it can also identify the problem to be addressed. Reflexive learning is good for open-box thinking, helping to think about new ways and perspectives—which is very important when there is uncertainty and complexity. Bargaining learning can be used to assess perspectives, get a sense of interests and get buy-in. In hierarchical learning, the subordinate or implementer learns instructions from the superior, but also reports any implementation problems so that the superior can act accordingly. All these assumptions need empirical confirmation. On the other hand, retrospective approaches, such as those used in this study, enables researchers to select case studies that fit the conceptual or theoretical framework they are using. It would then be valuable to further explore the heuristic power of Dunlop and Radaelli’s framework through prospective studies.

Conclusions

Our study contributes to the emerging body of literature on the ‘learning health system’ and ‘learning for UHC’ concepts, offering new insights into the role of learning in complex policy processes. We have applied a policy learning framework to the case of translating the concept of strategic health purchasing into concrete policies and practices in Burkina Faso. Our findings indicate that the inability to mobilise the full range of learning modes as outlined in Dunlop and Radaelli’s framework has resulted in suboptimal policy outcomes. Conversely, the more consistent use of these learning modes under the auspices of SPARC has contributed to a more effective operationalisation of strategic health purchasing. The emergence of national or international policy learning entrepreneurs, in a role akin to that of SPARC, coupled with the application of effective facilitation techniques, appears to represent a promising avenue for fostering the advent of pluralistic learning. It is incumbent on each country to assume full ownership and responsibility for establishing the conditions that will make this a reality. This suggests structural investments in learning health systems and, more generally, in the national knowledge ecosystem.