An Exploration of the Unintended Consequences of Performance-Based Financing in 6 Primary Healthcare Facilities in Burkina Faso

Background: Performance-based financing (PBF) is promoted to improve the quality and quantity of healthcare services in low-income countries. Despite the complexity of the intervention, little attention has been given to studying its unintended consequences. Our objective is to increase evidence on the unintended consequences of PBF in Burkina Faso. Methods: Using the diffusion of innovations theory, we conducted a multiple case study. The cases were 6 healthcare facilities in two districts. Between April 2015 and 2016, we collected data through 101 semi-structured interviews, discussions, observations, and documents. We conducted thematic analysis using a hybrid deductive-inductive approach. Secondary data was used to illustrate the evolution of reported services. We conducted a cross-case synthesis to identify the results arising independently from more than 1 case. Results: A desirable unintended consequence of PBF was that 3 facilities limited the sale of non-prescribed medication to encourage patients to consult. Undesirable unintended consequences were found in the majority of facilities including fixation on measures rather than on underlying objectives, the pursuit of narrow and less relevant performance indicators, gaming, and teaching trainees improper practices. Providers in all facilities deliberately manipulated medical registers and documents, such that the reported quantity and quality of care differed from what was actually delivered. While most participants indicated that PBF was more advantageous than previous practices, the long payment delays were a source of dissatisfaction and demotivation across all facilities. Dissatisfaction also emerged in relation to the distribution of subsidies and the non-attribution of quality points for services delivered by certain staff considered “unqualified” in guidelines. Results in many facilities revealed suboptimal planning, a perception of the intervention as “budgetivorous,” as well as tensions related to the principle of managerial autonomy. Conclusion: PBF led to numerous unintended consequences that could undermine the intervention’s effectiveness. The findings contribute to a more comprehensive picture of the consequences of implementing PBF. Policy-makers can use the results of this study to devise effective strategies before, during and after the implementation of the intervention to minimize undesirable unintended consequences and promote desirable ones. The study took place in 2 rural districts of Burkina Faso where improving the healthcare system’s performance remains a challenge. The low quality of healthcare is often characterised by the staff’s inhospitality, insufficient equipment/medication and lack of training. 27 In 2011, the government of Burkina Faso, with World Bank support, conducted a pre-pilot PBF test in 3 districts to address generalized quality deficiencies and improve healthcare system performance. 28 According to Steenland et al, 29 this intervention changed the previous financing system by defining a package of key health services to be targeted at contracted facilities, and issuing payments based on quantity and quality for these services. However, this pre-pilot PBF test did not include some recommended elements of PBF including an increase in health facility autonomy and the introduction of improved management tools. 29 Thus, in 2014, the intervention was modified to incorporate these elements and expanded to 12 more districts in order to conduct an experimental impact evaluation funded by the World Bank. The specific objectives of the intervention were to (1) increase the utilisation of healthcare services; (2) improve the quality of healthcare services; (3) improve the efficiency of the healthcare system; (4) ensure equity in access to healthcare services; (5) reinforce the motivation of personnel; (6) improve community participation; (7) reinforce the health information system; (8) consolidate public – private partnership; and (9) reinforce the governance of the healthcare system. Four intervention arms were implemented combining PBF modalities with different unit fees for service and equity measures (Supplementary file 1). The intervention model is available online. 30 Each month, a PBF auditor from the contractualisation and verification agency counted the number of reported healthcare services in registers to establish the subsidies. In total, 23 indicators were subsidized for the quantity of care at the primary care level. Every trimester, a team composed of district management team members (eg, doctor, midwife, nurse and pharmacist) assessed the facilities’ technical quality of care by sampling records from various medical registers and observing the facilities’ environment. Quality scores were reported in a 113-item grid (covering 28 domains) and were used to determine the bonuses to be paid (Supplementary file 2). Community verifications were also supposed to be conducted every trimester. Subsidies and bonuses were used to pay for facility expenditures and premiums to motivate staff. Providers were required to use an index tool every month to update the facilities’ revenues, plan expenditures, and determine the distribution of premiums. Each trimester, providers were also required to produce a performance improvement plan to set objectives and plan activities. We conducted a multiple case study with several embedded levels of analyses, using both qualitative and quantitative data. 31 Case studies are useful to investigate contemporary phenomenon in depth and within their real-life context when the boundaries between the phenomenon and context are not clear. 31 Evidence from multiple cases is considered more robust because it enables replication. 31 For this study, the cases were 6 primary healthcare facilities, called Centres de santé et de promotion sociale (CSPS) located in 2 districts. We selected 2 of the 12 districts involved in PBF in Burkina Faso. We selected these 2 districts because they (1) were located in diverse regions, (2) represented the normal context of the healthcare system, and (3) were relatively safe for data collectors. The first district has 21 primary healthcare facilities and 1 medical center with surgical satellite services. The second district has 56 primary healthcare facilities and 1 medical center with surgical satellite services and 1 regional hospital. Case selection for healthcare facilities followed a multistage screening procedure described elsewhere. 32,33 Table 1 describes each primary healthcare facility included. Abbreviations: PBF, performance-based financing; IHW, itinerary health worker. a Data available online: http://www.fbrburkina.org/data. Participants included a wide range of stakeholders, including providers (ie, nurses, midwives, itinerary health worker), support staff (ie, drug manager, janitors, guards), patients (eg, seeking care for curative consultations or maternal and child health), and community representatives (eg, members of the facility management committee) (see Table 2). Participants were purposefully selected based on their ability to provide relevant information and their accessibility. In each facility, we selected all the providers, support staff and volunteers for semi-structured interviews. Then, following the snowball approach, some participants referred us to other people who could shed light on the intervention. 34 This strategy was used to identify potential participants who were knowledgeable about or had a particular experience with the intervention (eg, auditors and administrative staff at the district level). Overall, we conducted 101 semi-structured interviews. Abbreviations: COGES, facilities’ management committees; CHWs, community health workers; PADS, program to support health development. We adopted a broad, exploratory approach in order to capture all changes that were not initially targeted by program planners. Through observation, semi-structured interviews and informal discussions, we collected data on various dimensions of the healthcare system including service delivery, governance, human resources, medication, health information system and financial management. Data were collected during 3 sequential phases, with each informing methods for the next. In the pilot phase (April 2015), the first author conducted one week of fieldwork in 2 facilities in the same district (facilities A and B). These served as pilot case studies to validate the feasibility of the methods. In phase 1 (January–April 2016), the first author conducted 3 months of fieldwork, examining 4 facilities in another district with longer field visits and more participants, for greater depth (facilities 1–4). Each facility was visited for 2 weeks. The first week primarily served to conduct observation within the facilities and the second week served to conduct semi-structured interviews with participants. The first author lived in the facilities which enabled her to conduct observation as well as informal discussions around the clock. The first author also attended a 6-day annual national PBF review meeting for 2015. In phase 2 (May 2016), the second author conducted 20 days of fieldwork in those facilities and neighbouring ones, to deepen the assessment of community verifications. Of the 101 semi-structured interviews conducted, 11 were in the pilot phase; 76 in phase 1; and 14 in phase 2. Our semi-structured interview guides 35 built upon previous questionnaires used for research on diffusion of innovations 36,37 but were tailored to our objectives and participants (see Supplementary file 3). They enabled us to assess how factors related to the social system, characteristics of the members, and the nature and use of the intervention interacted to produce consequences over time. Interviews were recorded and local research assistants made verbatim transcriptions. In total, 258 observation sessions were recorded in research diaries. Observations sites included health facilities and villages. During observation, we collected a wide range of intervention documents (eg, quantity and quality verification reports, index tools) to fuel our analyses. We also used publicly available secondary quantitative data on service delivery (http://www.fbrburkina.org). These longitudinal data were collected monthly in each facility for PBF audits. Healthcare workers reported the quantity of healthcare services that were recorded in the medical registers. Then, PBF auditors verified the reported data by manually recounting the quantity of services. They entered the data into an electronic platform. We accessed and used data collected between 2014 and 2016 on the number of integrated household visits (IHVs) and the number of people who underwent voluntary HIV screening. We used several strategies to increase the trustworthiness of findings including (1) prolonged engagement on the field, (2) peer debriefing with members of the research team, (3) collection of audio recordings and photographs to test findings, (4) triangulation between sources of information and methods, and (5) member checks with stakeholders to confirm results. 38 The primary unit of analysis was each healthcare facility. We combined deductive and inductive thematic analysis. 39 We began by developing a template of themes based on our theoretical framework. Then, we carefully read the transcripts of the recorded semi-structured interviews and field notes to assign the raw data to the predefined themes. At the same time, we derived new themes that emerged from the data and were judged relevant to our research topic. Data were triangulated by comparing different information sources. 40 QDA Miner, a qualitative data analysis software, was used to code and retrieve text segments. We integrated the results from all data collection phases and used a cross-case synthesis to draw general conclusions. Following a replication logic for multiple case studies, we considered results arising independently from more than one case to be more powerful than those from a single case, and gave the former more importance in the results. 31 To avoid cherry-picking results within the rich material, we only present unintended consequences that emerged in multiple healthcare facilities. We organized a member check in Burkina Faso to confirm the researchers’ data interpretation, triangulate results, and validate conclusions. 41 Further member checks were conducted subsequently on specific elements. We selected verbatim quotations from participants and field notes to enable readers to access the raw data and assess the credibility of results. 38