Stakeholder perceptions and context of the implementation of performance-based financing in district hospitals in Mali

Background: To improve the performance of the healthcare system, Mali’s government implemented a pilot project of performance-based financing (PBF) in the field of reproductive health. It was established in the Koulikoro region. This research analyses the process of implementing PBF at district hospital (DH) level, something which has rarely been done in Africa. Methods: This qualitative research is based on a multiple, explanatory, and contrasting case study with nested levels of analysis. It covered three of the 10 DHs in the Koulikoro region. We conducted 36 interviews: 12 per DH with council of circle’s members (2) and health personnel (10). We also conducted 24 non-participant observation sessions, 16 informal interviews, and performed a literature review. We performed data analysis using the Consolidated Framework for Implementation Research (CFIR). Results: Stakeholders perceived the PBF pilot project as a vertical intervention from outside that focused solely on reproductive health. Local actors were not involved in the design of the PBF model. Several difficulties regarding the quality of its design and implementation were highlighted: too short duration of the intervention (8 months), choice and insufficient number of indicators according to the priority of the donors, and impossibility of making changes to the model during its implementation. All health workers adhered to the principles of PBF intervention. Except for members of the district health management team (DHMT) involved in the implementation, respondents only had partial knowledge of the PBF intervention. The implementation of PBF appeared to be easier in District 3 Hospital compared to District 1 and District 2 because it benefited from a pre-pilot project and had good leadership. Conclusion: The PBF programme offered an opportunity to improve the quality of care provided to the population through the motivation of health personnel in Mali. However, several obstacles were observed during the implementation of the PBF pilot project in DHs. When designing and implementing PBF in DHs, it is necessary to consider factors that can influence the implementation of a complex intervention.

The study took place in Mali, in the Koulikoro region. In this country, the health system pyramid has 3 levels (Figure S1, Supplementary file 1). Table S3 outlines socio-demographic and health characteristics of Mali and the Koulikoro region. Data collection and process analysis were carried out using the Consolidated Framework for Implementation Research (CFIR).25 According to this framework, 5 dimensions should be studied to understand the implementation of a health intervention (Figure): (i) Characteristics of PBF intervention, (ii) Outer Setting to DH, (iii) Inner Setting to DH, (iv) Characteristics of Individuals, and (v) Process (of PBF). Each of these dimensions includes several constructs. Thanks to a preliminary analysis of the CFIR conceptual framework by the research team, we were able to classify the selected constructs and their descriptions as well as the constructs not retained and their justification (see Table S4, Supplementary file 1). Conceptual Framework, Adapted From Damschroder et al.25 Abbreviation: PBF, performance-based financing. We adopted a qualitative approach, based on a set of explanatory and contrasting multiple case studies with nested levels of analysis26 corresponding to DHs and participants to the implementation of PBF. The conceptual framework outlined above (CFIR) guides our case studies. As per Yin’s 5 components of a case study, our study used: (1) a research question: how is PBF implemented in DHs?; (2) one main proposition in relation to the research question: the characteristics of the PBF intervention, the outer and inner setting of DHs, the characteristics of individuals, and the processes embedded in the intervention affect the implementation of PBF; (3) units of analysis (detailed in the paragraph below); (4) logic linking of the data to the propositions; and (5) the following criteria for interpreting the findings: CFIR’s 5 dimensions (ie, characteristics of the PBF intervention, outer setting of DH, inner setting of DH, characteristics of individuals, and PBF process) guiding data collection (eg, interview guides) and analytical approach. We chose 3 of the 10 DHs in the Koulikoro region. The characteristics of the 3 health districts and their DH are summarised in Table S5, Supplementary file 1. These figures take into account our resource constraints, but also feature an adequate representation of the diversity of contextual situations conducive to the analytical generalisation process specific to case.26 Our study is part of a wider research programme entitled: “Results-based financing for equitable access to maternal and child health care in Mali and Burkina Faso.” The cases were those identified for this research programme. Several districts were removed from the eligible cases because they did not represent the regular context of the health system that the intervention aims to improve (strong interventions by several NGOs), or are not accessible for security reasons. District 1 was selected for its urban character and the presence of a medical assistance scheme (Régime d’assistance médicale [RAMED]). In Mali, RAMED provides medical care for the poor and other vulnerable populations. District 2 was chosen for featuring an articulation of PBF with a community-based insurance was envisaged. This was intended to assess the extent to which PBF could serve as a basis for the accreditation process of health facilities, a process that predates the development of social protection systems (compulsory health insurance, community-based insurance, and RAMED). Finally, we selected District 3, a landlocked and agricultural area in which a community experiment to identify the poorest was also tested. Of the 3 districts selected, only District 3 experienced the PBF pilot phase. Purposeful sampling was used to select participants, so as to ensure external diversification. This selection makes it possible to contrast the different points of view of actors who occupy different positions within a group, in order to have an overall analysis that can be generalised.27 For each district, it was necessary to identify the different categories of actors involved in the implementation of PBF at the level of the 3 DHs. Different stakeholder profiles (Table 1) were selected to compare points of view.27 Participants were recruited based on their availability to answer interview questions. Abbreviations: PBF, performance-based financing; DH, district hospital. The research was conducted from December 2016 to January 2017. We stayed for 12 days in each of the 3 DHs. Actors’ perceptions and practices was the centre of the interview questions. Three semi-structured interview guides were prepared for the District Health Manager (Médecin chef de District [MCD]), staff (medical doctors, nurses, and other health workers and staff), and circle council members. The contents of the 3 guides have been adapted with the selected CFIR constructs. We conducted 36 semi-structured interviews. We also conducted 24 non-participant observation sessions as well as 16 informal interviews with 2 caretakers, 5 physicians, 1 pharmacist, 1 health information officer, 1 social development officer, 2 nurses, 2 interns, 1 midwife, 1 nurse woman, and 1 hygienist. Personal notes taken during non-participant observations and informal interviews were recorded in a journal. In the non-participant observation sessions, we focused on the following topics: (i) work environment (hygiene in hospital outbuildings, treatment rooms, waiting areas, washrooms, etc); (ii) technical tools in DHs (ie, availability and filling of clinical records, attendance books); and (iii) work performance (ie, quality of reception, quality of orientation, guard system). Data processing was done iteratively. All interviews and notes written were classified by site. Research assistants transcribed verbatim all the recordings. The transcribed data was reviewed and coded using a codebook derived from our theoretical framework.28 We coded the data using the QDA Miner Lite software. Based on these codes, we conducted data analysis using the CFIR. This method allowed for an analytical approach that followed a deductive-inductive logic, based on the CFIR dimensions, and allowed empirical themes to emerge that may be relevant to better understand the implementation of PBF in Mali. Our results are presented using the 5 dimensions of the CFIR framework, following a comprehensive and logical flow as recommended by a conceptor of the CFIR.25

• zitti, tony
• gautier, lara
• coulibaly, abdourahmane
• ridde, valéry