Using experience to create evidence: a mixed methods process evaluation of the new free family planning policy in Burkina Faso

Background: In 2019, Burkina Faso was one of the first countries in Sub-Saharan Africa to introduce a free family planning (FP) policy. This process evaluation aims to identify obstacles and facilitators to its implementation, examine its coverage in the targeted population after six months, and investigate its influence on the perceived quality of FP services. Methods: This process evaluation was conducted from November 2019 through March 2020 in the two regions of Burkina Faso where the new policy was introduced as a pilot. Mixed methods were used with a convergent design. Semi-directed interviews were conducted with the Ministry of Health (n = 3), healthcare workers (n = 10), and women aged 15–49 years (n = 10). Surveys were also administered to the female members of 696 households randomly selected from four health districts (n = 901). Results: Implementation obstacles include insufficient communication, shortages of consumables and contraceptives, and delays in reimbursement from the government. The main facilitators were previous experience with free healthcare policies, good acceptability in the population, and support from local associations. Six months after its introduction, only 50% of the surveyed participants knew about the free FP policy. Higher education level, being sexually active or in a relationship, having recently seen a healthcare professional, and possession of a radio significantly increased the odds of knowing. Of the participants, 39% continued paying for FP services despite the new policy, mainly because of stock shortages forcing them to buy their contraceptive products elsewhere. Increased waiting time and shorter consultations were also reported. Conclusion: Six months after its introduction, the free FP policy still has gaps in its implementation, as women continue to spend money for FP services and have little knowledge of the policy, particularly in the Cascades region. While its use is reportedly increasing, addressing implementation issues could further improve women’s access to contraception. The free FP policy was introduced as a pilot in the Cascades and Centre-Ouest regions in June 2019 by the Government of Burkina Faso (see Fig. 1). These areas comprise a total population of ~ 2.5 million, mostly (> 80%) located in rural areas, and present a fertility rate of around six children per woman [21, 22]. The policy applied to all public health facilities and covered 100% of the cost of FP consultations and counseling, tests and examinations, and contraceptives themselves (injectables, implants, copper intrauterine devices, emergency contraceptive pills, condoms, surgical methods, and a range of natural methods). Management of side effects and transport to the reference health facility for medical evacuations were also covered, as were all FP-related medical procedures (e.g., installation and removal of implants) and consumables (gloves, syringes, swabs, disinfectants, etc.). The goal is for women who are sexually active to pay nothing for any aspect of FP. Map of study area. The four health districts are displayed in blue. Main roads are shown as grey lines The introduction of free FP followed a national policy implemented in July 2016 that removed user fees for maternal and child healthcare services [23]. Both interventions use the same third-party reimbursement mechanisms and follow similar administrative and reporting procedures. As such, the free FP policy was conceptualized as a functional scale-up of the national user fee removal policy [24]. The implementation process was therefore facilitated and consisted mainly of informing health personnel through official channels of the extension of free procedures to family planning-related services. Dissemination activities in the population were also planned through radio messages and awareness campaigns in the communities by healthcare providers. This study was conducted between November 2019 and March 2020 in two separate phases (see Fig. 2). This study was conducted with the use of a mixed method design as per Creswell and Clark [25]. A mixed methods design was chosen in order to benefit from knowledge that comes from both qualitative and quantitative research as well as the integration of these two approaches [25]. First, a qualitative exploratory phase was undertaken to gain insight into the policy implementation process and refine research questions and instruments. Second, a qualitative and quantitative field data collection phase was conducted. Data was triangulated during the analysis using a convergent mixed method design, which was used to assess the coverage and implementation level of the free FP policy among the targeted population (Objective 2). Study design and collection phases The other two objectives, to investigate the presence of obstacles or facilitators to implementation and to assess the policy’s influence on the perceived quality of FP services, were pursued qualitatively. This was intentional, since it was necessary to gain an in-depth understanding of these topics and explore emerging themes—which is particularly suitable to qualitative research [26, 27]. To attain a variety of perspectives, qualitative investigations focused on three levels of policy implementation: (i) the central level, with the Ministry of Health (MoH); (ii) the peripheral level, with HCWs; and (iii) the community level, with direct beneficiaries of the policy. Moore’s conceptual framework for process evaluations of complex interventions guided this implementation study (see Additional file 1: Appendix S1) [28]. The specific components of Moore’s conceptual framework studied in this evaluation are process (Objective 1), fidelity and reach (Objective 2), and outcomes (Objective 3). The study took place in the context of rising insecurity in the country caused by terrorist attacks [29]. It was also conducted shortly after a nationwide strike had paralyzed non-essential activities in health facilities for several weeks. In October 2019, official documents (national planning and implementing strategy textbooks, information guides for health authorities, policy statements) were collected to gather as much information as possible before conducting interviews. In November 2019, semi-structured individual interviews (n = 3) were conducted in the capital Ouagadougou with program planners within the MoH involved in developing the free FP policy. The participants were conveniently selected with the assistance of a well-known knowledge broker for health matters in Burkina Faso. Interviews took place in MoH actors’ offices. They were conducted in French by LB, were supervised by a local researcher trained in qualitative research (AB), and lasted 30–60 min. Field notes were taken during the interviews, which were audio-recorded. After three interviews, information collected was deemed sufficient to form a good understanding of how implementation of free FP had been planned. Data was interpreted based on the review of planning documents, field notes and audio recordings, and with feedback from the other research team members. The fact that many researchers on the team had been involved for several years with the MoH in studies on free healthcare policies enriched this exploratory phase. This study’s quantitative component was embedded in another ongoing research project that aimed to evaluate the impacts of the national policy that removed health center user fees for pregnant women and children under five in 2016. Questions specific to the free FP policy were added to the original survey. The quantitative component was mostly designed to pursue Objective #2 of the present study: assessing the coverage and implementation level of the free FP policy among the targeted population. The coverage dimension was explored by assessing the beneficiaries’ knowledge of the FP policy and its associated factors, while the implementation level was examined by considering the presence of residual costs related to FP. The sampling procedures were derived from those of the USAID Demographic and Health Surveys program. A two-stage cluster sampling was carried out in four out of 10 health districts: Leo and Tenado (Centre-Ouest) and Sindou and Banfora (Cascades). These districts were purposively selected based on two criteria: They contained health facilities located in rural areas, and they were secure (not having experienced any attacks since at least 2016). Using the enumeration areas as defined by the Demographic and Health Surveys Program in these four districts, 29 were randomly selected with a probability proportional to the size of their population. In the second stage, 24 households per unit were randomly selected with equal probability. The target sample size of households was 696. Only households with ≥ 1 woman aged 15–49 were eligible. Ineligible households and households that could not be found were replaced by the nearest one. The survey took place in March 2020, after a five-day training for the interviewers. A questionnaire adapted from the standardized Demographic and Health Survey was administered to all consenting women aged 15–49 years from the selected households. It was administered in the local language by female interviewers with prior experience in community-based surveys. Although the standardized questionnaire already covered the participants’ sociodemographic characteristics and use of FP services, some questions were added to record out of pocket payments for FP services covered by the policy and the participants’ knowledge of the free PF policy. Responses were collected electronically on tablets using Commcare software (Dimagi, Cambridge, USA). Questionnaire data was automatically uploaded to a secure server then extracted and cleaned using Stata 14.0 (StataCorp, College Station, TX). Descriptive analysis was performed on three key variables related to the implementation of the free FP policy. The coverage of this policy was first assessed by estimating the proportion of targeted women who know that FP was now officially free of charge at health facilities. Secondly, contraceptive use (and moment of procurement) was measured by categorizing women according to their current use of contraceptives (yes/no) and the moment they last procured them, i.e., before or after the policy had been implemented. Thirdly, the costs associated with the respondent’s last FP visit to the health facility were analyzed by the time of the visit (before or after the introduction of the free policy) and broken down by type of service. All statistical analyses were conducted in open-source R statistical software V3.5.2. Maps were produced using the open-source software QGIS v3.8.1 Zanzibar. A multivariable logistic regression model with robust variance estimators was used to investigate factors related to knowledge of FP policy. The difference in the proportion of participants who paid for FP services before and after the introduction of the free policy was estimated by Chi-square tests of homogeneity of variance. The threshold for statistical significance was set at 0.05 (bilateral tests). The qualitative component consisted of semi-directed individual interviews with health personnel and female community members. For convenience and logistical reasons, the qualitative research took place only in the district of Banfora, where five public health facilities were selected based on their location (accessible rural areas) and their type (health facilities without maternity services were excluded). In each health center, the head nurse and another member of the health staff (preferably a midwife) were individually interviewed (n = 10). With their assistance, households or groups of households with women of reproductive age were identified in the catchment area, and two women were selected in the community surrounding each health facility (n = 10). Selection was stratified by age, with one female community member < 20 and the other one > 20; these participants were from different households. To be selected, female community members had to be currently using FP. Data was collected in January 2020. For community members, the interviews took place outside of their home, in a secluded location that guaranteed the confidentiality of the respondents. For health personnel, interviews took place in a private room at the health facility. An interview guide specific to the type of the participant and with open-ended questions was used during the discussion (see Additional file 1). The interviews were conducted in French or in Djoula (depending on the participant’s preference) by a single female researcher with extensive training in qualitative research. Interviews with health personnel and women lasted 30–50 min and 15–30 min, respectively. They were recorded, transcribed verbatim and translated into French by an assistant. The field researcher listened to the original audio recordings and validated the transcripts. The research team members met once. Ten interviews were conducted with each type of participant and, with feedback from the researcher and her field notes, data saturation was considered to be reached. A content analysis technique was carried out on the qualitative data. The transcripts were read several times for thorough understanding. The text was entirely segmented, and a mixed inductive and deductive coding was used: deductive because the coding grid was first established based on Moore’s conceptual framework, and inductive because new codes were created for emerging themes [28]. The concepts from the framework which were used in the deductive coding grid were: process (obstacles and facilitators to the implementation), fidelity (payment for services or methods of FP), reach (knowledge of the intervention), and outcomes (perceived quality of care) [28]. Double coding was performed by two authors on a sample of the material to confirm the reliability of the final coding grid. Coding was conducted by LB and CB, as well as the thematic analysis. Peer debriefing was conducted with members of the research team, LB, CB and SC to derive themes from the codes. The interview results were triangulated with the data collected in the qualitative exploratory phase. The analysis was performed using QDA Miner software (QDA Miner 5.0). Integration of qualitative and quantitative data for the convergent design was carried out specifically for Objective 2 of our study: assessing coverage and implementation level of the free FP policy among the targeted population by investigating knowledge of the FP policy and residual costs related to FP. For this objective, quantitative and qualitative results were analyzed in parallel to study the same object before being integrated. A resulting comparison strategy was used by comparing the qualitative and quantitative components and identifying and interpreting divergences and convergences [25]. This interpretation was carried out to expand our understanding of conclusions on the free FP policy. For example, qualitative results were used to confirm our quantitative finding and explain specific quantitative results (expenditures on FP and knowledge level of the new policy). All participants provided informed written consent for both the qualitative and quantitative data collections. For the quantitative phase, consent was recorded on the tablet where the questionnaire was conducted. The questionnaire and the interviews were administered individually in a secluded area to preserve the confidentiality of participants. Participants aged 15–17 years old were considered mature minors and consented as adults. All the study procedures, including those for obtaining consent, were approved by the Comité d’éthique de la recherche en sciences de la santé at University of Montreal (Certificate #CERSES-20-146-D) and by the Comité d’éthique pour la Recherche en Santé in Burkina Faso (Deliberation #2018-6-075). The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript.