Menu
Menu
Menu
Menu
Background: Delays in accessing skilled delivery services are a major contributor to high maternal mortality in resource-limited settings. In 2015, the government of The Gambia initiated a results-based financing intervention that sought to increase uptake of skilled delivery. We performed a midline evaluation to determine the impact of the intervention and explore causes of delays. Methods: A mixed methods design was used to measure changes in uptake of skilled delivery and explore underlying reasons, with communities randomly assigned to four arms: (1) community-based intervention, (2) facility-based intervention, (3) community- and facility-based intervention, and (4) control. We obtained quantitative data from household surveys conducted at baseline (n = 1423) and midline (n = 1573). Qualitative data came from semi-structured interviews (baseline n = 20; midline n = 20) and focus group discussions (baseline n = 27; midline n = 39) with a range of stakeholders. Multivariable linear regression models were estimated using pooled data from baseline and midline. Qualitative data were recorded, transcribed, translated and thematically analyzed. Results: No increase was found in uptake of skilled delivery services between baseline and midline. However, relative to the control group, significant increases in referral to health facilities for delivery were found in areas receiving the community-based intervention (beta = 0.078, p < 0.10) and areas receiving both the community-based and facility-based interventions (beta = 0.198, p < 0.05). There was also an increase in accompaniment to health facilities for delivery in areas receiving only community-based interventions (beta = 0.095, p < 0.05). Transportation to health facilities for delivery increased in areas with both interventions (beta = 0.102, p < 0.05). Qualitative data indicate that delays in the decision to seek institutional delivery usually occurred when women had limited knowledge of delivery indications. Delays in reaching a health facility typically occurred due to transportation-related challenges. Although health workers noted shortages in supplies and equipment, women reported being supported by staff and experiencing minimal delays in receiving skilled delivery care once at the facility. Conclusions: Focusing efforts on informing the decision to seek care and overcoming transportation barriers can reduce delays in care-seeking among pregnant women and facilitate efforts to increase uptake of skilled delivery services through results-based financing mechanisms.
The Gambia is one of the smallest countries in West Africa, with a population of approximately 1.9 million in 2013 [17]. The maternal mortality ratio and neonatal mortality rate are both high at 433 deaths/100,000 live births and 22 deaths/1000 live births respectively [18]. Maternal deaths accounted for 36% of all deaths among women aged 15–49 in 2013 [18]. Less than 60% of births were attended by skilled providers, and only 63% of births occurred in health facilities [18]. The overarching goal of the national health policy is to reduce morbidity and mortality through the provision of equitable, affordable and quality health services and related services [19]. Regarding skilled delivery, this encompasses ensuring the availability of skilled attendants for the provision of basic essential obstetric and newborn care within a “functioning healthcare setting” as well as 24-h access to comprehensive emergency obstetric and newborn care [20]. Historically, TBAs have played an important role in assisting women during pregnancy, delivery and the post-partum period. In January 2015, the Ministry of Health and Social Welfare redesignated TBAs as “community birth companions” (CBCs) and redefined their role. This change was based on the recognition that despite the importance of their support to women, the inability of TBAs to provide emergency obstetric care has hampered efforts to reduce maternal mortality. Their new role was envisaged as referring and/or escorting women during labor or childbirth, with or without complications, to health facilities within their catchment areas and providing antenatal and postnatal care [19]. With this shift in policy, the government advised that all deliveries should take place within health facilities and be attended by skilled personnel. The Maternal and Child Nutrition and Health Results Project (MCNHRP), supported by the World Bank and the Health Results Innovations Trust Fund, was an initiative undertaken by the government of The Gambia to improve maternal and child health outcomes, with one objective being to increase uptake of skilled delivery. It was designed to be implemented in the three regions of The Gambia with the poorest health indicators: the Upper River, the Central River and the North Bank West Regions (URR, CRR and NBR-W). One-third of the total population of The Gambia lives in these regions. Preliminary assessments conducted to guide the design of the MCNHRP identified both supply-side and demand-side obstacles to achieving the desired health outcomes, and suggested that interventions should be purposefully integrated into existing community structures and the government health system. Taking into account these considerations, it was determined that the MCNHRP would incorporate RBF mechanisms into two core intervention packages: one implemented at the health facility level, and the other implemented at the community level. The facility-based intervention package was largely intended to address supply-side barriers identified in the preliminary assessments. It incentivized the provision of specified maternal and child nutrition and health services, including facility-based delivery by a skilled practitioner, as well as incentivizing service quality. It was implemented at hospitals, major health centers and minor health centers. Quarterly payments were issued to facilities on a fee-for-service basis. The incentivized service that we report on in this study, skilled delivery at a health facility, was rewarded with a payment of approximately 525 Gambian dalasis (US$ 12.50) per skilled delivery performed. The facilities could earn an additional amount up to the equivalent of 100% of the quarterly payment for full compliance with quality standards. The quality assessment tool addressed a range of issues such as cleanliness, quality of record-keeping, and availability of staff and supplies. Each health facility could use the RBF payments that it received to finance items in its quarterly business plan, such as materials and equipment, drugs, training, consulting services, and other operating costs. A maximum of 40% of payments could be allocated to staff bonuses. As a part of the facility-based intervention, women were invited to enroll in a conditional cash transfer (CCT) program in which they received one payment for attending an initial ANC visit in the first 12 weeks of pregnancy and a second equivalent payment for completing at least three more ANC visits during the course of the pregnancy. The community-based intervention package sought to encourage community mobilization activities and social and behavioral change communication (SBCC) to overcome demand-side barriers identified in the preliminary assessments. It provided quarterly incentive payments to village development committees (VDCs) and village support groups (VSGs) for achieving specified community targets relating to maternal and child health and nutrition, including a target for the number of pregnant women referred to health facilities for delivery. VDCs serve as the lowest local level of governance in The Gambia, and VSGs conduct health and nutrition education and mobilization activities. A VSG is comprised of three men and five women, including a village health worker and community birth companion, all of whom carry out their duties on an unpaid basis. The overall purpose of the VSGs is to increase knowledge and awareness of maternal and child health and nutrition, promote the adoption of healthy behaviours and good nutrition practices, and encourage appropriate care-seeking. In communities that received the community-based intervention package, VSGs were asked to carry out SBCC activities intended to increase knowledge and awareness of pregnancy, labor and delivery, as well as to promote institutional delivery. Eighty percent of each incentive payment that a community earned was allocated to the VDC for implementing community development activities elaborated in a quarterly business plan. The VSG received the remaining 20% for distribution among its members as incentive payments for conducting SBCC activities. In the three regions of The Gambia where MCNHRP implementation took place, there are 22 health centers and two hospitals serving a total of approximately 800 communities. The interventions were first pilot tested in three health centers and their catchment areas in North Bank West Region. Following pilot testing, the interventions were then refined for implementation in the remaining 19 health centers and their catchment areas, with rollout staggered to allow for comparisons between communities reached at different times. Facility- and community-based interventions were independently randomized preceding implementation. Lots were drawn in a public ceremony to select the ten health facility catchment areas that would receive the facility-based intervention in the first phase of the study. Participation in the community-based intervention was similarly determined with the random selection of communities from each health facility catchment area. The National Nutrition Agency and Ministry of Health and Social Welfare jointly implemented the interventions in collaboration with regional health directorates, health facilities and communities. We used a randomised 2 × 2 study design to measure the community-level impact of three intervention arms compared with the control arm: the facility-based arm, the community-based arm, and the combination facility-based and community-based arm. (Table 1). Study design Prior to the launch of MCNHRP activities, it was determined that two-stage cluster sampling would be carried out to obtain evaluation data. Accordingly, in the first sampling stage, six communities from the catchment area of each of the 19 non-pilot health facilities were randomly selected to participate in the evaluation, with the most recent census estimates used as the basis for probability proportional to population size sampling. These communities would be visited in each round of data collection. Due to a technical failure with the tablets used to collect information, data from one community were lost at baseline. Thus, the final sample included 112 communities at baseline and 113 at midline. Baseline data were collected in October–November 2014, shortly before the launch of program activities. Midline data were collected approximately halfway through the overall study period, in July–August 2016. At the time of midline data collection, the facility-based component of MCNHRP had been implemented at 10 non-pilot health facilities for approximately 18 months. Their catchment areas collectively encompassed 60 evaluation communities (Table 2). Additionally, the community-based component of MCNHRP had been implemented in 37 evaluation communities across the 19 health facility catchment areas for approximately 15 months. Twenty of the 37 evaluation communities were in catchment areas where the facility-based intervention also was taking place, and the remaining 17 were in catchment areas that did not receive the facility-based intervention. The 36 evaluation communities that were not exposed to either intervention served as controls. Distribution of evaluation communities among study arms A qualitative study was embedded within the quantitative study to allow for triangulation across data sources. All quantitative data used in this analysis were drawn from household surveys administered at baseline and midline to heads of households and mothers of children under the age of five. Survey instruments were adapted from the household survey in the World Bank’s impact evaluation toolkit for results-based financing [21]. Experienced enumerators from the Gambia Bureau of Statistics were trained on and then administered these surveys with technical support from the evaluation team. The surveys collected information on household demographics, socioeconomic variables, healthcare utilization and health outcomes. All quantitative survey data were collected on tablets with real-time data quality checks. A target of 2400 households was chosen in order to be able to detect an improvement of 10 percentage points in the main outcome variables with power 0.9 and an intra-class correlation coefficient of 0.05. For skilled deliveries at facility, baseline prevalence in the randomized areas (19 non-pilot health facilities) was 0.42. The study was powered to detect a 16 percentage point increase with power 0.8, and an 18 percentage point increase with power 0.9. Despite the use of baseline data, the ex-post design effect was large (DEFF = 4.1) for the facility-based intervention due to the small number of clusters (N = 19) in the study, somewhat limiting the ability to detect project impact. Community randomization was done at the cluster level. For variables with a baseline prevalence of 50% (such as institutional deliveries), the study was powered to detect a 10 percentage point increase with power 0.8, and a 12 percentage point increase with power 0.9. For each round of the household survey, two-stage cluster sampling was used to identify a random sample targeting 120 households from each of the 19 health facility catchment areas included in the evaluation. As explained previously, evaluation communities were selected in the first sampling stage. Immediately prior to each round of survey activities, each evaluation community was visited by mapping teams to develop a list of eligible households. From all eligible households listed, 20 households were randomly selected for the survey. If fewer than 20 households in a given community were eligible, all would be selected. A household was eligible for inclusion if it had at least one woman aged 15 or older and at least one child under the age of five. A mother was eligible to participate if she was at least 15 years of age and her child was under five years of age. Within each household, questionnaires were administered to the head of the household and the mother of the youngest child. The mother was asked to respond to questions about delivery practices in relation to her most recent delivery. This analysis was restricted to households with a birth within the 450 days (approximately 15 months) preceding the survey (whether baseline or midline), which approximately restricts the midline analysis to women who gave birth since the program “fully” launched. It utilises six indicators to assess whether the intervention benefitted maternal and child health by encouraging the uptake of health services and behaviors: referred to delivery, accompanied to delivery, transported to delivery, community birth companion, facility delivery, and skilled delivery at facility, each of which is described below: Two indicators, ‘referred to delivery’ and ‘skilled delivery at facility’, were directly incentivized by the MCNHRP interventions. Multivariable linear regression models for all six indicators were estimated using pooled baseline and midline data. Regression models included indicator variables for midline survey (time trend) and each of the three intervention arms at the time of data collection (facility-based intervention only, community-based intervention only, and facility-based intervention plus community-based intervention). To account for spatial differences, settlement fixed effects were also included in all models. Standard errors were clustered at the facility level using Huber’s cluster robust variance estimator [22]. Beta coefficients and standard errors are presented to show intervention impact across different study arms. Based on an embedded mixed methods design, primary qualitative data were collected from focus group discussions (FGDs) and key informant interviews to help gain nuanced insight into people’s experiences of the project as well as the reasons underlying the project’s performance (Table 3). Study communities were purposively selected to reflect a mix of levels of performance across quantitative indicators as well as regional diversity. Different guides were used for each type of participant but core themes explored included: understanding of the MCNHRP, perceptions of the project, health and nutrition-related behaviors within the community, and how these behaviors may have changed since project inception. Qualitative participants and sample sizes a‘Vulnerable’ was not specifically defined but was determined by community leaders during data collection. Typically it included widows and unmarried adolescents FGDs lasted from 90 to 150 min, while interviews lasted from 45 to 90 min. Nobody refused participation or dropped out partway through a FGD or an interview. Three local qualitative researchers – one nurse as team lead (female, currently working for a sexual and reproductive health community-based organisation) and two community health nurses (one female, currently working at the national family planning association; one male, currently working as a monitoring and evaluation officer for an NGO) – collected the qualitative data in the language of the respondents’ choice. All team members had prior experience working in the field of maternal and child nutrition and health, and carrying out interviews and/or FGDs. Prior to each round of data collection, the team was trained on the interview and FGD guides. Instruments were field tested and amended as necessary before data collection began. There was no prior contact between the researchers and study participants. Participants were purposively sampled and recruited through face-to-face contact with community gate-keepers. Researchers introduced themselves to participants, explained the purpose of the study, answered any questions and sought informed consent before data collection. Participants helped to decide the most appropriate location for each interview/FGD to ensure a ‘safe’ space without the presence of anyone besides the participants and researchers. Researchers took field notes throughout data collection. The same process was used at baseline and midline with no attempt to find the same participants at midline as had been included at baseline. All interviews and focus group discussions were recorded, transcribed verbatim and translated to English. The transcripts, along with the researchers’ field notes, were entered and analyzed in NVivo 10. Data were double-coded and thematically analyzed using a framework derived originally from the literature and then refined as themes emerged in the data. The same code tree was used for the baseline and midline analyses (available on request). Quotes that are illustrative of the main themes that emerged are provided. Once independent analysis of the quantitative and qualitative data was complete, data were triangulated, which we understood to mean “a process of studying a problem using different methods to gain a more complete picture”. Areas of convergence, complementarity and divergence were identified, and the explanatory value the qualitative data could afford quantitative findings explored [23]. Ethical approval for the impact evaluation was obtained from The Gambia Government/MRC Joint Ethics Committee (R014036v2) as well as the Ethics Review Committee of the University of Southern California (HS-14-00688). Study participation was voluntary, and informed consent was obtained from all study participants after they were told about the study objectives and about how their information would be used.
N/A
