The mixed effects of a package of multilevel interventions on the health and care of pregnant women in Zinder, Niger

Background Anaemia is prevalent among pregnant women in rural Niger and antenatal care (ANC) attendance is suboptimal. We designed a programmatic intervention including community-based behaviour change communication, provision of essential drugs (including iron folic acid (IFA) supplements) and quality improvement activities at selected integrated health centres (IHCs). Objective To assess the impact of the programmatic intervention on: (1) utilisation of ANC, (2) adherence to daily IFA supplementation and (3) prevalence of adequate gestational weight gain (GWG) and anaemia among pregnant women in Zinder, Niger. Methods Using a quasi-experimental study design comparing a cohort of women at baseline to another cohort of women at endline, 18 IHCs and surrounding villages were randomly assigned to time of enrolment over 1 year. A baseline survey was implemented among randomly selected pregnant women in 68 village clusters. Subsequently, the intervention was rolled out and an endline survey was implemented 6 months later in the same villages. Results Mean age in the baseline (n=1385) and endline (n=922) surveys was 25.8±6.4 years. The percentage of pregnant women who reported attending any number of ANC and an adequate number of ANC for their gestational age, respectively, was not significantly different between the endline and the baseline surveys. Pregnant women in the endline survey were more likely to have received IFA (60.0% vs 45.8%, OR: 2.7 (1.2, 6.1)); and the proportion of pregnant women who reportedly consumed IFA daily in the previous 7 days was significantly higher in the endline than in the baseline survey (46.4% vs 32.8%, OR: 2.8 (1.2, 6.5)). There was no impact on the prevalence of adequate GWG or anaemia. Conclusions The programmatic intervention resulted in a modest increase in the number of pregnant women who reported receiving and consuming IFA supplements as recommended, but did not affect ANC attendance and nutritional status. The NiMaNu project was approved by the National Ethics Committee in Niamey, Niger and the Institutional Review Board of the University of California, Davis, USA. In the presence of a neutral witness, consent materials were presented by multiple enumerators in both written and oral formats. A standardised informed consent procedure was used and documented with a written signature or a fingerprint. Participants received a locally appropriate gift such as a bar of soap to compensate for their time and effort. The trial was registered at www.clinicaltrials.gov as {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01832688″,”term_id”:”NCT01832688″}}NCT01832688. The NiMaNu project was designed as a quasi-experimental descriptive study comparing preintervention and postintervention cohorts of pregnant women. A multistage clustered sampling design was used for the baseline and endline surveys. Two health districts (Mirriah and Zinder) in the Zinder region were selected in collaboration with regional and health district personnel for convenience based on their accessibility, distance to Zinder, and the limited number and scope of interventions being implemented in the catchment area of the integrated health centres (IHCs) within these districts to avoid overlapping interventions conducted in the same area. In Niger, the health system in a district is organised at three levels: (1) health posts provide very basic outpatient curative and preventive care and are staffed primarily by community health workers, (2) IHCs serve several villages, provide outpatient and inpatient care, including labour and delivery, and are staffed by nurses, midwifes and sometimes physicians, and (3) a district hospital, provides diagnostics services, outpatient and inpatient curative care, and Caesarean sections, and is the level of referral for IHCs.17 18 In the health district of Zinder, all eight IHCs serving rural communities were selected, and the remaining two IHCs were excluded because they only served urban areas. In Mirriah, 10 out of 36 IHCs were selected based on convenience sampling. A statistician at the University of California, Davis stratified individual IHCs in groups of three by selected indicators (total population; district; distance to Zinder; number of health posts), and then randomly assigned individual IHCs to the order of enrolment. Within each IHC, 10 villages were randomly selected after stratification by distance from IHC and population size. First, the village where the IHC was located was automatically chosen for participation and one village containing a health post was randomly selected from all those with a health post that were within the catchment area of the IHC. Second, among the remaining villages in the catchment area of the health centre, four villages ≤10 km and four villages >10 km from the IHC were randomly selected and randomised to the order of enrolment. To assess the impacts of the programmatic intervention, a rolling baseline survey was implemented in 12 IHCs over the course of 1 year among randomly selected pregnant women in participating villages. Once the baseline survey was completed in the catchment area of an IHC, the programmatic intervention was implemented. Subsequently, the endline survey was implemented in each village 6 months after the completion of the village’s baseline survey. The baseline survey was extended in parallel with the endline survey for 6 months to serve as a pre-and-post and time-matched comparison and to account for the impact of season (figure 1); thus the remaining 6 IHCs participated in the baseline survey only. Timeline of the baseline and endline surveys. All the resident pregnant women in the selected villages (ie, women who lived in the study area for the 6 months prior to enrolment) and had no plans to move out of the study area within the next 2 months, were eligible to participate in the surveys and measurements, after providing written informed consent. Pregnant women who had severe illnesses warranting immediate hospital referral and those who were unable to provide consent due to impaired decision making ability were excluded. The target enrolment was 77 women per IHC, to reach the sample size described in more detail below. Pregnant women were identified using a random walk method19 until 16–20 women per village were enrolled or until there were no more eligible pregnant women in the village. Randomly selected villages were visited by order of randomisation until the target sample size per IHC was reached. Participants in the baseline and endline surveys were not the same. A local guide accompanied the survey teams to help identify pregnant women during the random walk process. The random walk method was conducted independently for each survey. The baseline and the endline surveys started in March 2014 and November 2014, respectively, and both continued until September 2015. Each participant was contacted twice during the baseline phase to enable the determination of GWG. During the first contact, the participants were visited at their homes, where written consent was obtained. Information on last menstrual period, parity, maternal age, marital status, education and other indicators of socioeconomic status was collected. KAP relating to diet, health, pregnancy and ANC were assessed using a structured interview. Height, weight, MUAC and fundal height were measured using anthropometric equipment calibrated daily and by multiple enumerators who participated in regular standardisation sessions. Lightly clothed pregnant women were weighed to 50 g precision (SECA 874) in duplicate. Women’s height (SECA 213, Seca, Hamburg, Germany), MUAC (ShorrTape Measuring Tape, Weigh and Measure, Olney, Maryland, USA) and symphysis-fundal height (SFH; ShorrTape Measuring Tape, Weigh and Measure, Olney, Maryland, USA) were measured in duplicate to 0.1 cm precision. If the two measurements were >0.2 kg (weight) or >0.5 cm apart (height, MUAC and SFH), a third measurement was taken and the mean of the two closest measurements was calculated. Each participating pregnant woman was invited to a gathering space such as an IHC or school, or the houses of a village chief for a follow-up assessment ~1 month later, where anthropometric measurements and the KAP interviews were repeated and weight gain over the elapsed period was calculated. Capillary blood samples were collected for on-site assessment of haemoglobin (Hb) concentration by HemoCue Hb 201+ (Hemocue, Lake Forest, California, USA). The survey questionnaires were developed in French and revised after pretesting. Interviews were conducted in Hausa (the predominant local language) or French. Gestational age was estimated as a weighted average of the following obtained information: reported last menstrual period (by estimated number of months, lunar cycles and/or proximity to a religious or cultural event), time elapsed since pregnant woman started to feel or perceive fetal movements, and two fundal height measurements taken ~1 month apart.20 21 Trimester of pregnancy was defined as follows: first trimester: <13 weeks, second trimester: ≥13 weeks to <27 weeks and third trimester: ≥27 weeks.20 Household socioeconomic status was estimated using three proxy indices (housing quality, household assets and household livestock holdings), as previously described.22 Household food insecurity, defined as the limited or uncertain access to adequate food of sufficient quality, was assessed using the household food insecurity access categories.23 Pregnant women who reported consuming at least 5 of 10 defined food groups in the previous 24 hours (using a list-based food frequency questionnaire) were considered to have met the minimum dietary diversity for women (MDD-W) requirement.24 To control for the effect of season on study outcomes, each month 4–10 villages in the catchment area of a new IHC were visited for the baseline survey to enrol new participants monthly for 12 months. Once all the follow-up data were collected in the villages, the respective IHC was eligible for the programmatic intervention as described below. The endline survey was identical in structure and data collection strategies to those of the baseline survey. After the initial 12-month baseline period, the baseline survey continued in an additional six IHCs. The surveys were temporally staggered to ensure that there was a baseline village to compare with an endline village at each time point and allow time-matched comparisons between intervention and control villages. The latter six control IHCs did not benefit from the programmatic intervention described below but received donations of essential medications at the end of the project. As soon as the baseline survey was completed in an IHC, the complete programmatic intervention was intended to start immediately in that IHC and its catchment area, and the endline survey was implemented 6 months after the completion of the baseline survey. Each month another IHC followed the same schedule (figure 1). However, due to various challenges encountered at different levels, the individual programmatic activities started with some delays (figure 2). The intervention consisted of three main strategies: (1) bimonthly behaviour change communication (BCC) activities in each participating village, (2) provision of essential drugs and supplies and (3) quality improvement (QI) activities at participating IHC, as previously described in more detail.16 The programmatic interventions were targeted at all pregnant women visiting 1 of the 12 participating intervention IHCs and/or residing in participating intervention villages. Implementation of the selected components of the programmatic intervention. □ Selected component of the intervention was implemented over 6 months. ∆ Selected component of the intervention was implemented <6 months due to delays. O Selected component of the intervention was not implemented due to delays. In each participating village, four community health workers (two men and two women) were identified in collaboration with the respective IHC to conduct BCC sessions. Selection criteria for these individuals included that they were already active volunteers in their village, that at least two among each group of four could read and write and that they were interested and available to conduct two BCC sessions per month. During a 3-day training, the community health workers were trained in BCC, focused on the importance of adequate nutrition during pregnancy, IFA supplementation, the overall importance of ANC for the mother and the fetus and of seeking ANC early in the first trimester of pregnancy, prevention of malaria, recognition of danger signs during pregnancy, and the need for a birth plan.16 25 26 Each community health worker received visual aids for use during the BCC activities. The BCC activities included discussion sessions and/or theatre performances with different target groups (men only, women only or both together) and the sites were opportunistic, and included gathering spaces such as the markets, mosques, schools and the houses of village chiefs. The applied theoretical framework of the BCC strategy is based on the recognition that individual behaviours are shaped by personal choices, social norms, environmental and economic constraints.27 The potential pathway of the BCC strategy was outlined in a programme impact pathway (online supplementary figure 1). bmjgh-2018-001200supp001.pdf At the beginning of the present project, a facility assessment conducted in 12 IHCs identified large gaps in the availability of essential commodities and supplies.16 Thus, one of the intervention strategies was to procure and distribute needed drugs and supplies, which included IFA supplements, impregnated bednets, deworming medication, health cards for mothers and infants, cotrimoxazole, sulfadoxine pyrimethamine, amoxicillin, magnesium sulfate, malaria rapid diagnostic kits, urinalysis rapid diagnostic multistix, thermometers, stethoscopes, blood pressure monitors, scales and stadiometers. These donations were distributed to the selected IHCs in collaboration with the leadership of the regional and local health districts. The implemented QI activities adopted a modified strategy of a previous quality assurance project conducted in Niger from 1993 to 1997, relying on formative supervision and coaching of key IHC personnel.28 Prior to the implementation of the programmatic intervention, a second programme impact pathway was developed for the health-centre level interventions (online supplementary figure 2). The present QI strategy was implemented in close collaboration with the regional and local health districts; one coach was identified in each of the two health districts and one at the regional level. All QI activities were implemented jointly by the project QI staff and the local QI coach. One training session was organised for health staff at participating IHC to explain the QI strategy and how to work together with the coaches towards improvements in ANC services. bmjgh-2018-001200supp002.pdf After the initial training, the QI strategy was rolled out in each participating IHC as follows: first the QI team visited each IHC to identify difficulties related to the ANC activities. Identified difficulties were discussed with the IHC health staff and strategies for improvement were developed jointly. After 2–3 weeks, the QI team revisited the IHC for an evaluation and encouraged each IHC to further develop best solutions to solve their specific difficulties, and meetings occurred on a monthly basis thereafter. Over the course of the project implementation, operational research at multiple levels was used to assess programmatic activities. For community-based BCC activities, the community health workers’ knowledge was assessed before and after the BCC training. In addition, a supervisory team used a structured form to monitor the content and the quality of each BCC session, which was used to provide constructive feedback to improve future sessions. At IHCs, operational research was used to monitor QI activities and the availability of essential drugs and supplies. Direct observations of ANC sessions were conducted followed by postobservation interviews with pregnant women. Health agents in IHCs were also interviewed to assess the health centre’s infrastructure including number of buildings, water and power supply, staff, and available drugs and supplies prior to, and over the course of, the project. The sample size of the present study was designed to be able to detect a difference of 10% in the prevalence of anaemia between the baseline and endline surveys. In the absence of information on the magnitude of the intracluster correlation, the design effect was arbitrarily set at 2 to account for the cluster sampling design. Assuming an initial anaemia prevalence of 50%, a significance level of 0.05, power of 0.80 and a design effect of 2, a sample size of 768 was needed, which was then inflated by 17% for attrition, yielding a target sample size of 925 pregnant women for both the baseline and endline survey. The baseline survey was extended concurrently with the endline survey for 6 months to serve as a pre-and-post and time-matched comparison and to account for the impact of the effect of the season of participants’ enrolment on study outcomes. Based on the same assumption as above, the additional sample size required for extending the baseline survey was estimated at 77 pregnant women per IHC (ie, per month) resulting in a total of 463 over 6 months. Thus, in total 2313 pregnant women were required for the whole study (1388 baseline, 925 endline). Patients and the public were not involved in the design and the conduct of this study. Data were double-entered using EpiData V.3.1. All statistical analyses were performed using the SAS System software V.9.4. Analyses were based on complete case intention-to-treat. Descriptive univariate statistics were calculated for all variables and variables not normally distributed (eg, GWG per week, MUAC and gestational age in weeks) were transformed to achieve a more normal distribution prior to analysis. Individual participant surveys (from each of the two time points—baseline and endline) were the unit of analysis. A detailed statistical analyses plan is available online.20 One primary outcome was attendance of an adequate number of ANC, defined as whether a woman reportedly participated in an adequate number of ANC visits at any health facility during her current pregnancy, based on her estimated gestational age. This was defined for the purposes of this study as: at least one reported ANC visit for women with a gestational age of 14–27 weeks, at least two reported ANC visits for women with a gestational age of 28–34 weeks, at least three reported ANC visits for women with a gestational age 35–39 weeks and at least four reported ANC visits for women with a gestational age of 40–42 weeks. Other primary outcomes were reported adherence to the recommended IFA supplementation (ie, a woman reported consuming IFA supplements daily in the previous 7 days if she reported receiving any IFA supplements during her current pregnancy); knowledge of the benefits of IFA (ie, the woman reported knowledge of an illness where it is said that pregnant women do not have enough blood (local definition of anaemia) and responded that IFA supplements were used to prevent/treat anaemia); and prevalence of anaemia (Hb <110 g/L in the first and third trimesters and Hb <105 g/L in the second trimester).4 Secondary outcomes were: mean GWG per week (maternal weight measured ~1 month apart and GWG calculated as the difference between measurements at visit 2 and visit 1 divided by number of elapsed days and multiplied by 7); mean MUAC and mean Hb concentration; prevalence of adequate GWG (defined as a GWG of 0.35–0.58 kg/week, using the US Institute of Medicine recommendations for underweight and normal weight women combined) and adequate MUAC (MUAC ≥23 cm).29 Because the BCC activities were implemented at the village level, the village was considered the cluster in the present study. Twenty-four of the 92 villages had small numbers of participants (ie, <5 women). To avoid unstable estimates due to small sample sizes per cluster, villages with <5 participants were combined with villages of >5 participants and which had similar characteristics such as belonging to the same IHC, and located approximatively at the same distance from the IHC. Consequently, the analyses included 68 clusters. The study was designed such that the baseline survey continued during the endline survey. A preintervention and postintervention analysis comparing the baseline cohort to the endline cohort was performed, and time of assessment was considered in the analysis by including year and season in all models as described in more detail below. The PROC GLIMMIX procedure in SAS was performed to estimate the adjusted OR (for binary outcomes) or the adjusted mean difference (for continuous outcomes). A p value <0.05 was considered statistically significant. In order to account for the cluster design of this quasi-experimental design, robust SEs were used in all models, using the EMPIRICAL=FIRORES option in PROC GLIMMIX, with random intercepts specified for each cluster via a RANDOM statement. The level of randomisation for the robust SEs is the village. Both minimally and fully adjusted models were used in the analysis. Minimally adjusted models included year, season and village to account for the effect of season on the study outcomes. Covariates in these analyses included age of the pregnant women as a binary variable (adolescent (≤19 years old) vs adult (>19 years old)), ethnicity (Hausa vs other minority ethnic groups), woman’s principal occupation (housewife vs non-housewife), marital status (married vs separated/divorced or widow), gravidity (primigravida vs multigravida), age at first pregnancy (as continuous variable), outcome of last pregnancy (living child vs child not born alive / born alive and had since died), attended any ANC during last pregnancy (yes vs no), attended at least four ANC during last pregnancy (yes vs no), health facility delivery during last pregnancy (yes vs no), household head’s education level (some primary education or higher vs no education / literacy training only / koranic school), principal occupation of the household head (farming-related occupation vs non-farming-related occupation), level of household food insecurity (food secure vs mildly / moderately / severely food insecure). Fully adjusted models included all the covariates in addition to year, season, village and a variable indicating if a participant was enrolled in the baseline or in the endline survey. We also examined whether specific demographic characteristics (eg, maternal age, education, occupation, housing quality, household assets, household livestock, reported danger signs during current pregnancy, last pregnancy outcome, quantity of food consumed per day compared with the time when the woman was not pregnant, number of meals per day, age of gestation) were effect modifiers to the impact on all the outcomes, using interaction terms.