Improving maternal and newborn health services in Northeast Nigeria through a government-led partnership of stakeholders: A quasi-experimental study

Objectives This study aimed to quantify change in the coverage, quality and equity of essential maternal and newborn healthcare interventions in Gombe state, Northeast Nigeria, following a four year, government-led, maternal and newborn health intervention. Design Quasi-experimental plausibility study. Repeat cross-sectional household and linked health facility surveys were implemented in intervention and comparison areas. Setting Gombe state, Northeast Nigeria. Participants Each household survey included a sample of 1000 women aged 13-49 years with a live birth in the previous 12 months. Health facility surveys comprised a readiness assessment and birth attendant interview. Interventions Between 2016-2019 a complex package of evidence-based interventions was implemented to increase access, use and quality of maternal and newborn healthcare, spanning the six WHO health system building blocks. Outcome measures Eighteen indicators of maternal and newborn healthcare. Results Between 2016 and 2019, the coverage of all indicators improved in intervention areas, with the exception of postnatal and postpartum contacts, which remained below 15%. Greater improvements were observed in intervention than comparison areas for eight indicators, including coverage of at least one antenatal visit (71% (95% CI 62 to 68) to 88% (95% CI 82 to 93)), at least four antenatal visits (46% (95% CI 39 to 53) to 69% (95% CI 60 to 75)), facility birth (48% (95% CI 37 to 59) to 64% (95% CI 54 to 73)), administration of uterotonics (44% (95% CI 34 to 54) to 59% (95% CI 50 to 67)), delayed newborn bathing (44% (95% CI 36 to 52) to 62% (95% CI 52 to 71)) and clean cord care (42% (95% CI 34 to 49) to 73% (95% CI 66 to 79)). Wide-spread inequities persisted however; only at least one antenatal visit saw pro-poor improvement. Conclusions This intervention achieved improvements in life-saving behaviours for mothers and newborns, demonstrating that multipartner action, coordinated through government leadership, can shift the needle in the right direction, even in resource-constrained settings. This study was conducted in Gombe state, a predominately rural (80%) and sparsely-populated state in Northeast Nigeria,25 where the burden of maternal and neonatal mortality is higher than the national average and stood at 1549 maternal deaths per 100 000 live births in 2015 and 33 neonatal deaths per 1000 live births in 2017.26 27 In 2016, when this study started, 29% of women gave birth in a health facility,25 principally at primary health centres (PHC).28 29 Primary healthcare services are predominately delivered by the Gombe State Primary Health Care Development Agency, with little private sector provision, unlike some other regions of Nigeria.30 In 2016, 460 PHCs in the state provided antenatal care (ANC), birth and intrapartum services,28 mainly delivered by community health extension workers (CHEWs), junior CHEWs and community health officers, but very few nurses, midwives or doctors.31 From 2016 to 2019, the government led a maternal and newborn health partnership to improve access, use and quality of maternal and newborn health services. Within this partnership, non-governmental organisations (NGOs) implemented a package of evidence-based interventions that spanned the six WHO health system building blocks32 (figure 1). Components aimed to enhance uptake and provision of life-saving interventions at three interacting levels (individuals and families; community organisations; and the health system). At the individual and family level, interventions aimed to improve knowledge, attitudes and practices to increase enhanced home-based practices and increase demand for routine professional care; for example, a community-based village health worker home visit scheme was initiated to improve knowledge about and linkages between families and health services.33 At the community organisation level, interventions aimed to improve trust and accountability between the family and health system levels; for example, through supporting community-based mothers groups to interact with their local primary health services.34 Interventions at the health system level aimed to improve the supply of safe, effective and high quality care; for example, working with government to strengthen the supply chain for essential drugs in PHCs. Underpinning these three levels of engagement were interventions designed to raise public awareness about maternal and newborn health across the state through mass media and advocacy events.35 Intervention components by health system strengthening building block. CHEW, community health extension worker; HMIS, Health Management Information System; HSS, health system strengthening; MNH, maternal and newborn health; MPDSR, Maternal Perinatal Death Surveillance and Response; PHC, primary health centre; VHW, village health worker. This package of interventions was deliberately coordinated by government as a pathway towards improved maternal and newborn outcomes.36 The government, NGOs, partners and the funder met every six months to review monitoring data, trouble-shoot implementation challenges, course-correct and reinvigorate communal purpose towards a shared goal. To facilitate learning, the package of interventions was initially implemented in an intervention area, with a view to scaling-up to the entire state. The intervention area was defined as 57 subdistrict level wards (half of the state’s 114 wards), purposively identified by government. Community-based demand generation activities were implemented in these wards, and one centrally located PHC within each ward was chosen to implement the activities designed to improve health service quality: the rationale being that it was preferable to have one well-functioning PHC per ward, rather than a larger number of less well-functioning facilities. Residents of the state’s remaining 57 wards (comparison area) continued to receive their usual care, with the exception of mass media components which were statewide (figure 1). We used a quasi-experimental plausibility study design37 to explore the association between the intervention and indicators of use and quality of maternal and newborn health services, comparing changes observed over time in intervention areas to those in comparison areas. Repeat cross-sectional household and linked health facility surveys were undertaken. Survey methods were replicated each time. Data collection tools were informed by existing large scale survey tools such as the Demographic and Health Surveys38 and Service Availability and Readiness Assessment Surveys.39 We conducted four annual household surveys during July/August 2016–2019. The household survey consisted of a modular household questionnaire: (1) A household module which asked about characteristics of the household and ownership of commodities as proxy markers of household socioeconomic status, and during which a household roster of all usually resident people was generated; (2) A women’s module which asked all resident women about the healthcare available to them, their recent contact with health services and their recent birth history; and (3) A mother’s module which asked all women who reported a birth in the last 12 months about their contact with health services across the continuum of care from pregnancy to postnatal care. In a household with a resident recently delivered woman the questionnaire took approximately 90 minutes to complete. A random sample of 80 clusters was selected: 40 each from intervention and comparison wards. Clusters were segmented enumeration areas as defined by the National Population Commission. Within each cluster, all households were visited. During each survey, we aimed to survey a total of 6000 households across the 80 clusters. This was expected to result in interviews with 1000 women with a live birth in the previous 12 months (ie, 500 each in intervention and comparison). Sample size calculations assumed a design effect of 2.5, 95% probability and 80% power. For each survey, where comparison prevalence of indicators ranged between 20% and 60%, this sample size was sufficient to detect changes of 15 percentage points in intervention areas. During the same period, we conducted seven facility and birth attendant surveys at six monthly intervals. Four surveys were done concurrently with the household surveys (figure 2). At each time point, a facility readiness assessment was carried out plus an interview was conducted with the available birth attendant who had attended the most recent delivery recorded in the maternity register. The facility survey took approximately 120 minutes to complete all sections. All 57 intervention PHCs were surveyed plus one PHC selected at random from each of the 40 comparison clusters sampled during household surveys. For each survey, where prevalence of indicators in comparison facilities ranged between 5% and 70%, this sample of 97 facilities was sufficient to detect changes of 20 percentage points in intervention areas. Study timeline. MNH, maternal and newborn health. All recruited interviewers were from Gombe state and attended a one week training course at each survey. Eight survey teams were recruited in total including a supervisor, four household interviewers, one facility and birth attendant interviewer, one mapper who listed households and segmented enumeration areas as necessary and one data support member. Questionnaires were translated and back-translated between English and Hausa languages to ensure consistency and were pretested. As part of the week long training, the full study protocol was pilot tested in two clusters to identify and correct any operational or language problems. All data were collected using hand-held digital devices and synchronised with a supervisor laptop each day. Automated summary reports were produced to identify and address any internal inconsistencies. Eighteen indicators measuring the coverage of life-saving commodities and behaviours were selected a priori for their known association with maternal and newborn health outcomes (table 1). Changes over the four year study period were compared between intervention and comparison areas. Analyses were at the individual level for household survey indicators and at the health facility level for facility and birth attendant indicators. Data were analysed in Stata V.15 (StataCorp, 2017). Indicator definitions Data were collected for women aged 13–49 years. However, no women <15 years of age had a live birth in the 12 months before the survey. ANC, antenatal care. Household and health facility characteristics were summarised using appropriate summary statistics. We used survey commands (svy) to account for clustering. We identified variables which differed significantly between intervention and comparison areas in 2016 using a design-based F test and included their cluster-level means in regression models. Percentage point differences and 95% confidence intervals (CIs) in indicator coverage between 2016 and 2019 were calculated. Effectiveness of the programme on pre-specified indicators was estimated through logistic regression models. A likelihood ratio test comparing models with time as a continuous variable to one where each time point was included separately, determined whether time was included as a continuous or categorical variable. Models included fixed effects for area (intervention vs comparison), time (at all time points) and the interaction between area and time, to describe any additional effect in 2019 in the intervention areas compared with 2016 in the comparison areas. Cluster level means of variables which differed between intervention and comparison areas in 2016, identified using a design-based F-test, were included in all models. We included a random effect for cluster to account for clustering in logistic regression models of household survey indicators. To examine equity, principal components analysis was used to generate an index of household wealth, based on asset ownership.36 Using this, households were categorised into quintiles from poorest to least poor. Where there was evidence of greater improvement in the intervention areas compared with the comparison areas, we examined the difference of change over time by household socioeconomic quintile in the intervention areas. We tested for interaction between time and socioeconomic status quintile. Patients and the public were not involved in the design, conduct, reporting or dissemination plans of our research.