Effect of a participatory multisectoral maternal and newborn intervention on birth preparedness and knowledge of maternal and newborn danger signs among women in Eastern Uganda: A quasiexperiment study

Background: Knowledge of obstetric danger signs and adequate birth preparedness (BP) are critical for improving maternal services utilization. Objectives: This study assessed the effect of a participatory multi-sectoral maternal and newborn intervention on BP and knowledge of obstetric danger signs among women in Eastern Uganda. Methods: The Maternal and Neonatal Implementation for Equitable Systems (MANIFEST) study was implemented in three districts from 2013 to 2015 using a quasi-experimental pre-post comparison design. Data were collected from women who delivered in the last 12 months. Difference-in-differences (DiD) and generalized linear modelling analysis were used to assess the effect of the intervention on BP practices and knowledge of obstetric danger signs. Results: The overall BP practices increased after the intervention (DiD = 4.7 p < 0.05). The increase was significant in both intervention and comparison areas (7-39% vs. 7-36%, respectively) with a slightly higher increase in the intervention area. Individual savings, group savings and identification of a transporter increased in both intervention and comparison area (7-69% vs. 10-64% 0-11% vs. 0-5% and 9-14% vs. 9-13% respectively). The intervention significantly increased the knowledge of at least three obstetric danger signs (DiD = 31%) and knowledge of at least two newborn danger signs (DiD = 21%). Having knowledge of at least three BP components and attending community dialogue meetings increased the odds of BP practices and obstetric danger signs' knowledge respectively. Village health teams' home visits intervention area residence and being in the 25+ age group increased the odds of both BP practices and obstetric danger signs' knowledge. Conclusions: The intervention resulted in a modest increase in BP practices and knowledge of obstetric danger signs. Multiple strategies targeting women in particular the adolescent group are needed to promote behavior change for improved BP and knowledge of obstetric danger signs. The study was conducted in Pallisa, Kamuli, and Kibuku districts in Eastern Uganda. The estimated population in this area was 1,075,242 [18]. The three districts had a total of 104 health facilities, 33 in Pallisa, 17 in Kibuku, and 54 in Kamuli [19]. The study employed a quasi-experimental pre–post comparison design. The study was implemented at health sub-district level. A health sub-district (HSD) is an adminsitrative structure within the decentralized health system in Uganda [20]. It consists of a cluster of health facilities of varying levels of care, usually headed at the highest level of care, which could be a health center IV or a hospital. The intervention arm comprised three HSDs (Pallisa, Kibuku, and Buzaya), while the comparsion arm had two HSDs (Bugabula and Butebo). The intervention and comparison areas were selected purposively in consultation with the district leaders during the formative stage of project (January–December 2012) [20]. The MANIFEST project had two main components: (1) community mobilization and empowerment to stimulate demand for services, and (2) health provider and management capacity building to strengthen the delivery of quality maternal and newborn health services. The community mobilization and empowerment component strategies aimed to increase awareness about birth preparedness and to increase access to household financing for maternal health and access to transport. Such strategies included: (1) the use of community health workers, also referred to as village health teams (VHTs), to do home visits; (2) radio spot messages, talk shows, and community dialogues; (3) promotion of savings groups and other saving methods; and (4) linking of local transporters with saving groups. VHTs were responsible for conducting two home visits during pregnancy and one visit after delivery, and counseling mothers on essential maternal and newborn care practices, safe delivery, and birth preparedness. They identified women and children with danger signs, and those identified were referred to the health facility for further screening and care. VHTs were also supposed to encourage women to save money either in savings groups or as individuals. Saving groups are local financial clubs where individuals meet regularly with the goal of saving money together. This money is then used for various reasons, which include health and non-health needs. The health-related needs include transport to health facilities, purchase of birth items, or medication, among others. In addition, some of the groups engage in income-generating activities. The VHTs were responsible for conducting community dialogue meetings quarterly. The radio spots were aired three times a day, while the talk shows were held once every month. The capacity-building component strategies were emergency obstetric and newborn care refresher training, mentorship and support supervision of primary health workers, training in health services management for health managers, and recognition of best performing facilities and managers. The implementation of the study was led by the district and sub-county officials and supported by the Makerere University research team. Details of the study implementation are outlined in the study protocol in this special issue [20]. In this paper, a variable on intervention setting was included, which consisted of two categories of participants in the intervention and comparison areas. VHT home visits and community dialogue meeting attendance variables were also included, which described if the respondents were visited by a VHT and attended community dialogue meetings while pregnant, respectively. In addition, several socio-demographic characteristics were assessed, including age, religion, education level, occupation, marital status, and wealth index. The wealth index was derived through principal component analysis of household assets and housing material. The first principal component was used to generate the wealth quintile scores. The outcomes assessed were birth preparedness practices and knowledge of maternal and newborn danger signs. Overall, the respondent was considered to have prepared for birth if she practiced at least three birth preparedness components. The birth preparedness practices included buying birth items, identifying a health provider and a transporter to facilitate access to delivery services, and saving money individually or through group savings for maternal health. The respondent was considered knowledgeable about birth preparedness practices if she mentioned at least three birth preparedness components. A woman was considered knowledgeable of maternal and newborn danger signs if she could mention at least three pregnancy and labor danger signs and at least two newborn danger signs. The study sample size was determined using an individual randomized trial sample calculation formula [21], with 80% statistical power, a 5% significance level, and 1.5 design effect. The proportion of women who delivered in a health facility was used to calculate the sample size. The assumption was that after 3 years (2013–2015) of implementation, skilled deliveries in the intervention arms would increase from 38% to 58% in Kibuku, from 62% to 72% in Pallisa, and from 68% to 78% in Kamuli. These targets were informed by the Uganda Newborn study that was implemented in an area with a context similar to the current study [22]. Calculation of sample size based on all these assumptions yielded a sample size of 2,293 women. Two-stage sampling was applied per district for each of the study areas. It was estimated that 119 villages would be sufficient for the sample size to be achieved. Hence, 119 villages (52 in Kamuli, 46 in Pallisa, and 21 in Kibuku) were selected out of a total of 847 villages (257 in Kamuli, 346 in Pallisa, and 244 in Kibuku) using probability proportionate to size sampling techniques. Thereafter, all households in each district were listed to identify eligible study participants. During listing, 3,456 and 3,199 women who delivered in the 12 months preceding the baseline and end line, respectively, were identified. Women whose pregnancies were terminated before 20 weeks and non-resident women who had not stayed in the community for at least 1 year were excluded from the study. In addition, women with severe illnesses at the time of the survey and those who refused to consent were also excluded. As a result, 2,237 (1,101 in the comparison group and 1,136 in the intervention group) and 1,946 (920 in the comparison group and 1,026 in the intervention group) women were identified as eligible and were interviewed at baseline and end line. Additional details about the sample size calculation and participant selection can be found in the study protocol in this special issue [23]. Twenty-four research assistants (RAs) with tertiary education, who had previous experience in data collection and were fluent in local languages spoken in the study area, were recruited and trained. They formed two teams, each with 12 RAs, an editor, and a field supervisor. The RAs conducted face-to-face interviews with eligible women using a structured questionnaire. The questionnaire collected information on participants’ socio-demographic characteristics, health facility utilization, birth preparedness, and newborn care practices. The questionnaire was translated into local languages (Lusoga in Kamuli, Ateso in Pallisa, and Lugwere in Kibuku). Prior to data collection, the questionnaires were pretested in Wakiso district to check for any flaws and to increase the research team’s familiarity with the questions before the actual data collection. After the pretest, some questions were revised for better understanding and ease of administration while keeping their intended meaning. A data management manual detailing data collection, storage, and entry procedures was developed. During data collection, the data editors checked the questionnaires for any errors and made necessary corrections while in the field. Each supervisor sampled and re-interviewed randomly selected respondents each day in order to check the consistency of the information collected. Lastly, an independent quality control team also visited the team to ensure that the data collection was proceeding as planned in the data collection manual. The data were entered using Epi info 7 software. Ten percent of the questionnaires were double entered in order to check the consistency of the data entered. The Epi info database was backed up, and the data were transferred to Stata v13 (StataCorp LP, College Station, TX) for statistical analysis. To understand the project’s counterfactual, difference-in-differences (DiD) analysis was used [24] (see Equation 1), controlling for individual background characteristics, which included age, education level, occupation, parity, and religion. Yit are the study outcomes. The variables T and t are treatment and time parameters, respectively. T and t were dummy variables: 1 = treatment group and 0 = non-treatment group, and 0 = before intervention and 1 = after intervention, respectively. β3 is the difference-in-differences estimator, which indicates whether the expected mean change in outcome before the intervention and after the intervention were different in the intervention and control group. xit are covariates such as age, education, parity, and occupation, and χi represents covariates’ estimators; μit is an error term. A significant coefficient of the interaction term implies that the outcomes differed by groups over time. Propensity score matching was also used, where individual socio-demographic characteristics (age, education level, occupation level, and religion) in the treatment group were matched with those in the comparison group using a nearest-neighbor matching method. The generalized linear model (GLM) with binomial logit link function was also used to assess the predictors of birth preparedness and knowledge of obstetric danger signs. In each model, an interaction of study area and VHTs visits was introduced in order to assess whether VHT home visits increased the level of knowledge about maternal and newborn danger signs in the intervention compared to control areas. Bivariate analysis was performed using ulogit command in Stata to assess the likelihood of variables affecting the study outcome. Variables whose p-value was ≤25% were considered for multivariate analysis. The collinearity was also assessed using the collin command in Stata, and the presence of multicollinearity was considered for variables whose variance inflation factor (VIF) value was >2.