Effectiveness of checklist-based box system intervention (CBBSI) versus routine care on improving postnatal care utilization in Northwest Ethiopia: a cluster randomized controlled trial

Background: The period after childbirth poses a substantial risk both to the mother and the newborn. Yet, this period received less attention as compared to the cares provided during pregnancy and childbirth. Hence, this study aimed to assess the effectiveness of checklist-based box system intervention on improving three postnatal care visit utilization. Methods: A double blind, parallel group, two-arm cluster randomized controlled trial design was used to assess effectiveness of checklist-based box system intervention on improving third postnatal care visit. Pregnant mothers below 16 weeks of gestation were recruited from 15 intervention and 15 control clusters, which were randomized using simple randomization. Data from baseline and end line surveys were collected using open data kit and analyzed using STATA version 15.0. The status of three postnatal care visit between intervention and control groups over time was assessed using difference in difference estimator. The predictors of the outcome variable were then analysed using mixed effects multilevel logistic regression model. Result: Of 1200 mothers considered from each of the baseline and end line studies, this study included data from 1162 and 1062 mothers at baseline and end line surveys, respectively. As it is shown from the difference-in-difference estimation (14.8%, 95%CI 5.4–24.2%, p = 0.002) and the final model (AOR 4.45, 95%CI 2.31–8.54), checklist-based box system intervention was effective on improving third postnatal care visit. In addition, institutional delivery (AOR 1.62, 95%CI 1.15–2.28) and knowledge on danger signs during postnatal period (AOR 5.20, 95%CI 3.71–7.29) were found to be significant predictors of the outcome variable. In the contrary, mothers who got influenced by older generations of individuals were (AOR 0.32, 95%CI 0.18–0.59) less likely to attend three postnatal care visit. Conclusions: The implementation of checklist-based box system intervention was found to be effective in improving utilization of the recommended three postnatal care visits. The contribution of the trial on improving third postnatal care visit can be enhanced by minimizing practical level challenges, as well as expanding health messages to reach unreached mothers and significant others who can influence the mother’s decision. Trial registration: ClinicalTrials.gov, NCT03891030, Retrospectively registered on 26 March, 2019, https://clinicaltrials.gov/ct2/show/NCT03891030.

The trial protocol for this study was published [16] and the trial was retrospectively registered on ClinicalTrials.gov, with trial identifier {“type”:”clinical-trial”,”attrs”:{“text”:”NCT03891030″,”term_id”:”NCT03891030″}}NCT03891030 on March 26, 2019. A double-blind, parallel-group, two-arm cluster randomized controlled trial was conducted to assess the effectiveness of checklist based box system interventions on improving utilization of three postnatal care visits. This trial was conducted in Debre markos, Gozamin and Machakel districts of East Gojjam Zone, one of the administrative zones of Amhara region, located in North-western Ethiopia. In Ethiopia, women commonly receive maternal health services from primary health care units (PHCU), which are composed of a health center and satellite health posts. Health centers commonly staffed with Midwives and Nurses, while health posts are staffed with health extension workers. This PHCU linkage provides preventive and curative services to 25,000 people altogether [17–19]. By the time the trial begins health centers were staffed with a minimum of 2 midwives and more than 50% of health posts were staffed by two health extension workers. The sample size calculation for this trial followed a recommendation for cluster randomized controlled trials with equal sized and fixed numbers of clusters [20]. Assumptions: three postnatal care visits in the control group was 16.0% [21], number of clusters available—30 (15 per arm), 95% confidence interval and 80% power, intra-cluster correlation coefficient of 0.04849 [22]. Then the sample size was calculated to determine number of observations per cluster using two sample comparisons of proportions using normal approximations and STATA 13 was used to run the calculation. Assuming individual randomization, the sample size per arm was 194, then allowing for cluster randomization, the average cluster size required was 40, and the final sample size was 1200 pregnant mothers (600 for intervention and 600 for control). Then 1200 mothers at the baseline and 1200 mothers during the end line survey were considered. The minimum detectable difference was calculated to be 12%. Debere-markos, Gozamin, and Machakel were chosen from among the sixteen districts available in Esat Gojjam Zone based on confirmation that none of the three districts had received an intervention/project aimed at improving maternal health service utilization. This trial has both community and facility level intervention. As the community level intervention would result information contamination, health posts/kebele under selected districts were used as a randomization units. After the inclusion of health posts/clusters, the study team used SPSS generated random sequence to assign the available 30 clusters to intervention and control arms in a 1:1 allocation ratio. In this study, identification and recruitment of health posts/clusters and identification of mothers for the baseline survey were done before the randomization procedure took place, which helps in minimizing identification bias. Mothers who were pregnant and below 16 weeks of gestation, who were living in the selected districts were identified as participants for this trial. Mothers with severe clinical complications who required special ANC follow-up, on the other hand, were excluded from the study. Mothers in the intervention group passed a two stage screening process to be enrolled in the study. First, community level pregnancy screening was done using Stanback et al. [23] pregnancy screening checklist, this community level screening identified suspected pregnant mothers. Then suspected pregnant mothers received a referral slip to be received by the nearby health center, for confirmation of pregnancy. Second, facility level screening took place at health centers using beta-human chorionic gonadotropin (HCG) urine test. Mothers who had confirmed pregnancy using the HCG test were enrolled in the study and received the first ANC on the same day. This community level survey, which was done by health extension workers (HEWs), was supported by a family folder: a documentation containing family profile of the residents at health post. Unlike to the intervention group, health centers under the control group recruited self-referred pregnant mothers below 16 weeks of gestation and provide the first ANC on the same day. Study participants (mothers) and outcome assessors were blinded to the intervention. Despite receiving all the listed packages of the intervention, mothers were unaware that they are in a different treatment group and receiving different intervention. The study team used the advantage of cluster randomization, where clusters rather than individuals are randomized. In this type of randomization, mothers who lived in the same cluster received the same intervention. Similarly, mothers who were under the intervention cluster of this study and were eligible for the service received CBBSI, which was common to all. In addition, health extension workers who were already familiar with the study’s catchment population delivered the intervention. Outcome assessors, data collectors, were blinded to the intervention. This study used a separate group of health care providers to provide the intervention and to collect the data. Health extension workers and health care providers who provided the intervention were involved in community level referral, health education sessions, managing referral slips, tracing dropouts, and providing a list of packages to be implemented in the intervention. This group of professionals was made aware of the intervention. However, a different group of health care workers meets mothers at their homes to collect data on their experiences during pregnancy, delivery, and the postnatal period for their index pregnancy. This group of health care providers who collected the data were unaware of, which group is receiving the intervention and which group is not. Furthermore, they are unaware of the reason for the intervention and the goals it is intended to achieve. Checklist based box system intervention introduced a two pronged approach: community level demand creation and service utilization monitoring components. The trial was started with the work of HEWs, though community level identification of pregnant mothers. Mothers who had suspected pregnancy were linked to the nearby health center with a referral slip. On the same visit mothers were asked about their reasons of not attending maternal health services, and again to prioritize the top three reasons they had. These lists of reasons were documented using the reason picking card and placed at health posts’ health education scheduling box. Health extension workers provided individual health education based on local evidences: based on the mothers own reason for not utilizing maternal health services. Besides the person centered health educations, HEWs placed reason picking cards at health education scheduling boxes placed at health posts. These numbers of reason picking cards placed in each compartment determined the topic for health education that was delivered by HEWs in different platforms including pregnant mothers’ conferences. At the receiving health center, mothers who have confirmed pregnancy received the first ANC care on the same day. Then mothers were expected to smoothly pass through the service utilization monitoring box placed at health centers: from ANC 1 to PNC 3. Mothers who fail to follow this smooth transition were identified by health care providers at health center and respective HEWs where the dropped mother belongs were communicated. Timing of postnatal visits followed the a recommendation from WHO and the nationally implemented three visit model: within 2 days of delivery (PNC 1), within 7 days of delivery (PNC 2) and within 42 days of delivery (PNC 3). The published protocol of this study includes a detailed description of the intervention [16]. Mothers in the control arm received the existing routine care provided by the government. Accordingly, clusters in the trial arm II didn’t have community level pregnancy screening, health education scheduling box and person centered health education for health posts and service utilization monitoring box for health centers. The intervention followed previously developed intervention package, where each and every step of the trial with the necessary documentation was provided with. The intervention was supposed to be implemented in 10 months, but it took 20 months from the date the first mother with a confirmed pregnancy enrolled in the study and received the first ANC at the health center on January 3rd, 2019, to the end of follow-up for the last rounds of mothers on August 27th, 2020. This extended implementation period was following the practical level challenges on the ground, which are discussed further below. For a significant number of the cases, the family folder (used to guide community level pregnancy screenings) wasn’t updated, staff turnover and trained staff turnover at health posts and health centers was also a common problem. As it is clearly indicated in the protocol, for the purpose of follow-up of the identified suspected pregnant mothers, referral slip was filled in two copies and the first was given to the mother and the second was remained with the visiting HEW. Then these slips reached to health centers, where the mothers was referred using the weekly meeting platform between health posts and health centers. This weekly meeting went irregular and together with the above challenges the planned time for the trial was delayed. The questionnaire used to collect data was designed incorporating different steps of the PRECEED-PROCED model as a base [15, 24]. The PRECEDE part focuses on the assessment of desired health outcome from different perspectives in four phases. Phase 1: Social assessment to identify desired outcome, Phase 2: Epidemiological assessment to set priority outcomes and assess determinants that stand in the way of achieving postnatal care three visit, Phase 3: Ecological assessments to identify predisposing, enabling and reinforcing factors that affect behaviour of a mother in receiving postnatal care and Phase 4: Administrative and policy assessment to identify administrative and policy issues that can influence the implementation of CBBSI. The PROCEED part also follows four phases: Phase 5: Design and implementation of CBBSI, Phase 6: Process evaluation to ensure the trial is implemented as planned, Phase 7: Impact evaluation to evaluate whether the intervention has contributed for the improvement of postnatal care utilization and Phase 8: evaluation to ensure if CBBSI is leading to the improvement in postnatal care utilization [25]. Then the tool was translated to the local language version of the study area. The study tool was pretested on mothers living outside of the chosen study area prior to data collection. Mothers who had their last child 1 year or less ago were chosen to take the pretest. A pretest assists the research team in re-examining and correcting the study tools (selected questions) for clarity, simplicity, and ordering. Then, the final tool was uploaded to kobo tool box and open data kit (ODK) was used to conduct face to face interview with selected mothers both for the baseline and end line study. The primary goal of this trial was to assess the effectiveness of a checklist-based box system intervention in increasing maternal health care utilization (ANC, skilled delivery and postnatal care). The first two primary outcomes had already been reported and were planned to be published. The findings of this study focused on reporting the effectiveness of the intervention on three postnatal care utilization. Accordingly, attending three postnatal care visits was the outcome variable for this study. Mothers were asked about the number of postnatal care visits or check-ups they had for their index delivery. In this study, Kebele was identified as a level variable and the independent variables were categorized as level 1 (individual level variables) and level 2 (community and facility level variables) variables (Table ​(Table11). Description of study variables, East Gojjam Zone, Northwest Ethiopia, January 2019–September 2020 Before the actual field data collection training for a team of data collectors and supervisors took place, a manual explaining each and every question with its response categories and on how to work on ODK was prepared and provided for data collectors (BSC holder midwifes) and supervisors (MPH holder). In addition, the use of open data kit for data collection helped in settling questions as required (to avoid unanswered questions), in settling range checks for selected data values and allowed to deal with field editing before leaving the respondent. Data collection was managed centrally from Jimma University where, all study data bases were secured with password-protected access system. After the field data collection was over, data were exported from kobo-tool box and imported to STATA MP Version 15 for analysis. Data from the trial followed an intention to treat analysis: participants were assigned to the cluster where they were resident when the trial begins. Allocation of clusters to the intervention and control arms followed simple randomization techniques, following this the treatment effect was analysed using the Difference in Difference (DiD) estimation [26]. Similarly, the status of utilization of three postnatal care visits between the intervention and control arms was tested using chi-square test for the significance of the association. Then to identify factors affecting utilization of three postnatal care visits, first bivariate analysis was conducted to test the association between each independent variable and the outcome. Then, variables having p < 0.25 were included in the multivariable model. The analysis employed multilevel mixed effect logistic regression; this model was selected because the participants of this study were nested in kebeles, which violates the assumption of independence of the ordinary logistic regression. Multilevel analysis allows the simultaneous examination of the effects of group level and individual level variables on individual level outcomes while accounting for the non-independence of observations within groups. Also, multilevel models took into consideration the dependency nature of individual probability in the areas of residence where the participants belong. This dependence on the context was accounted to obtain correct regression estimates. First, an empty model was fitted to check for the presence of cluster level variability affecting the three postnatal care utilization and to measure the proportion of total variance that is attributable to cluster level, the intra-cluster correlation coefficient was calculated [27]. Additional measures of variation, median odds ratio (MOR) to measure unexplained cluster heterogeneity [27–31] and proportional change in variance (PCV) [32] were calculated. Models fitness for the multilevel was checked using the log likelihood ratio (LR) test. Four models: an empty model: to evaluate the extent of cluster variation affecting three postnatal care utilization, second model: controlled for individual level variables, third model: controlled for community level variables and the fourth model controlled for both individual and community level variables were constructed. Variables with p value of < 0.05 in the second and third model were included in the final model. The p value of < 0.05 was used to define statistical significance, AOR together with 95% CI were used to show the strength of association and level of significance respectively.

N/A

• Dima Health