Study site
The study was conducted in Koibatek, one of six sub-Counties in Baringo County, Kenya. The sub-County has an approximate total population of 125,637 with 30,203 women of reproductive age (15–49) and 4799 children under 1 year. The EBF rate for under 6 months in 2015 was 37%, below the national rates of 61% [40].
The prevalence of HIV infection among pregnant women in Baringo County in 2015 was estimated to be 3% with a total of 313 mothers living with HIV [41]. The County was ranked 25 out of 47 counties in Kenya in terms of contribution to new HIV infection. In addition, it ranked 21 out of 47 counties in Kenya on HIV prevalence and having small improvement in HIV infection reduction of less than < 50% as compared to 14 counties which made progress of > 50% HIV reduction in the same year [42, 43].
Study design
The study was part of a bigger community-based cluster randomized trial implemented between October 2014 and July 2017 to test the feasibility and effectiveness of BFCI in rural Kenya in order to inform scale up [44]. Training of healthcare providers and community health volunteers (CHVs) and baseline data collection were conducted January to March 2015, while prospective data collection was carried out from April 2015 to December 2016. Qualitative formative data were collected in August 2014 and the findings used to inform the intervention design.
Cluster randomization was used as opposed to individual randomization as this would minimize ‘contamination’ of information across the study groups. The BFCI is also inherently a community-based program including community and mother support groups meaning that the intervention needed to be allocated at the level at which it would be implemented.
Community units (CUs), established under the MOH community health strategy, were the unit of clusters. Each CU has a population of approximately 5000 and is served by 25 CHVs. There are several villages within one CU and each CHV is allocated between 100 and 150 households depending on the distance and geographical location [45]. The 13 clusters were randomly allocated into the BFCI intervention arm (n = 6) and control arms (n = 7) by a biostatistician with no information on the CUs based on a ratio of 1:1 using Microsoft Excel. The additional extra CU for the control was also allocated randomly.
Sample size determination
We designed the study to be able to detect an 18% increase in the prevalence of EBF, based on effects observed in prior similar studies testing effectiveness of community based interventions on EBF [46]. With a two-sided 5% significance level and a power of 80%, a sample size of 780 mother and child pairs in total was required. An additional 10% at the intervention and control was considered respectively to account for loss to follow-up. A design effect of 4 based on an intra-cluster correlation of 0.05, and previous research on breastfeeding practices in some settings in Kenya and an average cluster size of 62.5 was used [44]. The target was to recruit 400 mothers in each of the study arms. Nine Hundred and one (901) women were ultimately recruited out of the targeted 800. Twelve CUs were required for the estimated sample size. The 13th CU was randomly assigned to the control arm.
Study eligibility
Any pregnant woman (15–49 years) in their first or second trimester residing in the 13 CUs who intended to stay in the CUs for 6 months or more after delivery was potentially eligible to take part in the study. Potential participants were identified by the CHVs at the community or through the facility by the community health extension worker (CHEWs). They were then referred to the data collector for eligibility screening, consent, and baseline interview.
Blinding
The investigator and the CHVs were aware of the interventions given to the experimental groups. The hypothesis of the study could not be concealed from the CHVs since they were offering education and counselling to mothers. Field workers knew about control and intervention groups but were independent of the intervention delivery. The mothers were not told of the different treatment groups.
Training and mentorship of study staff
A five-day BFCI training was conducted by six master trainers who were MOH employees and accredited trainer of trainers (TOTs). A total of 25 participants composed of the sub-County health management team (SCHMT), health workers from the sub-County, and two CHEWs from each of the 6 intervention units were trained. The training content was based on the BFCI implementation package for Kenya that had been developed by the MOH, adopted from the WHO/UNICEF integrated infant and young child feeding counselling course [34, 47]. The key messages for counselling were covered in each of the 8 steps for BFCI, including infant feeding in the context of HIV [48]. Other key messages in the training that were revealed to be important through formative work for this study and elsewhere [21] included male involvement, and the common tradition of giving prelacteal feeds.
The CHVs were trained by the CHEWS under the supervision of master trainers. The CHVs are volunteers providing community health services as established under the community health initiatives [49]. They conduct home visits and other community-based activities e.g. community dialogues, action days and also collect community-based data for health information systems. A total of 25 CHVs from each of the 6 intervention CUs underwent the training conducted at the community level either in a nearby church or a school.
The CHVs were also trained on their roles and responsibilities in the study including counselling skills, establishment and conducting of monthly mother-to- mother support group (MTMSG) meetings and nutrition education to mothers. Thirteen field research assistants and one supervisor also underwent a seven-day training including field experiences on data collection. The questionnaires were subjected to multiple pre-testing during their development and finally pilot tested and refined for clarity and accuracy.
Upon completion of the training, the SCHMT, CHEWS and CHVs developed a joint work plan for rollout of the BFCI. The rollout included: formation and facilitation of community mother support groups (CMSGs) targeting pregnant and lactating mothers including influencers, provision of counselling and information and education materials, development of a framework for mentorship and supportive supervision, promotion, protection and support of breastfeeding.
The paper presents findings from a community-based intervention in which trained CHVs were used to offer community-based intervention on MIYCN to pregnant mothers from their first or second trimester to 6 months post-delivery. We describe the outcomes in terms of EBF rates for those in the intervention and control groups, stratified by HIV status of the mother, which was ascertained after mothers were recruited to either the intervention or the control groups. We also report on knowledge and attitudes regarding EBF for HIV-exposed infants. The primary goal of the study was to generate evidence that would be used to change breastfeeding practices for HIV negative and HIV positive mothers, and improve knowledge for community support towards elimination of mother to child transmission (eMTCT) in addition to informing policy.
Activities in the BFCI intervention group
The mothers in the BFCI intervention group received a minimum of 12 home-based counselling sessions by trained CHVs. The CHV visited the pregnant woman after she had enrolled and completed the baseline survey. Visits occurred approximately once every month up to week 37 of pregnancy, and thereafter every 2 weeks through the first month postpartum, then once a month until the infant was 6 months old. The times for visits were scheduled between the CHVs and the mother at her convenience.
At these visits, the CHVs used counselling cards with the content structured around the eight-step point plan for BFCI which included good maternal nutrition and early antenatal care (ANC), importance of EBF including early initiation, feeding of colostrum, attachment and positioning, prevention of mother to child transmission (PMTCT), solving breastfeeding difficulties and family support. The CHVs also had a counselling checklist to ensure all areas were covered depending on the stage of counselling.
Other activities included weekly scheduled education sessions on infant and young child feeding at the maternal child health clinic which targeted mothers who attended the clinic that particular day within a CU catchment area, monthly MTMSG meetings, and bi-monthly baby friendly community gatherings targeting all influencers which were conducted by established community mother support groups members. Breastfeeding corners/rooms were also established at the primary health care facility where mothers would meet and share experiences when they came for ANC clinic. Simplified information, pictures, illustrations and messages translating the policy into local languages were placed in the breastfeeding corners and mothers were mostly taught by CHEWs or CHVs in these corners.
Ongoing mentorship and supportive supervision by the SCHMT to the CHEWS and CHVs took place every month in the first 6 months of study commencement and thereafter quarterly. Based on outcome of the mentorship and supportive supervision including gaps identified, actions were agreed upon by both the mentee and the mentor and review of the implementation followed up in the subsequent visits.
Control group
The control group received standard health education at the facility. Topics on MIYCN were part of other topics given within the schedule. As such, a woman could entirely miss receiving any breastfeeding information during her ANC clinic visits if during her attendance such topics were not in the schedule for that day. They also received standard counselling delivered by CHVs not trained on MIYCN at home during the CHV routine monthly visits. The routine visits were conducted once a month for the pregnant or lactating mother. Routine data were also collected during the visits for reporting purposes. Standard care did not differ by HIV status. However, mothers diagnosed with HIV would usually receive individualized counselling on infant feeding at the PMTCT clinic.
Data collection
The HIV status of all mothers enrolled in the study was collected from the facility records. Each individual mother enrolled was allocated a unique number during baseline recruitment by the field worker. The same number was presented to the facility during the ANC visit and recorded against each mothers ANC profile register. The HIV status could be determined from the same records for ANC profile register as it is a policy that all mothers should be tested. The data from the ANC profile were collected and linked to the same data collected in the community during home visits by data collectors.
The primary outcome was the EBF rates over the 6-month postpartum period. Breastfeeding behaviour was determined using 24-h recall interviews at each of these time points. The secondary outcome was maternal knowledge and attitude regarding breastfeeding for HIV-exposed infants which was assessed at baseline and the final data collection point at 6 months. The knowledge and attitude questionnaire was formulated from the current policy guidelines for maternal infant and young child feeding in the context of HIV for Kenya [17, 18, 50].
Knowledge was determined through answers about the recommended period of EBF, duration of continued breastfeeding and practices about breastfeeding for HIV-exposed infants. Attitudes towards breastfeeding were determined through responses to a Likert scale questionnaire with five points ranging from strongly agree to strongly disagree. The topics covered included length of EBF and overall duration of EBF.
Data analysis
Data analysis was done using Stata Version 12.1 (Stata Corporation LP, College Station, TX). The characteristics of the study population, and outcomes such as infant feeding practices, EBF rates and maternal knowledge and attitudes regarding EBF for both HIV negative and positive mothers in the study were presented in frequencies and means. T-test was used to determine differences between intervention and control groups for continuous variables, while chi-square tests and cluster-adjusted relative risks were used to test relationships between categorical variables with the main outcomes.
Comparison of outcomes between BFCI intervention group and the control group by HIV status was done using multivariate logistic regression and multinomial logistic models that account for clustering. Statistical significance was set at p < 0.05.
Ethical considerations
A study permit was obtained from the National Commission for Science, Technology, and Innovations. Ethical clearance for PhD research was obtained from Kenyatta University Ethical Committee (Ref KU/R/COMM/51/678), while ethics approval for the main study was obtained from the Kenya Medical Research Institute (KEMRI) SERU (Non-SSC Protocol No. 443). In addition, approval was given by Baringo County Government to carry out the study in Koibatek sub-County. Informed consent was sought from the participants before involving them in the study. An outline of randomization of participants into intervention and control is given in Fig. 1.