Study setting and design
We conducted our study in Bunyala, a rural sub-county in Busia, Western Province, Kenya. We selected Bunyala for two primary reasons: (1) the MMR and NMR are much higher than national estimates, and (2) the MOH demonstrated strong interest and support of AMPATH’s programs and collaborations. Women and infants in Western Province suffer from the second highest maternal and neonatal mortality rates nationally.2,22,23 In Busia County, the most recent estimate for infant mortality rate (IMR) is exceedingly high at 125.9 per 1,000 live births.24 In Bunyala, MNCH activities, including antenatal and postpartum care led by the GOK and supported by AMPATH, exist across 16 community units and 8 MOH health facilities.
To evaluate our primary and secondary outcomes of interest, we used a prospective, matched-cohort study design. We compared outcomes between a cohort of Chamas participants recruited during their first ANC visits at public health facilities in Bunyala and controls receiving the standard of care identified through health facility registers, matched for age, parity, and prenatal care location. We followed both cohorts prospectively for one year and recorded outcome data 12 months following enrollment for all participants.
We used a facility-based recruitment strategy to enroll women to our intervention cohort. We invited all pregnant women attending their first ANC visit at an MOH-sponsored health facility in Bunyala between October-December 2012 to enroll in the Chamas program and to participate in this study (Figure 1). We did not exclude women based on any sociodemographic or reproductive health factors including age, education-level, employment-status, marital status, parity, or prior history of facility delivery.
To recruit our control cohort, we retrospectively identified pregnant women who attended the same health facilities for their first ANC visits in the three months preceding Chamas enrollment (July-September 2012) from clinic registers. We matched controls based on three criteria: age, parity, and prenatal care location (health facility). We tasked CHVs with approaching eligible women at their homes, if they provided an address; we then enrolled women who were successfully located and agreed to participate (Figure 1). Women in both cohorts provided baseline sociodemographic and reproductive health data at the time of enrollment and consented to complete a 12 month follow-up survey (6-12 months postpartum).
Community Health Volunteers in Kenya
Chamas leverages CHVs to deliver health and microfinance education in a safe and familiar setting. As delineated by Kenya’s community health strategy, CHVs are members of the community, nominated from within, who are tasked with improving the community’s health and well-being as well as linking individuals to primary health care services.25 CHVs are considered part-time government volunteers and are supervised by Community Health Extension Workers (CHEWs), salaried frontline healthcare providers integrated within government health facilities.26 CHV facilitators across both study arms were connected to eight health facilities, specifically: five dispensaries, two health centers, and one sub-county level hospital.
Nationally, the Kenyan government delineates a CHV’s scope of work to include: monthly household visits within a defined catchment area of 20 households in rural areas and 100 households in urban areas.27 During routine visits, CHVs collect basic health information, identify health problems, and refer individuals needing additional services to health facilities. All CHVs are required to complete a 10-day, MOH-led basic training session prior to beginning work during which they are introduced to a broad array of health topics, including MNCH. With regard to MNCH, CHVs are provided with a handbook that covers basic information on caring for mothers during and after pregnancy, instructions on facilitating the creation of an individualized birth plan, and lists of specific health behaviors they are encouraged to promote (i.e. attend ANC, deliver in health facilities, adopt family planning).27 They are also expected to recognize danger signs during pregnancy as well as perform basic nutritional assessments, aid in growth monitoring, and recognize when infants require further evaluation for malnutrition. This basic training is often supplemented by technical training that aligns with local priorities; however, technical sessions are variable and often implemented by local governments or non-governmental organizations.27
In September 2012, we selected 32 GOK sponsored CHVs to participate in an additional four-day technical training session on Chamas, sponsored by AMPATH and the Busia County MOH. During these sessions, we trained attendees on how to deliver our evidence-based health curriculum using an illustrated flipchart, facilitate participatory group discussions, and equip program participants with basic microfinance literacy and skills. In addition to conducting didactic sessions, CHVs also received additional training on basic health interventions, such as taking vitals, assessing for hemorrhage and signs of infection at the 48-hour postpartum home-visit, supporting mothers in exclusively breastfeeding, counselling participants on options for family planning, and adopting safe sleep practices. Throughout the year, CHVs attended regularly scheduled check-in meetings (at months 1-4, 6, 9, and 12) with AMPATH implementation leads to provide feedback, as well as receive additional mentorship and support.
Women attending Chamas convened twice per month for 12 months to attend a total of 24 CHV-facilitated group health education and microfinance sessions. Each group was typically comprised of 15-30 women and each session consisted of a 60 to 90-minute participatory lesson on one health (i.e. antenatal care, family planning) and one social (i.e. intimate partner violence, microfinance literacy) topic (Table 1). CHVs used an illustrated flip-chart with an accompanying discussion guide to facilitate sessions. Upon joining the program, women agreed to practice key MNCH behaviors, namely to: deliver in a health facility, attend at least four ANC visits, EBF to six months, receive a CHV home visit within 48 hours of delivery, consider a long-term method of FP, ensure their infant received OPV0, and save money to finance health expenditures. Each group also delineated personal goals they wished to accomplish during the program.
Following lessons, members elected to participate in a table-banking program called “Group Integrated Savings for Health and Empowerment” (GISHE). GISHE is an adaptation of the Catholic Relief Services’ Savings and Internal Lending model, which encourages a savings-led, group-based microfinance scheme.28 We deemed participation optional to avoid excluding women that could not afford to contribute the minimum 50 KSH (0.50 USD) share per meeting. Members contributed up to ten times the amount of the minimal share at each Chamas session. The group provided loans that amounted to a multiple of the individual member’s savings and returned a dividend payment based on interest accrued at the end of the year. Profits generated were distributed to the entire group in amounts proportional to individual shares contributed.
We designed Chamas in collaboration with the GOK and county-level MOH representatives to ensure support and investment from local community members. The Chamas curriculum was designed by a diverse group of stakeholders including AMPATH researchers, community members, and local MOH representatives. The curriculum was designed with the intent to highlight evidence-based recommendations by international authorities (i.e. WHO), bolster training provided through the current CHV handbook, and respond directly to the needs of and questions asked by the local community. We sought feedback throughout curriculum development through conducting focus group discussions with community representatives. This pilot study served as a debut for this curriculum.
Our control cohort received the current standard of care as delineated by the MOH (described under Community Health Volunteers in Kenya). In contrast to Chamas participants, they received monthly, individual CHV household visits, but did not participate in structured, evidence-based health education and microfinance sessions nor experience the group-based format offered by the program.
Data collection and study variables
We collected baseline and outcome data at two time-points for all participants using paper-based, structured, data collection forms (Additional File 1). We tasked AMPATH research assistants trained in data entry with collecting data at both time-points. We recorded baseline data on sociodemographic and reproductive health information at enrollment and collected outcome data at 12 months follow-up (6-12 months postpartum). Interview location depended on the study time-point and cohort assignment. We collected baseline data on intervention participants at health facilities and on controls at participant homes on the day of enrollment. We collected all outcome data at participant homes. During both time-points, we made every effort to collect data individually and privately so as to minimize potential for response bias.
Our primary outcome was the odds of facility-based delivery. Our secondary MNCH outcomes included: the relative proportion of women who attended at least four ANC visits, received a CHV home-visit within 48 hours postpartum, EBF to 6 months, adopted a modern FP method, and adopted a long-term or permanent FP method. We additionally assessed the relative proportion of infants that received OPV0 at birth across cohorts. Where possible, we extracted data from Maternal and Child Health (MCH) booklets. If women did not have their MCH booklet available or if booklets missed data, we asked participants to self-report answers.
To assess the modifying effect of covariates we collected sociodemographic and reproductive health information, including: age, education level, employment status, marital status, parity, prior facility delivery (among those who previously delivered), and facility location of first ANC visit. Maternal age may worsen maternal and fetal outcomes, increasing the propensity of older women to seek care or establish contact with health facilities earlier in pregnancy.29 Sociodemographic characteristics such as education level, employment status and marital status may impact the likelihood of facility delivery as these variables serve as proxies for socio-economic status. We defined “employment” as earning the national daily minimum wage of 450 Kenyan Shillings and allowed participants to select a categorical descriptor (i.e. housewife/unemployed, self-employed, agricultural worker, other).26 Previous studies demonstrate women of lower socio-economic status or lower levels of education are less likely to deliver in facilities.30 Further, reproductive health characteristics such as parity and prior facility delivery may positively or negatively impact a woman’s likelihood of returning to facilities, based on experiences with the health system.31,32 Lastly, we collected first ANC visit facility location to address potential area-level variance on the likelihood of facility delivery.
Though not powered to detect significant differences, we assessed pregnancy-related morbidity and mortality outcomes as well as microfinance data using program monitoring logs recorded by CHVs. These outcomes specifically included: the gestational age (GA) at delivery, the incidence of miscarriage (defined as loss of fetus less than 28 weeks gestation) and stillbirth (defined as loss of fetus between 28 weeks and delivery), as well as the incidence of infant and maternal mortality. We restricted microfinance outcomes to the Chamas cohort and these included: the proportion participating in GISHE, individual loans received, group savings accumulated, and general categories of investment (i.e. school-fees, health expenditures, small businesses). CHVs reported these data monthly to trained research assistants, who electronically transcribed and uploaded outcomes to an encrypted database.
Sample size determination
To calculate our estimated sample size, we assumed 55% of women who attended at least one ANC visit delivered in a health facility and an intra-class correlation coefficient of 0.34, which accounts for population-level variance due to area-level effects (i.e. contact and proximity to the health system and CHVs, clustering by health facility catchment area).2,33 With these assumptions, we determined a 2:1 sample of 240 (156 Chamas and 84 Control) participants adequate to detect a 20% difference in the proportion of facility deliveries between intervention and control groups, with a type I error rate (α) of 0.05 and power of 85%. We assumed a 10% loss to follow-up and established a final target sample size of 267 total participants.
We tabulated frequencies and calculated descriptive statistics comparing socio-demographic and reproductive health variables between Chamas participants and controls. For all bivariate analyses, we used student’s T tests for continuous variables, Mann-Whitney U tests for continuous variables with non-normal distributions, two-sample Z-score tests for proportions, and Chi-square tests for categorical variables.
Multivariable nested models were used to test the association between Chamas participation and facility delivery independently, with successive inclusion of covariates namely: age, education level, employment status, marital status, parity, and prior facility delivery. We examined age as a continuous variable. We collapsed education level into a three-level categorical variable (none-some primary, completed primary, some-completed secondary), and employment (unemployed vs. employed), marital status (single/separated/divorced vs. married), parity (nulliparous vs. multiparous), and prior facility delivery into dichotomous variables. We performed complete case analyses and excluded records with missing data on the primary outcome variable or covariates.
Random effects models, employing the same nested-inclusion technique described above, tested for significant area-level variance as determined by prenatal care location. We additionally ran an interaction model with ANC attendance (dichotomous variable, <4 visits vs. ≥4 visits) and Chamas participation as we hypothesized mothers attending at least four ANC visits were more likely to deliver in a health facility than those who attended fewer than four visits.32 We decided a priori to conduct an additional sensitivity analysis restricting our intervention sample solely to Chamas women who participated in GISHE to examine the impact of combined effect of health education and microfinance participation on MNCH intervention uptake. We conducted all statistical analyses using Stata version 13.1 (StataCorp, College Station, Texas) with α set to 0.05.
Ethical consideration and trial registration
Our study received ethics approval from the Institutional Research Ethics Committee at Moi Teaching and Referral Hospital (IREC/2013/76), the Office of Research Administration at Indiana University (#1306011628), and the Research Ethics Board at the University of Toronto (# 2907). We obtained written informed consent from all participants prior to data collection. We retrospectively registered this study with ClinicalTrials.gov (NCT03188250). No substantial changes were made to the study design or outcomes following participant enrollment.