A cluster randomised controlled effectiveness trial evaluating perinatal home visiting among South African mothers/infants

Background: Interventions are needed to reduce poor perinatal health. We trained community health workers (CHWs) as home visitors to address maternal/infant risks. The Institutional Review Boards of University of California Los Angeles (UCLA), Stellenbosch University, and Emory University approved the study, whose methods have previously been published [16]. We received written informed consent from all study participants. Three independent teams conducted the assessment (Stellenbosch), intervention (Philani), and data analyses (UCLA). This cluster randomised control trial is registered with ClinicalTrials.gov ({“type”:”clinical-trial”,”attrs”:{“text”:”NCT00996528″,”term_id”:”NCT00996528″}}NCT00996528). The protocol for this trial and supporting CONSORT Checklist are available as supporting information; please see Protocol S1 and Checklist S1. Aerial maps, observations, and street-intercept surveys of residents were conducted in order to match township neighbourhoods [16] outside Cape Town, South Africa on the types of housing (formal/informal), presence of electricity, running water, type of sanitation, the number of households and density, counts of alcohol bars (shebeens), child care resources, distance to clinics, length of residence, and original homeland area. UCLA randomised 26 neighbourhoods within matched pairs to either the intervention or the control arm using simple randomisation. One matched pair was eliminated after six months of recruitment due to low numbers of pregnant women (n = 13 combining both neighbourhoods, compared to n = 38–44 on average), leaving 24 study neighbourhoods [16], [17]. Because we were training CHW as generalists, we identified an analytic strategy that included multiple indices as the primary outcome, considering the base rate of each composite measure in each measure. Sample size calculations were conducted to determine the minimum number of pregnant women that would need to be recruited per clinic to achieve 80% power to detect a standardized effect size of 0.40 between women from the 12 intervention and 12 control neighborhoods on one overall summary measure, considering the anticipated base rate on each measure included in the index. Pregnant women were identified by recruiters conducting house-to-house visits every other month to all households in one intervention and one control neighbourhood. Potential participants were pregnant women at least 18 years old living in the neighbourhood from May 2009 to September 2010. Recruiters obtained consent-to-contact and then scheduled transport to a research site for interviewers to obtain informed consent and a baseline assessment. Transportation was also provided for the post birth interviews at 2 weeks post-birth, 6 months and 18 months. Pregnant women were recruited at an average 26 weeks of pregnancy (range, 3–40 weeks). Only 2% of pregnant women refused participation. Figure 1 summarises participant flow through the study. We assessed 1238 women at baseline. Assessments were conducted post-birth at two weeks (92%; mean = 1.9 weeks; SD = 2.1 weeks; median = 1.1; range = 0.1–14.9); six months (88%; mean = 6.2 months, SD = 0.7; median = 6.0; range = 4.2–11.7); and 18 months (84%; mean = 19.1 months; SD = 3.0; median = 18.0; range = 13.6–34.4). All assessments were completed by 83% of mothers; 7% completed no follow-up reassessments; and 10% completed one or two reassessments. Although 84% of mothers completed the 18-month assessment, fewer infants were reassessed at 18 months, as mothers did not consistently bring their children to assessment interviews. As described in an earlier publication [17], the neighbourhoods and pregnant women were highly similar across conditions. After initial recruitment, we appeared to have fewer pregnant women in the control clusters. Recruiters re-canvassed all households in each control neighbourhood and identified 94 additional women pregnant during the recruitment period (included in Figure 1 and follow-up rates above). These “late-entry” controls (16% of the control sample) were from 10 of the 12 control neighbourhoods (median of 7 late-entry participants per neighbourhood; range, 3–24). The late-entry mothers received at least two assessments. The first assessment included the questions from the baseline, post-birth, and six month interview, and abstracted data from the infant Road to Health card. The first assessment was conducted when infants were a mean age of nine months old (median = 8.9; range, 1–18 months). In addition, all late-entry mothers/infants received the 18-month assessment. We recruited, trained, and certified township women to interview participants, entering responses on mobile phones (Nokia E61i and 2630) programmed by Mobenzi (http://www.mobenzi.com/researcher/). Interviewers recorded infants’ physical and developmental status, and gathered data from the infant’s government-issued Road-to-Health card. Supervisors monitored and gave feedback on the data quality weekly. Data collection concluded in October 2012. HIV-related prevention included maternal HIV status (both self-reported and indicated on the infant Road to Health card) and disclosure of serostatus to partners; asking partners to test for HIV; and consistent condom use (on 10 of the last 10 sexual encounters). Among women living with HIV (WLH), managing one’s health and stopping transmission to others requires knowing one’s CD4 cell count (or not), adherence to antiretroviral medications (ARV) over the last week, and complete regimens to Prevent Mother-to-Child Transmission (PMTCT). To PMTCT, mothers must adhere to ARV starting from week 28 of pregnancy; take ARV during labour; provide ARV to infants post-birth; test infants for HIV and retrieve results at 6 weeks post-birth; use one feeding method (either breastfeeding or formula) for the first 6 months; and exclusively breastfeed. Child health status was assessed by LBW (<2500 grams) and by converting infant anthropometric data (collected from birth records and growth monitoring) to z-scores based on the World Health Organisation's (WHO) age-adjusted norms. A z-score below -2 standard deviations was considered a serious health deficit [18]. Mothers reported the length and exclusivity of breastfeeding and risky alcohol use [19]. The features of a Fetal Alcohol Spectrum Disorder (FASD) were screened and assessed in two stages at 18 months [20]. The Bayley Scales of Infant and Toddler Development were administered at 18 months, with age-adjusted scores on cognitive and motor development [21]. Healthcare was assessed by self-reports of antenatal clinic visits, post-birth complications (heavy vaginal bleeding, malodorous discharge, fever, persistent cough, breast infection), and tuberculosis (TB) testing. Immunisation data were extracted from Road-to-Health cards. Depressive symptoms were reported on the Edinburgh Postnatal Depression Scale (EPDS), using a cut-off of >13 to indicate depressed mood [22], [23]. Social networks’ size and frequency of contact and paternal acceptance of the child were self-reported. Receipt of the government child grant was documented. Standard clinic care in Cape Town is accessible and provides free HIV testing, dual regimen therapies for WLH, consistent access to milk tins (formula), TB and CD4 cell testing, co-trimoxazole for infants until HIV testing, HIV polymerase chain reaction (PCR) testing for infants at six weeks, postnatal visits at one week, treatment for WLH, and HIV testing for partners of WLH (http://www.westerncape.gov.za/eng/directories/services/11500/6389). In our sample of women who gave birth in Cape Town, approximately 79% of women gave birth in a hospital, 20% gave birth at a non-hospital facility, and 1% gave birth at home. In addition to clinic care, CHWs provided home visits to participants. CHWs were women with 10th–12th grade education around 40 years old (range 34–59) who were trained for one month in cognitive-behavioural change strategies and roleplaying. They also watched videotapes of common situations that CHWs might face. CHWs were women selected to have good social and problem solving skills, having raised healthy children through their own coping skills, and were trained to provide and apply health information about general maternal and child health, HIV, alcohol use, and nutrition to township women. CHWs were certified and supervised biweekly with random observations of home visits. CBW from Philani implemented the intervention. Eight health messages were delivered regarding a healthy pregnancy, HIV/TB testing and PMTCT, reducing alcohol use and malnutrition, and encouraging breastfeeding, with the aim to deliver these messages in at least four antenatal visits and four post-natal visits within the first two months of life [16]. The intervention dose delivered (i.e., the number of home visits, the visit duration, and content) by CHWs was monitored by CHWs’ entries on mobile phones that included a time stamp and summary visit reports. On average, CHWs made six antenatal visits (SD = 3.8), five postnatal visits between birth and two months post-birth (SD = 1.9), and afterwards about 1.4 visits/month (range: 0.1–6.4 visits/month). Sessions lasted on average 31 minutes each. We first looked for significant differences in baseline demographics between conditions at baseline and among those re-assessed (or not) at post-birth, six and 18 months. To control for multiple comparisons and measure the intervention’s overall effect on well-being, our primary analysis of the intervention’s impact was conducted using one overall test which compared 32 different outcomes simultaneously. On many of the outcomes, almost all mothers would have accomplished the task, without an intervention (e.g., immunize their children). The potential benefits of an intervention are relatively small for such outcomes. Comparing 32 outcomes, chance would lead one to observe up to three significantly different outcomes between the control and the intervention conditions. The binomial test evaluates the number of significant differences between the control and intervention conditions to determine if there is a significant overall difference between conditions. Thus, a binomial test evaluated the number of significant effects favouring the intervention among 32 correlated binary outcomes tested at a one-sided, upper-tail alpha = 0.025 (performed in R, version 2.11.1; please see Appendix S1 for analysis details). Exploratory analyses compared individual measures between intervention and control at a two-sided alpha = 0.05 using logistic random effects regression models adjusting for neighbourhood clustering in SAS PROC GENMOD (version 9.2; SAS Institute Inc., Cary, North Carolina, USA). As the binomial test was the primary analysis, we considered our analyses of individual outcomes to be exploratory and retained the model p-values in lieu of further multiple-testing adjustments. To examine if early intervention impact was sustained for the 12 outcomes that were created by combining data from multiple time points (ex: “Discussed HIV status with sexual partner at six and 18 months”), a second exploratory analysis compared these measures between intervention and control using 18-month data only, using the regressions described above. An exploratory analysis examined the as-treated outcomes, investigating the association between each outcome and the number of CHW visits received (using the regressions described above). We included the number of visits between each assessment point as a covariate in order to systematically control for the number of home visits. Thus, post-birth outcomes were a function of the number of antenatal visits; six month outcomes were a function of postnatal visits between birth and the six month assessment; and 18-month outcomes were a function of the number of home visits between six and 18 months post-birth. By definition, women in the control arm had zero visits. Intervention mothers who received zero CHW visits were excluded from the as treated analysis (3%, n = 18). Late-entry participants’ data were included in all analyses. Based on their age at the late-entry assessment, data from infants of late-entry mothers were split between post-birth (0–4 months old, n = 19) and 6-month (>4 months old, n = 75) outcomes. Overall, results were similar whether or not late-entry participants’ data were included; results are available upon request.