Evaluation of an Early Childhood Development Intervention Among Children Born to HIV-Infected Women in Eswatini

With increased use of effective strategies to prevent HIV transmission from women to their infants, the number of HIV infected children has dramatically decreased. This has led to an increased number of children who have been exposed to HIV, but who are not themselves infected. There is concern that exposure to HIV can lead to poorer health outcomes, including less optimal neurodevelopment. Because of this possibility in a growing population of HIV-exposed, but uninfected (HEU) children, there is a need to identify interventions to optimize their growth and development. This study was undertaken to evaluate the impact of a two-generation intervention, jointly targeting child neurodevelopment and maternal psychosocial wellbeing, integrated into antenatal care/prevention of maternal to child transmission (ANC/PMTCT) clinics. The intervention was implemented by mothers2mothers (m2m), an organization based in South Africa with a long-standing program using peer mentors (Mentor Mothers) to support HIV-infected women and their families. The m2m ECD intervention provided by specially trained Mentor Mothers included home- and community based activities meant to assist mothers and other caregivers in building resilience and gaining skills to support their children's development. The primary intervention was an intensive program of home visits beginning during pregnancy and continuing until the child reached two years of age, with up to 46 home visits. At each visit, the ECD Mentor Mothers helped parents/caregivers to learn about ages and stages of growth (including pre-birth), and to facilitate responsive parenting and early learning opportunities for their children. Mentor Mothers also conducted parenting information play groups (PIPs) in each community for ECD clients and ECD information was integrated into the existing facility support groups provided at each health clinic. The intervention covered general areas of maternal and child health, nutrition, social services, support for primary caregivers and stimulation for early learning. Since the intervention was targeted at two generations, the primary aims were two-fold: to assess the impact on pediatric neurodevelopment and on maternal psychosocial function. Secondary aims included assessment of maternal and child retention in care, maternal HIV outcomes and child mortality, health, growth and nutrition. Children were enrolled into one of two study arms: 1) the intervention arm included children from clinics where the m2m ECD program was being implemented, and 2) the comparison arm included children from government clinics that offered standard of care services but no m2m or ECD program. Mothers were enrolled during antenatal care, in the third trimester of pregnancy. Children and their mothers were followed though 18 months of age and their evaluation included an assessment of development (Mullen Scale of Early Learning) at 9 and 18 months, as well as assessments of child growth, maternal psychosocial well-being, parenting practices and retention in care.

Intervention Methods ECD Intervention. Mothers2mothers (m2m) was selected to pilot-test a parenting intervention integrated within the Eswatini standard integrated antenatal care and PMTCT program developed according to the World Health Organization (WHO) Option B+ recommendations. Each clinic selected for the pilot test had already implemented a m2m mentor mother (MM) peer support program at the facility, as well as defaulter follow-up in the community. The parenting intervention strategy, known as "uMntfwana Wami, uMntfwana Wetfu"( "My Child, Our Child") was focused on peer support for pregnant women and parents/caregivers of young children to assist them in building resilience and gaining skills to support their children's development. M2m's facility-based Mentor Mothers (MMs) recruited pregnant women at the ANC/PMTCT clinics. Both HIV-positive and HIV-negative women were eligible to participate in the m2m intervention. Community-based ECD MMs, who received five weeks of program-specific training, provided the parental training intervention. Evaluation Methods Evaluation Design. The independent evaluation was designed as a quasi-experimental comparison between clinic sites (and their surrounding communities) implementing the intervention and similar clinic sites not implementing the intervention. Sample size calculations. A sample size of 416 participants (213 intervention arm, 203 comparison arm) was estimated to be sufficient to detect a clinically meaningful difference of 0.5 standard deviation (SD) on an age-normed Mullen scale, after allowing for potential 10% child mortality and 20% loss to follow-up, and accounting for clustering at the clinic level. Enrollment of participants. At intervention arm clinics, m2m facility MMs referred ANC clients in the third trimester of pregnancy, residing in the clinic catchment area, and interested in participating in both intervention and study to the study team. At comparison arm clinics, nurses referred interested third trimester pregnant women to the study. A study research assistant based at each clinic conducted eligibility screening and obtained informed consent. Although women could participate in the parental training intervention irrespective of HIV status, the evaluation was designed to assess its effect only on the HIV-positive subgroup of mothers. Participant inclusion criteria were: a woman, aged at least 18, with confirmed HIV-positive status recorded on a clinic chart or register, visiting the PMTCT clinic in the 3rd trimester of pregnancy, and verified resident within the clinic-defined catchment area. Participants who planned to move outside the catchment area or switch HIV care to another site, did not provide written consent to participate in the study or did not complete the enrollment questionnaire were excluded. Data collection and follow up. Study participants were interviewed at enrollment to collect demographic and contact information, abstract patient-held HIV and maternal health cards, and carry out baseline assessments. Follow-up interviews and abstraction of the patient-held HIV, maternal health and well child care cards were targeted for two weeks after the baby's birth and when the baby reached three, six, nine and 18 months of age. A home visit and observation were conducted when the baby was between 12-15 months of age. If the mother was not reachable, the baby's primary caregiver was interviewed. Primary outcomes assessment. The primary pediatric outcome of cognitive development at nine and 18 months of age was assessed using the Mullen Scales of Early Learning, administered by two assessors. The items were translated into siSwati as needed and adapted for and tested on emaSwati children in the relevant age groups. At the 18-month time point, inter-rater reliability was assessed by independently scoring a sample of 40 children. The primary maternal outcome of psychosocial well-being was assessed using the Edinburgh Postnatal Depression Scale (EPDS), categorized dichotomously as depressive symptoms present (EPDS score \>13) or absent (EPDS score ≤13). The cutoff was selected based on work in neighboring South Africa by Tomlinson et al (2017). Parenting self-confidence was also assessed, using the Maternal Self-Report Inventory (Short Form) (MSRI). These instruments were administered to women at enrollment and at nine and 18 months postpartum. Secondary outcomes assessment. Length, weight, head circumference and mid-upper arm circumference were measured at the child's 9- and 18-month visits and compared to norms established by the WHO Multicentre Growth Reference Study..\[ref 49\]. Three separate anthropometric measurements were taken and the median value used. Retention in care and pediatric HIV testing was assessed based on health care visits and outcomes abstracted from multiple sources including: participant-held maternal health/ANC and HIV cards and child health and HIV cards, postnatal care and child welfare registers at the clinics, and the client management information system (CMIS) in some clinics, as well as self-report. Compliance with feeding recommendations was obtained by maternal or caregiver report. Mortality information was obtained by report from families. Potential mediators. Information on routine caregiving practices and household characteristics was obtained by observation using the Home Observation for Measurement of the Environment (HOME) Inventory \[refs2 3\] during a home visit at 12-15 months of age. Households were instructed to carry out everyday household activities during this observation visit. Study Research Assistants (RAs) completed classroom training on this instrument and then pilot-tested it in the field using non-participating households. Inter-rater reliability was assessed by having multiple observers score the pilot and then discuss discrepancies in scoring. Key, non-routine child stimulation practices were assessed by interview using the Family Care Indicator, a series of standardized questions developed as part of the UNICEF Multiple Indicator Cluster Surveys (MICS) version 3 (2006), which was administered at baseline to all participating mothers who at the time had children under age five, and to all mothers and caregivers at the 9-month and 18-month assessments. Potential confounding factors. Eswatini was affected by a drought during 2016-2017 and its impact was assessed at home visits (age 12-15 months), and at the 18-month interview using the Household Food Insecurity Access Scale (HFIAS). Contamination assessment. Participants in both arms were interviewed about their contact with m2m at each visit and at the 18-month assessment. Statistical methods. Mullen scores: Scores from each domain (Visual Reception, Fine Motor skills, Receptive Language, Expressive Language and Gross Motor skills) were analyzed separately. For each domain, the within-sample means and standard deviations of the raw scores by arm were calculated. To facilitate understanding of the effect sizes, the Mullen scores in each domain were then transformed into within-study sample standardized scores (Z-scores), for which effect sizes represent unit changes in the distribution SD. Multivariable linear mixed effects regression models with a random effect for enrollment site were applied to test the difference in the standardized mean scores between the two study arms at each time point. Adjusters selected based on the literature included the child's actual age at the assessment date, the child's gender, and an indicator for stunting (length-for-age Z-score \< 2) as a proxy for chronic malnutrition. Additional adjusters were included in the multivariate model based on their strong (p\<=0.2) effects in bivariate analyses: the size of the mother's household at enrollment, the length of residence of the mother at her current location at enrollment, household asset-based socio-economic status of the mother at enrollment, the mother's number of other living children at enrollment, her number of previous miscarriages, and her years of education. Other measures. Growth: Percentages of children with Z-scores of the WHO-standardized anthropometric measurements \< -2 at the 9 month and 18 month assessments were reported. Maternal retention in HIV care: Time in care was calculated as days from the date of enrollment in the study to the last documented visit as found in the patient-held HIV cards, the chronic care file of the HIV clinic at the clinic where they enrolled, the CMIS and from self-report. A log-rank test for equality of mean retention between the two arms was conducted. Socio-economic status: Principal components analysis was used to develop a composite score of socio-economic status based on reported household assets, which was divided into quintiles for analysis. Family Care Indicator. Performance on individual items was assessed and as well as through a combined score (possible range: 0-6) of activities that someone in the household was reported to have performed with the child: (1) "Read books or look at picture books" (2) "Sing songs" (3) "Tell stories" (4) "Play with" (5) "Take \[the child\] outside the home, compound, yard or enclosure" and (6) spend time naming, counting and/or drawing things." Mediation analysis. The multivariable linear regression model used in the primary analysis of the 18-month timepoint was extended into a generalized structural equation model to evaluate a potential mediating effect of book-sharing with the child on the relationships between study arm and expressive language scores, and between study arm and visual reception scores. Robust variance estimation was used to adjust for site clustering The presence of book sharing was defined as a positive response to item (1) on the Family Care Indicator at the 18 month timepoint.