Home Visiting and Antenatal Depression Affect the Quality of Mother & Child Interactions in South Africa

Abstract

Objective

To examine whether maternal depressed mood at birth moderated the protective effect of a home-visiting intervention on the quality of caregiving for children growing up in a low- and middle-income country.

Method

Almost all pregnant mothers in 24 Cape Town neighborhoods were recruited into a cluster randomized controlled trial matched by neighborhoods to the Philani home-visiting condition (HVC) or the standard care condition (SC). At 3 years after birth, the quality of mother-child interactions between HVC and SC mothers with and without antenatal depressed mood was assessed in a representative subset by rating videotaped observations of mother-child interactions on 10 dimensions of caregiving.

Results

As predicted, maternal depressed mood at birth moderated the effect of the HVC on the quality of mother- child interactions. Among nondepressed mothers, mothers and their children in the HVC scored significantly higher on 5 of the 10 dimensions of the matemal-child interaction scale than mothers in the SC: mothers exhibited more maternal sensitivity, talked more, had more harmonious interactions, and had children who paid more attention and exhibited more positive affect. However, being in the HVC did not significantly affect the mother-child interaction scores among mothers with depressed mood. Among HVC children, those with mothers with depressed mood showed significantly less positive affect and talked less with their mothers than children with nondepressed mothers. SC children with mothers with depressed mood were more responsive and paid attention to their mothers than children with nondepressed mothers.

Conclusion

Home visiting resulted in a better quality of caregiving for mothers without depressive symptoms. Future interventions need to specifically target maternal depression and positive mother-child interactions.

The quality of close relationships early in life has a lasting impact on children’s development.^1,2 The first significant interaction is between parents and their child. Caregiving that is “sensitive,” or responsive to a child’s needs and engages the child, with limited hostility, predicts a range of positive child outcomes.^{1, 2,3,4} Sensitive caregiving refers to a parent’s ability to notice, interpret, and respond appropriately to a child’s cues with limited negative verbal comments or coercive and intrusive behavior.⁵ It also involves active engagement with the child and effective support of a child’s interest.⁵ Research on caregiving in low- and-middle income countries (LMICs) suggests that sensitive and responsive caregiving might be a particularly important factor mediating the adverse effects on child development while growing up in poverty with many health risks.^3,5 Specifically, maternal sensitivity and responsiveness help foster resilience in children to cope with later challenges. ^{1, 2} However, a lack of sensitivity and responsiveness can result in insecure attachments and child cognitive and emotional difficulties.⁶ The present study examined the impact of the Philani Program, a home-visiting program in the townships of Cape Town, South Africa,⁷ on the quality of caregiving by studying the interactions between mothers and their children at 3 years after birth.

The goals of caregiving are not only child survival, but also supporting children to thrive.^8,9 In LMICs, mothers raise their children with overwhelming risks associated with poverty and violence,^10,11,12 risks that challenge the foundation of caregiving: protection and nourishment. In the townships of Cape Town, baseline data from the present cohort showed that most mothers lived in informal dwellings or shacks in periurban settings (69%), with only half of all mothers reporting access to water (53%) or a flush toilet on site (55%).⁷ Unemployment rates were high, with fewer than 1 in 5 mothers being employed and more than half of all mothers reporting household monthly incomes less than 2,001 Rand (US$294; 53%).⁷ Only 36% of mothers reported being married and living with their partner.⁷ In consequence, most mothers served as the primary or the only caregiver to their children in impoverished households, with limited support from partners.

In the context of these adverse living conditions, the health challenges are overwhelming. Approximately 35% of all mothers experience depressed mood during and after pregnancy and 26% of pregnancies are in mothers living with human immunodeficiency virus (HIV; MLHs) and who therefore are concurrently at risk for tuberculosis (TB).¹³ MLHs must adhere to burdensome medical regimens lifelong and to antiretroviral therapy for the welfare of their children and themselves. MLHs also are more likely to experience depressed mood.¹⁴ In high-income countries, maternal depression is consistently associated with negative consequences for the mother and child, consequences that can persist for a lifetime.¹⁵ However, less is understood about the patterns of maternal depression and their consequences on maternal and child health in LMICs. Recently, maternal depression has been linked to poverty, age, income, having unintended children, HIV status, and intimate partner violence in LMICs.¹⁶ Evidence from the present cohort in South Africa suggests that maternal depression is associated with limited structural resources, including lower rates of employment, education, and income and higher rates of problematic consumption of alcohol, food insecurity, and likelihood of being HIV seropositive.¹⁷ Thus, there is evidence that health risks, specifically depression and HIV, together with poverty, limit South African mothers’ physical and psychological resources to care for themselves and their children.

Although poverty is consistently liked with poor growth outcomes in children, maternal depression also is associated with children’s growth delays emotional difficulties.^17,18 With rates of depressed mood during and after pregnancy reaching almost 50% in some regions of South Africa,^19,20 mothers face parenting with limited psychological resources. Indeed, a meta-analysis on the negative impact of maternal depression on parenting behavior cited the strongest effects for mothers and their children living in disadvantaged environments.²¹ Maternal depression is associated with a lack of engagement with the child and insecure attachments in high-income countries and LMICs, including South Africa.^4,21 Given the importance of maternal depression for caregiving and child development, we focused on this risk factor in examining caregiving in this study.

Home-visiting interventions in LMICs have the potential to support families and secure children’s positive developmental trajectories.²² With the growing shortage of health care professionals globally, alternative strategies are needed to address the public health challenges in LMICs. These challenges include malnutrition, infectious diseases, and maternal behaviors in the context of widespread poverty.^23,24 Home-visiting programs supporting maternal health and caregiving result in better child growth outcomes, infant attachment, and cognitive and behavioral health in children living in LMICs,^25,26 including sub-Saharan Africa.²² Specifically, task shifting from doctors to nurses and non-physician clinicians in LMICs, including sub-Saharan Africa, results in equal or better quality of health care and these interventions are cost effective.²⁷ The Philani Program in the townships of Cape Town is unique because it shifts health care tasks to local township women who are trained to be community health workers known as Mentor Mothers (MMs).

MMs are positive peer deviants²⁸ or positive role models in their communities and are selected because they have healthy children of their own and strong communication and problem-solving skills. The Philani Program trains, monitors, and evaluates MMs as generalists to address multiple health challenges, including malnutrition, HIV, alcohol use, and malnutrition, and healthy daily routines during home visits.⁷ The primary outcomes of the program include benefits for the mothers and children receiving home visits by MMs, including increased exclusive breastfeeding, infants with better growth trajectories and recoveries from malnutrition, and increased adherence to tasks to prevent vertical HIV transmission and to adhere to a single feeding method for MLHs.^29,30,31 Although Philani does not explicitly focus on or train MMs to specifically address caregiving, they support interventions for mothers with multiple health challenges and encourage them to bond with their infants. As a result, exclusive breastfeeding and adherence to medical regimens are significantly higher among mothers visited by MMs compared with peers, especially among MLHs.^29,30,31 Thus, mothers receiving home visits from MMs are more likely to complete recommended childcare tasks.

Research in high-income countries suggests that child gender also might be associated with maternal sensitivity.³² Mothers tend to be more sensitive in their interactions with their daughters compared with their sons from the first months of life.³² Although gender biases are pervasive in the townships of South Africa,³³ less is understood about how these gendered ideas can influence a mother’s interactions with her sons versus her daughters. Although early research focused only on maternal factors in explaining gender differences found in caregiving styles, more recent examinations also consider the child’s contribution, including their temperament, level of activity, and responsiveness in interactions with their mothers.^2,34 Therefore, the present study examined mother-child interactions based on maternal and child behavior and affect while controlling for child gender.

The quality of mother-child interactions is often assessed in older children and their caregivers with structured play tasks, guided tasks in which a toy is introduced and taken away (eg, “the barrier task”), or a snack is introduced that is likely to elicit the parent’s involvement (eg, a sticky chocolate cake or an orange).³⁵ The quality of caregiving, as measured by these interactional tasks, is associated with children’s positive language, cognitive, social, and emotional development.^{1, 2} Within the context of this randomized controlled trial (RCT),⁷ mothers and children were followed from pregnancy, and at 3 years, videotaped ratings of mother-child interactions of mothers and children eating sticky cake were compared as part of a secondary analysis between mothers receiving home visits from MMs in the home-visiting condition (HVC) or in the standard-care condition (SC). Ten dimensions of mother-child interactions were rated.³⁵ Based on the benefits in maternal and child health of the intervention,^29,30,31 HVC mothers and their children were predicted to exhibit a higher quality of interactions than mothers and children in the SC. Given the evidence for the negative impact of maternal depression on caregiving and child development, we predicted that maternal depressed mood during pregnancy would moderate the effects of the intervention on mother-child interactions. Specifically, children with nondepressed mothers were predicted to exhibit a higher quality of interactions than children with mothers with depressed moods.

Method

The institutional review boards of the University of Califomia-Los Angeles (UCLA), Stellenbosch University (Stellenbosch, South Africa), and Emory University (Atlanta, GA) approved the study whose methods were previously published.⁷ Written, voluntary, informed consent was received from all study mothers. Three independent teams conducted the assessment (from Stellenbosch University), intervention (from the Philani Program), and randomization and data analyses (UCLA). This cluster RCT is registered with ClinicalTrials.gov (NCT00996528; Neighborhood Alcohol & HIV Prevention in South African Townships; https://clinicaltrials.gov/ct2/show/NCT00972699).

Participants

All pregnant mothers in 24 Cape Town neighborhoods were recruited into a cluster RCT; 450 to 600 households were identified and matched based on housing type, presence of electricity, water, sanitation, size and density, alcohol bars, child care resources, distance to clinics, and length of residence. Buffer neighborhoods prevented cross-neighborhood contamination. UCLA randomized neighborhoods in 6 blocked sets of 4 neighborhoods each for 12 Philani HVC neighborhoods (n = 644) and 12 SC neighborhoods (n = 594). A subset ofmother-children interactions was assessed in the HVC (n = 12 neighborhoods, n = 275 mothers) or SC (n = 12 neighborhoods, n = 305 mothers).

Recruiters

Township women conducted house-to-house visits every other month from May 2009 to September 2010 to identify all pregnant mothers and refer to the study; only 2% refused participation. Assessments were conducted at a research office in the township, with transport provided. One thousand two hundred thirty-eight mothers were assessed at baseline. Follow-up assessments were conducted at 2 weeks (92%), 6 months (87%), 18 months (91%), and 36 months (85%) after birth. Maternal and child deaths were excluded from analyses (n = 109); most deaths happened at birth (66%; n = 69).

SC and HVC

Standard Care Condition

Standard clinic care in Cape Town was accessible within 5 km of each study neighborhood. Each antenatal clinic provided comprehensive maternal and child health services and prevention of mother-to-child transmission services, including HIV testing.

Intervention Condition (HVC)

In addition to access to standard clinic services, home visits were conducted. The Philani Program, a nongovernmental organization, trained township women, who typically never worked outside the home and who had less than a high school education, as MMs. For 1 month, MMs were trained in cognitive-behavioral change strategies. Philani selected MMs for having good social and problem-solving skills and raising healthy children using their own coping skills. They were trained as generalists to provide and apply health information about HIV/TB prevention, strategies to prevent-mother-to-child transmission, the consequences of alcohol use/abuse, the importance of breastfeeding, and how to avoid malnutrition. Videotaped models and role plays depicting common challenging situations that MMs might face were provided. MMs were to deliver these messages in at least 4 antenatal visits and 4 postnatal visits within the first 2 months of life. They were certified and supervised biweekly in real time with random observations of home visits once every 2 weeks.

On average, MMs made 6 antenatal visits (standard deviation [SD] 3.8), 5 postnatal visits from birth to 2 months after birth (SD 1.9), and approximately 1.4 visits per month (range 0.1–6.4 visits per month) until 6 months after birth. Sessions lasted on average 31 minutes. Visits were biannual after 6 months, when MMs primarily checked for growth and/or abuse.

Measures

Maternal Measures at Baseline During Pregnancy

Depressive Symptoms

Mothers self-reported depressive symptoms using the Edinburgh Postnatal Depression Scale (EPDS),³⁶ with a cutoff score higher than 13 to indicate depressed mood.³⁷ Demographic characteristics were self-reported at baseline, including mother’s age, education, marital status, employment, monthly household income, current housing conditions including type of housing (formal or informal/shack), access to water, flush toilet, and electricity. Food insecurity was assessed using 1 item (“How many days in the past week have you gone hungry?”) from the Household Food Insecurity Access Scale (HFIAS). This item has been found to be highly correlated with the 9-item scale used to distinguish food-insecure from food-secure households across different cultural contexts.³⁸

Health

Mothers self-reported whether they were primipara, their mean number of livebirths, whether they attended at least 4 antenatal clinic appointments, and whether they were ever tested for TB and had any positive results.

HIV Status

All mothers reported their and their partner’s serostatus.

Quality of Mother-Child Interaction

As part of the 3-year assessment, at the research center in 1 township, a subset of mothers and their children were videotaped interacting in a private room according to the method used by Murray et al.³⁵ The researcher brought in a tray with refreshments, including a “messy” food (eg, a sticky chocolate cake or an orange), that would likely elicit the mother’s involvement and control of the child. Mother and child were left to have the refreshment together for a 10-minute period and a video recording was made of the interaction. The videotapes were rated by 1 trained researcher who was trained by L. Murray, the task’s developer.³⁴ The researcher was unaware of maternal condition (HVC or SC), psychiatric state, postpartum period, or any other data concerning the child. Murray and the researcher co-rated a random sample of 10% of the 5-minute tapes of the mother and child interacting and rated the interactions on 10 dimensions, based on the research protocol: 5 maternal, 3 child, and 2 joint behaviors. The trained researcher was certified reliable. The physical data for interobserver reliability were unfortunately lost. However, the laboratory’s policy is to co-code 10% of all dataset offline and data are randomly selected from across the study videotapes to ensure that there is no drift over time. The researcher was trained and certified a reliable coder by the team that not only developed the scale and but also completed several projects and published results using the same measure.^39,40 This interaction scale was devised after a review of the current literature at the time and previous scales were used to examine mother-child interactions.^39,41 This scale has been found reliable (0.70–0.89) and sensitive to a range of maternal psychiatric disorders and the presence of economic and social adversity.^40,42,43,44

Maternal behavior was rated on 5 scales: how attuned the mother was to her child’s needs and how consistently and appropriately she responded to the child’s signals (eg, distance between mother and child, eye-to-eye contact possible, picking up on child’s cures for help or to explore alone; maternal sensitivity); how much the mother used scaffolding or assisted the child in self-feeding (eating facilitation); the degree of the mother’s engagement or withdrawal from her child, psychologically and/or physically (remote/engaged); the ability of the mother to manage the child appropriately and respond to the child’s requests (maternal control); and the degree of the mother’s positive emotional affect or the appearance of enjoyment during the interaction (maternal emotion).

Child behavior was scored on 3 scales: how responsive the child is to the mother’s actions, gestures, and commands (child responsiveness); the degree of the child’s sustained attention during the activity including whether the child is paying attention to the food and acting on the cues of the mother (child attention); and the child’s degree of positive emotional affect or the appearance of enjoyment during the interaction (child emotion).

Joint behaviors were rated with 2 items that assessed the degree of harmony or disharmony in the overall interaction (atmosphere) and the amount of verbal communication between mother and child during the interaction (conversation).

The sum of scores on all 10 dimensions (summary score) represented an overall quality of mother-child interaction designed to take the mother’s and child’s behavior into account to allow for mothers with more difficult children who might require firmer parental behavior to attain similar sensitivity ratings as mothers with less difficult children. A higher sum represented a higher overall quality of mother-child interaction.

Analyses

At 36 months, a subset of mothers and their children from the home-visiting RCT⁷ completed an interactive activity as originally reported by Murray et al.³⁵ The first analysis compared mothers in the present subset with mothers in the original RCT sample⁷ across all baseline measures using fixed-effects regression models to assess whether the subset was representative of the full sample of mothers. Logistic regression was used for binary variables (ie, formal housing, electricity) and linear regression was used for continuous variables (ie, age, highest education level). The same approach was used to compare mothers in the HVC and SC of the present subset across all baseline measures.

Then, the effects of the HVC on each dimension of mother-child interactions and the summary score were assessed using linear regression models with intervention condition as a fixed effect (SC as the reference group). The following models assessed the effects of intervention condition and the predicted moderating effect of maternal depressed mood (EPDS score >13 versus ≤13) on mother-child interaction scores. Then, contrasts were run for any dimensions that had significant interactions between the intervention condition and depressed mood status. All models were adjusted for neighborhood clustering, maternal age, non-primipara status, marital status, and having a flush toilet. Regression models were fit in IBM SPSS Statistics 20 (IBM Corp., Armonk, NY).

Results

Comparison of Mothers in Present Subset With Original RCT Sample

The present subset of mothers was similar to the original sample of mothers from the RCT,⁷ including depressed mood (EPDS score >13; 35%), mean EPDS score (mean 10.8, SD 7.0), married or living with a partner (57%), highest education level (mean 10.3, SD 1.8), ever employed (19%), monthly household income over 2,000 Rand (US$294; 47%), formal housing (31%), water (53%) or flush toilet (55%) on site, electricity (90%), food insecurity (50% mothers, 29% children), number of livebirths (mean 1.6, SD 1.0), at least 4 antenatal clinic visits (77%), tested positive for TB (9%), living with HIV (26%), and partner tested for HIV (11%; Table 1). For any differences, mothers in the present subset tended to be older (26.37 versus 26.89 years; p < .01) and were more likely to be non-primipara (69% versus 66%; p < .05).

Table 1.

Comparisons of Baseline Characteristics in the Full Cluster Randomized Controlled Trial (RCT)7 Versus the Present Subset

	Home-Visiting Condition		Standard Care Condition		Total
	Full RCT (n = 339)	Present Subset (n = 305)	Full RCT (n = 319)	Present Subset (n = 275)	Full RCT (n = 658)	Present Subset (n = 580)
Demographic characteristics
Age (y), mean (SD)	26.5 (5.5)	26.9 (5.6)	26.3 (5.6)	26.9 (6.1)†	26.4(5.6)	26.9 (5.8)**
Highest education level, mean (SD)	10.4(1.9)	10.3 (1.7)	10.3 (1.8)	10.3 (1.8)	10.3 (1.8)	10.3 (1.8)
Married or lives with partner	59%	65%†	55%	55%	57%	60%
Ever employed	20%	21%	18%	20%	19%	20%
Monthly household income >2,000 Rand	46%	48%	49%	49%	47%	49%
Formal housing	31%	35%	33%	35%	31%	33%
Water on site	52%	50%	55%	56%	53%	53%
Flush toilet	53%	50%	59%	60%	55%	55%
Electricity	88%	87%	92%	92%	90%	89%
Mother hungry in past week	48%	50%	51%	52%	50%	51%
Children hungry in past week	27%	30%	32%	33%	29%	32%
Maternal health
EPDS score >13	37%	36%	33%	32%	35%	34%
Mean EPDS score	11.2 (6.9)	10.9 (6.8)	10.4(7.1)	10.2 (6.9)	10.8 (7.0)	10.6(6.8)
Non-primipara	66%	69%	66%	70%	66%	69%*
Number of livebirths, mean (SD)	1.5 (0.9)	1.6 (1.0)	1.7 (1.1)	1.8 (1.2)†	1.6 (1.0)	1.7 (l.l)†
Antenatal clinic appointment	78%	78%	75%	80%	77%	79%
Test positive for TB, lifetime	8%	7%	9%	8%	9%	7%
Mothers living with HIV	26%	24%	27%	24%	26%	24%
Partner HIV+	10%	8%	12%	11%	11%	9%

Note: EPDS = Edinburgh Postnatal Depression Scale; HIV = human immunodeficiency virus; TB = tuberculosis.

^*p < .05;

^**p < .01;

^†p < .10

Present Subset

Table 1 presents a summary of all baseline measures for the present subset of mothers in the HVC and SC. In the subset participating in the mother-child interaction task, more than one-third of all mothers (34%) reported antenatal depressed mood (EPDS score >13). HVC and SC mothers were similar on most main baseline measures, such as demographics including age (~26 years), education (mean 10.3, SD 1.8), employment (20%), and household income (51% earned <2,000 Rand per month [~US$294]); type of housing (33% formal), source of water (53% on site), presence of flush toilet on premises (55%), and electricity (89% on site); and general health, including food insecurity (51% of mothers and 32% of children), visiting antenatal clinics at least 4 times (79%), prevalence of being HIV seropositive (24%), and rates of TB infection (7%). However, HVC mothers were more likely to be married or living with a partner (65% versus 55%; p < .05) but less likely to have a flush toilet on the premises (50% versus 60%; p < .05) than SC mothers.

Mother-Child Interaction Scores

The mother-child interaction scale demonstrated good internal consistency (Cronbach α = 0.88). All 10 mother-child interaction scores were significantly correlated with the summary score (r = 0.42–0.84, p < .001 for all comparisons).

Table 2 presents a summary of all estimates resulting from the fixed-effects linear regression models used to assess the average effects of the intervention on all 10 mother-child interaction dimensions and the sum score of all dimensions. When including mothers with and without antenatal depressed moods, the HVC effect was not significant for any of the mother-child dimensions. Because a moderating effect of maternal depressed mood was predicted, the following models included maternal depressed mood. Table 3 presents a summary of the estimates resulting from the fixed-effects linear regression models used to assess intervention condition, maternal depressed mood, and the interaction between intervention condition and depressed mood on all 10 mother-child interaction dimensions and the sum score of all dimensions.

Table 2.

Summary of Estimates of Fixed-Effects Models on the Present Subset Regressed on Intervention (Home-Visiting Versus Standard Care Condition; N = 580)

	Estimate	SE
Maternal dimensions
Maternal sensitivity	0.15	0.11
Eating facilitation	−0.12	0.13
Remote/engaged	0.04	0.12
Maternal control	0.08	0.12
Maternal emotions	0.05	0.10
Child dimensions
Child responsiveness	−0.03	0.11
Child attention	0.26	0.11
Child emotions	0.12	0.10
Joint dimensions
Atmosphere	−0.17	0.10
Conversation	0.01	0.12
Sum of all 10 raw measures	0.85	0.78

Note: Based on comparisons of mothers between the home-visiting condition and standard care condition on baseline characteristics, these models were adjusted for maternal age, non-primipara status, marital status, having a toilet, child gender, and neighborhood clustering. SE = standard error.

Table 3.

Summary of Estimates of Fixed-Effects Models on Present Subset Regressed on Intervention, Depressed Mood, and Interaction of Intervention and Depressed Mood (N = 580)

	HVC		Depressed Mood		HVC × Depressed Mood
	Estimate	SE	Estimate	SE	Estimate	SE
Maternal dimensions
Maternal sensitivity	0.19†	0.11	−0.06	0.14	−0.13	0.19
Eating facilitation	−0.07	0.14	0.15	0.17	0.08	0.37
Remote/engaged	0.11	0.12	−0.07	0.16	−0.21	0.21
Maternal control	0.09	0.12	0.21	0.14	−0.10	0.20
Maternal emotions	0.04	0.10	−0.05	0.12	−0.02	0.17
Child dimensions
Child responsiveness	0.08	0.12	0.30*	0.15	−0.32†	0.19
Child attention	0.32**	0.11	0.31**	0.14	−0.37*	0.18
Child emotions	0.21*	0.10	0.11	0.12	−0.37*	0.16
Joint dimensions
Atmosphere	0.21*	0.08	0.07	0.13	−0.23	0.18
Conversation	0.23†	0.12	0.22	0.16	−0.47*	0.21
Sum of 10 raw measures	1.8l†	0.94	1.01	1.16	−2.10	1.58

Note. These models were adjusted for maternal age, non-primipara status, marital status, having a toilet, and neighborhood clustering. Gender did not have a significant effect on mother-child interactions, so it was excluded from these models. HVC = home-visiting condition; SE = standard error.

^*p < 05;

^**p < .01;

^†p < .10

Among nondepressed mothers, mothers and children in the HVC tended to score higher than their counterparts in the SC on the matemal-child interaction summary score (estimate 1.81, standard error [SE] 0.94, p = .050).

When considering each of the 10 dimensions separately, nondepressed HVC mothers tended to exhibit more maternal sensitivity (estimate 0.19, SE 0.11, p = .063), had more harmonious interactions (estimate 0.21, SE 0.08, p = .049), and tended to talk more (estimate 0.23, SE 0.12, p = .067) with their children. There were no significant interactions for the maternal dimensions and neither intervention condition nor depressed mood significantly affected how much the mother helped the child eat, how engaged she was, or her level of control of positive emotional affect during the interaction.

Significant interactions were found between intervention condition and maternal depressed mood for all 3 child dimensions. Then, contrasts were estimated for the 4 groups (depressed mood versus nondepressed in the HVC, depressed mood versus nondepressed in the SC, HVC versus SC among nondepressed mothers, HVC versus SC among mothers with depressed mood) to assess these effects (Table 4). Among children with nondepressed mothers, children in the HVC tended to pay more attention (estimate 0.32, SE 0.11, p = .004), exhibited more positive affect (estimate 0.21, SE 0.10, p = .03), and talked more with their mothers (estimate 0.23, SE 0.12, p = .066) during the interaction. Among children with mothers with depressed mood, children in the HVC did not tend to score significantly differently than children in the SC.

Table 4.

Summary of Contrasts of Fixed-Effects Models on Present Subset (N = 580) Regressed on Intervention (Home-Visiting Condition [HVC] Versus Standard Care Condition [SC]), Depressed Mood, and Interaction of Intervention and Depressed Mood (N = 580)

	Depressed Mood vs. Nondepressed in HVC		Depressed Mood vs. Nondepressed in SC		HVC vs. SC Among Nondepressed		HVC vs. SC Among Depressed Mood
	Estimate	SE	Estimate	SE	Estimate	SE	Estimate	SE
Child dimensions
Child responsiveness	−0.01	0.12	0.30*	0.14	0.08	0.11	−0.24	0.15
Child attention	−0.06	0.11	0.31*	0.15	0.32†	0.11	−0.06	0.15
Child emotions	−0.26*	0.10	0.11	0.13	0.21*	0.10	−0.16	0.13
Joint dimensions
Conversation	−0.24**	0.13	0.22	0.16	0.23†	0.12	−0.24	0.16

Note: Contrasts were run for only the dimensions that had models with significant interactions (Table 4). These models were adjusted for marital status, having a toilet, maternal age, non-primipara status, and neighborhood clustering.

^*p < .05;

^**p < .01;

^†p < .10.

Among children in the HVC, children with mothers with depressed mood showed significantly less positive emotional affect (estimate −0.26, SE 0.10, p = .010) and tended to talk less with their mothers (estimate −0.24, SE 0.13, p = .065) than children with nondepressed mothers. Among children in the SC, children with mothers with depressed mood were more responsive (estimate 0.30, SE 0.14, p = .038) and paid more attention (estimate 0.31, SE 0.15, p = .039) to their mothers than children with nondepressed mothers.

Discussion

Home visits not only support health outcomes^29,30,31,45 but also improve mother and child interactions for mothers who are not depressed during pregnancy. A high quality of caregiving is especially important for mothers and children facing a myriad of challenges to their health and well-being.^{1, 2} Most mothers in the present cohort were unemployed, living in informal housing, and had low household incomes. Moreover, 24% of the sample was composed of MLHs and experienced additional stressors trying to stay healthy while concurrently supporting their children. The present results suggest that home visiting aimed at addressing multiple health issues in a LMIC also could support a high quality of caregiving, without specifically targeting these behaviors among mothers without depressive symptoms during pregnancy.

The HVC intervention did not target depression. Although the intervention targeted HIV, alcohol, and malnutrition, the MMs were not trained to address depression. HVC mothers without depressed mood during pregnancy scored significantly higher on the mother-child interaction scale than nondepressed mothers in the SC. Specifically, nondepressed HVC mothers tended to display more maternal sensitivity, talk more, and have overall more harmonious interactions with their children than nondepressed mothers in the SC. They also had children who paid more attention and were happier than their peers. Nondepressed HVC mothers scored significantly higher on most, but not all, dimensions of caregiving. Specifically, nondepressed HVC mothers did not help their child eat the snack, were more engaged with or exhibited more control over the child, had a more positive affect, or had a more responsive child during the interaction than nondepressed mothers in the SC. This suggests more targeted support by MMs is needed to improve maternal caregiving, in particular, positive maternal caregiving. Specifically, interventions that support positive emotional expressions, vocalizations, or increased conversation, physical contact, and cooperation have shown benefits for mother-child relationships and for children’s language and cognitive development and behavior.^3,46,47 In the same community, Cooper et al.³ found that shifting patterns of interactions between mothers and infants required a significant number of home visits (≥161-hour visits) focused almost solely on improving the mother-infant relationship.

Notably, more than one-third of all mothers reported depressive symptoms during pregnancy. This rate is consistent with previous research in periurban settlements near Cape Town (30%)⁴⁸ and is higher than in many other LMICs.¹⁶ Research suggests that maternal depression might influence parenting behaviors resulting in caregiving difficulties, decreased sensitivity and warmth, and impaired mother-child interactions.⁶ However, a meta-analytic review on maternal depression and positive parenting reported that the overall effect is small, because other maternal, child, and family factors can moderate this complex relationship.^21,49 In the present cohort, South African mothers faced many additional challenges to depression, because many cared for their children without a partner (40%), lived in poverty, and had multiple heath challenges. Thus, mothers in LMICs such as South Africa might need specific and consistent support for addressing depressive symptoms and positive parenting skills.

The HVC addresses multiple health challenges by training MMs to teach township mothers about healthy daily routines to improve their well-being and support their children’s healthy development. Although the HVC training does not specifically address maternal depression and caregiving, mothers receiving home visits exhibited positive caregiving behaviors, including higher rates of exclusive breastfeeding and better adherence to preventing vertical transmission and to a single feeding method for MLHs.^29,30,31 The children of antenatally depressed mothers in the HVC also were less likely to be stunted or cognitively impaired at 6²⁹ and 18³⁰ months after birth. The present results extend these benefits to include better interactions between nondepressed mothers and children receiving home visits. A possible limitation is that some of the effects for the individual dimensions had p values near the .05 threshold for statistical significance. If corrections for post hoc multiple comparisons are considered, then some of these effects might not be statistically significant based on an amended threshold. However, all analyses were based on a priori hypotheses and, as predicted, benefits in the quality of caregiving were not found for mothers who experienced depressive symptoms during pregnancy. These findings suggest that additional support might be needed to address mothers’ psychological challenges, in addition to targeting specific positive parenting behaviors.

An imbalance resulted in mother-child interactions for mothers reporting depressive symptoms in the HVC and SC. Despite home visiting, mothers with depressed mood during pregnancy did not score higher than nondepressed mothers in the SC on any maternal dimension of caregiving. However, children with mothers reporting depressed mood in the SC were more responsive and paid more attention to their mothers during the interaction than children of nondepressed mothers. Results from behavioral and biological paradigms examining mother-child interactions indicate that mothers and their children can influence attunement observed behaviorally and physiologically.⁵⁰ Evidence from the full RCT cohort suggests that maternal depression is associated with limited structural resources including lower rates of employment and income and higher rates of problematic consumption of alcohol and food insecurity.¹⁷ Thus, SC mothers experiencing depressed mood in the present subset also might experience more severe resource deprivations, limiting their physical and psychological resources to care for themselves and their children.

These findings support specific needs for children and their mothers experiencing depressed mood during and after pregnancy. Mothers experiencing depressive symptoms during pregnancy in LMICs need additional support to provide positive parenting. Future interventions can consider training MMs in specific cognitive-behavioral skills adapted for mothers experiencing depressive symptoms and in specific dimensions of high-quality caregiving.