Could cash and good parenting affect child cognitive development? A cross-sectional study in South Africa and Malawi

The sample included children between the ages of 5 and 15 years and their primary caregivers. Data were collected between 2013 and 2014 as part of the Child Community Care project, a study tracking the development of children and families affected by HIV attending established community based organisations (CBOs) across South Africa and Malawi. Eleven partner organisations (AIDS Alliance, Stop AIDS Now, Diana Memorial Fund, Firelight Foundation, Bernard van Leer foundation, REPSSI, World Vision, Comic Relief, Help Age, Save the Children and UNICEF) provided a list of all their funded CBOs. The list comprised 588 CBOs (524 in South Africa and 64 in Malawi). All 588 CBOs were stratified by funding partner and geographical location and 28 (24 in South Africa and 4 in Malawi) were randomly selected. All 28 CBOs agreed to participate in the study. Ethical approval was obtained from the ethics boards of University College London Research Ethics Committee (reference number 1478/002) and Stellenbosch University Health Research Ethics Committee (reference number N10/04/112) and authorised by each of the funding partners of the various community-based programmes. Caregivers received full information on the study and gave written consent for their own and their child’s participationon a specially developed informed consent form translated into local languages. Children were given information about the study in child-friendly local language and provided written assent on an assent form by writing their names or making another mark.

Data on the children were gathered by a combination of self-report and caregiver report. Questionnaires (for the child and caregiver) included a range of questions and standardised measures related to child’s health, education, psychosocial wellbeing, cognitive functioning and socio-demographic information. Questionnaires were translated into Zulu and Xhosa and converted to mobile phone technology for ease of data collection and to allow for live monitoring [24]. Children and caregivers were interviewed separately by trained data collectors and all data were entered live into mobile phones and captured via the Mobenzi system into a database. The cash transfer questions were available at time 2 of the data collection exercise (2013-2014) and were utilised in this analysis. At recruitment refusal rates were low (.7%).

Good parenting was operationalised based on a composite index of 10 items with a binary yes/no score. Children were asked four questions - whether they felt they belonged with the people at home, received praise, received treats and whether adults hugged as well as praised them (drawn from items of the Child Status Index tool [26]). Caregivers reported on 6 items – the use of positive discipline styles (explaining to the child when they did wrong deeds, taking away privileges as opposed to harsh punishments, and beatings), provision of consistent care, and absence of physical or emotional violence towards the child (drawn from items of the Parent-Child Conflict Tactics Scale [27]). A scale ranging from 0 to 10 was generated with 0 being the lowest score and 10 the highest score. The good parenting measure was then dichotomised to those scoring above 8 (n = 101) reflecting “good-enough parenting” and those scoring 7 or below (n = 732). This cut-off was chosen to reflect a high enough standard of parenting, as no participants scored 10, and only 1 caregiver scored 9 [28].

Five cognitive measures were employed in this study. Two were based on standardised tests which were administered by a fully trained objective data collector. Three were based on caregiver report according to a standardised disability inventory. These included the Draw-a-person (DAP) Test, a screening test used as an indicator of nonverbal cognitive ability based on children’s drawings of human figures [29]. Children were asked to draw a picture of themselves, a man, and a women. Drawings were then assessed using the Draw-a-Person Quantitative Scoring System (QSS), which analyses 14 different aspects of the drawings, such as specific body parts and clothing, for various criteria, including presence or absence, detail, and proportion. Overall, there are 64 scoring items for each drawing. All drawings were coded and marked by a researcher who was blinded to the child’s identity at the time of assessing the drawings. An age-standardised score was recorded for each drawing, and mean scores were calculated (scale ranges 40-130). There are few cognitive screening tools for young children in Sub-Saharan Africa and this test was considered the most appropriate. This revised version of DAP has been previously used in African countries [30‐32]. Additionally, the use of a nonverbal, quick and easy-to-administer task has the advantage of eliminating potential sources of bias, including primary language, verbal skills, or communication difficulties. The Digit Span Test is a subtest of the Wechsler Intelligence Scale for Children (WISC-IV) and measures attention and working memory [33]. The test consists of repeating dictated series of digits (e.g., 4 1 7 9) forwards and other series backwards. Series begin with two digits and keep increasing in length with two trials at each length. A total scaled score for the two recall conditions was computed (range 0-20). The scaled score is an age-based, norm referenced score for each child, based on a large nationally representative norm sample of South African children [34]. Primary caregivers were asked to report on child functioning and disability in three cognitive domains: learning, remembering new things, and comprehension. These questions were taken from a newly developed disability measure [35] for use in low and middle income settings. Ratings were in a 3-point difficulty scale: 0 (no difficulty), 1 (some difficulty), 2 (a lot of difficulty), 3 (cannot do at all). Mean scores were computed for each domain, and a total score was calculated for all 3 domains combined.

A five-stage analysis strategy was carried out in IBM SPSS 22.0. First, we looked at differences between those receiving a cash grant (at least one of six possible grants into the family) and those who received no grant at all on demographic variables and five cognitive measures: non-verbal cognitive ability (assessed using draw-a-person test), short-term memory/attention (measured using digit span test), and difficulty or disability in three cognitive domains: learning, remembering new things, and comprehension. Second, we examined associations between quality of parenting and child cognitive outcomes. Third, a cumulative “cash and good parenting” scale was hypothesised: no support (0), cash grant receipt(1), good parenting (based on existing evidence of impacts of positive parenting) (2), integrated cash and good parenting (3), and coded both as ordinal and as dummy variables for use in regression models. A series of ANOVA analyses tested associations between types of provision (cash, good parenting or both) and all five cognitive measures. Fourth, a series of linear regression models were used to further examine associations of cash, good parenting, and combined provision (represented by dummy variables, taking “none” as the reference category) with cognitive outcomes. Model 1 shows unadjusted associations between types of social protection and cognitive outcomes and Model 2 included potential co-factors predicting either cognitive development or receipt of social protection (child gender, age, HIV status functioning or disability, and number of household assets). Draw-a-person and digit span tests are age-adjusted, thus child age was not included as a covariate in multivariate regression analyses. Fifth, regression analyses disaggregated by HIV status and using interaction terms were used to examine whether receiving cash support, having good parenting or both had differential effects on cognitive outcomes of HIV positive and HIV negative children.

Data from a total of 854 children in South Africa (n = 708) and Malawi (n = 146) were analysed. 52.3% were female, and ages ranged from 5 to 15 years (M = 10.19, SD = 2.81). Primary caregivers reported that 13.5% of children (n = 115) were HIV positive. Of those, 112 (97.4%) were receiving medical treatment. Overall, 108 children (13.3%) were living in informal dwellings and most households lacked essentials such as a refrigerator or a stove (mean of 3.90 out of 10 household assets). Of the six possible grants available to families, 60.9% of caregivers reported they received just one grant (n = 520), 7.4% received two, and only 0.2% received three. 73.1% of caregivers (n = 624) reported receiving at least one cash grant; yet, 26.9% reported no cash grant at all, despite the fact that socio-economic status indicators showed high levels of deprivation.

Grant receipt according to HIV status of the child showed that HIV positive children were less likely to get a cash grant compared to HIV negative children (60.0% versus 75.3%, X²(1) = 11.89, p = 0.01). Differences between children residing in households receiving a grant and those not receiving are set out in Table 1 below.

Cognitive outcomes were measured for all children using the digit span test, the draw a person test and three items from the UNICEF disability inventory (learning, remembering new things and comprehension). The mean score for the Draw-a-Person test was 91.25 (SD = 17.28) which falls within the norm group scores (ranging between 90 and 109). A total of 361 children (43.3%) had scores below the normative scaled score mean of 90. The mean Digit Span scaled scores for the entire group was 8.97 (SD = 3.56). Less than half of children (44.8%, n = 371) had scores at or below the normative scaled score mean of 10 [33]. Children scored low in the severity scale for the three cognitive disability domains: mean for learning difficulty was 0.20 (SD = 0.47), mean for remembering new things difficulty was 0.34 (SD = 0.58), and mean for comprehension difficulty was 0.04 (SD = 0.24). Children in households receiving grants showed better cognitive outcomes as set out in Table 1 below.

A total score on 10 dimensions of parenting provided for a working definition of good parenting with 0 being the lowest score and 10 the highest score. The mean score of the parenting scale was 6.46 (SD = 0.98), and higher scores were significantly associated with better cognitive outcomes. More specifically, higher parenting scores were associated with better performance on draw-a-person test (B = 1.98, 95% CI: .79, 3.17, p = .001), and on digit span test (B = .37, 95% CI: .13, .62, p = .003). Higher scores on the parenting scale were also positively associated with less severity in learning difficulty (B = −.049, 95% CI: −.08, −.02, p = .003), and less severity in remembering difficulty (B = −.06, 95% CI: −.10, −.20, p = .003). There was no difference according to parenting score on comprehension difficulty score. For the purpose of the next set of analyses, good parenting was dichotomised to those scoring above 8 (n = 101) seen as good parenting group, and those scoring 7 or below (n = 732) as not good parenting, and consequently a cut-off of 8/10 was chosen to reflect ‘adequate parenting’ as no caregivers scored 10/10 and only 1 caregiver scored 9/10.

Of the total sample, more than half of children lived in households receiving cash support (55.4%, n = 473), only 6% of children (n = 51) received care above the cut off point for good parenting but no cash, 17.7% (n = 151) received combined cash support and had good parenting, and 179 (20.9%) received none of those. A series of univariate ANOVA analyses tested associations between types of social protection and five cognitive measures: non-verbal cognitive ability (assessed using draw-a-person test), short-term memory/attention (measured using digit span test), and difficulty or disability in three cognitive domains: learning, remembering new things, and comprehension. For all cognitive outcomes, apart from the comprehension difficulty score, cash plus parenting above the cut-off was associated with better outcomes. Statistically significant associations are illustrated in Figs. 1, 2 and 3. As shown in Figs. 1 and 2, as provision increased from no support to cash plus good parenting, child cognitive performance improved. Cash plus good parenting access was also positively associated with less severity in two cognitive difficulty/disability domains: learning and remembering new things (see Fig. 3).

Unadjusted linear regressions examined associations of cash, care, and combined cash plus good parenting (Table 2) (represented by dummy variables, taking “no support” as the reference category) with all cognitive outcomes measured (Model 1). Compared with no support, cash receipt was associated with better performance on draw-a-person test (scaled scores ranged between 40 and 130) (B: 15.57; 95% CI 12.81-18.33, p < .001) and cash plus good parenting was associated with greater performance (B: 18.66; 95% CI 15.17 - 22.15, p < .001). Cash receipt was also associated with higher scores on digit span test (scaled scores ranged from 0 to 20) (B: 1.33; 95% CI: .72-1.95, p < .001), and cash plus good parenting was associated with an almost twofold improved score (B: 2.13; 95% CI 1.35-2.90, p < .001). Compared to no support, receiving cash was associated with lower scores in learning difficulty (B: −.17; 95% CI: −.25, −.09, p < .001), and cash plus good parenting was associated with the lowest level of difficulty (B: −.24: 95% CI: −.34, −.14, p < .001). Receiving cash plus good parenting was also associated with lower scores in remembering difficulty (B: −.21; 95% CI: −.34, −.09, p = .001). When combining the three indicators into an overall score of cognitive difficulty, we found that receiving cash was associated with lower difficulty scores (B = −.27, 95% CI: −.45, .09, p = .003), and that cash plus good parenting was associated with a greater reduction in cognitive difficulties (B = −.47, 95% CI: −.70, 95% CI: −.70, .25), p < .001).

In multivariate linear regressions (Model 2, Table 2), after controlling for factors predicting cognitive development or receipt of cash plus having good parenting (child gender, age, HIV status, functioning or disability, and number of household assets), combined cash plus good parenting remained a strong predictor. Children receiving cash plus having good parenting had higher scores, both on draw a person test (B: 16.01; 95% CI12.45-19.57, p < .001) and digit span test (B:1.73; 95% CI.94, 2.51, p < .001). Being HIV positive and having a disability also remained significant predictors of cognitive performance. After adjusting for significant cofactors, receipt of cash was no longer associated with cognitive difficulties, but combined cash and good parenting was significantly associated with lower scores of cognitive difficulties (B: −.30, 95% CI: −.53, −.07, p < .001), and in particular with lower severity scores in learning difficulty (B: −.17; 95% CI: −.28, −.06, p = .02) and difficulty in remembering new things (B: −.13, 95% CI: −.27, −.001, p = .04). No significant effect for comprehension was found.

HIV positive children had a significantly poorer performance in cognitive tests and greater difficulty/disability scores compared to the HIV negative group. In a series of linear regressions using interaction effects, we tested whether the effects of cash or/ and good parenting on cognitive outcomes differed by child HIV status (Table 2). For draw-a-person test and compared to no support, receiving cash was associated with better performance in both groups. Good parenting had a positive impact on performance for the draw-a-person test, particularly amongst HIV positive children (B = 9.83, (95% CI: -1.25, 20.92) compared to HIV negative children (B = 5.89, 95% CI: 5.89, 95% CI: .35, 11.43)p = 0.036. Cash plus good parenting had an additive effect on cognitive performance in both groups. Receiving cash was also associated with better performance in the digit span test, in particular for the HIV negative group (B = 1.34, 95% CI: 1.34, 95% CI: .67, 2.01) compared to HIV positive children (B = .90, 95% CI: -2.63, 2.46), p = .02. For the cognitive components in the disability measure (learning, remembering and comprehension difficulty), as provision increased from no support to cash plus good parenting, difficulty severity scores were reduced for both groups. We also noted that good parenting was associated with lower comprehension difficulty for the HIV negative children (B = .02, CI: −.04, .09) compared to the other group (B = .10, 95% CI: −.18, .38), p = .008, and also a lower overall cognitive difficulty score, particularly amongst the HIV negative group (p = .03).

Vulnerable children (Table 3) were defined as being HIV infected, boys and girls living in informal housing, and those with a disability. For receipt of cash alone, there were no differences by gender and disability, but higher likelihood of cash receipt amongst children in South Africa (66.8%, p < .001), informal dwellers (69.2%, p = .001) and younger children (aged 5 to 9) (59.8%, p = .04). HIV positive children were significantly less likely to live in households receiving a cash grant (45.2%, p = .02); yet they were more likely to receive better care (good parental practices) (10.2%, p = .03). Overall, only 151 children (17.7%) received combined cash support and good care. Children with a disability were more likely to receive cash plus care (19.4%), but there were no differences amongst other risk groups (HIV infected, informal dwellers, or younger age).