Infancy and early toddlerhood are critical periods for the development of motor abilities, and nearly all children acquire specific motor skills due to brain and neuromuscular maturation (Purves, 1994). Engagement in daily routines supports the development of fine and gross motor skills in young children (Kariger et al., 2005). On the other hand, restriction in motor performance may interfere with optimal functioning in several spheres of life.
Individual variations in motor development are evident from an early age and a child’s progression across time may be unstable or steady (Pollitt & Triana, 1999). The age and sequence of motor development varies both within and across individuals perhaps due to caregiving practices and the opportunities for practice. Across genders, sex differences in motor abilities begin to be seen early in childhood (Rademeyer & Jacklin, 2013) and may be attributed to physical differences among boys and girls or to the influence of cultural socialisation. For example, boys may be pushed towards more vigorous outdoor activities, while girls may be encouraged to engage in more quiet indoor activities. Nutritional status influences on motor development are seen directly through their effects on the structural and functional development of the brain, or indirectly through their influence on children’s experiences and behaviour (Grantham-McGregor et al., 2007). In modelling the potential mechanisms of the influence of nutrition on child outcomes, Prado and Dewey (2012) explicate that, ‘undernutrition affects motor development, which in turn may influence brain development through both caregiver behaviour and child interaction with the environment.’
With maturation and as children’s environments become increasingly differentiated, the importance of biological factors decreases as other factors within the environment begin to be more influential. In line with Bronfenbrenner’s bioecological theory (Bronfenbrenner & Morris, 2006), factors which are closest to the child may exert the strongest influence while those that are more distal to the children may show less impact. Some of these associations may be stronger during infancy while others are strongest during toddlerhood, suggesting that changes occur in the pattern of influences as the child matures. In view of the variety of child-related and environmental factors that may influence the course of motor development at this age, an investigation of their impact will be useful in identifying the risk factors for poor outcomes in resource-constricted settings.
The assessment of infants using tools that have been validated in context will provide an accurate picture of child functioning at this age. The Kilifi Developmental Inventory (KDI:
Abubakar, Holding, et al., 2008) provides a valid and reliable assessment of motor functioning in early childhood among rural African populations. The tool has been extensively applied in Kenyan
83
studies of typically developing children and HIV-infected populations (Abubakar, Holding, Newton, van Baar, & van de Vijver, 2009; Abubakar et al., 2013; Abubakar, van de Vijver, et al., 2008). Its use has also been reported in studies among infants in several sub-Saharan African contexts including Malawi (Prado et al., under prep.), Ghana (Prado et al., under prep.) and South Africa (Mathe, 2011). However, there are currently no studies reporting the use of the KDI in longitudinal studies. Our study therefore set out to test two hypotheses. The first hypothesis tested whether or not within-person changes occurred in mean KDI scores over time. The second hypothesis tested whether or not there were variations in intra-individual influences on KDI scores over time.
Method
Detailed information on the study setting, sample, data collection tools and procedures is presented in Chapter 3.
Study Setting
The study was conducted in Msambweni District.
Study Sample
The sample consisted of 231 infants across the 3 ages, 12, 18 and 24 months.
Data Collection Tools
The KDI was used to obtain information on psychomotor functioning in infants at the three time points.
Analysis
Stability of scores across time was examined using correlation coefficients. We correlated test scores at 12, 18 and 24 months with one another.
To detect if there were any overall differences in changes in mean scores across all the ages, an ANOVA with repeated measures was performed with time points as independent variables and KDI scores at 12, 18 and 24 months as the dependent variables. Only the scores of the children who were seen at all three time points were included in this analysis (n = 150). We tested the homogeneity of variance assumption for the KDI scores using Mauchly’s Test of Sphericity. Main effects were tested using the Bonferroni correction.
To identify correlates of psychomotor development at each time point, I first identified child characteristics pertaining to proximal influences (child gender and anthropometric status), key variables that reflected maternal demographics (maternal age and partner’s contribution) and other more distal influences such as household socioeconomic status (parental education and occupation and home ownership). I fit separate, univariate models between each of these variables and KDI scores at 12, 18 and 24 months. Effect sizes were measured using partial eta squared.
84
In the multivariate analysis, I then constructed regression models that included all variables that had effect sizes of more than .01 in the univariate analysis. Using the stepwise method, the best predictors were entered into the model until no more variables met the entry criteria. Given the potential for collinearity between the anthropometric measures, I included stunting and wasting, but not underweight in the mulitivariate model if all three variables had effect sizes > .01 in the univariate analysis. Stunting and wasting reflect chronic and acute undernutrition, respectively, but underweight does not distinguish between the two and is correlated with both. Significance was based on a p-value of < 0.05.
Results Descriptives
At infancy, the mean age at follow-up was 12.15 (SD = .31. range: 11.56, 14.09). During early and late toddlerhood, the mean age at follow-up was 18.13 (SD = .32, range: 16.72, 19.68) and 24.14 (SD = .29, range: 23.46, 25.82), respectively. The proportion of boys (n = 102, 48.3%) and girls (n = 106, 50.2%) was nearly equal. One-tenth (n = 22) of the sample was wasted while 16% (n
= 33) was stunted. Nine children (4.3%) were both stunted and wasted. Mean WAZ and HAZ were -.5259 (SD = 1.17, range: -3.86, 4.29) and -.6802 (SD = 1.24, range: -3.73, 4.51), respectively.
The majority of mothers were aged between 20 and 34 years (n = 148, 70%). Nearly all the mothers (92.9%) received financial support from a partner. Almost two-thirds of the mothers (61.6%) compared to less than half of fathers (41.2%) had attained less than 8 years of education.
More than half of the mothers (54.5%) compared to 13.3% of the fathers were not employed.
Slightly more than half of the sample (50.2%) lived in their own homes.
Retention Rates
A total of 231 children were seen at 12 months, 184 at 18 months and 186 at 24 months. The number seen at each time point represents an attrition rate of 8.7% at 12 months, 20.3% at 18 months and 19.5% by the end of the study. There were no differences by gender between those who completed the KDI and those who did not at 12 months (χ2(1, n = 208) = .259, p = .611), 18 months (χ2(1, n = 181) = .179, p = .672) and 24 months (χ2(1, n = 183) = .010, p = .920). Reasons for loss to follow-up included travel outside the study area and partners’ refusal for continued participation. Although none of the mothers indicated that they no longer wished to be part of the study, repeated unsuccessful visits (at least 3) to some homes suggested loss of interest in continued participation in the study.
Stability of Test Scores
KDI scores at 12 months correlated moderately with the scores at 18 months (r = .306, p <
.001). The correlation between scores at 18 and 24 months was also moderate, r = .284, p < .001.
85
Scores between 12 and 18 months had significant but weak correlations, r = .222, p = .004.
Repeated Measures Results
Mean KDI scores differed significantly across time points, F(1.678, 250) = 757.477, p <
.001, ƞp
2 = .836. Mauchly’s Test of Sphericity indicated that the assumption of sphericity had been violated, χ2(2) = 31.592, p < .001, and therefore a Greenhouse-Geisser correction was used. Post hoc tests using the Bonferroni correction revealed that the increase of KDI z-scores between 12 and 18 months (27.18 ± 2.72 vs. 35.00 ± 2.3, respectively) and between 18 and 24 months (35.00 ± 2.3 vs 39.67 ± 4.23, respectively) were statistically significant (p < .001).
Effects of Background Variables
A summary of the differences in KDI scores according to the various background characteristics is presented in Table 20. Effect sizes of these differences are depicted in Figure 4.
Child variables
Gender. Boys had consistently higher scores than girls at all ages, and these differences became significant at 24 months, F(1,181) = 6.62, p = .011. Medium effect sizes were also seen at 24 months (ƞp2 = .035).
Wasting. Children with wasting had significantly lower scores at 12, 18 and 24 months, and effect sizes for these differences were strongest at 24 months (ƞp2 = .049).
Stunting. Children with growth retardation had lower scores than their counterparts who were not stunted across all ages and this effect was significant, F(1,177) = 4.211, p = .042, and strongest (ƞp
2 = .023) at 18 months.
Underweight. Children who were underweight had lower scores across all time points.
These differences were significant at 12 months, F(1,202) = 10.468, p = .001, with medium effect sizes (ƞp
2 = .049).
SES variables
Maternal age. Effects of maternal age on test scores increased as children became older although differences seen were not significant. The pattern of differences in scores was not consistent. At 12 months, children of the youngest mothers had the lowest scores. At 18 months, children of mothers aged between 20 and 34 years had the lowest scores while at 24 months, children of the oldest mothers had the lowest scores.
Support for upkeep. Children whose mothers reported that they did not have partners or did not receive any support from their partners had higher scores across all ages. Although not significant, the effects of this variable were strongest at 18 months.
86
Figure 4. Effect sizes of associations between background variables and KDI scores
Maternal education. The influence of maternal education varied at the different time points.
Differences in test scores according to maternal education were only seen at 18 months, with children of mothers with between 8 and 11 years of education attaining the highest scores.
However, these differences were not significant.
Paternal education. Children whose fathers had less than 8 years of education obtained lower scores across all ages, and the effects of this differences were strongest at 12 months (ƞ p
2 = .023).
Maternal occupation. There was no pattern to the differences seen in children’s scores according to maternal occupation at any of the ages although the strongest (but small) effects were seen at 18 months (ƞp
2 = .010).
Paternal occupation. At 12 months, children of fathers who were not employed had the highest scores while at 24 months, children of fathers employed at Level 3 performed better than children whose fathers were at employed at other levels. The effect sizes of these differences, although small across all the ages, were strongest at 24 months ((ƞp
2 = .027).
Home ownership. Children who lived in homes shared with members of the extended family had the highest scores across all ages. The effect of these differences, although not significant at all ages, was strongest at 12 months ((ƞp
2 = .011).
Multivariate Results
A summary of the results of the regression analysis at the different time points is presented in
87
Tables 21, 22 and 23. None of the child-related variables was associated with KDI scores in the multivariate analysis at 12 months. For instance, although wasting was significantly associated with child outcome in the univariate analysis at 12 months, this significance was not retained in the multivariate models. Only home ownership and paternal education were significant predictors of KDI scores at 12 months, explaining 6.1% of the variance observed. The final regression model (Model 2) was significant, F(2, 163) = 5.317, p = .006. Being coded ‘1’ for home ownership (those living in homes shared with members of the extended family) increased estimated KDI scores by .969 while being coded ‘0’ for paternal education (having less than 8 years of education) reduced KDI scores by .851.
At 18 months, the final model (Model 2) was significant, F(2, 176) = 5.009, p = .008. HAZ and paternal occupation explained 5.4% of the variance observed. Children who were not stunted scored higher on the KDI by .379 points compared to their counterparts who were stunted. Being coded ‘1’ on paternal occupation (Level 1 – low income and uncertain) predicted an increase of .690 on KDI scores.
The final model (Model 3) was significant at 24 months, F(3, 142) = 6.680, p < .001.
Gender, maternal age and stunting explained 12.4% of the variance observed on KDI scores.
Having a mother aged between 20 and 34 years increased estimated KDI scores by 1.791. For females, the predicted KDI scores at 24 months were nearly 2 points lower than those for males.
HAZ had the smallest effect, and for every unit increase, there was an increase of .833 on KDI scores.
Discussion
The undesirable consequences of poor motor development in early childhood may only be detected at later ages when skills become more differentiated. As many of these impairments may be persistent, early identification of risk factors for poor motor skills in children is important, providing an opportunity to minimise their impact on overall child development through implementation of targeted preventive intervention programmes. The current study set out to identify risk factors for poor motor outcomes between the ages of 12 and 24 months, and whether or not the effects of the identified risk factors changed or remained the same.
The attrition rates seen in the current study were within acceptable ranges. The reasons for loss to follow-up suggest that attrition rates may be related to social factors, such as the availability of support from a partner. Future studies should therefore consider factors such as long-term residency and partner involvement when recruiting participants into similar studies. This is especially so in patriarchal societies where male members of the family are the chief decision- makers on many issues affecting the family.
88
The strength of the correlation of scores between 12 and 18 months, and between 18 and 24 months was of a similar magnitude. As expected, scores for adjacent time points were more highly correlated than for those that were not. This finding suggested that the change in infants’ scores between 12 and 18 months, and between 18 and 24 months was stable, and that scores increased in tandem. However, the low correlations between initial and final scores suggested that the trajectory of motor development from 12 to 24 months was unstable. These findings are in line with previous studies which have found that motor developmental trajectories vary considerably among healthy children (Darrah, Hodge, Magill-Evans, & Kembhavi, 2003; Roze et al., 2010).
Mean KDI raw scores increased substantially more between 12 and 18 months than between 18 and 24 months, suggesting that they were sensitive to maturational changes in children. This finding demonstrates that rapid changes take place in motor development between 12 and 18 months whose rate slows down from then until the age of 2 years. In line with what Darrah et al., (2003) have suggested, it could be that the longitudinal motor performance of typically developing infants is variable and non-linear, rather than constant.
Effects of Background Variables
Boys consistently scored higher than girls, with the magnitude of gender differences in KDI scores increasing with age, corroborating findings from other studies. For example, in a review of the sex differences in the activity levels of infants, Campbell and Eaton (1999) consistently found that boys were more active than girls, even from a young age. These differences in activity levels, an important component of motor skill development, were attributed to biological processes and socialisation experiences. The same may apply to the current study context. Mothers’ expectations of their children’s performance of various tasks may show a gender bias (Mondschein, Adolph, &
Tamis-LeMonda, 2000) as they may encourage boys to engage in vigorous outdoor play activities while girls are pushed toward more quiet indoor activities (Lever, 1976).
The finding that mean KDI scores were lower in undernourished children than their well nourished counterparts reinforces findings from previous studies linking poor nutritional status to impaired development (Y. B. Cheung, Yip, & Karlberg, 2001; Kuklina, Ramakrishnan, Stein, Barnhart, & Martorell, 2006; McDonald et al., 2013; Walker et al., 2007). Past research however does not clarify the precise mechanism through which undernutrition is linked with poor motor development during infancy. We surmise, like has been done in past studies, that undernutrition may directly affect motor development through its impact on the central nervous system and brain maturation processes; or indirectly through decreased opportunities for interaction with the environment and with caretakers (Kuklina et al., 2006). Taking anthropometric measurements at each time point that children were tested on the KDI allowed us to capture the pattern of the
89
changing influence of nutritional status over time. The pattern of change in effect sizes observed in the current study suggests at the influences of underweight, stunting and wasting were strongest at 12, 18 and 24 months, respectively. The findings of the current study demonstrated that each of these indices of anthropometric status (HAZ, WAZ and WHZ) exert independent effects on outcome at different ages during infancy and toddlerhood. Such information is useful in timing nutritional interventions during infancy.
The current study did not find significant effects of maternal age on child outcome, even though the magnitude of its effects increased with child age. This finding suggests that interventions should be put in place earlier rather than later. The null effects of maternal age on child outcome have been reported previously in low-income settings (Kuklina et al., 2006). On the contrary, some studies in high-income settings have reported that higher maternal age predicts low developmental scores among infants (Alvik, 2013).
Surprisingly, children whose mothers reported the lack of a partner from they received any support obtained higher scores, and more so at older rather than at younger ages. This finding is contrary to research which indicates that growing up in a family structure headed by a single parent has negative implications for child development (Bain, Boersma, & Chapman, 1983; Downey, 1994). The availability of other forms of social support that such mothers may rely on from members of the extended family may explain our findings. It could be that other family members provide numerous opportunities for motor stimulation through their interaction with the child on various levels (de Paiva et al., 2010). The increase in the influence of family support at older ages supports the notion that environmental influences become stronger as children move away from exclusive caregiving by the mother. Access to a broader support network in such contexts may thus ameliorate the negative effects of the absence of a supportive partner on a mother, as the members of this network provide a mother with direct (caregiving) or indirect assistance (advice on child- rearing practices), which in turn impacts a child’s developmental status.
The lack of significant associations between maternal education and KDI scores across all ages, as well as small effect sizes, confirmed previous findings among infants in both low- (Abubakar, Holding, van de Vijver, Newton, & van Baar, 2010; Kuklina et al., 2006) and high- income settings (Ravenscroft & Harris, 2007). However, other studies have reported a strong association between maternal education and fine (but not gross) motor skills at 18 months (Koutra et al., 2012). These inconsistencies may be related to the manner in which maternal education was conceptualised (none/some vs number of years of schooling) or to the age of the child at which these measures are taken. On the other hand, paternal education seemed to exert stronger effects on child motor outcomes, more so at younger ages, suggesting that having an educated father may