The effect of psychosocial stimulation on cognition and behaviour at 6 years in a cohort of term, low-birthweight Jamaican children
Professor Susan P Walker at Epidemiology Research Unit, Tropical Medicine Research Institute, The University of the West Indies, Mona, Kingston 7, Jamaica. E-mail: firstname.lastname@example.org
Aim The aim of this study was to determine whether psychosocial stimulation up to the age of 2 years benefits cognition and behaviour at age 6 years in low-birthweight, term-born (LBW-T) children (gestational age ≥37wk, birthweight <2500g), and to compare LBW-T and normal-birthweight (NBW) children.
Method LBW-T Jamaican infants were randomized at birth to a control group or an intervention group. Children in the intervention group received psychosocial stimulation for 2 years. LBW-T infants were also compared with NBW infants born in the same hospital. IQ, cognitive function, and behaviour were measured at age 6 years in 109 LBW-T infants. The LBW-T group were divided into the intervention group (55 out of 70 enrolled, 32 females, 23 males; mean birthweight 2190g, SD 200g; and the control group (54 out of 70 enrolled, 33 females, 21 males; birthweight 2240g, SD 180g]. These were compared with 73 out of 94 enrolled NBW infants (38 females 35 males; birthweight 3130g, SD 330g).
Results Among the LBW-T children performance IQ scores were higher in the intervention group than in the control group (regression coefficient [B] 4.06, 95% confidence interval [CI] 0.01–7.98) as were visual–spatial memory scores (B 1.12, 95% CI 0.45–1.87). Children in the intervention group also exhibited fewer behavioural difficulties (B −2.21, 95% CI −4.13 to −0.10) than children in the control group. Compared with NBW children, LBW-T children in the control group had poorer selective attention (B=−3.35, 95% CI −5.59 to −1.26) and visual–spatial memory (B=−0.76, 95% CI −1.54 to 0.00), but there were no differences in IQ, language, or behaviour.
Interpretation Stimulation had sustained benefits in LBW-T infants. Finding few differences between LBW-T and NBW school-aged children concurs with results from other developing countries.
List Of Abbreviations
Low birthweight, term
Middle Childhood Home Observation for the Measurement of the Environment
Peabody Picture Vocabulary Test
Strengths and Difficulties Questionnaire
Low birthweight infants born at term (LBW-T) are below the 10th centile of weight for gestational age and account for 11% of births in developing countries.1 Evidence from industrialised countries suggests that LBW at term is associated with cognitive and behavioural deficits in children and adults.2,3
In developing countries, LBW at term is often associated with poverty and inadequate stimulation in the home, which would be expected to increase children’s risk of poor development. Developmental deficits associated with term LBW have been reported in infants in India4 and Brazil,5 and up to age 3 years in Guatemala.6 However, follow-up in later childhood in Brazil and Guatemala showed no significant difference in cognition compared with normal-birthweight (NBW) children.6–8
Studies from the USA have shown that intensive intervention combining home visits and centre-based education has sustained benefits in preterm LBW infants.9 We are unaware of any intervention with stimulation in LBW infants in developing countries.
We conducted a study in Kingston, Jamaica, to determine the benefits of psychosocial stimulation among LBW-T infants. Stimulation was provided up to age 24 months through weekly home visits and was found to benefit the children’s development at 7, 15, and 24 months.10,11 We also compared the development of LBW-T infants with that of NBW infants. Development up to age 24 months was poorer in LBW-T infants than in NBW infants.11
The objectives of this study were to determine whether psychosocial stimulation had sustained benefits on cognition and behaviour in LBW-T children at 6 years of age and whether LBW-T children still exhibited deficits in cognition and behaviour compared with NBW children at this age. Ability at this age, when children start primary school, may be especially important, as it is likely to affect school progress.
Infants were recruited from the main public maternity hospital in Kingston, Jamaica, from March to October 1999. A total of 140 LBW-T infants (gestational age ≥37 completed weeks, birthweight <2500g) and 94 NBW infants (gestational age ≥37 completed weeks, birthweight 2500–4000g) were enrolled. Inclusion criteria were residency within the Kingston metropolitan area and maternal education below three passes on national secondary level examinations. The maternal education criterion was used to recruit a predominantly low-income sample and reduce socio-economic differences between LBW-T and NBW infants. Infants were excluded if they were twins, had congenital abnormalities, were admitted to the special care nursery for more than 48 hours, or their mother had been tested and found to be HIV positive. NBW infants were matched for sex to two of every three LBW infants. The next eligible NBW infants born following the LBW infants were enrolled. The LBW infants were randomly assigned to control (n=70) or intervention groups (n=70). A sample size of 63 had 80% power to detect a 0.5SD difference in developmental quotient between the intervention and control group at the 0.05 significance level.11 A total of 130 LBW-T infants (63 in the intervention group and 67 in the control group) and 88 NBW infants completed the initial study, which continued up to age 2 years.
The follow-up at age 6 years was conducted from November 2005 to August 2006. We attempted to locate all children who completed the initial study. All children located were invited to participate in the follow-up, and the child’s mother or guardian gave written informed consent. Ethical approval for the study was given by the University of the West Indies ethics committee.
Infants in all groups were visited weekly from birth to 24 months by community health workers to obtain information on infant feeding and morbidity. The intervention was delivered to infants and mothers in the LBW-T intervention group during the weekly home visits for the first 8 weeks and from age 7 to 24 months. Before each phase, the community health workers received 2 weeks training in child development and the intervention. In the first phase, visits lasted 1 hour and focused on improving the mothers’ responsiveness to their infants. Mothers were encouraged to ‘converse’ with their infants, respond to their cues, show affection, and focus their attention on the environment. From age 7 to 24 months visits lasted 30 minutes, during which the community health worker demonstrated play techniques to the mother and involved her in a play session with her child. Praise and positive reinforcement were encouraged and physical punishment discouraged. Toys made from commonly available recyclable materials were left in the home, and mothers were encouraged to include play activities in their daily routines. Further details of the intervention are available in earlier reports of the study.10,11
Measurements at age 6 years
The children were assessed on the Wechsler Preschool and Primary Scale of Intelligence, 3rd edition (WPPSI-III) 12. The verbal and performance subscale scores and full-scale score are reported. Receptive vocabulary was measured with the Peabody Picture Vocabulary Test (PPVT) 13.
Auditory short-term memory was assessed with the digit span forwards test and visual–spatial memory with the Corsi blocks test.14
Attention was measured using the Test of Everyday Attention for Children.15 The test is designed for children aged 6 to 16 years. After piloting, three subtests were selected that were at an appropriate level for our population: two tests of selective attention (map search and sky search) and a test of attentional switching (opposite–same worlds).
Early reading ability
Reading was assessed with the Early Reading Assessment.16
Behaviour was assessed by maternal report using the Strengths and Difficulties Questionnaire (SDQ).17 Subscales for emotional symptoms, conduct problems, hyperactivity, and peer relationships were summed to give a total difficulties score.
Stimulation in the home
Stimulation and support provided at home was assessed with the Middle Childhood Home Observation for the Measurement of the Environment (MC-HOME).18
Maternal and social background variables
Maternal education and occupation were obtained by questionnaire. A measure of socio-economic status (SES) was derived by factor analysis of crowding in the home (persons per room), ratings of water supply and toilet facilities, and number of household possessions from a list of 14 items. The mothers’ vocabulary was measured on enrolment with the PPVT.
Height and weight of the children were measured using standard procedures; height-for-age and body mass index z-scores were calculated with the Centres for Disease Control references.19
Information on birthweight, gestational age, and birth order was available from the original study. 11
Tests and questionnaires which we had not used previously were piloted. We made minor modifications to the wording of some of the SDQ items in which words were unfamiliar to low-income Jamaican mothers. For example the word ‘clingy’ was not clear so the item ‘Nervous or clingy in new situations’ was changed to ‘in new places is nervous or holds on to you/stays next to you’. Factor analysis of the SDQ confirmed the five-factor structure, although there was some overlap between the emotional and peer problems scales. The internal consistency of the subscales was lower than previously reported. We had not hypothesized that any specific behaviour problem would be more affected by the intervention, and we wanted to reduce the number of statistical tests, and, therefore, reported only the total difficulties score.
For the MC-HOME, we changed the cut-off for the item concerning children’s books from 10 books to 5, as few homes had more than 10 books. We also made minimal changes to the WPPSI in the matrix reasoning subtest because two items were culturally inappropriate. For example, for the item ‘bird and worm’ the original answer of ‘squirrel and acorn’ was changed to ‘dog and bone’. Test–retest reliability (intraclass correlation coefficient, [R]) for all cognitive tests and the SDQ was determined after 2 weeks in 15 non-study children. One attention test, sky search, had low reliability (R=0.58) and was dropped. Test–retest reliability was 0.69 for the map search test, 0.73 for Corsi blocks, and for the others ranged from 0.80 (digit span) to 0.96 (WPPSI full scale).
All tests were administered by one research assistant and the SDQ and MC-HOME interviews by another. Both were blind to the children’s group assignment. Children were tested at our research unit accompanied by their mother or other familiar adult. MC-HOME observations and maternal interviews were conducted in a single home visit.
Dependent variables were checked for normality. Log transformation was used for attentional switching (opposite–same) and square root for the early reading assessment. The characteristics of the groups were compared using analysis of variance (ANOVA) or, in the case of categorical variables, χ2 tests. Associations between the cognitive and behavioural outcomes and potential covariates – characteristics of mother, home, and child – were determined using correlational analyses and ANOVA. For the analyses of the intervention trial, we considered the covariates to be possible confounders if they differed between the intervention and control groups and were correlated with the outcomes. None of the variables examined met these criteria. Intervention effects were, therefore, determined among the LBW-T infants only using separate univariate regression analyses for each outcome, with group codings of intervention=1, control=0. We did not correct for multiple comparisons.20
A second set of analyses was conducted to derive adjusted estimates of the difference between the two LBW-T groups and the NBW group. Two dummy variables were used to compare the LBW-T groups with the NBW group. Variables associated with two or more outcomes or which differed among the groups were included as covariates. All identified covariates (child age, sex, and birth order, gestational age, mother’s PPVT and post-school training [yes/no], SES, MC-HOME, and amount of time child had attended preschool) and the group variables were entered. To achieve a more parsimonious final model, covariates with p≥0.20 were removed. For both sets of models (intervention effects and comparison of birthweight groups), the bootstrap method was used to produce bias-corrected accelerated centile confidence interval estimates of the difference between the groups.
A total of 190 children were tested (LBW-T intervention, n=57; LBW-T control, n=55; NBW, n=78; 81.1% of enrolled). Two children in the LBW-intervention group, one in the LBW control group, and five in the NBW group were not included in any analyses because some social background data were missing. There was no significant difference in enrolment characteristics between those tested and those lost to follow-up (lost or with missing data), nor was there any consistent direction in differences with one group being better than the other. Mean birthweight (lost 2600g, tested 2580g), gestational age (lost 38.9wk, tested 38.9wk), and home stimulation score (lost 33.7, tested 33.5) were very similar; mean SES score was slightly higher in those lost (lost 0.14, tested 0.04), and mothers’ verbal ability slightly lower (lost 93.1, tested 96.6). The percentages of mothers <19 years at child’s birth (lost 25.0%, tested 20.3%) or with less than grade 10 education (lost 38.4%, tested 43.4%) were similar. There was no difference among the groups in the proportion of children lost to follow-up.
The characteristics of the tested children, their caregivers, and home background are shown in Table I. There was no difference in birth characteristics between the LBW intervention and control groups. The gestational age of the LBW-T group was lower than that of NBW infants, but the difference was small (mean difference 0.87wk, p<0.001). At age 6 years, the LBW-T children were shorter and thinner than the NBW children. However, the LBW-T children’s height-for-age was generally within the normal range, and only four children were stunted (height-for-age less than −2SD).
Table I. Child, caregiver, and home background characteristics of low-birthweight term (LBW-T) and normal birthweight (NBW) groupsa
| Females/Males, n||32/23||33/21||38/35||0.57|
| Birthweight (g)||2190 (200)||2240 (180)||3130 (330)||0.001|
| Gestational age (wks)||38.5 (0.9)||38.6 (0.9)||39.4 (0.8)||0.001|
| Birth order=1, n (%)||29 (52.7)||22 (40.7)||23 (31.5)||0.054|
| Age (y)||6.82 (0.12)||6.83 (0.11)||6.84 (0.10)||0.54|
| Height-for-age (z-score)||−0.21 (1.10)||−0.41 (0.95)||0.35 (0.97)||0.001|
| Body mass index (z-score)||−0.70 (1.28)||−0.76 (1.13)||0.29 (1.10)||0.001|
| Months at preschool||35.7 (7.7)||35.0 (8.4)||36.4 (7.7)||0.64|
|Type of school, n (%)|
| None||1 (1.8)||1 (1.9)||0||0.24|
| Preschool||1 (1.8)||4 (7.4)||4 (5.5)|| |
| Primary (government)||49 (89.1)||49 (90.7)||62 (84.9)|| |
| Preparatory (private)||4 (7.3)||0||7 (9.6)|| |
| PPVT||92.4 (24.0)||90.3 (26.3)||104.5 (21.7)||0.001|
| Age ≤18y at birth of child, n (%)||17 (30.9)||8 (14.8)||12 (16.4)||0.064|
| Education (n, %)|| || || ||0.23|
| ≤Grade 9||22 (40.0)||28 (51.9)||27 (37.0)|| |
| Grade 10–11||33 (60.0)||26 (48.1)||46 (63.0)|| |
| Post-school training, n (%)||6 (10.9)||8 (14.8)||15 (20.5)||0.33|
|Caregiver’s occupation, n (%)|
| Unemployed||24 (43.6)||23 (42.6)||26 (35.6)||0.8|
| Unskilled||5 (9.1)||7 (13.0)||7 (9.6)|| |
| Semi-skilled||18 (32.7)||20 (37.0)||30 (41.1)|| |
| Skilled/highly||8 (14.5)||4 (7.4)||10 (13.7)|| |
| Socio-economic statusc||−0.11 (1.08)||−0.09 (1.00)||0.10 (0.95)||0.41|
| MC-HOME||33. 5 (6.7)||34.0 (7.0)||33.5 (6.0)||0.9|
Most of the children attended government primary schools, and 93% were in grade 1. There were no significant differences among the groups in school type or amount of time the children attended preschool. Mothers of the NBW children had higher PPVT scores than the LBW-T groups (p<0.01). There was no other significant difference among the groups in caregiver or home background variables.
Cognition and behaviour at age 6 years
The children’s scores on the cognitive tests and the SDQ are shown by group in Table II, and the analyses for the effect of intervention are shown in Table III. The intervention group had significantly better scores in performance IQ and visual–spatial memory (Corsi blocks) and fewer behaviour difficulties than the LBW-T children in the control group (Table III). Additional regression analyses were conducted for performance IQ, Corsi blocks, and behaviour difficulties to examine whether there was an interaction between maternal PPVT score and intervention effects. The interaction term was not significant.
Table II. Cognitive test scores (mean and 95% confidence interval) and SDQ-total difficulties scores at age 6 years
| Full-scale||81.0 (78.5–83.5)||78.7 (75.9–81.6)||81.8 (79.6–84.1)|
| Performance||80.8 (78.0–83.7)||76.7 (73.8–79.7)||80.2 (77.7–82.8)|
| Verbal||82.3 (80.5–84.2)||81.9 (79.6–84.2)||83.8 (81.9–85.7)|
| Corsi blocks||6.7 (6.2–7.2)||5.6 (5.0–6.2)||6.6 (6.1–7.2)|
| Digit span||5.3 (4.9–5.7)||5.4 (5.0–6.0)||5.8 (5.4–6.2)|
| Map search||22.1 (20.1–24.2)||20.7 (18.9–22.5)||24.2 (22.6–25.8)|
| Opposite–same (log)a,b||1.40 (1.33–1.47)||1.46 (1.40–1.51)||1.48 (1.43–1.52)|
|PPVT||43.5 (39.7–47.3)||45.4 (41.2–49.6)||46.4 (43.4–49.4)|
|Early reading (square root)||5.46 (4.98–5.93)||5.33 (4.80–5.86)||5.58 (5.12–6.04)|
|Total difficulties (SDQ)a||14.5 (13.1–16.0)||16.8 (15.2–18.3)||16.2 (15.1–17.4)|
Table III. Estimates of effects of psychosocial stimulation on cognition and behaviour in low-birthweight children born at term
| Full-scale||2.28||−1.58 to 5.46||0.014||0.20|
| Verbal||0.40||−2.29 to 3.41||0.001||0.79|
| Corsi blocks||1.12||0.45–1.87||0.070||0.003|
| Digit span||−0.14||−0.76 to 0.48||0.002||0.67|
| Map search||1.42||−1.08 to 4.15||0.010||0.29|
| Opposite–same (switching) (log)a||−0.06||−0.14 to 0.03||0.016||0.17|
|PPVT||−1.93||−7.38 to 3.55||0.004||0.50|
|Early reading (square root)||0.12||−0.59 to 0.85||0.001||0.73|
| Total difficulties||−2.21||−4.13 to −0.10||0.040||0.029|
The results of analyses comparing the two LBW-T groups with the NBW children are shown in Table IV. Visual–spatial memory scores were lower in the LBW-T control group than in NBW children (Corsi blocks; p=0.051), as were selective attention scores (map search; p=0.003). Covariates in the models for visual–spatial memory were child age and sex, SES, and maternal PPVT and post-school training, and for selective attention they were child age and sex, SES, maternal post-school training, and child’s time at preschool. There was no significant difference between the LBW-T intervention group and the NBW group; however, in attentional switching (opposite–same) there was a tendency for children in the LBW-T intervention group to perform faster than the NBW children (p=0.053).
Table IV. Regression analyses of the effects of LBW-T compared with NBW on cognition and behaviour
| Full scale||−0.81 (−3.68 to 2.27, 0.60)||1.20 (−1.84 to 4.49, 0.44)||0.265||Sex, months at preschool, mother’s PPVT, SES|
| Performance||−1.45 (−4.88 to 1.85, 0.40)||2.49 (−1.17 to 5.92, 0.18)||0.206||Mother’s PPVT, SES|
| Verbal||0.27 (−2.13 to 3.08, 0.83)||0.38 (−2.11 to 2.86, 0.75)||0.217||Age, months at preschool, mother’s PPVT|
| Corsi blocks||−0.76 (−1.54 to 0.00, 0.051)||0.42 (−0.34 to 1.13, 0.25)||0.187||Age, sex, mother’s PPVT, SES, training|
| Digit span||−0.16 (−0.69 to 0.46, 0.59)||−0.31 (−0.91 to 0.20, 0.28)||0.118||Age, sex, months at preschool, mother’s PPVT|
| Map search||−3.35 (−5.59 to −1.26, 0.003)||−1.94 (−4.33 to 1.03, 0.14)||0.130||Age, sex, months at preschool, SES, training|
| Opposite-same (switching)(log)c||−0.02 (−0.09 to 0.04, 0.58)||−0.08 (−0.17 to −0.00, 0.053)||0.034||Age, sex|
|PPVT||2.46 (−1.89 to 6.76, 0.28)||0.14 (−4.11 to 4.18, 0.95)||0.260||Months at preschool, mother’s PPVT, SES, MC-HOME|
|Early reading (square root)||0.07 (−0.54 to 0.65, 0.82)||0.20 (−0.41 to 0.82, 0.52)||0.187||Mother’s PPVT, age, sex|
| Total difficultiesc||1.02 (−0.89 to 3.33, 0.34)||−1.39 (−3.28 to 0.62, 0.17)||0.158||Age, sex, gestational age, SES, MC-HOME|
The intervention benefited LBW-T infants in performance IQ, memory, and behaviour at age 6 years, with effect sizes (mean difference/SD) of 0.38SD for performance IQ, 0.40SD for total behaviour difficulties, and 0.53SD for visual–spatial memory. This is the only psychosocial stimulation intervention with LBW-T children that we are aware of from a developing country. The intervention was much less intensive than those from the USA carried out in preterm LBW infants,9 and comprised weekly home visits by paraprofessionals who trained mothers to be more responsive to their infants and to provide more stimulation. This approach would be feasible to conduct on a larger scale, especially in situations where community health workers are part of the existing health system. However, the programme would still require investment in training and supervision of the health workers, and would probably need to be targeted to children at highest risk of poor development.
The benefit of intervention on full-scale IQ in this study was less than that seen at age 7 years in stunted Jamaican children who participated in a similar intervention at a slightly older age,21 and there was no benefit on verbal ability. The smaller cognitive benefits in the LBW-T children compared with that found in the previous Jamaican study could be a result of intervention differences. The visits were of shorter duration than previously, and the LBW-T children were younger at the beginning and end of the intervention. It is possible that, if intervention had continued until the children attended preschool, more benefits would have persisted.
In contrast, in the present study, intervention was found to reduce behaviour problems, whereas this was not found in the previous study of stunted children at age 11 years.22 A possible explanation for the behavioural improvement is that we added a focus on improving maternal responsiveness to the intervention in the first 2 months of life.
Comparison of NBW and LBW-T children
Although the LBW-T children in the control group had significantly lower developmental levels than the NBW cohort at age 2 years, we found few differences in cognitive ability at age 6 years. This pattern of early developmental deficits compared with NBW children diminishing in later childhood concurs with the findings of other studies from developing countries of the later ability of LBW-T children.6,8 The finding of little difference between NBW and LBW-T children is contrary to our original hypothesis, which anticipated greater deficits in situations of poverty. It is possible that, for children growing up in poverty in developing countries, multiple deprivations overwhelm the relatively small effect of term LBW.
We found that selective attention scores were significantly lower in the control LBW-T group than in NBW children (effect size 0.49SD), as were visual–spatial memory scores (effect size 0.34SD). Attention has been reported to be one of the areas in which term infants who are small for gestational age experience difficulties in adolescence,23 but no such differences were found in a study of preschool children.24 Associations between intrauterine growth retardation and short-term memory difficulties at age 9 years25 and between birthweight and spatial working memory at age 8 years have been reported.26
Several social background variables affected the children’s development, including measures of the home environment (SES and MC-HOME), maternal characteristics (verbal ability and post-school training), and preschool attendance. The gestational age of the LBW children was slightly lower than that of NBW children, which might also be expected to contribute to differences; however, the range in gestational age was limited and it was not significantly associated with development. The findings suggest that, among disadvantaged poor children, LBW-T is associated with only a small additional risk of long-term poor development, and may be associated with deficits in specific domains rather than a global deficit. However, longer-term follow-up of the LBW-T children is needed to be certain.
The reduction in sample size will have affected the power of the study. However, differences between the birthweight groups were very small or absent, thus reduced power is unlikely to explain the lack of a statistically significant difference. The mean WPPSI scores in all groups were lower than the standardized norms. This is probably because the groups comprised only children from very poor backgrounds. It is also possible that, although we piloted the tests and made some modifications, cultural differences reduced the children’s performance. The scores are similar to levels found in poor Jamaican children on the Stanford–Binet and the WISC in previous studies.21,27 A further limitation, which is true of most studies comparing LBW and NBW children, is that the mothers of LBW children had lower vocabulary scores and may also have been less consistent and accurate in reporting information than mothers of NBW children.
In conclusion, a psychosocial intervention up to age 2 years had significant benefits to performance IQ, visual–spatial memory, and behaviour at age 6 years in LBW-T children. Findings to date suggest that in developing countries the impact of term LBW on cognition and behaviour in mid childhood is small.
What this paper adds
- • Psychosocial stimulation benefits cognition and behaviour at age 6 years in LBW-T children.
- • LBW-T children exhibited deficits in attention and visual–spatial memory but no global deficit in IQ compared with NBW children.
- • In conditions of poverty, other disadvantages may overwhelm the effect of term LBW.
We thank Sydonnie Pellington for testing the children and Mitchelle Archat for conducting the interviews. The study was supported by a grant from the Nestlé Foundation. The authors have no conflict of interest. The funders approved the design of the study. They had no involvement in the execution of the study, analysis, and interpretation, or the decision to publish.