Effect of socioeconomic status on psychosocial problems in 5- to 6-year-old preterm- and term-born children: the ABCD study

This study is part of the ABCD study, an ongoing, multi-ethnic, population-based, prospective birth cohort to examine and determine factors in prenatal and early life that might explain the later health of the child and differences in health between children (http://​www.​abcd-study.​nl). The cohort study design has been described previously [17].

Figure 1 shows the study procedure and inclusion in the current analyses. For this study we included only those children for whom gestational age, maternal educational level and SDQ data from the mother or the teacher were available. The SDQ was part of a larger 5-year questionnaire filled in by the mother. Twins were excluded from the ABCD study at an earlier stage. Of the 12,373 women approached, 8266 women filled out the pregnancy questionnaire (response rate: 67 %). 6161 mothers who gave permission for follow-up were approached for the 5-year follow-up measurement of their child. For the present study, the final sample consisted of 4553 children (response rate 74 %); mean age 5.2 years, 50.2 % boys and 48.8 % girls. Of these participating children, 4336 were term born (mean gestational age 39 + 5 weeks) and 217 were preterm born (mean gestational age 34 + 3 weeks). Of the 217 preterm-born children, 24 were born very preterm (<32 weeks’ gestation, mean 29 + 1 weeks) and 193 were born moderately preterm (≥32 and <37 weeks’ gestation, mean 35 + 0 weeks).

Psychosocial problems were measured with the Dutch version of the SDQ [18]. The SDQ 25 items divided into five subscales: emotional symptoms, conduct problems, hyperactivity/inattention problems, peer problems and prosocial behaviour. Each item has three response options (not true = 0, somewhat true = 1 and certainly true = 2).The summed score of the first four subscales provides a total difficulties score. A higher total difficulties score indicates a higher risk for psychosocial problems [19]. The Dutch cut-off score for the total difficulties score for the parent-rated SDQ for 7- to 12-year-old children is ≥11; this cut-off score is commonly used in practice in Youth Health Care in the Netherlands [20]. A score of ≥11 means that possible psychosocial problems need further follow-up. SDQ scores can also be used as continuous variables [21]. In the ABCD study the SDQ data were handled following standard procedures; this means that a subscale score was calculated for all children with less than two items missing on that subscale. Further information on the SDQ and scoring is available at http://​www.​sdqinfo.​com. The validity and reliability of the total difficulties score of the parent and teacher SDQ are satisfactory for detecting psychosocial problems in children in a multi-ethnic society [22]. The subscales of the SDQ should be interpreted with caution because the reliability of the SDQ subscales is poorer compared to the total difficulties score, especially in children with a non-Western ethnic background and in 5- to 6-year-old children [21, 22]. In our sample the mother-reported total difficulties score of the SDQ had a Cronbach’s alpha of 0.69, the subscales had alphas of 0.55 for emotional symptoms, 0.46 for conduct problems, 0.75 for hyperactivity/inattention problems, 0.54 for peer problems and 0.63 for prosocial behaviour. These alphas were in line with those found in previous studies [21, 23].

The self-reported maternal education level and perceived income adequacy were used as separate SES indicators [10]. These data were collected from the 5-year questionnaire. In 42 children, missing information on maternal education level in the 5-year questionnaire was supplemented from the pregnancy questionnaire.

Maternal education level was classified as: high (degree higher vocational education/university), mid (degree higher vocational secondary education/academic secondary education/intermediate vocational education) or low (primary school/technical secondary education/lower vocational secondary education) [24]. Perceived income adequacy was classified as inadequate, adequate or more than adequate income to live.

Variables known to be associated with psychosocial problems were analysed as possible confounders. Potential confounders were: gender, ethnic origin, maternal age, parity, smoking during pregnancy, family structure and maternal stress and anxiety. From the pregnancy questionnaire we obtained information on parity (nullipara or multipara), smoking during pregnancy (yes or no) and ethnic origin (Dutch, Turkish, Moroccan, African or other). Ethnic origin was defined by the country of birth of the mother of the pregnant woman [25]. From YHC registration we obtained gender, birth weight and gestational age; small for gestational age was scored based on gender and parity-specific reference curves [26]. From the 5-year questionnaire we obtained age of the mother, family structure (living with both parents or one parent, number of siblings) and maternal anxiety and stress; these were assessed by the Depressive Anxiety and Stress Scale 21 (DASS21) [27].

Next, we calculated mean and standard deviations (SD) for SDQ total difficulties scores as well as subscale scores across levels of maternal education and perceived income adequacy for term and preterm children. We analysed total difficulties scores and subscales as continuous data, but also calculated % of children scoring above cut-off of the mother-reported SDQ. SDQ differences (total difficulties scores, as well as the subscales) between preterm- and term-born children were assessed with linear regression. SDQ differences between maternal education levels and income levels were assessed with one-way ANOVA (total difficulties scores, as well as the subscales). We performed linear regression models to examine the association between preterm birth and total difficulties score for the total group, followed by stratification by SES level. Linear regression models were only performed on the total difficulties scores, because of concerns regarding the reliability of the subscales of the SDQ in 5- to 6-year-old children [22]. Adjustments were made for gender, number of siblings, one-parent household, Dutch or non-Dutch ethnic origin, age of the mother, smoking during pregnancy, small for gestational age and the DASS21 total score (model 1). In model 2 we additionally adjusted for the other SES indicator (income adequacy or maternal education level). Finally, we assessed whether preterm birth and low SES had multiplicative effects on SDQ scores by adding the preterm birth * SES interaction term. Analyses were performed for SDQ reported by mother and teacher independently. Statistical analyses were performed in SPSS version 21, statistical significance was set at p < 0.05 level.

Table 1 shows the pregnancy, child and family characteristics of the children in this study. The preterm group included significantly more children of non-Dutch ethnic origin (p = 0.023) and low maternal education (p = 0.016). The preterm children were more often first born and had less siblings at 5 years of age, than term-born children.

Responders (n = 4553) to the 5-year questionnaire were compared with non-responders (n = 1604), as shown in an additional table in Online Resource 1. Non-responders were more often of non-Dutch ethnicity (p < 0.001), had less years of maternal education (p < 0.001), had a shorter gestational age in weeks (p = 0.005) and their children had a lower birth weight (p < 0.001). Prematurity was not more common in the non-responder group (p = 0.293).

Table 2 shows the unadjusted mean SDQ scores for term- and preterm-born children. Overall, the mean total difficulties score reported by mothers was 5.3 ± 4.1. Mothers reported a higher mean score on the total difficulties scale for preterm children compared to term children (p = 0.003). Mothers reported higher SDQ subscale scores in preterm children on emotional symptoms (p = 0.004), hyperactivity/inattention (p = 0.011) and peer problems (p = 0.009). Teacher SDQ scores followed almost the same trends, but differences between preterm children and term children were non-significant. For term-born children 10.1 % of the mothers reported scores that were above the Dutch cut-off (≥11) for the total difficulties score; for preterm-born children 16.1 % of the mothers reported scores above cut-off (p = 0.004). The difference in SDQ total difficulties score between very preterm children (<32 weeks’ gestation) and term children (Δ1.8, 95 % CI 0.1–3.5) was larger than the difference between moderately preterm children (≥32 and <37 weeks’ gestation) and term children (Δ 0.7, 95 % CI 0.1–1.3).

All differences in mean SDQ scores of mothers and teachers according to level of maternal education were significant (p ≤ 0.001). Mothers with a low education level or perceived income inadequacy reported more psychosocial problems in both the term and preterm group (Fig. 2). The highest SDQ scores were reported in preterm children with a low maternal education level or inadequate income. Table 3 shows the differences between term and preterm children on SDQ total difficulties score after the addition of the covariates in the two models. After adjustment for gender, siblings, one-parent household, ethnic origin, age of the mother, smoking during pregnancy, small for gestational age and DASS21 score (model 1), the difference between the mean SDQ total difficulties score of mothers in preterm children and term children became non-significant (0.5, 95 % CI 0.1–1.0). Differences in SDQ total difficulties score between preterm and term children were largest in mothers with a high education level. After adjustment, only differences between preterm and term children in the highly educated group remained significant (Δ0.9, 95 % CI 0.2–1.5). Differences between preterm and term children from low educated mothers appeared larger when the teacher reported psychosocial problems; however, the difference was not significant. When comparing differences in SDQ total difficulties score of mothers by level of income adequacy, there was a significant difference between preterm and term children in the ‘more than adequate income’ group (Δ1.0, 95 % CI 0.3–1.6); this difference remained significant after adjustment (Δ0.9, 95 % CI 0.3–1.6) (Table 4).

The interaction between preterm birth and perceived income adequacy on SDQ total difficulties score reported by mothers was borderline significant (p = 0.072). No other preterm birth * SES interactions were found (maternal education SDQ mother p = 0.320, maternal education SDQ teacher p = 0.497, income adequacy SDQ teacher p = 0.631).

A strength of this study is the longitudinal information on (risk factors in) pregnancy, birth and first years of life. Also, we used both parental and teacher reports to assess psychosocial problems. We found an interrater correlation of 0.40, which is comparable to the interrater correlation found by Goodman (r = 0.46) [36]. This low to moderate agreement is common and indicates the importance of multiple informants, because children may show different problematic behaviour in the home or the school situation, depending on the situational context and the person they are interacting with. Apart from different behaviour, the informants also have different possibilities to observe specific behaviour in their context [37]. There is no gold standard available for the integration of multiple assessments in the diagnostic process. If information from parents and teachers is available, instead of looking at discrepancy effects, it seems more important to consider additive informant effects as a predictor of outcome [38].

Furthermore, the SES measures we used are frequently used indicators of SES. Because maternal education and income adequacy do not completely overlap, analysing these two SES indicators separately provides additional information [39]. In our study, only 26 % of the mothers with a low education level had perceived inadequate income. We found higher SDQ scores (indicating more psychosocial problems) for children with low maternal education than for children with inadequate income at home. There was a larger difference between total difficulties scores of children with low and highly educated mothers than between children with an inadequate income and more than adequate income. This confirms the hypothesis that maternal education is linked to more behavioural influences on child development than income adequacy [10].

Our study also has some limitations. First, our non-response analysis showed selective response, with a higher participation rate among highly educated mothers. Although this may not have biased our results, our reported socioeconomic inequalities may be an underestimation of the actual effects. Second, the ABCD cohort had a low percentage (4.8 %) of preterm children compared with the preterm birth prevalence in the Dutch population (7.7 %) [2] because twins were excluded from the study at an earlier stage. The preterm group was relatively small (217 preterm children); 21.8 % of the preterm children had mothers with low maternal education and 17.5 % with inadequate income. Our data did not allow to further analyse differences between very preterm and moderately preterm children because of low power (n = 24 and n = 193). Third, in our study we used maternal education level and perceived income adequacy at age 5–6 years when SDQ was rated. This might be different from the situation during pregnancy for some women. Finally, the SDQ must be seen as a screening tool, rather than a diagnostic instrument for psychosocial problems. Further evaluation of psychosocial problems is necessary. We choose to show corrected models for the total difficulties scores only, because there are concerns regarding the reliability of the subscales of the SDQ in the 5- to 6-year-old children [22].

This study underlines the disadvantageous effects of low SES on psychosocial development. However, we found no evidence to support a change in the existing YHC guideline that describes detection of psychosocial problems in preterm children [40]. Early detection and treatment of psychosocial problems may lead to considerable benefits regarding child development, well-being and health. Further research is needed to explore whether suitable interventions are available for children with low SES to prevent psychosocial problems. Based on the present study, we recommend YHC to be extra alert for psychosocial problems in low SES families, irrespective of term or (moderately) preterm birth.