Page 76 - Sample Moderate prematurity, socioeconomic status, and neurodevelopment in early childhood

Chapter 7
had some trouble filling in the ASQ. Nevertheless, the effects on study outcomes were likely to be small, given the relatively small number of this group.
We assessed behavioural and emotional problems at the age of four using the CBCL. This questionnaire has good psychometric properties, is widely used in research and clinical settings, and is easily filled in by parents.81 Some psychometric properties, for example, are the high test-retest reliability (r=0.85) and good predictive validity, with a correlation of 0.64 for the total problems score over a period of six years.81 Most scores on the CBCL syndrome scales at pre-school age significantly predict higher CBCL scores at school age, with sometimes even better correlations over a longer time span.81 Pre-school scores on the withdrawn syndrome scale, for example, correlate better with outcomes at the age of nine than at younger ages. Therefore, pre-school CBCL scores seem to reflect underlying problems that predict long-term psychosocial functioning.81 From other studies we know that behavioural and emotional problems in children tend to persist into adolescence and even into adulthood.82-84
Clinical meaning of statistical significance and interaction
In the studies reported on in this thesis, we found statistically significant associations of gestational age and SES with developmental and behavioural outcomes. Statistical significance, however, is not always equal to clinical significance. For measuring clinical significance, effect size is the appropriate measure. By way of example, we measured the effect sizes of associations between moderate prematurity and behavioural and emotional problems, a measure which is relevant for clinicians taking care of children who have elevated CBCL scores. The effect sizes for elevated total, externalizing, and internalizing problems were 0.34, 0.27, and 0.50, respectively, meaning small (0.34 and 0.27) to moderate effects (0.50). This means that the threshold to initiate individual treatment for internalizing problems was reached. When looking at population level, the effects of all outcomes are rather important, given the prevalence of moderate preterm birth, and imply a major global burden of disease.
To determine the effect of SES on the association between moderate prematurity and neurodevelopmental problems, we measured statistical interaction between moderate prematurity and low SES. In logistic models, we found that SES did not interact with moderate prematurity with regard to most neurodevelopmental outcomes. This meant that the effects of SES and gestational age were multiplicative, as is the case for any logistic model without significant interactions. This absence
126