Growing Up In Scotland: Maternal mental health and its impact on child behaviour and development

Appendix A: Methods in detail

Assessing maternal mental well-being over time

This report uses data from the birth cohort respondents only (the younger cohort in GUS⁹). At sweep 4, the cohort children were aged 46 months.

GUS has measured maternal mental health using two different scales: at sweeps 1 and 3 (ages 10 months and 34 months respectively), the SF12 Mental Health Component Score ( MCS) was used, whereas at sweeps 2 and 4 (ages 22 months and 46 months respectively) selected items from the Depression, Anxiety and Stress Scale ( DASS¹⁰) were used. ¹¹ Both scales are widely used and well validated.

The use of two different scales at alternative sweeps potentially presents some problems in relation to comparability, particularly as the MCS aims to assess the impact mental health problems have on everyday life, while the DASS is a more symptomatic measure. In order to calculate the consistency of the measurement of maternal mental health problems between the scales, Pearson's correlation coefficient was calculated for the raw scores at each sweep. There were statistically significant and fairly strong correlations between the scales at each sweep, particularly between the MCS at 34 months and the DASS at 46 months. This suggests that each measure is capturing a set of similar difficulties or symptoms among respondents at each sweep. ¹² Table 3.1 shows the correlations between the two measures for the GUS sample.

Table A.1 Correlations between SF12 (MCS) and DASS scores at sweeps 1 to 4

** Correlation is significant at the <0.01 level (2-tailed).

Neither the SF12 ( MCS) nor DASS have threshold scores which define "caseness" - that is, whether an individual's score indicates that they are suffering from a psychiatric disorder which could be classified using standardised diagnostic criteria such as DSM- IV (2000). For these analyses, maternal mental health is defined in terms of an individual's score in relation to the mean score for the cohort as a whole. Preliminary analyses suggested that scores which were more than one standard deviation from the mean should be used.

A respondent is defined as having "poor" mental health at sweeps 1 and/or 3 if she has a score on SF12 ( MCS) which fell more than one standard deviation below the mean population score for that sweep. At sweeps 2 and 4, poor mental health is defined in relation to DASS scores which fall more than one standard deviation above the mean for the cohort.

A "brief" mental health problem was defined as a score more than one standard deviation from the mean at any one sweep - on SF12 ( MCS) at sweep 1, DASS at sweep 2, SF12 ( MCS) at sweep 3 or DASS at sweep 4. A "repeated" mental health problem was defined as a score more than one standard deviation from the mean at two or more occasions. Table A-2 shows the range, mean and standard deviation scores for SF12 ( MCS) and DASS at each survey sweep.

Table A.2 Maternal mental health scores by Survey Sweep

We have no information about the duration of any mental health problems, including the timing of any onset or fluctuations over time. Our measures are based on score results on the SF12 ( MCS) or DASS at the time of the survey interview. As the data were collected at discrete points in time, we do not have access to continuous measures of respondents' mental health which means that we cannot assume that we are identifying chronic - that is, long-standing - mental health problems.

The relationship between maternal mental health and child outcomes

Previous research has found that exposure to poor maternal mental health in the early years can have a range of impacts on child behavioural, emotional, social and cognitive outcomes, and that there may be differences in outcomes for those exposed to brief or long-standing maternal mental ill health. In order to explore whether such associations can be discerned within the representative GUS sample of mothers and children, multivariate analysis attempted to examine the impact of brief and repeated exposure to poor maternal mental health on each of these areas, while controlling for possible confounding effects of other known socio-demographic and environmental factors.

Scores on four scales from Goodman's "Strengths and Difficulties Questionnaire" at age 46 months (sweep 4) were used to assess impact on emotional, social and behavioural outcomes. The Strengths and Difficulties Questionnaire ( SDQ) is a brief behavioural screening questionnaire designed for use with 3-16 year olds. The scale includes 25 questions which are used to measure five aspects of the child's development: emotional symptoms, conduct problems, hyperactivity/inattention, peer relationship problems and pro-social behaviour. A score is calculated for each domain, as well as an overall 'difficulties' score which is generated by summing the scores from all the scales except pro-social. For all scales, except pro-social where the reverse is true, a higher score indicates greater evidence of difficulties. The data were obtained via parental report, normally the mother, in the computer assisted self-completion module of the sweep 3 interview. Scores on each of the emotional symptoms and peer relationships sub-scales were examined individually whilst the hyperactivity/inattention and conduct problems scales were combined to explore the impact on behaviour. For all these measures, a higher score indicates a greater level of difficulties.

Cognitive ability was measured in the GUS birth cohort at age 34 months via two assessments: the naming vocabulary and picture similarities subtests of the British Ability Scales Second Edition ( BAS II). These two assessments measure, respectively, language development and problem solving skills. Each subtest is part of a cognitive assessment battery designed for children aged between 2 years and 6 months and 17 years and 11 months (Elliott, 1996). The assessments are individually administered. Numerous tests of ability and intelligence exist, but the BAS is particularly suitable for administration in a non-clinical setting as is the case for a social survey like GUS (see Bromley, 2009 for a discussion of the cognitive measures and their application within GUS). ¹³

Binary variables were created for all child outcome measures. Assignment to "positive" and "negative" child development outcome categories was based on mean or median scores for the various tests used. For the SDQ measures (social, behavioural and emotional), a child was deemed to be in the "negative" group if he/she had a score above the mean score for the cohort. The analysis reported here for the cognitive assessments used normative BAS scores, derived from the standard BAS tables and defined with reference to the standardisation samples used in developing the assessments. These normative scores were converted into T-scores based on the values in the standardisation sample for the applicable age band. T-scores range from 20 to 80 and have a mean of 50. A child with a T-score of 50 is therefore placed at the mean value for their age. Higher scores on either scale denote an increase in cognitive ability and conversely, lower scores indicate a reduced level of ability.

In order to establish if there were differential effects of brief or repeated maternal mental health on child behaviour, emotional, social and/or cognitive outcomes which were independent of socio-demographic and environmental factors, separate models were run for each of the outcomes. In a further stage, we explored whether access to social and other supports moderated the impacts of poor maternal mental health on child outcomes by adding various social support factors into the model. Variables used to explore access to support included: whether the mother is partnered, the presence of a maternal grandmother, use of childcare facilities, existence of friend and kinship networks, and the strength of the couple relationship, if applicable. As some of these were found not to be significant in the first stage of models, they were not included in the forced entry model.