Official Statistics

Methodology and supporting information: children living with parents in emotional distress, 2021 update

Published 22 March 2021

Purpose

Children living with parents in emotional distress is an annual official statistic. It consists of statistical commentary and data tables with indicators for England. This supplement supports the use and understanding of this official statistic. It includes:

• a summary of relevant background literature

• overview of survey used

• how the survey measure emotional distress

• method used to calculate the statistics

It aims to ensure transparency of terminology, definition and method used. It also outlines limitations the users should consider when using these indicators.

The indicators within this statistic look at common mental disorders (CMD). This includes conditions such as anxiety or depression.

Background

Poor parental mental health

Studies estimated that between 2014 and 2015 around one in 6 adults aged 16 to 64 years in England had a CMD. The prevalence of other mental health conditions is less common. For example, estimates for psychotic disorders are:

• 1 in 100 adults with schizophrenia

• 1 in 100 adults with affective psychosis

• 1 in 50 adults with bipolar disorder

Common mental health problems are more likely to occur in women than men. Since 2000, the prevalence of CMD in women has increased, while for men it remained stable. Evidence suggests most mental disorders have their onset in childhood, adolescence or young adult life.

Differences in the prevalence estimates of poor parental mental health exist. The estimates depend on severity of the mental health conditions and definition applied. Studies report that around 68% of women and 57% of men with mental health problems are parents. In 1,000 women giving birth it is estimated that throughout the perinatal period:

• 2 women will experience postpartum psychosis

• 2 severe mental illness (SMI)

• 30 severe depression

• between 100 and 150 mild to moderate depressive illness and anxiety states

• 30 post traumatic stress disorder (PTSD)

• between 150 and 300 adjustment disorders and distress

A study found that 38% of first-time fathers are concerned about their mental health. Between 5% and 10% of partners report mental health difficulties in the perinatal period.

Role of employment and worklessness

Common mental health problems are less common in employed than unemployed or economically inactive adults. Data from the 2014 Adult Psychiatric Morbidity Survey shows that common mental health problems are at:

• around 14% for adults in full-time employment

• 29% for unemployed adults

• 33% for economically inactive adults

Employment can positively affect mental health. Employment provides financial security, social status, identity, and social interaction. Poor working environment and stress within the workplace can negatively impact mental health. There is strong evidence of a causal relationship between employment status and psychological wellbeing.

Longitudinal studies use repeated measures to follow individuals over prolonged periods of time. Longitudinal analysis of the British Household Panel Survey for the period 1991 to 2007 examined the role of employment on psychological wellbeing. It reported that moving from employment to worklessness predicted lower psychological wellbeing. This impact was still present after accounting for other factors.

Moving from employment or seeking work into permanent sickness also predicted lower psychological wellbeing. This effect was stronger than permanent sickness on its own. This analysis also reported that moving into employment from unemployment improves psychological wellbeing. However, the change is less than the negative effects of job loss.

A 2004 survey of the mental health of children and young people (CYP) in Great Britain examined the role of family work status on CYP mental health. The prevalence of mental disorders was higher in CYP in families with neither parent working compared to where both parents worked.

Impact on child outcomes

Studies confirm the link between maternal and paternal depression and an increased risk of later behavioural and emotional difficulties in children.

Results for a longitudinal study show that children of mothers with repeated mental health problems were more likely to have poorer relations with peers at age of 3. This was compared with children whose mothers remained mentally well or who had only brief episodes of poor mental health.

A study of fathers with persistent depression (in the antenatal and postnatal periods) found that:

• children had a higher risk of emotional and behavioural problems at age 3.5 after controlling for other factors

• the associations between fathers’ mental health and child behavioural outcomes were no longer statistically significant at age 7

The period and frequency of parental mental health problems will differently impact on children wellbeing. Evidence suggest that children of mothers with poor mental health show:

• adverse emotional and cognitive outcomes where mother reported mental health problems once during the 4 years of the survey (defined as a brief exposure)

• further adverse behavioural outcomes where mother reported repeated occurrence of mental health problems over several years (defined as a prolonged exposure)

Research suggests that poor maternal mental health during pregnancy affects outcomes in middle childhood. Children whose mothers experienced high levels of anxiety in late pregnancy, or after birth, had a higher prevalence of behavioural or emotional problems at age 7. A longitudinal research investigating how parental mental health relates to adolescent child happiness found that that maternal and paternal mental distress predicts unhappiness in girls but not boys.

Potential transmission mechanisms

Current evidence on the mechanisms involved in the transmission of mental health problems to poorer outcomes in children is mixed. Some children of parents with mental health problems are at increased risk of poorer cognitive, emotional and social development. Yet many children will not suffer adverse effects. Research highlights the role of child resilience and protective factors on poor outcomes in later life. The impact may also differ depending on the age of the child. Evidence on the impact on boys and girls is mixed. The impact might differ for boys and girls. Whether the father or mother has mental health difficulties also plays a role.

Different biological characteristics, sociocultural contexts and psychological processes are likely to interact. They can act as protective or risk factors for both parents’ and children’s mental health. Some of the possible mechanisms by which parental mental health may impact on children include:

• direct exposure to symptoms such as experience of unpredictable, irrational behaviour or neglect

• the influence of mediating factors such as disrupted parental or couple relationship, and domestic violence

• disruptions to parenting

• parental genetic factors

• the interaction of genetic and environmental influences

The emotional environment within the home affects child’s development. This includes:

• the quality of relationships between parents

• the support available to the family

• the health and wellbeing of the parents

Research shows a strong link between children’s behavioural problems and parents’ relationship. Poorer-quality relationships predict greater behavioural problems, especially among children in lower-income families. For parents themselves, social support and relationships are also important. Being happily married or in a stable relationship has been linked to physical and mental health benefits.

Some evidence suggests that depressed mothers may be less responsive to their infants’ attempts to engage with them. This in turn affects the strength of the child’s attachment.[The development of attachment behaviours and bonding are important to babies and young children. They play an important role in their wellbeing and development. Poor attachment is related to impaired cognitive functioning at 18 months.

There is limited research on how paternal mental health affects children. Some studies suggest that fathers with depression spend less time with their children. They also undertake fewer activities, so the quality of time is also reduced. Other studies found that self-reported paternal depression has a significant negative effect on parenting. This includes decreased positive and increased negative parenting behaviours in fathers.

Older children may have a greater ability to understand some aspects of a parent’s mental health problems. Therefore, they might be more tolerant of some disruptions to their relationship with the parent. They may also find their parent’s unpredictable behaviour difficult to cope with. Some children may take on a caretaking role to support their parents mental health needs.

Data source and screening tool

Understanding Society Longitudinal Study

Understanding Society study is a nationwide household survey. It follows 41,000 households across the UK from 2009 to 2010 onwards. The survey:

• captures important information about people’s social and economic circumstances

• collects data on their attitudes, behaviours and their health

• provides a rich data on families and their circumstances over time

Participants are assessed over an extended period. Repeated evaluations spread across several years. Periods of assessment are often called waves. The survey enables a longitudinal analysis on disadvantage and worklessness.

Understanding Society Study household members aged 16 or older are interviewed to take part in the study. The same individuals are re-interviewed in successive years to see how things have changed. Individuals become eligible for a full interview once they reach the age of 16. A subset of 10 to 15 year old children are also interviewed in the ‘youth survey’. However, questions differ from the main survey. Full questionnaires of current surveys currently can be found online.

The Understanding Society fieldwork within each wave is conducted over a 2 calendar year period. Participants are interviewed around the same time each year. It is important to note that the periods of waves overlap.

Attrition (known as loss of study participants from the study) reduces the Understanding Society sample size over time. This is a common issue with most longitudinal surveys.

General Health Questionnaire

The Understanding Society survey uses the self completed 12-item General Health Questionnaire (GHQ-12). The GHQ-12 is a condensed version of the GHQ. It is commonly used in social research and features in many household surveys.

The GHQ-12 is the most extensively used screening instrument for common mental disorders. It is also used as a general measure of emotional distress. It concentrates on how the respondent is feeling relative to normal. This includes breaks in normal functioning rather than life-long traits. Therefore, it covers disorders or patterns of adjustment associated with distress.

The GHQ-12 asks questions about the way an individual has been feeling over the last few weeks. This includes questions on sleep, self-confidence, worry, and concentration. Items on the GHQ-12 are rated on a 4-point scale: 2 responses are negative (where the respondent is feeling worse than usual) and 2 are positive (the same or better than usual). A score of one is given for a negative response and a score of 0 for a positive response. These 12 scores are added together so that each individual has a score which ranges from 0 to 12. A score of 4 or more has been shown to indicate that the individual has symptoms of minor psychiatric morbidity.

A high score on this scale does not necessarily indicate severe mental disorders. Severe mental disorders are characterised by deterioration of normal social functioning. This high score is more likely to indicate common mental health problems such as anxiety and depression.

The common use of GHQ-12 in research enables further comparison and analysis. For these purposes, using a self-reported scale such as the GHQ-12 is better than using questions that concentrate on whether a respondent has or had a diagnosis of depression or anxiety. The latter approach is likely to underrepresent the level of poor mental health in the population due to under diagnosis and under reporting.

Wave 1 (2009 to 2010) income information

There are some issues with the income information in the first Understanding Society survey wave covering 2009 to 2010. As result, income data from the first wave of Understanding Society are not comparable with later waves. It is also likely to be of lower quality. Therefore, the first wave is largely excluded from the analysis in this statistic. Parental separations between the first and second wave is included to increase the underlying sample size. The first wave of data on parental separation is considered of sufficient quality.

Waves 2 to 8 (2010 to 2017) retrospective changes

The Understanding Society wave 9 data release (November 2019) included some minor changes to data from previous waves. A correction was applied to trends in this publication using the new source data. The changes to the indicator values from 2010 to 2017 are negligible. However, there are some differences to the previously published figures.

Wave 8 (2016 to 2017) methodological changes

From 2016 to 2017 wave onwards, a mixed mode interview approach was used. This means that a larger proportion of interviews were administered through online interviews. The change in interview administration may affect how parents responded to the GHQ-12.

Some individuals may be less likely to answer honestly depending on the setting. The extent of this will vary between individuals. It is not thought that this change had a significant impact on the indicators in this publication. However, the user should consider this when looking at the observed changes between 2016 to 2017 values and those in previous years.

Methodology

Terminology within the statistics

For the purpose of this official statistic, Public Health England (PHE) uses the following terminology to define parental disadvantages and the relevant characteristics.

1. Poor parental mental health – defined as a measure of the proportion of children living with at least one parent reporting symptoms of emotional distress.

2. Emotional distress – which is a snapshot measure where a person has scored 4 or more on GHQ-12. This indicates minor psychiatric morbidity, such as anxiety or depression.

3. Workless family – used where no adult (neither parent nor guardian) in the family was in paid employment.

4. Lone parent – used for families where only one parent is living in the household with the child or children.

The terminologies are based on Understanding Society study definitions. It aligns with Improving Lives: Helping Workless Families publications.

Underlying sample population

For the poor parental mental health indicator, the sample used for the analysis includes children if they were present in any of the 4 most recent waves of the study. The analysis includes any child where at least one parent has responded fully to the GHQ-12.

For specific sub-sections, the analysis only includes children who have parents that make up the relevant measure. For example, only those children with fathers within the household are included when calculating the proportion of fathers with emotional distress.

The analysis could construct the statistics from the perspective of:

• the parents; this measures the proportion of parents reporting symptoms of emotional distress

• the child; this measures the proportion of children living with at least one parent reporting symptoms of emotional distress

As the analysis is interested in how many children are affected by poor parental mental health, the latter is used. A parent based indicator would count every parent, even if the same child was affected twice. Comparison of both approaches shows that trends are very similar for both measures.

For over 20% of children in study sample, information was missing for one of the parents. This analysis could include children where both parents provide responses to GHQ-12. Alternatively it could include children where at least one parent responds. Both options come with a bias. Excluding any children with at least one unknown parent biases the sample towards lone parents. This is because it is easier to be included if only the response of one parent is required. Lone parents are more likely to report symptoms of emotional distress. This construction would overestimate the overall prevalence of poor parental mental health.

Alternatively, including all children where at least one parent is known is likely to under report prevalence. This is because this approach is assuming that the ‘unknown’ parent does not have poor parental mental health.

Analysis conducted by PHE found that there is around 4% points difference between the 2 measures. Trends were very similar for both.

In conclusion, it was decided to base the total parental distress measure on where at least one parent is known. For the relevant breakdowns, PHE constructs the measure on whether the relevant parents (mothers, fathers, and both parents) are known. These measures exclude children if the relevant parent was missing.

The approach to method is designed to reduce the bias in the more granular breakdowns. It also gives a more refined representation of emotional distress and poor parental mental health.

Weights used and attrition

To assess changes in the prevalence over time this analysis uses cross-sectional weights. The use of weights ensures that the results are representative of the UK population. Self-completion weights were also considered. Self-completion weights adjust for non-response to questions. This is often used in research using self-completion questionnaire such as GHQ-12. Self-completion weights are not appropriate for the indicators in this official statistic. This is because this statistic reports parental emotional distress from the perspective of the child. PHE examined the use of cross-sectional and self-completion weights for indicators from the perspective of the parent. There was no significant difference between results using the different weights.

Rounding and suppression

This official statistic applies rounding to one decimal point for all indicators. As a result, differences may not sum exactly due to rounding. PHE supresses any proportions based on a sample population of 100 or less.

Limitations

This section outlines some of the limitations to this analysis. These are issues common to most survey based longitudinal analysis. We are confident that the findings are robust to any of the issues outlined here.

Standard limitations of using survey data

Surveys gather information from a sample rather than from the whole population. The sample is designed to be as representative of the general population. Sample design considers practical limitations such as time and cost constraints. Results from sample all surveys are estimates. This means that they are subject to a margin of error (sampling error). This can affect how changes in the numbers should be interpreted. Therefore, year-on-year movements in results should be treated with caution.

Surveys are also at risk from a systematic bias due to non-response or missing values. This is because:

• households invited for interviews do not respond to the survey

• some individuals within households may not respond although others do

• a respondent gives a full interview but refuses or gives a ‘don’t know’ answer to a question leading to a missing value

To correct for these biases, results are generally weighted to adjust for non-response. This method makes it clear where results within this statistic may be subject to bias. The analysis excludes results for children where at least one parent gives an unknown value. This may bias lone parent families, as it is easier for those families to give full responses.

Other non-sampling errors should be considered. Non-sampling errors are introduced by some systematic bias in the sample as compared to the population it represents. Such biases include inappropriate definition of the population, misleading questions, data input or handling errors. Non-sampling errors are difficult to control for and method to do so are limited.

Careful application of the appropriate survey techniques can minimise the risks. This should be considered from the questionnaire and sample design stages through to analysis of results. Non-sampling errors are more likely to be random. They are less likely to be related to some underlying characteristics of the individuals interviewed.

Definition of a family in longitudinal analysis

In this longitudinal analysis, the changes in children’s lives are followed and analysed from wave to wave. This is because the analysis is from the perspective of children’s experiences. Following families would be too complex, because they form or dissolve over time.

The definition of family is based around the adults (parents) that are living with the child in each wave. This means it is possible that a child could be living with completely different parents from one wave to the next. For example, a child that lives in a family that became workless could have been living with working parents or guardians, and then moved to live with a different set of workless parents or guardians. Indicating and accounting for these changes in the analysis would be confusing and misleading at times. This situation will also not apply to most children in the survey.

Definition of worklessness

The definition of a workless family is based on whether a child is living in a family where no adult is in paid employment. This is a binary indicator, and the analysis has not removed families where, both parents are retired or are students. Removing these families reduced the proportion of dependent children in workless families by around 0.5%. This did not affect the nature of the findings.

Transitions between waves

Survey respondents are interviewed annually. Whilst there is some information on changes in characteristics between waves PHE chose not to use this. This information on change includes for example, employment transitions, and relationship changes. Including this information added further uncertainty and complexity to the analysis. For example, it increased the number of unknowns. This change did not change results or the nature of the findings.