Maternal employment and indicators of child health: a systematic review in pre-school children in OECD countries
- Professor C Law, Centre for Paediatric Epidemiology and Biostatistics, UCL Institute of Child Health, 30 Guilford St, London WC1N 1EH, UK;
- Accepted 15 December 2008
- Published Online First 4 February 2009
Background: The aim of the study was to determine whether maternal employment is associated with two indicators of family health behaviour in pre-school children—childhood vaccination uptake, and childhood overweight including obesity—in OECD countries.
Methods: Narrative systematic review of nine medical/social science databases (1980–2007), relevant websites, retrieved article reference lists and consultation with experts. Relevant articles in English published from 1980 were included; studies of pregnancy, single occupational groups and non-OECD countries were excluded.
Results: 8924 abstracts yielded 21 eligible articles: 15 on vaccination and 6 on overweight. Meta-analysis was not possible. Vaccination uptake appeared at least as good or better for children of employed as unemployed mothers. Child overweight may be more prevalent with maternal employment, particularly if long hours are worked.
Conclusions: Maternal employment may have variable effects on pre-school children’s health. Policies promoting parental employment should monitor and evaluate the effect on the health and wellbeing of all members of the family.
The promotion of parental employment forms a central element of current UK policies to tackle child poverty.1 2 For low income couple families, this may mean both parents going to work. Support is provided for lone parents to find employment3 and there is active debate about whether to require them to seek employment when their youngest child reaches a certain age, and, if so, what that age should be.4 In the UK, employment of mothers of children under five years of age has increased markedly over the last two decades, from 27% in 19845 to 56% in 2005.6 In contrast, paternal employment has remained steady over this period. The view that ‘work … promotes good physical and mental health’ was asserted in 2008 in a consensus statement from more than 30 UK bodies of health professionals.7 Systematic review evidence shows that employment is associated with better health in men, and the limited research on women generally shows a neutral or beneficial effect.8
As unemployment, poverty and deprivation are all associated with poorer adult health8–10 and child poverty is associated with poorer child health,11 maternal employment might be expected to improve maternal and child health by increasing household income. Other potential benefits (to mothers) of paid employment include training and empowerment.12 13 However, employment may limit time for parenting, increase stress and alter childcare arrangements. Childcare may incur costs and could bring additional benefits or harms depending on quality and affordability. Thus, maternal employment could have adverse or beneficial effects on children’s health.
Forthright opinions on this topic are widely expressed in the media but do not appear to draw on a substantial research base. A synthesis of available evidence could inform a debate on policies on parental employment as well as highlighting areas for further investigation. To start this debate, a systematic review of studies describing an association between maternal employment and two indicators of health in pre-school children in Organisation for Economic Co-operation and Development (OECD) member countries was carried out.
To focus the review question, the literature on maternal employment and ten child health indicators from ‘CHILD’ was mapped.14 Following systematic searches to assess the scope and quantity of the literature,15 childhood vaccination and childhood overweight/obesity were selected for review. These were chosen because they: are important for pre-school children’s health; are sufficiently prevalent to allow adequate power in studies examining differences by maternal employment status; are relevant to health and social policy; and had been examined in a suitable number of papers for a review. In addition, they relate in contrasting ways to the financial and time resources available to the family.
In this review, the term ‘overweight’ is used to describe children meeting definitions of ‘overweight’ or ‘obese’, irrespective of the terminology chosen by the authors of the primary research.
Quantitative studies were included irrespective of design, if they: reported the relationship of maternal employment with any measure of child overweight (including obesity) or child vaccination uptake; included children below statutory school age for the country of residence; collected information on maternal employment after the birth of the index child; were published in or after 1980. Thus, cross-sectional, case-control, cohort, intervention and evaluative studies were eligible for inclusion, as well as reports of “natural” experiments. Studies were excluded: in which employment was only measured in pregnancy; which contained only qualitative research; from non-OECD countries; on single occupational groups; not reported in English.
The search was conducted from 1980–2007 (a period of rising maternal employment) in these databases: Medline, Embase, Psychinfo, Cochrane, Kings Fund (all using Dialog Datastar interface), Science Citation Index and Social Science Citation Index (via Web of Science), IBSS, and JSTOR, and the websites INTUTE16 and OECD.17 Papers containing terms related to all the following concepts were searched for: employment ‘and’ mothers ‘and’ children ‘and’ OECD countries. ‘And’ was used to combine these results with searches for each of the indicators of child health. For each concept sensitivity was maximised, making use of synonyms for text word searches, thesaurus terms and ‘explosion’ where possible. Full details are available on request.
For every article meeting the inclusion criteria, the citations generated by the ‘related articles’ feature of Pubmed,18 those from the ‘cited reference search’ facility of Web of Science and the reference lists of the paper were screened. The final reference lists were checked for obvious omissions with an expert in each field.
Study selection, data extraction and synthesis
Two reviewers independently screened the abstracts (agreement 95%) and the retrieved candidate papers for inclusion. Data were then independently extracted by these reviewers (MJM and RJ), with differences resolved by discussion and involvement of a third party (CL) if necessary.
Sociodemographic and behavioural factors may be associated with both maternal employment and child health. Factors likely to lie on the causal pathway were conceptualised as mediators, and the rest as confounders.19 Potential mediators of an association between maternal employment and the chosen health indicators included household income, time available to spend with children and childcare. Potential confounders, not lying on the putative causal pathway, included socioeconomic position, ethnicity, being a single mother, maternal education and maternal age.
Data were extracted on the date, setting and the population, definitions and measurement (of exposure, outcomes, potential confounders and mediators), and the analyses performed. The results of both unadjusted and adjusted analyses were noted, as they may offer different insights into the relationship between maternal employment and child health. As the direction or strength of any association might differ between social groups, the results of stratified analyses were noted and studies confined to single minority groups were included.
Wherever possible, the data were transformed into a common format, as odds ratios taking a value less than one where maternal employment was associated with a better health outcome (more complete vaccination, not being overweight). Results that could not be transformed to odds ratios were presented in their original form.
The Effective Public Health Practice (EPHP) quality tool (version downloadable in 2007)20 was designed for use in quantitative intervention and observational studies. This was used to assess internal validity, also developing its criteria for cross-sectional studies (not explicitly considered in the tool). At the time of the present study, the tool did not include a method for a summary score or global rating. To clarify the presentation, ratings of strong, moderate or weak internal validity were assigned, based on the three-point scores (strong, moderate and weak) in the domains of the quality tool and the reviewers’ judgement (see footnotes to tables). Applicability to the UK context was assessed using criteria based on those used at the National Institute for Health and Clinical Excellence (NICE).21 EPPI-reviewer software22 was used to track the references, and to extract and tabulate the data.
A narrative synthesis was conducted and it was also considered whether meta-analysis was possible. It was assessed whether relationships between maternal employment and indicators of child health varied by study quality, year of birth of participants, prevalence of maternal employment, proportion of single mothers, and the geographic, socioeconomic and ethnic composition of the study population.
The search of bibliographic databases generated 8924 citations, with a further 268 identified through the secondary search strategies. After screening, 21 full-text articles, reporting on 20 studies, met the inclusion criteria (figure 1).
The studies are described in tables 1 and 2, ordered by year of birth of the study children. Heterogeneity of analysis and presentation of the results precluded graphical presentation, or meta-analysis. All the studies were observational (8 cohort studies, 12 cross-sectional). Although many studies adjusted for socioeconomic position, only three studies used stratification to examine differences with socioeconomic position or ethnicity.23–25 The internal validity was rated as high for three articles,26–28 weak for three articles29–31 and moderate for the rest.
Fifteen articles, reporting the results of fourteen studies, included vaccination as an indicator of child health.25–27 30–41 (table 1). Only one was specifically designed to assess the relationship with maternal employment.37 Five were cohort studies and nine cross-sectional. Three studies were nationally representative,27 36 40 three overrepresented disadvantaged groups,25 32 35 and one overrepresented advantaged groups;39 in others the socioeconomic position of participants was mixed or unclear. Most studies were from the USA or Europe. Year of birth of the children ranged from 1970 to 2003. Only one paper stratified by ethnicity,25 and none by socioeconomic position.
Most studies showed a tendency to higher vaccination uptake in children of employed mothers, although this was not always statistically significant. There was some inconsistency within studies; for instance three studies showed a significant association in the unadjusted but not adjusted analysis,25 36 38 two studies found differences between results for measles-containing and pertussis vaccinations,36 38 two studies had results differing by the age of the subjects26 35 and the only study to stratify on ethnicity found differences between ethnic groups.25
The relationship between maternal employment and vaccination showed no consistent relationship with year of birth, percentage of mothers employed, socioeconomic or ethnic composition, proportion of single mothers, internal validity, or applicability.
Six articles,23 24 28 29 42 43 detailing the results of six studies, featured overweight or obesity as an indicator of child health (table 2). Two studies were designed to examine the relationship with maternal employment.23 24 Three were cross-sectional28 29 43 and three were cohort studies.23 24 42 Two studies were nationally representative,23 24 two were representative of smaller geographical areas,28 42 and two were in specific ethnic groups.29 43 Three studies were from the USA,23 29 43 and one each was from the UK,24 Germany28 and Japan.42 Year of birth of the children ranged from 1975 to 2003. Two studies stratified by socioeconomic position.23 24
Most studies showed a higher rate of childhood overweight where mothers were employed, though this was not always statistically significant. The most informative studies (Anderson et al,23 Hawkins et al24 and Lamerz et al)28 showed congruence between their research question and that of the present study, and so presented particularly relevant analyses, including analyses to assess a dose–response effect between maternal employment and childhood overweight.
Anderson et al23 presented data from the National Longitudinal Study of Youth, aged 3–11. The percentage of overweight (above 95th BMI centile) children was higher in mothers who had been employed, and higher in full-time than part-time employed mothers. Although the overall percentage of overweight children was greater in poorer families than wealthier ones, the adverse effect of employment was observed only for wealthier families and those with better educated or non-Hispanic mothers. Econometric analyses using probits examined effects of employment intensity and duration, adjusted for various sociodemographic factors. Overweight was more common in children whose mothers were employed for longer hours; an extra 10 hours per week of maternal employment increased the likelihood that children would be overweight by 1.2 percentage points. The association was stronger where mothers were more educated. After adjustment for sociodemographic factors, childhood overweight was unrelated to duration of employment.
Hawkins and colleagues24 analysed data from the UK Millennium Cohort Study at age 3 years. Prevalence of overweight was higher in children of employed than non-employed mothers. Overweight prevalence was higher in lower income groups, but stratification demonstrated that the association between maternal employment and childhood overweight was stronger in higher income groups. A dose–response relationship was shown for overweight with intensity of maternal employment (hours per week). Duration of maternal employment (weeks per year) was not significantly associated with childhood overweight.
Lamerz and colleagues,28 in their cross-sectional study of 5–7 year olds in Aachen, found lower rates of overweight in the children of employed mothers than non-employed mothers though this was not significant after adjustment. Overweight rates were higher in the children of full-time than part-time employed mothers, but this did not reach statistical significance.
These three studies suggest that increasing intensity of employment (working longer hours) may be associated with increases in childhood overweight, while increased duration (working more weeks) is not. Although childhood overweight rates are higher in disadvantaged groups, the effects of employment appear more marked in affluent households.
The other studies were consistent with these findings. Two small cross-sectional studies in specific ethnic groups (Mexican Americans43 and Hmong)29 primarily examined feeding practices in relation to overweight, but found a non-significant unadjusted relationship between maternal employment and child overweight. In the Toyama study,42 a cohort study in a single prefecture in Japan, overweight rates were significantly higher where mothers worked full-time.
The relationship between maternal employment and childhood overweight showed no consistent relationship with year of birth, country of study, percentage of mothers employed, socioeconomic or ethnic composition, percentage of single mothers, internal validity, or applicability.
There was some evidence that compliance with vaccination programmes is at least as high, or higher, in employed mothers compared with unemployed mothers. There was also some evidence that maternal employment was associated with greater risk of childhood overweight, an effect which was more marked in higher income households. Rates of overweight were higher amongst the children of mothers who were employed for more hours per week, after adjustment for potential confounding variables. Few studies investigated whether relationships between maternal employment and indicators of child health differed by socioeconomic position, or ethnic group.
Strengths and limitations of the review
Strengths of this review include the use of rigorous standard methodology and extensive searching, using databases from a range of disciplines,15 44 though it remains possible that we missed some studies. Any systematic review is dependent on the quality, conduct and relevance of the primary studies; this varied considerably, in common with much observational literature.45 46 Very few studies in the present review had been designed to examine the review question, making publication bias less likely but meaning that authors did not always employ designs or report (or possibly conduct) analyses that optimally addressed these associations. The quality ratings could not always distinguish between problems with methodology (in relation to the review question) and inadequate reporting.47 Publication bias remains possible; heterogeneity in the presentation of results precluded construction of a funnel plot to assess this.
Effect modification by setting is possible; some factors, such as social policy, which vary between countries, may affect the relationship between maternal employment and child health. For example, a comparative study has shown that unemployment is a more important factor in the health disadvantage of single mothers in the UK than in Sweden,48 though child health was not studied in this comparison. Unfortunately, effect modification could not be assessed further because of the limited data available. However, the nature of employment, its social meaning and the reasons for the participation of women with young children in the paid workforce are all likely to vary over time and between settings.
Studies also varied in their approach to confounding and mediating variables, further hindering synthesis of the findings. Multivariable analysis that specifically addressed the hypothesis that maternal employment was associated with vaccination or overweight could give clues as to causality.19 Studies that included maternal employment among a barrage of potential ‘predictors’ in an adjusted analysis risked underestimating the strength of the relationship if they adjusted for potential mediators (eg household income), and overstating it if they omitted potential confounders. Some of the variation between studies in the adjusted odds ratios may therefore be due to the variation in the factors included in multivariable analyses. Potentially illuminating analyses, such as stratification, were rarely performed. This limited the inferences which could be drawn about specific groups, such as single mothers, and more generally about the potential effect of maternal employment on health inequalities.
Fit with the literature
Most studies of the health effects of paid employment focus on men.8 The limited available evidence suggests that, for women, employment generally has beneficial or neutral effects on overall health.8 It is recognised that paternal employment or living in a workless household may also influence children’s health. Furthermore, work in the form of domestic and caring tasks is an integral part of parenting and may not be equally distributed between adults in a household; paid employment forms only a part of the work done by mothers. However, the present review focused only on the associations of maternal paid employment and indicators of child health.
Maternal employment was associated with vaccination uptake among children, which appeared to be at least as good, or better than, that in the children of mothers without paid work, but there were inconsistencies between and within studies which may be due to chance or differences in social context. Maternal employment might reduce financial obstacles to vaccination in countries where families have to pay, and employed mothers may be more motivated to avoid time off work for preventable illnesses, or have different attitudes to reports of vaccine ‘risks’ in the media. On the other hand, access to vaccination services could be reduced if mothers could not attend appointments because of employment commitments.49
Putative pathways linking maternal employment to childhood overweight23 include factors likely to be protective (eg higher income allowing purchase of a healthier diet or more opportunities for physical activity) and those likely to increase overweight (lack of time for active play or food preparation, use of prepared foods which tend to be more energy dense,50 and encouragement of sedentary behaviours). There is some evidence that mothers in employment tend to maintain time spent with their children at the expense of other activities, such as domestic or cooking tasks.51 This review found that increases in the numbers of hours mothers worked per week were associated with increased rates of childhood overweight. Increased duration of maternal employment did not show this association. The association between maternal employment and childhood overweight was stronger for those of higher income. These findings suggest that either lack of maternal time, or increases in time spent by children in non-maternal care, may make overweight more likely.
Implications for policy and practice
The findings suggest that recent changes in the prevalence of maternal employment are unlikely to have a general deleterious effect on child health. This is reinforced by evidence from large UK cohort studies suggesting maternal employment has little effect on pre-school children’s educational and behavioural outcomes.52 However, maternal employment may be contributing to trends in specific aspects of child health, such as the prevalence of childhood overweight, particularly if long hours are worked. More research is needed to explore this possibility further, and to find how to support employed parents to promote the health of their children.
The analytical approaches used in some of the studies highlight the importance of making the best use of observational data, using stratified analyses to reveal differential effects in specific social groups, and seeking dose–response relationships to test causality. Separate analyses of single and couple mothers might clarify whether different approaches are needed to promote the health of their children.
Policies that promote parental employment need to monitor and evaluate the effect of these policies on the health and wellbeing of all members of the family. Such evaluations need to recognise that the relationship between employment and child health may be different in different social contexts and groups, and that parental employment has potentially beneficial and adverse effects on children’s wellbeing.
What is already known on this subject
Maternal employment in developed countries has increased markedly from the 1970s to the present day.
UK government policy promotes parental employment as one strategy to reduce child poverty, but the effects on child health are unknown.
What this study adds
Overall compliance with vaccination programmes appears at least as good or better for the children of employed mothers compared to mothers without paid employment.
Childhood overweight may be more common in the children of employed mothers and there are indications of an increase with increased hours per week of maternal employment.
Policies promoting parental employment should monitor and evaluate the effect on the health and wellbeing of all members of the family.
Contributors: MJM and CL contributed to the conception, design and interpretation of the data. MJM conducted the searches. MJM and RJ screened the references for inclusion and extracted data from the included articles. MJM, RJ and CL contributed to drafting the article. All authors have seen and approved the final version. MJM will act as guarantor for the paper.
Funding: M Mindlin completed this work while on academic placement from the London Deanery Public Health Training Scheme. This work was undertaken at GOSH/UCL Institute of Child Health, which received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme. The Centre for Paediatric Epidemiology and Biostatistics also benefits from funding support from the Medical Research Council in its capacity as the MRC Centre of Epidemiology for Child Health. The study design, collection, analysis and interpretation of data, writing of the report, and the decision to submit the article for publication was conducted independent of the funding sources.
Competing interests: None.