OUP user menu

Parenting and health in mid-childhood: a longitudinal study

Andrea Waylen, Nigel Stallard, Sarah Stewart-Brown
DOI: http://dx.doi.org/10.1093/eurpub/ckm131 300-305 First published online: 17 January 2008

Abstract

Background: Parenting and parent–child relationships influence children's emotional and social development and evidence exists that they may be life-course determinants of health. This study tests the hypothesis that adverse parenting in the early years predicts poor health in mid-childhood. Methods: A prospective study using data from the Avon Longitudinal Study of Parents and Children cohort. Health data on over 8000 children (60% of those recruited) were available for analysis at 6.9 and 7.7 years. Exposures: self-reported maternal hostility, resentment and hitting/shouting in early childhood. Outcomes: maternal report of child's health in general and number of health problems when the child was 6.9 and 7.7 years, adjusting for socioeconomic factors. Results: Sub-optimal parenting, as measured here, was observed among 62, 80 and 83% of families for hostility, resentment and hitting/shouting, respectively. Resentment was more common among older mothers in owner-occupied housing. Resentment and hostility predicted health at both ages independently of socioeconomic circumstances. ‘Hitting/shouting’ was weakly predictive of number of health problems. A greater proportion of variance was explained by parenting variables than by socio-economic variables. Conclusions: Parenting and parent–child relationships in the early years predict health in mid-childhood in a way consistent with a causal role. If further studies replicate this finding, policies to improve parenting could be expected to have a modest beneficial impact on health as well as emotional and social development. As some aspects of sub-optimal parenting show reverse social class distribution, initiatives targeted at those living in social deprivation may not achieve the optimum impact on health.

  • ALSPAC
  • childhood health
  • longitudinal
  • mother–child relations
  • parenting

Introduction

Parenting and parent–child relationships influence multiple aspects of children's social and emotional development1–5 including internalizing (anxiety and depression) and externalizing disorders (conduct and hyperactive disorders),6 educational attainment1 and future social status.7 Evidence is emerging from longitudinal studies that parenting and parent–child relationships may impact on physical health both in childhood and adulthood.6,8–13 Aspects of optimal and sub-optimal parenting that have been implicated include warmth, hostility, negativity, conflict and discipline.12, ,14 Many of these are amenable to intervention15 and such interventions are a key component of current early-years policy in the UK.16

The aim of this study was to examine the extent to which child health in mid-childhood is predicted by self-reported sub-optimal parenting on the grounds that the latter was less likely to be affected by socially desirable reporting than self-report of optimal parenting. We used data from the UK-based Avon Longitudinal Study of Parents and Children (ALSPAC) where self-report measures of parenting were collected on multiple occasions during the first 4 years of life. The study reported here focuses on three derived measures of parenting and two measures of child health. Our hypothesis was that sub-optimal parenting in early childhood would predict poor health in mid-childhood after adjusting for socioeconomic factors.

Methods

Participants

ALSPAC (see www.alspac.bris.ac.uk)17 is a geographically representative, population-based study investigating social, environmental, biological and genetic influences on the health and development of children. All pregnant women in the former Avon Health Authority (UK) with an expected delivery date between 1 April 1991 and 31 December 1992 were invited to take part; ∼85% of the eligible population enrolled resulting in a cohort of 14 541 pregnancies. A total of 13 988 infants survived to their first birthday. Detailed information has been collected from mothers or main carers bi-annually about the health, behaviour and development of the study child.

Questionnaire data regarding child health was returned for 8502 children (61% of those surviving to their first birthday) at around 81 months (6.9 years) and 8255 (59%) around 91 months (7.7 years). (Demographic data for this study population is provided in table 1.) Compared to those who dropped out, mothers of children remaining in the study when the children were aged 7 were older (28.9 versus 27.3 years P < .001), reported less poor health (4.8% often or always unwell versus 6.3%, P = 0.006), had fewer financial difficulties (2.54 versus 3.54 from a maximum of 15, P < .001) and were more likely to own their own home (81.8% versus 61.1%, P < .001).

View this table:
Table 1

Demographic details of the study population [N (%)]

MaleFemale
Child's gender (age 7.7)4224 (51.4)3994 (48.6)
Financial difficulties (in pregnancy) mean—(range)2.9 (0–15)2.9 (0–15)
Maternal age at birth—mean (range)28.4 (15–44)28.3 (14–45)
Housing tenure (in pregnancy)Owner/mortgaged5002 (73.3)4658 (73.4)
Rented (council)1095 (16.0)1010 (15.9)
Rented (housing association)510 (7.5)434 (6.8)
Other218 (3.2)242 (3.8)
Maternal assessment of child's health (age 7.7)Very healthy, no problems2698 (63.8)2513 (62.9)
Healthy, minor problems1455 (34.5)2886 (35.1)
Sometimes quite ill62 (1.5)111 (1.4)
Almost always unwell9 (0.2)1 (0.03)
Number of health problems in last year (age 7.7) mean (range)1.4 (0–6)1.5 (0–6)

Parenting measures

Mothers reported their attitudes, behaviours and feelings toward the study child in 32 separate questions at several time-points from child age 4 weeks until 4 years. Correlations between these items and subsequent factor analysis (with variables loading >0.32) suggested three measures of sub-optimal parenting: hostility, resentment and hitting/shouting. Scores for each measure were computed by summing Likert scale responses for each contributing variable: each of the scores was divided at two cut points to give three groups representing high, moderate and low levels of sub-optimal parenting. The distribution of the hostility variable allowed us to divide the data into approximately equivalent tertiles. The distributions for the resentment and hitting/shouting variables were non-normal. The resentment variable was divided using the following cut points:—exactly how parents felt (high levels of sub-optimal parenting); how they felt sometimes (moderate levels) never or rarely felt (low levels of sub-optimal parenting). The hitting/shouting variable was split according to the following cut points: behaviours used every day or almost every day (high levels of sub-optimal parenting); sometimes used (moderate levels) and never or rarely used (low levels).

The variables contributing to these measures of parenting are shown in table 2 together with the frequency of high, moderate and low scores. For all three, high scores indicate least good parenting. All aspects of the study conform to the ethical regulations of both the ALSPAC Law and Ethics Committee and the research and ethics committees of the local National Health Trust.

View this table:
Table 2

Derived parenting variables, constituent items (high scores denote sub-optimal parenting) and frequency of occurrence in study population

Parenting variablesLevela%
    Constituent items
Hostility
    I often get irritated by this child (47 months): 0 = no, 1 = sometimes, 2 = yesLow (0–2)38
    I have frequent battles of will with this child (47 months): 0 = no, 1 = sometimes, 2 = yesModerate (3–4)28
    This child gets on my nerves (47 months): 0 = no, 1 = sometimes, 2 = yesHigh (5–6)34
Resentment
    I dislike the mess that surrounds my child (33 months): 1 = never feels, 4 = feels exactlyLow (4–8)20
    I really cannot bear it when the child cries (33 months): 1 = never feels, 4 = feels exactlyModerate (9–12)70
    I feel I have no time to myself (33 months): 1 = feels exactly, 4 = never feelsHigh (13–14)10
    I dislike the noise and mess that surrounds this child (47 months): 0 = no, 1 = sometimes, 2 = yes
Hitting / shouting:
    How frequently is the child smacked because of a temper tantrum?(18 months): 1 = never, 3 = oftenLow (4–10)17
    How frequently is the child shouted at?(24 months): 1 = never, 5 = every dayModerate (11–14)53
    How frequently is the child slapped?(24 months): 1 = never, 5 = every dayHigh (15–18)30
    How frequently is the child smacked for being naughty?(42 months): 1 = never, 5 = daily
  • a:Each aspect of parenting is categorized with mothers reporting either low levels of hostility, resentment or hitting/shouting (optimal parenting), moderate or high levels of each style (sub-optimal parenting).

Child health measures

Mothers assessed the child's health during the past year in separate questionnaires at 81 months (6.9-years old) and 91 months (7.7-years old) in the context of a series of questions about physical health. Response options were: very healthy with no problems (61.3%/63.4% for 81- and 91-month assessments, respectively), healthy but with minor problems (36.8%/35.1%), sometimes quite ill (1.7%/1.4%) or almost always unwell (0.1%/0.1%): high scores indicate poorer health. Mothers were also asked to record common health problems the child had suffered during the past year including temperature, cough, earache and asthma. Between none and 19 different conditions (from a possible total of 24) were reported when the children were 81 months and between none and six (from a total of 23) when the child was 91 months.

Socio-economic factors

Socio-economic risk was assessed on the basis of three variables selected because of their capacity to predict health problems in other studies of this type18–21: number of financial difficulties during pregnancy (ranging from 0 to 15); housing tenure during pregnancy [high score indicates higher socioeconomic status (SES)] and maternal age at birth. Analyses were also adjusted for gender of child (table 3).

View this table:
Table 3

Correlations between parenting variables, demographic factors and child health outcomes at 91 months

HostilityResentmentHitting/shoutingMother's assessment of child's health aNumber of health problems in last year
Hostility0.264***0.247***0.073***0.055***
Resentment0.137***0.070***0.039**
Hitting/shouting0.0150.035***
Mother's assessment of child's health0.297***
Mother's age at birth−0.041***0.100***−0.174***0.0130.024*
Child's sex b−0.008−0.007−0.174***0.0080.052***
Housing tenure c−0.026*−0.049*−0.038*−0.031−0.029**
Financial difficulties d0.075***0.083***0.080***0.047***0.011
  • *P < 0.05; ** P < 0.01 *** P < 0.001

  • a: 1 = very healthy, no problems, 4 = almost always unwell

  • b: 1 = male, 2 = female

  • c: 0 = other, 1 = rented (housing authority), 2 = rented (council), 3 = (owned/mortgaged)

  • d: 0 = no financial difficulties, 15 = maximum financial difficulties

Statistical modelling

Analyses were carried out using Intercooled STATA 9.0 on both the 81 (N = 7789) and 91 month (N = 8255) outcomes. As the pattern of results was virtually identical at both ages, only analyses based on 91-month outcomes are reported. Data for the 81-month analyses are available on the website of this journal if required. Correlations between variables were assessed using Spearmans Rho. The extent to which measures of parenting predicted overall health independently of socioeconomic factors was assessed using proportional odds regression because the health outcome measure was categorical but ordinal. For each category of health, this modelling process calculates the odds of being in a specific health category (e.g. healthy, but with minor problems) relative to all lower categories combined (sometimes quite ill or almost always unwell). Models for both 81- and 91-month data met the assumptions for proportional odds modelling (P = 0.07 and 0.09, respectively). Analysis of impact on condition counts was assessed using Poisson regression because of the skewed distribution.

Results

Sub-optimal parenting was reported for 62, 80 and 83% of children on measures of hostility, resentment and hitting/shouting, respectively during the first 4 years of life with high levels of sub-optimal parenting being reported for 34, 10 and 30% of children, respectively (table 2). Parenting measures were correlated with each other at around 0.25 with the exception of resentment and hitting/shouting (0.14). Correlations between measures of parenting and child health were small ranging from 0.02 to 0.07. Hitting/shouting and hostility were correlated in the expected direction with socio-economic risk: less hostility and less hitting/shouting were associated with higher maternal age, owner occupation and lack of financial difficulties. The relationship of resentment to socio-economic factors was more complex. While resentment increased with financial difficulties, it also increased with maternal age and owner occupation. Housing tenure and financial difficulties predicted health outcomes as expected. Younger maternal age was associated with poorer health as measured by the number of health problems reported by the mother, but not by her assessment of the child's health in general. Mothers reported more childhood health problems during the past year for daughters than for sons (table 3).

Table 4 presents unadjusted and adjusted univariable regression analyses estimating the impact of the parenting measures independently of each other on health in general and number of health problems at 91 months. In both analyses, children in poor health were more likely to have been exposed to high levels of resentment and hostility than those in better health. The health of children exposed to moderate levels was intermediate. Adjusting for socio-economic factors had little impact on the odds. Hitting/shouting increased the odds of reporting more health problems, but not the odds of poor health in general.

View this table:
Table 4

Univariable and multivariable proportional odds regression: parenting variables on child's health in general and number of health problems age 91 months

Child's health in generalNumber of health problems in last year
POR95% CIPIRR95% CI
Univariable
Unadjusted
    HostilityOverall<0.001<0.001
Moderate versus low0.0021.201.07–1.36<0.0011.091.04–1.15
High versus low<0.0011.431.28–1.60<0.0011.091.05–1.15
    ResentmentOverall<0.0010.014
Moderate versus low<0.0011.431.25–1.630.0111.071.02–1.13
High versus low<0.0011.651.37–2.000.0111.111.02–1.19
    Hitting/shoutingOverall0.154<0.001
Moderate versus low0.4330.940.82–1.090.0011.021.02–1.08
High versus low0.4471.060.91–1.24<0.0011.061.07–1.14
Adjusted (gender, maternal age, housing tenure and financial difficulties)
    HostilityOverall<0.001<0.001
Moderate versus low0.0061.191.05–1.350.0021.081.03–1.14
High versus low<0.0011.401.24–1.57<0.0011.101.05–1.15
    ResentmentOverall<0.0010.019
Moderate versus low<0.0011.451.27–1.670.0111.071.02–1.13
High versus low<0.0011.661.36–2.030.0171.101.02–1.20
    Hitting/shoutingOverall0.0600.044
Moderate versus low0.4790.950.81–1.100.3441.030.97–1.10
High versus low0.2471.100.93–1.310.0201.091.01–1.16
Multivariable (adjusted as above)
    HostilityModerate versus low0.0161.171.03–1.330.0011.111.04–1.17
High versus low<0.0011.301.15–1.480.0021.091.03–1.15
    ResentmentModerate versus low<0.0011.361.18–1.57
High versus low<0.0011.501.22–1.85
R20.007R20.002

In the multivariable analyses (table 4), where all variables are entered together, moderate and high levels of both hostility [odds ratio (OR) 1.17 (1.03–1.33) and OR 1.30 (1.15–1.48), respectively, P = 0.016] and resentment [OR 1.36 (1.18–1.57) and OR 1.50 (1.22–1.85), P < 0.001] independently predicted health in general. Hostility was the only parenting measure to independently predict number of childhood health problems at 91 months [IRR 1.11 (1.04–1.17, P = 0.001) and IRR 1.09 (1.03–1.15, P = 0.002) for moderate and high levels of hostility, respectively].

R2-values obtained in the stepwise regression on health in general were 0.002 for SES variables collectively, 0.007 for SES variables and hostility, and 0.010 for SES variables, hostility and resentment. Results for parallel models on number of health problems were consistent with these findings. Adding hitting/shouting did not increase the R2 in either model. Corresponding values for stepwise regression in which parenting variables were entered first were parenting variables collectively 0.007 rising to 0.010 when all three SES variables were added.

Discussion

In these analyses, the chances of a child experiencing poor health at 7 and 8 years were increased by exposure to maternal resentment and hostility between 2 and 4 years of age, with higher levels of both predicting greater ill health. The strongest predictor was resentment, a measure capturing dislike of the mess and noise associated with the child, the child's crying and lack of time to self. Hitting and shouting were the least predictive and hostility intermediate. As with the other measures of adverse parenting we analysed, the prevalence of resentment increased with financial difficulties. However, unlike others, resentment was more prevalent among older mothers and those whose houses were owned outright. The capacity of these measures to predict child health was not confounded to a significant degree by financial difficulties, maternal age, housing tenure or gender of the child. Results are consistent with, but fall short of, proving causality as they demonstrate a dose–response relationship and are independent of the effects of key socio-economic confounders.

They are also consistent with some11,,13 but not other8 studies of parenting and health in early childhood. These studies differ with respect to aspects of parenting studied and the way they were studied. Mantymaa11 used an objective measure of maternal–infant interaction in early infancy, examining health at age 2 years in a relatively small sample and concluded that socio-economic circumstances did not confound the relationship. Belsky8 used a range of objective measures from which were derived composite variables covering warmth, negativity and control (the latter two measures including hitting and shouting) in a large representative sample and showed that whilst parenting mediated the effect of socio-economic circumstances, the parenting variables were not independently predictive. Both studies were based on maternal report of child's health in general and objective measures of parenting. Self-report measures of parenting such as those used in this study are generally regarded as less robust than objective measures. It is likely that parents under-report difficulties, particularly those about which there has been negative publicity like hitting and shouting. However, it is less likely that parents who do not have problems with parenting report them. Self-report measures of hostility, resentment and hitting/shouting may therefore be more robust than self-report measures of more positive aspects of parenting. Self-report measures also have the advantage that they capture attitudes and behaviours across a longer time span than observations, which are necessarily based on intermittent and relatively short observation visits.

The results of the two studies discussed earlier and the study reported here are compatible with a stronger predictive effect for more subtle problems with parenting—sensitive interaction and resentment—which do not show a classic distribution with social inequalities, than for the more commonly studied aspects of parenting such as conflict, hitting and shouting. The story they tell with regard to the role of parenting and social inequalities in health is therefore complex. One message is that sub-optimal aspects of parenting are not confined to those living in poverty and may be more of a problem amongst older mothers living in better circumstances. Indeed, the minimal change in ORs after adjustment for socio-economic factors suggests that the effects of the aspects of parenting on health reported here may be relatively independent of socio-economic circumstances at birth. Another message is that parenting predicted health in mid-childhood to a greater extent than social circumstances. As this study focused on health as a categorical outcome, we were not able to assess the extent to which parenting mediated the influence of social circumstances on health. Other studies have suggested that it may.8

The results of these studies focusing on health in mid-childhood can be contrasted with those examining the effect of parenting on health in adolescence and adulthood. The latter have shown a consistent effect of parenting on health that is reduced but not abolished by adjustment for social circumstances.6 It is possible that the health impact of parental hostility gets stronger with age. As well as biological mechanisms likely to be pertinent at all ages,6 during adolescence and adulthood additional mechanisms, acting through impaired social competence or the adoption of unhealthy lifestyles, may increase deleterious health effects.

The R2-values in this study were modest, suggesting that neither parenting nor social inequalities, as measured here, play a large role in determining the health of 7- and 8-year-olds. However, in this respect, it is important to note the relatively small proportion of the population in the optimal parenting categories. A small increase in risk of a common problem applied to a large proportion of the population can be very significant in terms of population health and demand for health care. It is also of note that the study focused only on maternal parenting and other research has suggested that the impact of fathering on health is at least as strong as that of mothering.12

Limitations of the study

Like most other studies of this kind, the measurement of children's health was based on maternal reports of ‘health in general’ and the number of common health problems children had experienced in the last year. Responses to questions asking about ‘health in general’ tend to reflect physical rather than emotional health, especially when asked in the context of other questions relating to physical health. Whilst fulfilling the function of generic measures, which take into account the range of health problems potentially related to parenting,6 maternal report raises the possibility of confounding by, for example, mother's perception of illness severity or her own mental or physical health. Those that have investigated the validity of generic health measures in childhood by examining the correlation between physician- and mother-reported child health have, however, concluded that the measure is valid7 and a further study suggests that the influence of maternal mental health on reporting of child health status is negligible.22

In this study, we examined three aspects of parenting which were correlated with each other. We focused on sub-optimal rather than optimum aspects of parenting on the grounds that these were likely to be less susceptible to bias relating to socially desirable responding. These aspects of parenting are likely to be correlated with other aspects of the parent–child relationship, which are not considered here and may warrant further study.

Whilst recruitment to the ALSPAC cohort was very good, attrition was high during the first 7 years, with 60% of those recruited available for analysis at 7.7 years. As with all such studies, attrition has been differential with greater drop-out among the most deprived. It may be that the relatively modest effect of social inequalities on health reported here is partly attributable to loss from the study of the most deprived families. Differential attrition might also account for the low level of effect of hitting and shouting.

Conclusions

This study examined the predictive capacity of common aspects of sub-optimal parenting on child health and, although results fall short of proving causality, they are consistent with a causal role. The findings are potentially important in terms of child and public health more generally because of the large numbers of children affected and the potential for intervention. If the results prove robust, current policies16,,23 and guidance24 designed to improve parenting via both universal and targeted approaches could have a modest beneficial impact on physical as well as emotional and social development. At present, estimates of the potential benefits which could derive from wider provision of parenting interventions are based on their impact on the latter alone. Inclusion of potential health effects would weight cost benefit analyses in the direction of provision.

The apparent deleterious effects of sub-optimal parenting, particularly resentment, on health were not confined to young mothers and those living in poverty, indicating that parenting initiatives targeted on those living in social deprivation (as recommended for the prevention of crime and disorder) might not have an optimum impact on health.

Further research is needed to confirm the findings of this study and to define the aspects of parenting which are most detrimental to child health. In this study, hostility and resentment were more influential than hitting and shouting.

Supplementary data

Supplementary data are available at EURPUB online.

Acknowledgements

We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council, the Wellcome Trust and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and Dr Waylen will serve as guarantor for the contents of this article. This research was carried out with funds provided by the Department of Health to support the development of research in new medical schools. The data analysis described in this report was supported by a Fellowship provided by Warwick Medical School (Waylen).

Conflict of interest: None declared.

Key points

  • Parenting and parent–child relationships are potential life course determinants of health.

  • Much of the research in this area has been retrospective and therefore may be subject to bias.

  • In this prospective cohort, maternal resentment of and hostility towards the child in the early years predicted poorer health in mid-childhood.

  • Mothers who were older or who owned their own homes reported the highest feelings of resentment.

  • Universal initiatives aimed at all families with young children may be more effective than parenting initiatives targeted on those living in social deprivation.

References

View Abstract