Parental stress affects the emotions and behaviour of children up to adolescence: A Greek prospective, longitudinal study

Abstract

Systematic research about the continuity of mental health problems from childhood to adolescence is limited, but necessary to design effective prevention and intervention strategies. We used a population-based representative sample of Greek adolescents, followed-up from birth to the age of 18 years, to assess early influences on and the persistence of mental health problems in youth. We examined the role of peripartum, early development and parental characteristics in predicting mental health problems in childhood and adolescence. Results suggest a strong relationship between behavioural problems in childhood and adolescence for both genders, while emotional problems were more likely to persist in boys. Age and sex-specific models revealed significant positive associations between higher scores on the behavioural and emotional problems scales and higher frequency of accidents in preschool years, physical punishment in early childhood, lack of parental interest in child's school and activities, and perceived maternal stress in all children. Perceived paternal stress was associated with higher scores on the Total and Internalizing problems scales in the total population. Our results suggest that early interventions are necessary as mental health problems strongly persist from childhood to late adolescence. The adverse effects of parental stress and poor care-giving practices on child's psychopathology need to be recognised and improved.

Keywords::

• adolescence
• child
• parental stress
• Rutter's Parent Questionnaire
• sex differences
• Youth Self-Report

Introduction

In most children behavioural and emotional problems remain in later life, a progression often referred to as “continuity” of mental health problems. Studies on the continuity of behavioural and emotional problems from childhood to adolescence are crucial for understanding the natural course of mental health problems in young people, as well as for prevention and early intervention. With respect to the maintenance of relative mental health problems, previous studies suggest that conduct problems in childhood are a strong and consistent determinant of externalizing problems in adolescence, whereas a weaker association has been reported between emotional problems in childhood and later internalizing problems (Fischer et al. 1993; Koot 1995; Fergusson et al. 1996; Costello et al. 1999; Webster-Stratton and Taylor 2001; Pihlakoski et al. 2006). However, gaining knowledge in this area has been hampered by several methodologic limitations, such as reliance on clinical samples, use of retrospective designs and focus on a limited range of behaviours (Hofstra et al. 2000). Indeed, prospective, epidemiologic studies that provide information on the contribution of earlier psychopathology to the prediction of various problem behaviours based on representative, community samples of adolescents are limited (Achenbach et al. 1995; Ferdinand and Verhulst 1995; Caspi et al. 1996; Fergusson et al. 2005; Sourander and Helstelä 2005).

Furthermore, there is a growing need to understand the specific factors that influence child and adolescent emotional and behavioural problems (Mathijssen et al. 1999). Psychological health during childhood may be determined by peripartum, early development and parental influences. With respect to peripartum influences and early development, type of delivery, fetal distress and maternal distress in the perinatal period may affect externalizing problems in childhood (Allen et al. 1998; Raine 2002; Linnet et al. 2003). With respect to parental determinants, parent-child relationship, parenting strategies, family circumstances, such as low socio-economic status (SES), stressful life events, and family psychopathology may also be involved in the development of internalizing and externalizing problems in later childhood (Campbell 1995).

Finally, development of mental health problems may be associated with gender given that the prevalence rates and continuity of emotional and behavioural problems have been found to differ by gender (Cohen et al. 1993a,b; Visser et al. 2003; Sourander and Helstelä 2005; Pihlakoski et al. 2006; Kapi et al. 2007a,b). Since there is limited information related to the reasons behind these gender differences, identifying childhood antecedents of mental health problems for both genders separately would improve the understanding of the mechanisms underlying their development. Such valuable information can be gained from large cohorts of children by examining various forms of mental health problems in childhood.

However, the existing evidence on the relative importance of these factors on mental health problems in a longitudinal study of child development up to adolescence is weak, given that previously conducted longitudinal studies have often inadequately accounted for confounding variables or have not investigated multiple potential factors simultaneously, nor gender-specific issues.

In the present study, we examined to what extent parental reports of a child's emotional and behavioural problems in childhood can predict self-reported internalizing and externalizing problems in adolescence. In addition, multiple linear and logistic regression models were used to determine the combined effect of different peripartum, early development and parental variables on mental health problems during childhood and adolescence. We hypothesized that assisted delivery, fetal or maternal distress at birth and various parenting strategies and characteristics are related in different ways to emotional and behavioural problems in later years among boys and girls.

Methods

The methods of selection and follow-up of the cohort have been previously described (Tzoumaka-Bakoula 1987; Stefanis et al. 2004; Kapi et al. 2007a,b), and are summarized briefly here.

Study design and population

The Greek Birth Cohort is based on the Greek National Perinatal Survey, a prospective study of all the 11,048 births throughout Greece between the 1st and the 30th of April 1983. The enrolled participants represented 8% of the country's annual births and constituted a representative, population-based sample. In the year 1990, when the children were aged 7 years, attempts were made to identify all children at primary schools throughout Greece in order to collect participant postal questionnaires. Individuals were assured of confidentiality and anonymity. From the 9000 questionnaires sent out, a total of 8158 completed questionnaires were returned (91% response rate), of which 6643 questionnaires were successfully merged with corresponding data in 1983. In the year 2001, attempts were again made to locate participants, now aged 18 years, through 6200 high-schools throughout the country in order to collect adolescent and parental postal questionnaires once again. Of the 4675 postal adolescent/parent questionnaires distributed, 3500 responded (75% response rate). Individuals were assured that all procedures had been anonymized and that all information obtained was strictly confidential. For the 2001 survey the proportion of responders was estimated to be 32% of the initial birth cohort. Non-responders were either not identified as the questionnaires were kept anonymous, or were not attending one of the country's high schools. Others simply did not reply because they were invited to respond to the questionnaire shortly before their university entry examinations. Nevertheless, sensitivity analyses showed that even though the dropout in follow-up was quite high, the data collected in year 2001 was fairly well representative of the initial Greek birth cohort (Stefanis et al. 2004; Kapi et al. 2007a,b).

All studies were approved by the National Hellenic Research Foundation, the Greek Ministry of Education and Health, the Institute of Biological Research and Biotechnology and the National Privacy Principles Board.

Questionnaires and measures

Phase 1 (birth data)

In 1983 a questionnaire was completed by specially trained midwives and obstetricians in charge of labour. From this phase we retrieved information regarding: (1) method of delivery, grouped as spontaneous vaginal delivery, assisted vaginal delivery (i.e. use of forceps, vacuum, episiotomy and oxytocin administration) and caesarian section, and (2) presence of child distress and maternal distress at birth (based on reviewed obstetric records following delivery), and included them in the analysis to increase our understanding of the possible long-term effects of birth events on child mental health at 7 years.

Phase 2 (7-year data)

Parents and teachers completed a postal questionnaire when participants were 7 years of age. Data collected included information on the child's mental health extracted by the parent, as well as the recalled number of accidents during early childhood, the use of physical punishment in pre-school years and maternal interest in child school performance. The socio-demographic background of the children, based on the highest level of paternal and maternal years of education reported at this period, was also used for adjustment in the analyses, as it seemed to be the most relevant measure over the study time.

Phase 3 (18-year data)

Parents and adolescents responded to separate questionnaires in the 2001 follow-up study. The assessment included data on adolescents' mental health, parental interest in the child's school and leisure activities and mother/father's perceived stress status. The latter were self-reported in the parental questionnaire.

Outcome variable

Child mental health at age 7 years was assessed by asking parents to fill in the Rutter's Parent Questionnaire A2 (Rutter et al. 1970). This questionnaire consists of three subscales in the conduct, hyperactivity and emotional symptom domains. The total item score and the symptom scores for the conduct and emotional domains were used to indicate symptom loading.

A second psychometric assessment was done at the age of 18 years. The adolescents themselves completed the Youth Self-Report (YSR) (Achenbach 1991). This instrument includes items scores on a Total problems scale and two broad band scales: internalizing and externalizing problems.

The sex-specific cut-off point of about the 90th percentile of the distribution of the symptom scores in the present sample for both the total and the subscales scores on the Rutter's Parent Questionnaire A2 and the YSR was used to indicate a high level of symptom loading.

Statistical analysis

Statistical analyses were performed using SPSS 13.0 for Windows (Chicago, IL, USA). A significance level of 0.05 was used and all the statistical tests were two-tailed.

Spearman correlations analysis was used to evaluate the pairwise relationships between the Rutter A2 total and subscale scores (emotional and conduct) and the YSR total and subscale (internalizing and externalizing) scores. Univariate analyses, using Student's t-test and one-way analysis of variance (ANOVA), were carried out to examine bivariate associations between the peripartum and early parenting independent factors (type of delivery, child and maternal peripartum distress, number of accidents, physical punishment and maternal interest in child school performance) with the Rutter A2 total and subscale scores (emotional and conduct). Then, univariate analysis was carried out to examine bivariate associations between the parent/adolescent independent factors (parental interest in child's school and leisure activities and mother/father's perceived stress status) with the YSR total and subscale scores. The independent factors that showed significant associations with the Rutter A2 total and subscale scores (emotional and conduct) and with the YSR total and subscale scores were separately studied in linear regression analysis in the total sample. The linear regression models with outcome the YSR total and subscale scores included the corresponding Rutter A2 total and subscale scores as independent variables. We controlled for gender and parental years of education. In subsequent linear regression models, the interactions by gender for the significant terms were additionally included to investigate possible interaction effects. Next, we reran the linear regression models separately in boys and girls for the independent factors that showed significant associations with the Rutter A2 total and subscale scores, and with the YSR total and subscale scores, in univariate and linear regression analysis, in the total sample. This was done to further control for gender given our sample size of 1112 girls and 953 boys. The long-term influence of variables that referred to conditions before or at the age of 7 years on the YSR total and subscale scores was also examined. Finally, multivariable logistic analyses were performed to evaluate the relationship between the variables that were found to be significant in linear analyses and “scoring above cut-off point” for each dependent variable (total, emotional and behavioural problems) at 7 and 18 years of age.

Results

Participant characteristics

For the present study questionnaire data were examined for 2065 children (953 boys and 1112 girls) with available longitudinal information on youth's mental health problems at 7 and 18 years of age. Information was matched-up through the study phases using the following key-variables: date of birth, gender, place of birth and mother's place of residence prenatally. The background factors of the matched cohort members were compared to those of the children who did not respond later, or were not alive, or were not identified in the follow-up surveys (comparison group, N = 8894).

Children in the matched data and in the comparison group were equally likely to have an urban background (72% and 71.9%, Pearson's χ² test p = 0.93). In the matched data there were more children whose mothers had been married when the child was born (99% vs. 98%, p < 0.0001) than in the comparison group, but the bias in this respect is considered to be minimal because of the very low proportion of unmarried mothers in both groups. The mothers of the children in the matched data were slightly more educated and the fathers' social class was slightly higher (p < 0.0001 for both), but all levels of education and social classes were adequately represented in the matched data. Like in many studies, in our study females were over represented (54% in the matched data vs. 46% in the comparison group), but, since most of the outcomes were studied separately for boys and girls, this was unlikely to cause any bias in the results. Therefore, it can be assumed that no significant bias between the two samples was introduced.

Characteristics of the study sample are shown in Table I.

Table I.  Characteristics of the study sample at birth, 7 and 18 years of age.

	Number (%)
Child peripartum distress	393 (19.0)
Maternal distress during delivery	32 (1.6)
Type of delivery
Supported vaginal delivery	363 (17.6)
Caesarean section	112 (5.4)
Parental education >9 years	1322 (64)
Number of accidents in child's preschool years
One	487 (23.6)
Two or more	212 (10.3)
Physical punishment in child's preschool years
Occasionally	1086 (52.6)
Often	225 (10.9)
Maternal interest in child school performance	1654 (80.1)
Mother feels stress often/very often in her life	1026 (49.7)
Father feels stress often/very often in his life	966 (46.8)
Parents are interested in adolescent's school/leisure activities	1675 (81.1)

Measures of mental health problems and their correlations

To examine the relationships between the Rutter A2 total and subscale scores and the YSR total and subscale scores, pair-wise correlations were conducted, as shown in Table II. Results showed that the total scores on the Rutter A2 scale at the age of 7 years correlated directly with the total scores on the YSR and with externalizing problems at the age of 18 years. In addition, conduct problems at age 7 years correlated clearly with externalizing problems at the age of 18 years, while a weaker association was found with the total scale score at 18 years. Finally, emotional problems at 7 years correlated directly with internalizing problems and with the total scale score at the age of 18 years.

Table II.  Spearman correlations between 7 years Rutter Parent (A2) and 18 years YSR total and subscale scores.

	Rutter A2 scale (N = 2065)
YSR	Total	Emotional	Conduct
Total problems	0.079*	0.059*	0.050*
Internalizing	0.020	0.081*	− 0.030
Externalizing	0.092*	0.002	0.099*

*p < 0.05.

Risk factors associated with Rutter scale scores (age 7 years)

Results from the bivariate associations between the Rutter A2 total scale and subscale scores at age 7 years and the independent factors studied (type of delivery, child peripartum distress, maternal peripartum distress, number of accidents, physical punishment and maternal interest in child school performance) are shown in Table III. Linear regression models with the dependent variable Rutter A2 total scale and subscale scores all revealed a significant association with gender, number of accidents in child's preschool years and physical punishment (Table IV). In addition, assisted delivery was positively associated with conduct problems in childhood, while lack of maternal interest in child school performance was positively associated with both the total and the conduct problems scales in childhood (Table IV). The subsequent linear regression models including interaction terms with gender revealed significant relationships for being physically punished × gender (data not shown).

Table III.  Bivariate associations between the Rutter Parent (A2) total and subscale scores at 7 years and the examined factors.

	Total	Emotional	Conduct
Type of delivery
Assisted delivery	13.0 ± 0.3 (t = − 0.35)	1.3 ± 0.1 (t = 0.61)	1.8 ± 0.1 (t = − 1.98)*
Caesarian section	13.3 ± 0.5 (t = − 0.74)	1.5 ± 0.1 (t = − 1.01)	1.7 ± 0.2 (t = 0.04)
Child peripartum distress	13.3 ± 0.3 (t = − 1.36)	1.4 ± 0.1 (t = − 0.21)	1.8 ± 0.1 (t = − 1.60)
Maternal distress during delivery	13.0 ± 0.8 (t = 0.02)	1.2 ± 0.2 (t = − 0.55)	1.9 ± 0.3 (t = 0.57)
Physical punishment	(F = 14.81)*	(F = 20.91)*	(F = 44.32)*
Occasionally	13.1 ± 0.2	1.4 ± 0.0	1.7 ± 0.05
Often	17.1 ± 0.4	1.9 ± 0.1	3.1 ± 1.16
Number of accidents in child's life	(F = 32.74)*	(F = 13.34)*	(F = 19.89)*
One	13.6 ± 0.2	1.5 ± 0.1	1.8 ± 0.1
Two or more	15.3 ± 0.4	1.7 ± 0.1	2.4 ± 0.1
Maternal interest in child school performance	(F = 6.93)*	(F = 1.99)	(F = 6.99)*
Fairly	12.9 ± 0.13	1.3 ± 0.0	1.7 ± 0.0
None	15.8 ± 0.97	1.5 ± 0.2	2.7 ± 0.3

*p < 0.05. t, Student's t-test. F, one-way ANOVA.

Table IV.  Linear regression coefficients^a in the total study population and in both genders separately with outcome variable the 7 years Rutter Parent (A2) total and subscale scores.

	Total population			Boys			Girls
Rutter Parent A2 scores	Total	Emotional	Conduct	Total	Emotional	Conduct	Total	Emotional	Conduct
Gender girls vs. boys	− 0.103***	0.066**	− 0.142***	–	–	–	–	–	–
Parental education	− 0.045*	ns	ns	ns	ns	ns	− 0.110**	ns	− 0.096*
Assisted delivery	ns	ns	0.041*	ns	ns	0.078*	ns	ns	ns
Number of accidents in child's life	0.138***	0.104***	0.096***	0.159***	0.141***	0.119***	0.144***	0.085*	0.082*
Physical punishment
Occasionally vs. never	0.130***	0.054*	0.104***	0.114**	ns	0.104**	0.127**	0.089*	0.083*
Often vs. never	0.300***	0.161***	0.277***	0.280***	0.112**	0.268***	0.217***	0.123**	0.194***
Maternal interest in school performance
Fairly vs. very much	ns	ns	ns	ns	ns	ns	ns	ns	ns
None vs. very much	0.070**	ns	0.071**	0.079*	ns	ns	ns	ns	ns

ns, non-significant.

Note: ^a Standardized regression coefficients are shown.

*0.01 ≤ p < 0.05, **0.001 ≤ p < 0.01, ***p < 0.001.

Results from the separate linear regression models for boys and girls revealed that, for both genders, higher scores on the total Rutter A2, emotional and conduct scales were linearly associated with a higher frequency of physical punishment in pre-school years and a higher frequency of accidents experienced during the same period. In the girls' subsample neither the association between assisted delivery and conduct problems was significant, nor was the association between lack of maternal interest in child school performance and total or conduct problems scale scores. However, in the boys' subsample, lack of maternal interest in child school performance was significantly associated with the total Rutter scale score and assisted delivery was associated with a significant increase in Rutter conduct score. In girls, but not in boys, the level of parental education was significantly inversely associated with the total Rutter A2 score and with the conduct symptoms scale score (Table IV).

Risk factors associated with YSR scale scores (age 18 years)

Linear regression models with outcome the YSR total and subscale scores at age 18 years and independent factors were performed as follows: (1) the Rutter total and subscale scores, parental interest in child activities and mother/ father's perceived stress status (Model 1), with and without interactions by gender for the significant terms, (2) the independent factors significantly associated with Rutter and YSR total and subscale scores without controlling for the Rutter scores (Model 2), and (3) the independent factors significantly associated with the Rutter and YSR total and subscale scores including the Rutter scores (Model 3). For the total population, the initial model revealed a significant positive association between the YSR total and subscale scores and gender, mother being stressed in her life and lack of parental interest in child's school and leisure activities (data not shown). In addition, father's stress was positively associated with the total and internalizing YSR scale scores. The subsequent linear regression models including interaction terms with gender revealed significant relationships for total Rutter score × gender (data not shown).

Results from the separate linear regression models (Model 1) for boys and girls revealed a more constant positive linear association between the total Rutter A2 scores and the total YSR scale scores in boys, than in girls (Tables V and VI). In the models (Model 1) with the subscales, increasing scores on the conduct scale were associated with increasing scores on the externalizing problems scale in both genders, while increasing scores on the emotional scale at age 7 years were associated with increasing scores on the internalizing problems scale in boys, but not in girls. Lack of parental interest in child's school and leisure activities was associated with higher YSR total and subscale (internalizing and externalizing) scores in both boys and girls. Maternal stress was associated with an increased risk for higher total YSR scores in boys and girls, higher internalizing scale scores in boys and girls, and higher externalizing problems scores only in boys. The association between the YSR total and subscale scores with perceived paternal stress was insignificant in the analysis by gender.

Table V.  Linear regression coefficients^a in adolescent boys with outcome variable the 18 years YSR total and subscale scores.

	Model 1			Model 2			Model 3
YSR scores	Total	Internalizing	Externalizing	Total	Internalizing	Externalizing	Total	Internalizing	Externalizing
Parental education	ns	ns	ns	ns	ns	ns	ns	ns	ns
Rutter A2 scores
Total	0.100**	–	–	–	–	–	0.085**	–	–
Emotional	–	0.093*	–	–	–	–	–	0.093**	–
Conduct	–	–	0.112**	–	–	–	–	–	0.086*
Mother feels stressed this period of her life	0.132***	0.115**	0.090*	0.141***	0.079**	0.094**	0.094**	0.115**	0.088*
Father feels stressed this period of his life	ns	ns	ns	ns	0.052*	ns	0.064*	ns	ns
Lack of parental interest in adolescent's school/leisure activities	0.167***	0.174***	0.119**	0.172***	0.167***	0.124***	0.179***	0.174***	0.124***
Assisted delivery	–	–	–	ns	ns	ns	ns	ns	ns
Number of accidents in child's life	–	–	–	ns	ns	ns	ns	ns	ns
Physical punishment
Occasionally vs. never	–	–	–	ns	ns	ns	ns	ns	ns
Often vs. never	–	–	–	0.106**	ns	0.123***	ns	ns	0.100**
Maternal interest in child's school performance
Fairly vs. very much	–	–	–	ns	0.051*	ns	ns	ns	ns
None vs. very much	–	–	–	ns	ns	ns	ns	ns	ns

ns, non-significant.

Note: ^aStandardized regression coefficients are shown.

*0.01 ≤ p < 0.05, ** 0.001 ≤ p < 0.01, ***p < 0.001.

Table VI.  Linear regression coefficients^a in adolescent girls with outcome variable the 18 years YSR total and subscale scores.

	Model 1			Model 2			Model 3
YSR scores	Total	Internalizing	Externalizing	Total	Internalizing	Externalizing	Total	Internalizing	Externalizing
Parental education	ns	ns	ns	ns	ns	ns	ns	ns	ns
Rutter A2 scores
Total	ns	–	–	–	–	–	ns	–	–
Emotional	–	ns	–	–	–	–	–	ns	–
Conduct	–	–	0.134**	–	–	–	–	–	0.225**
Mother feels stressed this period of her life	0.114**	0.117**	ns	0.114**	0.119**	ns	0.094**	0.110**	ns
Father feels stressed this period of his life	ns	ns	ns	ns	ns	ns	0.064*	ns	ns
Lack of parental interest in adolescent's school/leisure activities	0.199**	0.168***	0.124**	0.257*	0.364*	0.257*	0.179***	0.152***	0.257*
Assisted delivery	–	–	–	0.292*	0.220*	0.267*	ns	ns	ns
Number of accidents in child's life	–	–	–	ns	ns	ns	ns	ns	ns
Physical punishment
Occasionally vs. never	–	–	–	ns	0.239*	ns	ns	− 0.103*	ns
Often vs. never	–	–	–	ns	ns	ns	ns	ns	− 0.167*
Maternal interest in child's school performance
Fairly vs. very much	–	–	–	ns	ns	ns	ns	ns	ns
None vs. very much	–	–	–	ns	ns	ns	ns	ns	ns

ns, non-significant.

Note: ^aStandardized regression coefficients are shown.

*0.01 ≤ p < 0.05, ** 0.001 ≤ p < 0.01, ***p < 0.001.

Results from the linear regression models examining the long-term influence of variables that referred to conditions before or at the age of 7 years (excluding the Rutter A2 scores, i.e. Models 2 in Tables V and VI) showed a significant linear association between being often physically punished in childhood and scoring higher on the total and externalizing problems scales in adolescence for boys. A significant positive association was also found between being occasionally physically punished in childhood and internalizing problems in adolescent girls. In addition, assisted delivery showed a significant linear association with all scale scores during adolescence for girls. However, when the Rutter A2 scores were added to the model (Model 3 in Tables V and VI), the above associations became insignificant or were reversed, which suggests that mental health status in childhood mediates or confounds the long-term association between both physical punishment and assisted delivery, and adolescent psychopathology.

Factors associated with Rutter A2 and YSR scores above the 90th percentile

A series of individual logistic regression models was conducted to test the odds of scoring above cut-off point on the problem scales (outcome) for the variables that showed a significant relationship with the outcome in linear analyses (Table VII). At 7 years of age, the odds for scoring above cut-off on almost all problem scales, in the total population and in boys and girls separately, increased with frequent physical punishment and with the number of accidents reported in pre-school years. Group differences in type of delivery (assisted delivery) were no longer statistically significant. At 18 years of age, participants with scale scores higher than cut-off point had mothers who reported feeling more stressed during that period of their life. Father's perceived stress status was associated with higher odds of scoring above cut-off point on the total problems scale for girls only. Both children and adolescents with increased problem scale scores experienced less parental interest towards their activities than their “under the borderline” counterparts.

Table VII.  Adjusted odds ratios (and 95% confidence intervals) from logistic regression analysis for the factors associated with mental health scores above cut-off point.

	Total population			Boys			Girls
Rutter Parent A2 scores	Total	Emotional	Conduct	Total	Emotional	Conduct	Total	Emotional	Conduct
Gender boys vs. girls	0.68*(0.50, 0.92)	2.17***(1.61, 2.91)	ns	–	–	–	–	–	–
Parental education
≤ 6 years vs. ≧13 years or more	1.70*(1.13, 2.55)	ns	ns	ns	ns	ns	2.03**(1.25, 3.32)	ns	1.73*(1.07, 2.80)
6–9 years vs. ≧13 years or more	ns	Ns	ns	1.71*(1.00, 2.92)	ns	ns	ns	ns	ns
10–12 years vs. ≧13 years or more	ns	0.59*(0.39, 0.89)	ns	ns	0.56*(0.34, 0.93	ns	ns	ns	ns
Assisted delivery	ns	ns	ns	ns	ns	ns	ns	ns	ns
Often physically punished	3.33***(1.96, 5.66)	2.36**(1.33, 4.22)	3.01***(1.83, 4.96)	2.21*(1.11, 4.39)	ns	2.04*(1.07, 3.88)	5.53***(2.62, 11.67)	4.47**(1.71, 11.68)	6.42***(2.98, 13.85)
Number of accidents in child's life
Two or more vs. none	2.56***(1.68, 3.88)	1.94**(1.29, 2.92)	2.33***(1.61, 3.33)	2.52**(1.46, 4.37)	2.09**(1.29, 3.38)	2.49***(1.50, 4.12)	2.99***(1.73, 5.15)	ns	2.22**(1.27, 3.86)
Two or more vs. one	1.59*(1.01, 2.52)	ns	1.75**(1.16, 2.70)	ns	ns	ns	1.96*(1.07, 3.60)	ns	2.14*(1.13, 4.05)
Maternal interest in child school performance
Very vs. none	0.25**(0.12, 0.55)	ns	0.28**(0.13, 0.61)	0.19**(0.07, 0.52)	ns	0.32*(0.12, 0.83)	ns	ns	0.29*(0.09, 0.92)
Fairly vs. none	0.25***(0.13, 0.51)	ns	0.34**(0.17, 0.69)	0.20***(0.08, 0.47)	ns	0.39*(0.17, 0.90)	ns	ns	0.32*(0.11, 0.93)
YSR	Total	Internalizing	Externalizing	Total	Internalizing	Externalizing	Total	Internalizing	Externalizing
Mother feels stressed this period of her life	1.42*(1.00, 2.06)	1.45*(1.06, 1.96)	ns	1.74*(1.03, 2.93)	ns	ns	ns	ns	ns
Father feels stressed this period of his life	ns	ns	ns	ns	ns	ns	1.55*(1.00, 2.38)	ns	ns
Parental interest in child's school/ leisure activities	0.47***(0.33, 0.68)	0.49***(0.35, 0.68)	0.50***(0.36, 0.69)	0.48**(0.28, 0.83)	0.53*(0.28, 0.99)	0.51**(0.34, 0.77)	0.42***(0.26, 0.67)	0.36***(0.24, 0.55)	0.57*(0.34, 0.98)

ns, non-significant.

Note: Unadjusted odds ratios and 95% confidence intervals are shown.

*0.01 ≤ p < 0.05, **0.001 ≤ p < 0.01, ***p < 0.001.

Discussion

This study further explored the continuity of mental health problems from childhood to adolescence and the association between peripartum, early development and parental influences and youth's psychopathology. Besides the continuity of emotional and behavioural problems from childhood to adolescence, we showed that mental health problems in childhood and adolescence are significantly associated with peripartum events, parenting strategies and parental stress.

We found a strong correlation between emotional and behaviour problems at the ages of 7 and 18 years, which confirmed findings from previous studies (Fergusson et al. 1996; Costello et al. 1999; Hofstra et al. 2000; Sourander and Helstelä 2005; Pihlakoski et al. 2006), despite the long period of observation and the use of cross-informant (i.e. from parent and child) rather than intra-informant (parent–parent) sources (Pihlakoski et al. 2006). The results showed a significant interaction between total Rutter scores in childhood and gender for the association between Rutter scale scores at 7 years and YSR scale scores at 18 years of age. The analysis stratified by gender revealed that the “persistence” of mental health problems was consistent across the behavioural problems scales for both boys and girls, but strong relationships between the total and emotional problems scales in childhood and adolescence, respectively, were found only for boys. The latter is contradictory to previous studies which found homotypic internalizing pathways from early school years to adolescence among boys and girls (Costello et al. 1999; Hofstra et al. 2000; Sourander and Helstelä 2005; Pihlakoski et al. 2006). A possible explanation is that cultural influences interfere with the complexity of continuity of emotional and behavioral problems, but cross-national comparisons are necessary to determine this. Additionally, the use of non-representative samples, differences in the age groups of the youths and other methodological differences can explain part of the differences. Furthermore, common genetic and environmental influences can lead to the stability of behaviour over time, while it is also likely that behaviour in childhood is another proxy for factors present in the child's adolescent environment, including peer rejection, and may not in itself be directly causal for adolescent behaviour (Fergusson et al. 2005). Nevertheless, proper path analysis using longitudinal data is needed to further test these claims.

Interestingly, our results confirmed those of earlier studies showing that assisted delivery influences child mental health (Paludetto et al. 2000; Taylor et al. 2000). Researchers hypothesized the presence of behavioural regulators programmed before and/ or at birth that influence postnatal responses to environmental stress (Paludetto et al. 2000; Taylor et al. 2000). As perinatal maternal and baby distress were strongly associated with assisted delivery (data not shown), we hypothesize that a brief but critical period around birth might influence child mental health in the long-term. The linkage between early child-mother interaction and development on the offspring has been shown in studies on mothers depressed during pregnancy (Lederman et al. 1981; Goodman and Gotlib 1999). Clearly, the current findings reinforce concerns about the long term impact of assisted delivery on conduct problems in childhood. At the same time, the linear analysis stratified by gender revealed an independent association in boys. It is possible that assisted delivery may have specific effects on personal adjustment in boys leading to an increased vulnerability to later problem behaviour. However, the analyses showed no significant association between conduct problems as a dichotomous measure and assisted delivery. It could be suggested that the choice of variable scaling may have influenced our findings, but further research on this issue is needed.

The findings from this study were also in accordance with previous longitudinal surveys which indicate that physical punishment in childhood is a risk factor for child and adolescent mental health problems (Fergusson and Lynskey 1997; Vitolo et al. 2005; Widom et al. 2007). Interestingly, we found a significant interaction between physical punishment in childhood and gender for the association between physical punishment in preschool years and mental health at 7 years of age. The analysis stratified by gender revealed that the effect on child mental health was consistent across all broadband scales for both boys and girls. However, we found the trend between being physically punished in childhood and emotional problems at 7 years of age to be more powerful in girls than in boys; and with conduct problems at the same age to be more powerful in boys than in girls. It is possible that such gender differences in the reaction of boys and girls to a stressful condition in childhood may be related to differences in coping with such problems. One should note, however, that the collection of information on this behaviour relied only on cross-sectional maternal reports, therefore this needs to be interpreted with caution. Finally, in the present study, the long-term effect on adolescent mental health was mediated or confounded by mental health status in childhood.

As expected, there was a strong association between number of accidents in a child's life and mental health physical problems in both boys and girls in childhood. Notably, number of accidents in a child's life is an indicator of the quality of parental supervision and care, parental stress and feeling of responsibility, and hence, may reflect sense of security of the child (Fleitlich and Goodman 2001; Ford et al. 2004; Kapi et al. 2007a,b). An increasing sense of security lowers levels of stress hormones in the child, which, in turn, improves cognitive and behavioural development (Fish et al. 2004). Posttraumatic parental stress and the behaviour of parents after the traumatic event may even be a more critical influence on children's psychopathology (Cox et al. 2008). However, in many cases pretrauma psychological problems were found to be the most consistent predictor of psychopathology after an accidental trauma (Cox et al. 2008). Although research indicates that posttraumatic females predominantly exhibit more internalizing problems and boys more externalizing behaviours (Winje and Ulvik 1998; Lalloo et al. 2003), in the present study a stronger association with emotional and behavioural problems was found in boys. Further research is necessary to understand the mechanisms underlying the influence of trauma experiences in childhood.

Attesting to this notion, we found that lack of parental interest in a child's activities was a significant contributor to poor mental health in childhood and adolescence. More exactly, lack of maternal interest in a participant's academic progress during childhood was associated with total and conduct problems reported at the age of 7 years, while lack of parental interest in a participant's school and leisure activities during adolescence was also associated with higher scores on all problem scales studied. From this, one could speculate that in the absence of parental involvement and monitoring in a youth's social life, whatever their age, children feel less “protected”, insecure and stressed. These results highlight the importance of parent nurture practices on parent–child interactions throughout childhood and adolescence. Therefore, attempts at preventing and reducing youth's psychological problems should include the provision of more attention and support. It is noteworthy, however, that the analysis stratified by gender for the 7-year-olds revealed that the association between maternal interest in child school performance and total problems on the Rutter scale was significant in boys, but not in girls. A possible explanation is that 7-year-old girls may be more aware of their role as a student, than boys of the same age. It could also be hypothesized that girls may be more willing or able to take on certain tasks at this age without feeling social or emotional distress.

Furthermore, we found strong associations between perceived maternal stress and behavioural and emotional problems in adolescents. The association between parental stress, particularly that of the mother, and psychopathology in children has been consistently described in earlier studies (Hackett and Hackett 1999; Fleitlich and Goodman 2001; Fish et al. 2004; Goodman et al. 2007). Other settings that might increase maternal anxiety, such as low socioeconomic circumstances (Goodman et al. 2007), family conflict (Loukas et al. 2001; Cummings et al. 2002), parental psychopathology (Nomura et al. 2002), or disruption of parent–child relationships (Olson et al. 2000), have also been shown to influence child psychopathology. Regardless of the underlying cause or structure, our results support the view that when mothers are stressed, children also become stressed, unhappy and insecure.

We noted that fathers play an additional role, as total problems and internalizing problems in adolescence were influenced by stress in fathers, too. Because maternal and paternal stress are intertwined and influence their perceptions and functioning in life, and are reversible, we believe that intervention at their level may prevent or improve child psychopathology. Therefore, training both parents to deal successfully with stressful situations may create a propitious home environment for the benefit of the entire family.

Our findings are based on a large, representative birth cohort studied prospectively. There are of course a number of limitations that need to be taken into account when interpreting the results. First, our findings were based on parental and child self-reports rather than on interviews by mental health professionals. However, both the Rutter Parent (A2) scale and the YSR are standardized and reliable psychometric tools that can be used in epidemiologic studies. Second, parental stress was studied with the use of self-reported measures, therefore perceived ideas, rather than psychometric tool measurements, were employed. One can argue that better validated measures may be necessary to capture the actual degree of stress in their life, however the possibility remains that parents in this cohort extensively reported the apparent level of stress in a self-completed questionnaire. Attesting to this notion, investigations that link stressors and physiological response have previously documented the importance of self-reported appraisal of stressful events (Kristensen 1996; Schwartz et al. 1996; Cohen et al. 1997) and its effect on physiological parameters (Goldman et al. 2005). In a small number of cases (9.8%) mothers and fathers were separated at the time the parental questionnaire was answered. However, this was unlikely to significantly distort our results, as this involved a small sub-sample of responses. Furthermore, the fact that mothers are considered accurate informants on family issues (Phares 1997) and responded to more than 90% of the questionnaires, supports the validity of our findings.

To conclude, the present study provides further evidence that there are certain perinatal, child and parental parameters that protect or predispose children to emotional and behavioural problems during childhood and adolescence. On this basis, special attention should be given to parental nurturing practices and their psychosocial health. The finding that these parental parameters have implications for a youth's mental health in later life prompts the need for further research on these issues in order to identify underlying mechanisms that might explain the epidemiological associations found. Such an approach may allow us to use the present findings in public health strategies to reduce the prevalence of mental health problems in childhood and adolescence.

Acknowledgements

We would like to thank all the respondents who participated in this study for their pivotal contribution. This study was supported in part by the European Commission Quality of Life and Management of Living Resources Program (Contract No. QLG1-CT-2000-01643). The Greek Birth Cohort study was further supported by the Academy of Athens and the Greek Ministry of Education.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.