Introduction

In the last decade, several large longitudinal studies on the epidemiology of child mental health and child psychopathology have been conducted [1–3]. However, prospective research on the psychopathology of preschool children has been lacking due to the late age of enrollment (mostly after school entry or even later in early adolescence) for most longitudinal studies on child mental health problems.

Dimensional studies of child psychopathology often focus on the dimensions of internalizing and externalizing problem behavior. Externalizing problems show considerable longitudinal stability, even from before the age of 4 years [4]. The assumption of a developmental sequence that starts in the preschool years with oppositional, hyperactive, and aggressive behavior, leading to conduct problems in later childhood [5] stands in contrast to findings that children with high degrees of externalizing problems are also vulnerable to internalizing symptoms of anxiety or depression [6]. Also, externalizing or disruptive behavior may be channeled into internalizing symptoms as a result of age-dependent and sex-stereotyped socialization in school [7]. In adolescents, anger and frustration have been found to be positively related to externalizing problem behavior [8], but also to psychosomatic disorders and depression [3], both of which constitute internalizing disorders. Such co-occurrence is a commonly described phenomenon [9, 10].

Studies of population-based preschool samples focusing on internalizing and externalizing symptoms show inconsistent results with regard to sex differences. In a population-based sample of children between the ages of 4 and 18 years [11], no differences were reported for internalizing problems between boys and girls in early childhood, and externalizing problems declined with age. Across all age groups, boys showed more externalizing problems than girls, and in adolescence, more internalizing symptoms were reported for girls than for boys [12, 13]. However, results from other studies have not provided any indications of sex differences at preschool age, neither for internalizing nor for externalizing problems [14, 15].

In addition to the particular risk of child mental health problems manifesting in boys, there are further risk factors for the onset of child mental health problems associated with the child’s family [16, 17]. Existing findings demonstrate that both structural factors, such as socioeconomic status and educational level [18], and changes in family relationships, such as separation or divorce, are associated with childhood psychopathology [19].

In summary, mental health studies on children spanning the transition from preschool to primary school are rare, and it is still not clear whether sex differences already exist at preschool age and what the significant factors in the development of childhood psychopathology are. The present study focused on the continuity of child mental health problems in the transition from preschool to primary school. In a four-year longitudinal design, a representative nonclinical community sample was examined at preschool age and 4 years later in primary school. The continuity of symptoms and the degree of internalizing and externalizing problems at both time points were examined. In order to control for the co-occurrence of the two broadband groups, a combined group evidencing both internalizing and externalizing problems was defined.

Methods

Sample and Procedure

This study was designed to prospectively explore the prevalence and changing patterns of mental health problems in subjects drawn from the general population of preschool children in a northern German city of 260,000 residents. Participation in both surveys was voluntary, and written consent was obtained from the parents of all children in the study.

The baseline survey was conducted 6 months prior to primary school entry in the context of a statutory pediatric health examination [20]. Four years later, the population of the baseline survey was re-examined in the fourth grade of primary school. The response rate of the 1,481 children from the baseline sample who were asked to take part in the second survey was 68.3%; of these children, 814 were adequately assessed to be included in the second wave of the longitudinal study.

For both surveys, the adult caregiver was asked to complete some questionnaires. Mothers were the primary respondents at both baseline (88 vs. 11% for fathers) and follow-up (90 vs. 9% for fathers). Four children lived in foster care (see Table 1). The sample was comprised of 377 boys (46.3%) and 437 girls, aged between 5 and 6 years, with a mean age of 5.9 years at baseline.

Table 1 Demographic features of the sample and the local population
Full size table

Drop-Out-Analysis

The participants did not differ significantly in age, gender, number of siblings, or parents’ education level and vocational situation relative to the eligible sample (see Table 1), although slightly fewer non-German children (3.6% vs. 7.4%) participated in the study.

Measures

The German version of the Child Behavior Checklist 4–18 (CBCL) [21] was completed by the parents or the primary caregiver. The CBCL 4–18 is a well-established screening instrument for childhood psychopathology [22]. The CBCL consists of 118 problem items that are scored from ‘0’ to ‘2’, based on the occurrence of the behavior during the preceding 6 months: ‘0’ if the item is not true, ‘1’ if it is somewhat or sometimes true, and ‘2’ if it is very true or often true. By summing the scores, eight syndrome scales, two broadband scales (Internalizing and Externalizing), and a Total score were computed. The good reliability and validity of the CBCL were confirmed for the German version of the measure [23]. Cronbach’s alpha for Total, Internalizing, and Externalizing scales ranged from 0.94 to 0.81. The test–retest reliability over a period of 5 weeks was 0.81. Four items concerning school, drugs, and sexual problems (items 73, 96, 101, and 105) were not considered applicable to preschool children and were excluded from the CBCL. This approach is in accordance with other studies of preschool children [6, 24].

Case Definition

The case definition is based on sex-specific T-scores according to the CBCL manual [23] and the national norms [25]. The cut-off point for the prevalence of child mental health problems was defined as a T-score of 64 or more on the Total score or on one of the broadband scales. For those children who met the criteria of child mental health problems, four subtypes of problem behavior were defined. Children who had a T-score of 64 or more on just one of the two broadband scales of internalizing or externalizing behavior were allocated either to the high internalizing (INT) or to the high externalizing (EXT) group. Those children who had a T-score of 64 or more on both broadband scales were considered to be in the combined group (COM) of high internalizing and externalizing behavior. Finally, children who met the criteria on the Total score, but not on any of the broadband scales, were allocated to the “other” category.

The grouping of COM and “other” were added to the already existing definition of INT and EXT [23]. The additional categories are necessary for a clear and unambiguous allocation of all possible patterns of psychopathology of those children meeting the case definition.

Statistical Procedures

The data analysis was performed using the Statistical Package for Social Sciences (SPSS for Windows, version 17). Differences in problem behavior between baseline and follow-up and between boys and girls were examined with a two-factor repeated-measures analysis by means of the general linear model procedure. Effect sizes for the ANOVA were computed in terms of partial eta-squared (η²). A conservative interpretation of η² is given by Cohen [26], with η² = 0.02 considered a small effect size, η² = 0.15 a medium effect size, and η² = 0.35 a large effect size.

The categorical distribution of broadband groups and the continuity over time were examined by means of cross-tabulation and Chi-square test. The risk of child mental health problems was calculated by means of a logistic regression analysis, with presence or absence of psychopathology at T2 as the dichotomous dependent variable. The odds ratios of each independent variable are reported as indicators of risk.

Results

CBCL Scores at Baseline (T1) and Follow-Up (T2)

The means of the CBCL scores at T1 and T2 and the post hoc statistics are listed in Table 2. In a repeated-measures analysis, the within-subjects effect of time of measurement on CBCL scores was significant, F (11, 802) = 9.70, p < .001. Also, the between-subjects effect of sex was significant, F (11, 802) = 5.05, p < .001. In terms of effect sizes, η² was .117 for the effect of time and .065 for the effect of sex, both representing small to medium effects. There was no significant interaction of time and sex, F (11, 802) = 0.99, p = 0.46.

Table 2 T-Scores of CBCL subscales at baseline (T1) and follow-up (T2)
Full size table

Effect of Time

The within-subjects effect of time on CBCL broadband scales and syndrome is reported in Table 2. Post hoc analyses revealed a significant increase of symptoms over time for the syndrome scales Somatic Complaints, Anxious/Depressed, Social Problems, and Attention Problems. With regard to the broadband scales, we found a significant increase in Internalizing symptoms and a minor decrease in Externalizing symptoms. There was no significant difference between T1 and T2 in Total Score.

Boys Versus Girls

The differences between boys and girls were significant on the broadband scale of internalizing symptoms and on two of the three internalizing syndrome scales, but not on externalizing scales. For all significant differences, boys had higher scores than girls: Withdrawn, 55.3 versus 54.3 (F = 7.13, p = 0.008, η² = 0.009); Anxious/Depressed, 56.2 versus 55.0 (F = 8.52, p = .004, η² = 0.01); Social Problems, 54.1 versus 53.2 (F = 7.46, p = .006, η² = 0.009) Attention Problems, 54.4 versus 53.0 (F = 16.56, p < .001, η² = 0.02); and Internalizing Behavior, 54.2 versus 52.5 (F = 9.70, p = .002, η² = 0.012).

Broadband Groups INT and EXT and the Need for Combined Group COM

At T1, 129 of 814 (15.8%) children had mental health problems. Of the 129 children with mental health problems, there were 98 children with pronounced internalizing problems and 43 with pronounced externalizing problems, including 13 children with both internalizing and externalizing problems. To differentiate between children with problems of a single broadband group and children with problems characteristic of both broadband groups, independent grouping into three categories was necessary. Of the total number of 129 children (100%) with mental health problems at baseline, 85 (66%) children belonged to the INT group of children with only high internalizing scores, 30 (23%) belonged to the EXT group of children with only high externalizing scores, and 13 (10%) children fulfilled the criteria for the COM group of children with both high internalizing and high externalizing scores. The symptoms of 1 (1%) child were assigned to the “other” group. There was no significant difference between boys and girls in the distribution of INT, EXT, and COM at T1 (Χ² (2, N = 128) = 1.49, p = .47). However, overall, significantly more boys (n = 73; 19.4%) than girls (n = 56; 12.8%) had mental health problems, Χ² (1, N = 814) = 6.51, p = .012.

At T2, 165 of 814 (20.3%) children showed mental health problems. Of those 165 (100%) children, 90 (55%) were classified as INT, 27 (16%) as EXT, 42 (26%) as COM, and 6 (4%) were identified as belonging to the “other” category. At T2, the distribution of INT, EXT, and COM between boys and girls was not significantly different (Χ² (2, N = 159) = 0.04, p = .98). However, as at T1, child mental health problems were significantly more common in boys (n = 93; 24.7%) than in girls (n = 72; 16.5%), Χ² (1, N = 814) = 8.41, p = .004. The number of children in each broadband group and the combined group, as well as the changes between T1 and T2, are shown in Table 3.

Table 3 Frequencies of Int, Ext, and Com at baseline (T1) and at follow-up (T2): continuity within specific broadband groups and within any broadband group
Full size table

Continuity of Behavioral and Emotional Symptoms

The overall concordance of child mental health problems between T1 and T2 was 81%. Out of a total of 129 children with mental health problems at T1, 71 (55%) children still had problems at T2. The continuity of having no mental health problem at T1 and T2 was at a high level: 86%.

The continuity of symptoms between T1 and T2 within the same problem group was 32 out of 85 (38%) children for INT, 4 out of 30 (13%) children for EXT, and 4 out of 13 (31%) children for COM. The continuity rates for INT (38%) and EXT (13%) were significantly different (Χ² (1, N = 115) = 6.1, p = .02). There was only a slight crossover move from INT to EXT (2%), a moderate move from INT to COM (12%) and from EXT to INT (13%), but a substantial move from EXT to COM (23%).

Predictors of Child Psychopathology at T2

The risk of child mental health problems at T2 was calculated by means of a logistic regression, with the following independent variables: child’s sex, the presence of child mental health problems at T1, educational level of the mother, family status (two-parent versus single-parent), and divorce or separation (see Table 4).

Table 4 Logistic regression predicting child psychopathology at follow-up
Full size table

Pre-existing child mental health problems at T1 (OR = 7.3; Wald = 87.85, df = 1, p < .001) had a significant effect on child mental health problems at T2. Also, parents’ separation or divorce (OR = 1.9, Wald = 2.80, df = 1, p = .09) and boys, as opposed to girls (OR = 1.6; Wald = 5.31, df = 1, p = .02), were associated with a higher risk of child mental health problems. However, no significant effect on the risk of mental health problems at T2 was found for the educational level of the mother (Wald = 3.56, df = 2, p = .17), nor for two-, as opposed to single-, parent families (Wald = 0.60, df = 1, p = .44).

Discussion

The results show significant effects of time and sex on CBCL scores. There was also considerable continuity of mental health problems in children with combined internalizing and externalizing problems as well as considerable continuity of children without any mental health problems. Children with pre-existing mental health problems at preschool age, children whose parents were separated or divorced, and boys had a significantly higher risk for child mental health problems at the four-year follow-up.

CBCL Scores at Baseline and Follow-Up

Over the period of four years, which included the important transition from preschool status to primary school, there was no significant change in the degree of child mental health problems on the global level of total scores, though there were changes in terms of the internalizing and externalizing subscales. In particular, significantly higher scores on Somatic Complaints and Anxious/Depressed represented the significant increase of internalizing symptoms, which is in accordance with some previous findings [11, 15]. Also, the shift from externalizing to internalizing symptoms over time has been reported elsewhere [11, 27].

The significantly higher scores for Social Problems and Attention Deficit Problems at follow-up were for neither internalizing nor externalizing symptoms and may have been related to the new demands on the children attending school at T2, especially in terms of social behavior, cognitive performance, focusing, and attention.

Grouping of Children with Mental Health Problems

By defining a combined group, COM, of co-occurring internalizing and externalizing symptoms, a new category was introduced as an addition to the two discrete broadband groups, INT and EXT. The definition of the COM group implied a high overall symptom load. The importance of co-occurring INT and EXT symptoms has been confirmed in several recent studies [28–30]. Already at T1, and even more prominently at T2, externalizing symptoms were seldom present on their own but were often combined with internalizing symptoms. This combination was represented in the COM group. A shift in psychopathology over time from externalizing symptoms to a broader scope of symptoms was also found elsewhere [11]. Therefore, the newly defined category COM has proven to be essential for a comprehensive classification of child mental health symptoms.

Continuity and Prognosis

The most impressive continuity from baseline to follow-up was found for the children in the COM group (70%) and for those children without any psychiatric problems (86%). Furthermore, the amount of continuity in the INT group (37%) was nearly twice as high as in the EXT group (19%), and the substantial crossover moves from EXT to INT (15%) and from EXT to COM (18%) demonstrate the shift from externalizing to internalizing symptoms. The crossover move from EXT to INT over time is in accordance with the results from the present study on the level of CBCL syndrome scales as described above. Similar results have been found in other studies [11, 27]. The higher risk of mental health problems for school children with pre-existing mental health problems at preschool age is an important finding—one that supports the assumption of continuity, aggravation, and chronic manifestation of mental health problems already evident in early childhood.

The effect of family background variables is controversial in the existing literature. Some studies have shown that children living in families with both biological parents had significantly lower frequencies of disturbance than almost any other family constellation [31, 32]. In the present study, we found no significant effect of single- versus two-parent families on child mental health, which is in accordance with Ford et al.’s findings [16]. However, there was a significant effect on child mental health problems in children who had experienced parental separation or divorce in the last 4 years. This closer link between parental separation or divorce and mental health problems has also been reported in other studies [24, 30, 33]. Separation or divorce and single-parent status are correlated, but only in a specific temporal sequence. The more recent state variable of separation or divorce explained a significantly higher amount of variance than the more enduring family background variables. However, further studies are needed to determine which effects of family variables are important for possible developmental pathways of child mental health problems.

Limitations of the Study and Conclusions

These results must be considered within the context of their limitations. The population-based sample was drawn from a city where university and public service employees are overrepresented. Consequently, generalization of the findings to national samples may be limited. Given the large sample size, a symptoms checklist was used to identify child psychopathology. A structured interview might have provided more specificity. Finally, the age of the children at baseline was between 5 and 6 years, which is at the upper limit of the usual definition of preschool. Nevertheless, all children met the preschool status due to the higher school entry age in Germany, which is comparable to Scandinavian countries and the Netherlands.

In conclusion, the results of the present longitudinal study have important research and clinical implications. In terms of research, further systematic investigation is needed on general development and on sex-specific aspects of mental health problems in early childhood. Our results, which show a significant increase of internalizing psychopathology in boys in conjunction with high degrees of general mental health problems during the 4 years of the follow-up period, require further prospective research on sex-specific early child mental health.

In clinical terms, the continuity of early child mental health problems from preschool age to the fourth year of primary school points to the important aspects of aggravation and chronic manifestation of mental health problems already present in early childhood. CBCL scores represent dimensional measures of symptoms and are not identical to statements of categorical diagnosis in our population sample. However, it has been shown that dimensional CBCL scores correlate significantly with categorical data of psychiatric diagnoses from structured interviews [34]. The increasing number of children with both internalizing and externalizing symptoms suggests a higher complexity in the development of child mental health problems with age. This study points to the need for early intervention at preschool age to prevent aggravation and chronic manifestation of psychopathology in later childhood. Specifically, the vulnerability for internalizing psychopathology in boys requires the particular attention of child mental health workers, not only at primary school age, but prior to school entry in preschool settings and in early childhood.

Summary

A sizeable number of studies have examined the development of mental health problems in older children, adolescents, and young adults, but only a few longitudinal studies have examined child mental health problems at preschool age. In the present follow-up study, changes in and continuity of behavioral and emotional problems were examined over a period of 4 years, from preschool age to the fourth grade of primary school. A total sample of 814 children was studied in kindergarten, 6 months prior to primary school enrollment, and at a follow-up 4 years later. Child mental health problems were assessed by means of the Child Behavior Checklist (CBCL). In addition to the CBCL broadband groups exhibiting either internalizing or externalizing symptoms, a third group of combined symptoms was defined.

There was a high degree of continuity of internalizing symptoms over the 4 years from baseline to follow-up. Also, there was a shift from externalizing symptoms in kindergarten towards a combination of internalizing and externalizing symptoms in primary school. The presence of mental health problems at follow-up was correlated to sex (higher among boys than girls), pre-existing mental health problems at baseline, and separation or divorce of parents.

The increased number of children with combined symptoms demonstrates the increasing complexity of child mental health problems in the developmental span from preschool to school age. The high level of continuity of early child mental health problems points to the need for intensive research on early child mental health disturbances and to the need for the provision of psychiatric services for young children.