To assess the degree to which preventive child health professionals (CHPs) identify and manage psychosocial problems among preschool children in the general population and to determine the association with parent-reported behavioral and emotional problems, sociodemographic factors, and mental health history of children.
The CHPs examined the child and interviewed the parents and child during their routine health assessments. The Child Behavior Checklist (CBCL) was completed by the parents.
Sixteen child health care services across the Netherlands that routinely provided well-child care to nearly all preschool children.
Of 2354 children aged 21 months to 4 years who were eligible for a routine health assessment, 2229 (94.7%) participated.
Main Outcome Measures
Identification and management of psychosocial problems by CHPs.
In 9.4% of all children, CHPs identified psychosocial problems. Two in 5 of the CHP-identified children were referred for additional diagnosis and treatment. Identification of psychosocial problems and subsequent referral were much more likely in children with a clinical CBCL total problems score than in others (identification: 29% vs 7%; odds ratio [95% confidence interval], 5.40 [3.45-8.47]; referral: 15% vs 3%; odds ratio [95% confidence interval], 6.50 [3.69-11.46]).
The CHPs frequently identify psychosocial problems in preschool children, although less than among school-aged children, but they miss many cases of parent-reported problems as measured by a clinical CBCL score. This general population study shows substantial room for improvement in the early identification of psychosocial problems.
Psychosocial problems, such as social-emotional and behavioral problems, are highly prevalent among children and adolescents and may severely interfere with everyday functioning of children and their families.1- 5 Interest in the occurrence of these problems among toddlers has grown recently,2,4- 7 whereas previous research mostly focused on school-aged children.1,8 As a consequence, relatively little is known of the occurrence of these problems among toddlers and the possibilities for early detection and subsequent treatment. Evidence in the general population is also lacking.
Early detection and treatment may improve the prognosis of psychosocial problems among preschool children. First, psychosocial problems in preschool children may continue in later stages of life. Lavigne et al4 found that more than 50% of children with psychiatric disorders at age 2 and 3 years continued to have a psychiatric disorder 3.5 to 4 years later, using a variety of measures, including the Child Behavior Checklist (CBCL), to assess a psychiatric disorder. Similarly, in a retrospective study of antisocial behavior in adolescence, Moffitt and Caspi9 showed that adolescents with childhood-onset problems already had behavior problems at age 2 or 3 years, whereas Aguilar et al10 showed social-emotional indices to be similarly predictive. Mesman and Koot6 found parent-reported internalizing and externalizing problems of preschool children to be predictive of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, counterparts 8 years later. Second, available reviews show that early interventions may be effective in preventing the continuation of psychosocial problems, especially antisocial behavior.11,12
In the Netherlands, preventive child health care offers an ideal opportunity for the early detection of psychosocial problems among preschool children, comparable to community pediatrics in the United States. In this system, child health professionals (CHPs) (ie, physicians and nurses) working in preventive child health care offer routine well-child care, including the early detection of psychosocial problems, to the entire Dutch population.13 Access is independent of insurance status, but the services do not provide treatment services, in contrast to the US system.
We previously reported on the degree to which Dutch CHPs identified and managed psychosocial problems in children aged 4 to 16 years.1 One or more psychosocial problems were identified in 25% of all children, and 1 in 5 of the identified children were referred for further diagnosis and treatment. Results further showed that identification of psychosocial problems in children and subsequent referral were 6 times more likely in the 8% with serious parent-reported problem behavior (measured by the CBCL). However, CHPs identified no psychosocial problems in 43% of these children and therefore undertook no action. We concluded that screening for psychosocial problems may be a promising option to reduce these problems among school-aged children but that accuracy of the identification should be enhanced,1 for instance, by training of professionals or by the use of short screening questionnaires.14,15
The aim of the present study is to assess the degree to which CHPs identify and manage psychosocial problems among children aged 21 months to 4 years in the general population. We also assess which child factors are associated with the identification of psychosocial problems by CHPs and with their referral for further evaluation and treatment. Finally, we compare our results with a similar study of older children.1
Trained CHPs interviewed a national sample of parents on psychosocial problems among their toddlers from October 1, 1997, to June 30, 1998.
The sample was obtained using a 2-step procedure. In the first step, a random sample of 16 of 65 child health services was drawn after stratification by region and degree of urbanization of their district. In the second step, each clinic provided a random sample of 75 children from their lists for 2 age groups: 2 years (±2 months) and 3 years and 9 months (±3 months), using random numbers. Of the 2354 eligible children, 2229 (94.7%) participated. The main reason for nonresponse was lack of interest in the objective of the study. The sample was representative for the entire Dutch population, except that infants living in big cities were relatively underrepresented due to the sampling procedure.
The data were collected in a regular way as part of the preventive health assessments to which all Dutch children are invited regularly. The CBCL/2-3 was mailed to parents along with the regular invitation to the preventive health assessment. The CBCL was completed by the parents and returned to the researchers in a sealed envelope. After each child's physical examination, the CHP obtained sociodemographic and mental health history information following a standardized interview with the parents. After each assessment, the CHP responded yes or no to the following question, "Does the child have a psychosocial problem at this moment?" The CHP then scored the type of identified problem(s) on a precoded list. Children who only had risk indicators for the development of psychosocial problems, such as parents with psychiatric problems or other family problems, had to be coded as no. If a problem was identified, the CHP was asked to rate the severity of the problem (mild, moderate, or severe) and to indicate how the problem was managed (precoded question). The design of the study was approved by the local medical ethical committee. It was similar to simultaneously performed studies on younger (1-6 months)16,17 and older (4-16 years) children.1,18
The CBCL/2-3 was used to assess the parent's report on the child's behavioral and emotional problems during the preceding 6 months.19 The good reliability and validity of the CBCL/2-3 established by Achenbach19 were confirmed for the Dutch translation.20,21 The CBCL/2-3 consists of 99 problem items and 3 additional items regarding handicaps and parental worries. Seven syndrome scales, 2 broad-band groups of syndromes designated internalizing and externalizing, and a total problems score were computed. Internalizing consists of the withdrawn/depressed and anxious syndrome scales, externalizing consists of the oppositional, aggressive behavior, and overactive syndrome scales, and sleep problems and somatic problems constitute separate syndrome scales. The overactive syndrome scale is specific for the Dutch version, whereas the further structure of the instrument is similar to the US version.21 Cases were subsequently allocated to a normal range or a clinical range of the scoring distributions for each separate scale based on the Dutch normative sample (H. M. Koot, PhD, written communication, April 28, 1999). Cutoffs were set at the 97th percentile for the 7 syndrome scales and at the 90th percentile for the total problems and internalizing and externalizing scales.
The sociodemographic variables assessed were sex, age, ethnicity, family composition, number of siblings living in the family at the time of study, educational level and employment status of the parent(s), degree of urbanization, duration of pregnancy, type of delivery, postdelivery hospitalization of the child, parity, and day care. Ethnicity was based on the native country of the child and both biological parents. At least 2 of them had to be born outside the Netherlands to qualify as non-Dutch. Parental educational level was used as a measure of socioeconomic status and based on the highest degree completed by the father or mother. Degree of urbanization was assessed using the postal code of the address of residence.22
Data on mental health history expressed whether the child was currently being treated (point prevalence) or had been treated (lifetime prevalence) for psychosocial problems. Response options included mental health professionals (eg, psychiatrist, psychologist), medical professionals (eg, general practitioner, pediatrician), and other professionals (eg, specialized family help, parenting support). In addition, life events in the previous year (such as hospitalization, death of family member, unemployment, divorce) were assessed in a standardized way.
First, we examined the prevalence of psychosocial problems as identified by CHPs. Second, we analyzed the management strategies used by CHPs and their relation to the severity of the problems. We used χ2 tests to determine the statistical significance of differences between distributions of categorical data. Third, we assessed which child factors (CBCL problem scales, sociodemographic variables, and [mental] health history variables) were related to the identification of psychosocial problems by CHPs (no or yes), using univariate and multivariate logistic regression analyses. We repeated these analyses regarding a referral for psychosocial problems by CHPs (no or yes). All independent child variables were dichotomized.
The regression analyses were performed using multilevel techniques because of the hierarchical nature of the data: characteristics of a CHP may have an impact on the assessments of all the children who are seen by the CHP. Multilevel models account for this clustering of individual data by the CHP (n = 108).23,24
Prevalence estimates presented in the tables and text are weighted by region and age to adjust for differences between the study population and the Dutch population. Test statistics, odds ratios (ORs), and 95% confidence intervals (CIs) were calculated on the basis of the unweighted data.
In 200 (9.4%) of all 2229 children, the CHP identified 1 or more psychosocial problems. The severity of the problems was rated as mild in 50.8% of these cases, moderate in 39.7%, and severe in 9.5%. At the time of the study, 1.0% of all children were being treated for psychosocial problems by a mental health professional. This group was excluded from all further analyses and 1 child with missing data on CHP-identified problems (the remaining number of nontreated children was 2205).
The CHPs undertook actions in 84% of the nontreated children with identified psychosocial problems. Various management strategies were used: advice or reassurance (72.4%); consultation with day care, colleagues, or official authorities (24.1%); follow-up (23.6%); and referral to another professional (40.7%). Management strategies varied according to the severity of the problems as rated by the CHP (Table 1). Follow-up, consultation, and referral were more frequent in those children whose psychosocial problems were rated moderate or severe.
Table 2 presents the number of nontreated children with a CBCL total problems score in the normal and clinical range who were identified by CHPs as having psychosocial problems and the management strategies used. Of the nontreated children, 6.1% had a CBCL total problems score in the clinical range (6.4% of all children). The CHPs identified psychosocial problems in 29.4% of the children with a CBCL total problems score in the clinical range (cutoff at 90th percentile). This percentage was 43.6 for those scoring above the 98th percentile of the CBCL total problems score and 47.6 for the children scoring above the 99th percentile. The CHPs identified psychosocial problems in 6.8% of the children with a CBCL total problems score in the normal range. These problems were rated as mild in 60.8% of the cases, moderate in 33.4%, and severe in 5.9%; the CHP rating of children scoring in the clinical range of the CBCL total problems score was 28.0%, 45.8%, and 26.2%, respectively (χ2 = 19.03, P<.001).
No actions were taken in 75.4% of all children with a CBCL total problems score in the clinical range; in 93.8% of these cases this was due to the fact that CHPs had identified no psychosocial problems. Referral to another professional was almost 6 times more likely in children with a CBCL total problems score in the clinical range (15.1%) than in those scoring in the normal range (2.6%).
Table 3 presents the association of CBCL problem scales with the identification of and referral for psychosocial problems by CHPs. Identification and referral were 5.4 (95% CI, 3.5-8.5) and 6.5 (95% CI, 3.7-11.5) times more likely in case of an elevated CBCL total problems score, respectively. Looking at CBCL broad-band scales, the ORs were much higher for elevated externalizing scores than for elevated internalizing scores, however, and this was reflected by the associations for the syndrome scales, with ORs being highest for the oppositional and overactive syndromes.
Concerning the sociodemographic variables, Table 4 shows that CHPs identified psychosocial problems relatively frequently in some groups: older children, children of single parents, and children with parents of low educational level. Referral depends on other characteristics, however, being more likely among non-Dutch children and children of parents of low educational level. Factors related to pregnancy and delivery had no relation with either identification or referral. Finally, parent-reported past treatments for psychosocial problems were significantly related to both the identification of and referral for psychosocial problems.
Table 5 provides the results of multiple logistic regression analyses. It lists those variables that were significantly related to the identification of psychosocial problems by CHPs (in the second column) and to referral for these problems (in the third column) after adjustment for the effect of all other variables. Due to the hierarchical structure of the CBCL scales in which problem items can contribute at the same time to syndrome scales, to the higher-order externalizing and internalizing scales, and to the total problems scale, 3 models were tested for both dependent variables. All models contained the sociodemographic and mental health history variables. Further, the first model contained the total problems scale, the second one contained the broad-band scales, and the third one the CBCL syndrome scales, as far as they univariately had a statistically significant relation with each outcome. These results show that the associations of CBCL and mental history variables with both identification and referral were much stronger than those of sociodemographic variables. Furthermore, the set of variables that predicts referral was much smaller than the one that predicts identification. Finally, day care was associated with less identification after adjustment for the other variables, whereas it was not in the univariate model. Additional adjustment for all other variables that had been included in Table 3 and Table 4 yielded similar results (not shown).
Physicians and nurses working in preventive child health care identified psychosocial problems in 9.4% of the general population of children aged 21 months to 4 years and mostly rated these cases as mild or moderate. The CHPs undertook actions in most of the identified cases of psychosocial problems, mainly by giving advice to parents but relatively frequently also by referring them. Two in 5 of the CHP-identified children were referred for further diagnosis and treatment. Identification of psychosocial problems and subsequent referral were much more likely (5.4 and 6.5 times, respectively) in children with CBCL total problems scores in the clinical range than in other ones. However, CHPs identified no psychosocial problems in most children (71%) with such a score in the clinical range.
Methodologic factors may partially explain why CHPs do not identify all children with parent-reported problems on the CBCL and vice versa, even though the CBCL is one of the best validated and most widely used parent-report questionnaires to assess psychosocial problems of children. First, CHPs identify a broad range of psychosocial problems, from rather minor behavioral problems to major psychiatric morbidity, whereas high CBCL scores typically occur among children who have been referred to psychiatric services. We thus expect CHPs to identify on the whole less severe problems and at the same time a lot more problems. Interestingly, the latter only partially applies, since CHPs identified problems in 9.4% of all children, whereas 6.4% of all children had a clinical CBCL total problems score. Second, the CBCL/2-3 is an imperfect gold standard. The sensitivity and specificity of the Dutch version at the cutoff for the borderline clinical range are 63% and 84%, respectively, using referral to mental health care as the criterion (OR, 8.720; for the US version, the OR was 9.1).19 Assuming a true prevalence of problems of 10%, this implies that two thirds of the CBCL cases produced false-positive cases for which the CHP was right in not identifying problems. Third, the CBCL only provides parent information, whereas CHP identification is also based on the child and on information from colleagues observing the same child. Disagreement between various informants on behavioral and emotional problems has been extensively documented. The mean Pearson correlation, for instance, between problem scales derived from different informants is only 0.22.25 This will again lead to less concordance between CBCL and CHP identification.
Other methodologic factors are unlikely to have biased our results significantly. The response in our study was very high (94.7%), and it concerned a representative sample of the Dutch population. Only children living in highly urbanized areas, which included many non-Dutch children, were underrepresented26 because of the sampling procedure. This may have led to some underestimation of the prevalence of CHP-identified problems, because they identified problems somewhat more frequently among these groups, although differences were not statistically significant. On the other hand, parents filled out the CBCL in advance and were thus forced to think about their child's problems, which may have lead to higher rates of reported problems. Finally, the prevalence of clinical CBCL scores in our sample was somewhat lower than in the Dutch normative sample.20 An explanation may be that the Dutch normative study concerned a relatively small sample (n = 420) from one province, including a metropolitan area.20
Our results show that both CHP identification of psychosocial problems and subsequent referral are strongly related to parent-reported problems on the CBCL and to previous treatment for psychosocial problems but less to sociodemographic background. Regarding the CBCL, especially troublesome behavioral problems that bring the child in conflict with the environment are highly predictive, as is shown by the high ORs for clinical scores on externalizing problems (7.65) and for contributing syndrome scales such as oppositional and overactive (19.8 and 10.2, respectively). A clinical score on the internalizing scale had no independent association with either identification or referral, and the same holds true for the syndrome scales withdrawn/depressed and anxious; these problems reflect inner conflicts of the child. Regarding sociodemographic background, relatively few characteristics contribute to identification and even less to referral, and characteristics of pregnancy and delivery contribute to neither of these. Thus, CHPs mostly base their identification of psychosocial problems on the mental health characteristics of the child and not on the child's background, and this holds even stronger for the decision to refer a child. Regarding this, the only background characteristic that predicts referral in the multivariate model, parental educational level, may also be interpreted as a proxy for parental skills to solve existing problems, but further research on this is needed.
The results of the present study show some interesting differences compared with our earlier study that was designed to study identification and management of problems among school-aged children (ages 4-16 years) in an identical way.1 First, rates of CHP identification of problems were much higher in the school-aged group (25% vs 9% in the preschool group). Next, the rates of referral of those identified and not yet under treatment were higher in the present study (41% vs 21%). True differences in the prevalence of problems between these age groups may account for these differences or a reluctance of CHPs to define mild behavioral problems among toddlers as psychosocial. Regarding the first explanation, it is interesting that in the present study CHPs identify relatively more problems among the oldest group (ages 3.5-4 years), whereas in the previous study this applies to those aged 4 to 11 years (compared with 12-16 years). This indicates a steady increase in problems until the end of primary school and a decrease afterward. A recent study by Bongers et al27 indeed shows mean CBCL total problems and externalizing scores to decrease from the age of 4 years until the age of 18 years, with the decreases being somewhat steeper for boys. Regarding preschool children, no data are available.
Second, among toddlers a clinical score on CBCL internalizing problems has a relatively small and not independent association with both CHP identification and referral, whereas among school-aged children the ORs for internalizing problems are similar to those for externalizing problems. For instance, among the latter the adjusted ORs (95% CIs) for identification are 2.49 (1.90-3.28) and 1.93 (1.48-2.53), respectively. An explanation may be that behavioral problems, covered by the externalizing CBCL scale, are really the major part of psychosocial problems among young children. An alternative explanation may be that in young children behavioral problems can be observed much more easily than emotional problems because of their limited verbal capacities. Probably both factors contribute.
Third, among school-aged children, the independent association of sociodemographic background characteristics with identification and referral was even smaller, with only age and urbanization being associated with identification and none with referral. Differences in sample size and thus in power to detect associations could also account for these differences if the present study had concerned more children, but in fact the sample size of the present study was approximately half the size of the study on school-aged children. Moreover, when only examining the strength of associations and not their statistical significance, similar results were obtained. Above all, it should be realized that in both age groups variables related to mental health and previous treatment dominate with regard to both CHP identification and referral, and this seems to be reinforced with increasing ages. This information can in all likelihood be obtained more easily in older children.
As far as we know our study is one of the first to examine CHP identification of psychosocial problems and subsequent management among preschool children and the first based on a population sample that also comprises children not seeking curative care. In 2 studies conducted in Monroe County, New York, in the late 1970s, Goldberg et al28,29 found that 3.3% to 3.5% of children aged 1 to 4 years who visited a pediatrician had mental health problems. Starfield et al30 found that 4.5% to 15% of children aged 1 to 4 years received a psychosocial diagnosis. Finally, Lavigne et al5 found a rate of pediatrician-identified emotional and behavioral problems similar to ours (9.3%) among children aged 2 to 3 years who visited a pediatrician from the US Pediatric Practice Research Group for any reason. In that study, sensitivity and specificity of pediatrician identification were 20.5% and 92.7%, respectively, using an assessment by a child psychologist as the gold standard. Because our study comprised all children in a general population, our prevalence rates are difficult to compare with these. However, our results regarding accuracy of identification confirm those of Lavigne et al.5 Thus, substantial room exists for improvement in the early detection of psychosocial problems among preschool children. This may, for instance, be reached by the use of short symptom checklists that parents fill out before the assessment31; training of professionals in the assessment of psychosocial problems, including the use of other informants such as day care providers and a separate developmental assessment to sort out developmental delay as a cause of behavioral problems; and reserving more time per visit. If a child has been referred, another problem may be engaging the child effectively in mental health services, especially if waiting lists are long.32 This makes it even more necessary to refer the children and families who are most in need.
Our results need confirmation in other community-based studies of preschool children, including the use of other criteria as a gold standard for child mental health, in addition to the CBCL. A next step should be the evaluation of various methods to improve early detection and effective early treatment and referral.
The prevalence of psychosocial problems is high among preschool- and school-aged children. Early detection and treatment improve the prognosis of these problems. Preventive child health care offers an ideal opportunity for the early detection and subsequent treatment of psychosocial problems. However, evidence on this early detection among preschool children in the general population is lacking. This study shows that psychosocial problems among preschool children are identified frequently by preventive child health care but that many cases of parent-reported problems are missed. Substantial effort should be invested in the improvement of this early detection.
Correspondence: Sijmen A. Reijneveld, MD, PhD, Department of Health Sciences, Groningen University, PO Box 196, 9700 AD Groningen, The Netherlands (S.A.Reijneveld@med.rug.nl).
Accepted for publication March 12, 2004.
This study was financially supported by the Dutch Ministry of Health, Welfare, and Sports, The Hague, the Netherlands; the Praeventiefonds (grant 28-2628-1), The Hague; and Zorg Onderzoek Nederland (grant 101004-12), The Hague.
We thank the personnel of the 16 Dutch child health care services who participated in this study.
Reijneveld SA, Brugman E, Verhulst FC, Verloove-Vanhorick SP. Identification and Management of Psychosocial Problems Among Toddlers in Dutch Preventive Child Health Care. Arch Pediatr Adolesc Med. 2004;158(8):811-817. doi:10.1001/archpedi.158.8.811