Abstract

Screening for psychopathology through administering generic questionnaires is common practice. Informants are possibly limited by the standardized formats of these instruments. The purpose of this study is to examine the clinical relevance of freely chosen descriptors of the child by translating these in terms of the Five Factor Model for personality. For over 15 years parents of clinically admitted children were asked to describe their child in their own words. In total 3436 youngsters (2150 males and 1286 females aged on average 10.2 years and 12.1 years respectively) were described on average by 8 descriptors. The thus acquired 27,774 parental free descriptors were then coded in terms of the Five Factor Model. The descriptors were reliably translated in proportional profiles on the Five Factor Model. Type of pathology was predicted by distributions of parental descriptors amongst the Five Factor Model for over half of the admitted youngsters. This study is the first to investigate the usefulness of free parental descriptors in relation to screening for psychopathology. Parental free descriptors appear to represent clinically relevant aspects of the child while at the same time offering a satisfying experience for parents.

1. Introduction

In juvenile psychiatric care it is common practice for clinicians to screen for psychopathology by relying on generic questionnaires. However, informants like parents, teachers, and the subjects themselves quite often appear to disagree on the pathological behaviour they perceive and how it should be clinically addressed ^{1, 2, 3}. Since theoretical frameworks on pathology are by default the foundation for these instruments, inevitably standardized items force the informant to narrow down a broad and unique view on the child. Likely to result in a mismatch in how problematic behaviour is perceived by those concerned.

Simply by asking the informant to describe the youngster in his own words, the informants’ unique and therefore valuable viewpoint could incorporated in clinical assessment of the child. According to the lexical hypothesis, first mentioned by Klages in 1926, it is assumed that “those individual differences that are most salient and socially relevant in people’s lives will eventually become encoded into their language; the more important such a difference, the more likely it is to become expressed as a single word” ⁴ (italics added by the authors). This premise resulted in the Five Factor Model of Personality ⁵. Extensive research confirmed the highly replicable and potentially universal status of the five dimensional personality factors Extraversion, Agreeableness, Conscientiousness, Emotional Stability and Openness to Experience ^{6, 7}.

Several scholars showed that parental free descriptors can be reliably converted in terms of this framework for personality. For over 80 % of parental descriptors originating from The Netherlands, Belgium and the USA could be reliably translated in terms of the Five Factor Model by a solid categorization system ^{8, 9, 10}.

Psychopathology can be understood in terms of the Five Factor Model as well ¹¹. Reference ¹² demonstrated that these personality dimensions are empirically and conceptually related to different types of psychopathology in child- and adulthood. Several types of Axis I pathology as described by the Diagnostic and Statistical Manual of Mental Disorders (4^th ed., text rev.) ¹³ were effectively differentiated by the Five Factor Model for a community sample and clinic-referred children and adolescents ¹⁴, ¹⁵. For example those suffering from anxiety disorders were characterized by significantly lower Extraversion, lower Agreeableness and lower Conscientiousness scale scores. Subjects suffering from major depression a main effect was found for lower Extraversion, higher Openness and lower Conscientiousness scale scores.

Moreover, by analyzing FFM descriptors at the highest level, three types of personality functioning (i.e. the undercontrolled, overcontrolled and resilient type could be differentiated ^{16, 17}. These types appeared to be associated to distinct DSM Axis I psychopathology ¹³ as well ¹⁸.

Thus, parental descriptions are effectively categorized within terms of the Five Factor Model of personality. This model itself meaningfully contributes to differentiating psychiatric disorders. To the best of our knowledge, the current study is the first to explore the clinical value of parental free descriptors for clinically referred children with regard to screening for psychopathology at time of admission.

The aims of the current study were (a) to assess whether parental free descriptors of their clinically referred children can be usefully categorized within the Five Factor Model of Personality and (b) to examine whether these categorizations are meaningfully related to the psychopathology of these subjects. It was expected that parental free descriptors could be satisfactorily coded within the Five Factor Model. Moreover, it was anticipated the personality dimensions on the Five Factor Model were differentially sensitive to the type of psychopathology.

2. Method

Descriptions were acquired for 3,436 children that were referred to our academic facility for paediatric psychiatry (Curium-LUMC) in the period between 1998 and 2013. Parental free descriptors were formulated by either one or both of the parents resulting in a maximum of ten descriptors per child. The total group of referred children consisted of 2,150 males ranging from 1 to 19 years of age (mean = 10.17; SD = 3.53) and 1,286 were female ranging from 2 to 19 years of age (mean = 12.0; SD = 3.95). At time of admission informed consent was acquired for anonymous data collection on demographic and clinical outcome variables of interest.

For over 15 years parents were required to complete a semi-structured questionnaire at time of admission. The questionnaire inquired about basic developmental, medical, school and family characteristics of the child. Furthermore, the informant was asked to describe the child in freely chosen adjectives, nouns, verbs or small sentences, to a maximum of ten. Psychopathology in terms of the Diagnostic and Statistical Manual of Mental Disorders (4^th ed., text rev) ²⁰, determined after careful multidisciplinary psychiatric and psychological (semi-structured) evaluation conform clinical diagnostic guidelines by academic staff, were retrieved from patient files. The wide array of Axis I classifications were condensed in terms of pathological phenotypes resulting in an ‘internalizing’, ‘externalizing’, autism spectrum’ and ‘remaining’ group.

Each single descriptor was coded within the Five Factor Model (FFM) framework according to system as presented by Kohnstamm et al. ⁹. First, the factor of the Five Factor Model corresponding to the descriptor was identified. Then it was determined whether the utterance resided at the positive or negative end of the dimensional factor. For example, according to the coding system 'Enthusiastic’ corresponds to the first factor (i.e. Extraversion) and was coded on the positive end (i.e. I+). Likewise, ‘Tends to shut oneself off’ was coded I-. The total coding system is presented in Table 1 (appendix). Inter-coder agreement was assessed by Cohen’s kappa coefficient (ҡ) for a randomly selected 1% of the in total 27,774 parental expressions.

A proportional coding profile per subject was created by dividing the acquired positive and negative ends for the five dimensional Factors by the total number of descriptors per child. Since the means and variances of proportion scores are dependent from one another, an angular transformation was performed, i.e. arcsin √p, where p is the proportion of descriptors ¹⁹.

Possible effects of gender and age on the angular transformed proportions of descriptors amongst the ten subcategories of the five dimensional factors of the Five Factor Model were inquired. Main effects for gender and age in relation to type of diagnosed psychopathology was assessed as well. The possible effect of age was clarified by subdividing the sample in distinct groups of successive three year intervals.

Mean angular transformed proportional distributions of descriptors in terms of the Five Factor Model were compared for the four types of psychopathology by MANOVA. When the mean angular transformed proportional distributions on the Five Factor Model were found to significantly differentiate the four types of psychopathology a consecutive discriminant analyses and ANOVA was executed.

All statistical operations and analyses were performed by IBM SPSS Statistics for Windows, Version 23.

3. Results

Parental free descriptions were obtained from 3,436 children, resulting in 27,774 descriptors, averaging 8.10 (SD = 2,52) descriptors per child. Number of descriptors produced by parents was independent of age (F = 1.24; p = .41) and gender (F = 1.14; p = .33).

Inter-coder agreement for 1% of the parental free descriptors resulted in a satisfactory Cohen's kappa coefficient (ҡ) of .82. Of the 27,774 descriptors, 92.2 % could be coded within the domains of the Big Five categories, leaving 6.5% of the descriptors to be coded in any of the remaining categories of the coding system. Another 1.2% of the descriptors were qualified as ambiguous and these descriptors were not coded. Table 1 shows characteristics of the different age groups and the arcsin transformed frequency means of the parental free descriptors.

MANOVA analyses showed gender of child and age to be significantly related to the distribution of arcsin transformed proportional means of parental descriptors on the high and low end of the Five Factors.

Boys were significantly more frequently described by descriptors representing the Extraversion+ subfactor (e.g. active, energetic) than girls (F = 11.22; p < .01). Moreover, girls were significantly more frequently described in terms of Extraversion- subfactor (e.g. quite, inhibited) than boys (F = 6.72; p < .01). The number of parental descriptors applying to the subfactors Conscientiousness+ (e.g. good concentration, responsible) and Openness+ (e.g. curious, interested in things) were significantly related to gender as well, resulting in an overrepresentation of girls (F = 38,67; p < .01) and boys (F = 5.47; p < .05) on these subfactors respectively. Table 2 presents an overview of the significant and non-significant subfactors of the Five Factor Model that were distinctive for gender in relation to arcsin transformed frequency means of parental descriptors.

Age was related to occurrence of descriptors representing the subfactors Extraversion+, Extraversion-, Conscientiousness+ and Conscientiousness- as is shown in Table 3. For example subcategory Extraversion+ appeared to become less relevant with age (F = 9.86; p < .01) whereas a reversed effect was found for the subfactor Extraversion- (F = 6.59; p < .01). Likewise with increasing age children were significantly more described in terms of Conscientiousness+ (F = 10.85; p < .01) and less in terms of Conscientiousness- (F = 4.75; p < .01).

DSM classifications ¹³ were available for 2250 subjects (i.e. 65.5% of the total sample) and were grouped according to type of pathology, i.e. externalizing (14.7%), internalizing (23.3%) and autistic disorders (21.8%) and a remaining category for a-specific, less frequent occurring classifications (14.5%). Chi-square test revealed a significant effect of gender on type of pathology (X² = 245.61, p < .01). Internalizing disorders were more frequently seen for girls (p < .05), whereas externalizing and autistic disorders were far more common among boys (p <. 05).

A significant effect of age on type of pathology was found as well (X2 = 100.89, p < .01). Within the classification category ‘autistic disorders’ initially the number of children increases (in the age from 3 to 6) to subside for older ages (see Table 5). The category ‘internalizing disorders’ shows a steady increase in number of children from the ages of 3 to 15 whereas the inverse applied for ‘externalizing disorders’ for subjects aged 6 to 18.

MANOVA analyses were performed for distinct groups of subjects categorized by age and gender due to their main effect on the outcome variables of interest. Arcsin transformed proportional means of parental descriptors on the high and low end of the Five Factors were chosen as dependent variable where the four distinct types of pathology were designated as the independent variable. The first half of Table 6 presents the subsamples wherein type of pathology was significantly associated to distributions of arcsin transformed proportional means on the high and low ends of the Five Factor Model. In the latter half of Table 6 consecutive outcomes for discriminant analyses (DA) are presented. By choosing type of pathology as dependent variable and distributions of the arcsin transformed descriptor on the Five Factor Model as independent variable, it was examined to what extent type of pathology could be reliably differentiated by parental descriptors.

For both sexes up to the age of 6 no significance for arcsin transformed proportional means on the high and low ends of the Five Factor Model in relation to type of pathology was found. However, for boys up to the age of 15 and girls into late adolescence, distinctive arcsin transformed descriptor means were significantly related to type of pathology (p <.01). Most frequently the high and low end of Extraversion, Conscientiousness, Openness and the low end of Emotional Stability were involved in differentiating types of pathology. For these subfactors discriminant analyses (DA) resulted in a variance accounted for by parental descriptors that steadily increased with age for both sexes to a maximum of 77.2% for boys aged 12 to 15 (Wilks’ Lambda = .78; p <.01). Consecutive quadric classification procedures resulted in maximum percentage of subjects correctly classified to their type of pathology by parental descriptors reached 52.9% for girls aged 15 years and over.

To understand how the extreme ends of the factors of the FFM are related to the four distinct types of pathology for boys and girls of consecutive ages additional ANOVA analyses were performed. Boys from 9 years and up and diagnosed within the realm of externalizing pathology were more frequently described by their parents in terms of Extraversion (+) and Conscientiousness (-) compared to boys suffering from internalizing, autism spectrum or a-specific pathology (p < .05). For girls aged between 6 and 9 and from 12 years up a likewise tendency was found for the subfactors Extraversion (+) and Conscientiousness (-) (p < .05). Moreover, for girls aged 12 and over it was found that girls diagnosed within the autism spectrum were more frequently described in terms of the subfactor Openness (-) compared girls suffering from other types of pathology (F (3)= 4.28; p < .05). Tables 7a and 7b provide an overview of arcsin transformed proportional means of parental descriptors for the extreme ends of the FFM factors and ANOVA results for boys and girls of successive ages respectively.

4. Discussion

The aim of the current study was to investigate whether free parental descriptions of clinically referred children accommodate any diagnostic value. As hypothesized, it was found that parental descriptors could be satisfactory coded within the Five Factor Model (FFM) for personality. Moreover, for both clinically referred boys and girls in the age of 6 till late adolescence, a predictive association was found between parental descriptors and type of psychopathology.

The association between the FFM and DSM classifications ¹³ has been widely established before. For example reference ²⁰ suggested that conceptualizing DSM-IV-TR ¹³ personality disorders in terms of the FFM provided a basis for integrating the abnormal and normal personality functioning across the life span.

Reference ¹⁶ and ¹⁷ identified three types of personality functioning associated to discerning DSM Axis I psychopathology ¹³. The undercontrolled type, indicative of problematic externalizing behavior, was associated with lower Agreeableness and Conscientiousness. Overcontrollers, indicative of internalizing problematic behavior, were significantly more frequently described by terms corresponding to the lower end of the factor Extraversion and the high end of the factor Conscientiousness ^{21, 22}. Moreover, both over- and undercontrollers had lower scores on emotional stability compared to the resilient type ¹⁷.

Within this study these recurring patterns of personality functioning could be largely replicated for the younger aged. For subjects suffering from externalizing disorders it was found these children were significantly more frequently described in terms corresponding to the high end of Extraversion (e.g. inexhaustible, spontaneous, brisk) compared to subjects diagnosed within the domain of internalizing disorders (e.g. withdrawn, detached, devious). A likewise tendency was found for the lower end of Agreeableness (e.g. impatient, nasty, selfish) and the high end of Conscientiousness (e.g. impulsive, unstoppable, unsettled).

In comparison to earlier publications on Dutch community samples an overrepresentation of parental descriptors on the low end of both the Agreeableness and Emotional Stability (e.g. hot-tempered, sensitive, agitated) factor was found. Parental descriptors coded at the low end of Agreeableness declined with increasing age of the child for the community sample ¹⁰. For the clinical sample descriptors coded at the low end of Agreeableness remained relatively high irrespective of the age. Moreover, with increasing age parents described their community based children more frequently in terms of the high end of the factor Emotional Stability and less in terms of the lower end ¹⁰. An inverse effect was found within the current clinical sample. Hence, it could be suggested that parental descriptors with relative high loadings on low agreeableness and low emotional stability represent an relevant indicator that the child is possibly at risk for developing problems of clinical significance.

With regard to the parental descriptors for the clinical sample, one of the factors to consider is that problematic behavior of the child affects the parent-child relationship and how parents perceive their child ²³. Possibly this explains the relative accent of parental descriptors coded in terms of the subfactors Agreeableness(-) and Emotional Stability(-) in the current sample. These subfactors might reflect the first thing that comes in to mind when asked to describe the youngster when it is not behaving up to the expectations of the parent. Moreover it could be relevant to analyze if and to what extend a hierarchy in parental descriptors can be found. For example starting with relative more positively formulated utterances might likely be related to more positive clinical perspectives.

Even though the current study offered an opportunity to analyze a substantive number of parental free descriptions for a large clinical sample, the sample is limited to one academic center for specialist psychiatric care for children and youth. Data were collected only at the time of admittance. Repeated data collection over time could reveal possible developmental influences and outcomes of treatment on how the characteristics of the child are perceived.

Our results support the importance of asking parents to describe their children referred for psychiatric evaluation in their own words. It was shown that these descriptors are meaningfully related to distinct types of most commonly classified types of clinical disorders. Such a for most parents pleasant and undemanding method ²⁴ for screening at time of admission might prove to be an effective and satisfying experience for parents when confronted with the realm of psychiatric care and offer worthwhile clinical impressions.

References