Health-related quality of life among children with mental health problems: a population-based approach

‘Health-related quality of life’ (HRQOL) can be described as a subjective, multidimensional and dynamic construct that comprises physical, psychological and social functioning and that is, among other things, influenced by the health condition of the particular person [1]. To date, more studies have been conducted evaluating the HRQOL of individuals with physical than with mental health conditions[2]. Additionally, more HRQOL studies have targeted adults than children[3].

In a recent review article [4], children with various mental health conditions were found to exhibit compromised HRQOL relative to healthy peers. The largest effect sizes (ES) have been identified for the total HRQOL score and various psychosocial scales, whereas the ES for physical scales generally have been smaller. Parent ratings of a child’s HROQL often were most affected in the psychosocial subscale most closely related to the particular mental health condition. One explanation for this observation is so-called item overlap, which is defined as content similarities between HRQOL items and the conceptualization of a particular mental health condition. When such overlap exists, HRQOL and mental health problems are inevitably related [5]. To date, most studies have failed to control for item overlap [4]. However, Sawyer et al. [6] analyzed this issue and observed similar relationships between mental disorders and HRQOL, even after controlling for item overlap. Nevertheless, the effects of this mechanism should be evaluated further [2, 5].

Besides non-consideration of item overlap, the following limitations of previous research were extracted in the above-mentioned review article [4]: first, many authors only used proxy ratings, even though the subjectivity of HRQOL actually demands self-rating [7] whenever the child’s cognitive abilities and particular mental health condition permit [5, 8]. Second, many studies failed to capture whether or not the child was receiving medication for the specific mental health condition, even though it is possible that such treatment affects HRQOL. Furthermore, other potentially-influential variables (e.g., severity of the health condition) were often not included in statistical analyses. Third, existing studies are limited by primarily relying on clinical samples. The utilization of such samples may lead to biased results, because mentally ill children who concurrently have compromised HRQOL may be more frequently referred to or treated in clinics than children with mental health problems lacking HRQOL impairment [6, 9]. Studies that use population-based samples should be used to verify any results obtained from clinical samples [4]. The few existing studies that have utilized population-based approaches have identified reduced HRQOL (especially within psychosocial (sub)scales) among children with mental health problems [6, 10]. Finally, only a limited number of studies have compared children with mental and physical health conditions with regard to their HRQOL (e.g., [6, 10‐14]). Even though the results of these studies are not completely consistent, Danckaerts et al. [8] summarized various studies and concluded that the overall HRQOL score is reduced to the same extent in children with mental and physical health conditions, whereas psychosocial HRQOL domains are more compromised in children with mental, and physical HRQOL domains in children with physical health problems.

Addressing existing research gaps and the aforementioned limitations, the aims of the present study were two-fold: First, we aimed to assess the influence of mental health problems on proxy- and self-rated HRQOL scores in a population-based sample. For comparison’s sake, we also intended to assess the influence of physical health problems on HRQOL. Hereby, other potentially-influencing variables (severity of symptoms, status of medication use) were taken into account. Second, we aimed to examine whether item overlap effects the association between mental health problems and HRQOL.

The purpose of this article was to evaluate the HRQOL of children with mental health problems in-depth. Hereby, some of the research gaps and limitations of existing studies were considered: most important was that 1) a large population-based sample was used; 2) the issue of item overlap was addressed; and 3) in addition to the CSHCN with mental health problems and healthy controls, children with physical health problems were included.

The descriptive statistics of our study are partially consistent with the pattern that was proposed in a review article [8]: That is, the lowest means for psychosocial HRQOL domains were identified among children with mental health problems, whereas the lowest means for physical HRQOL domains were apparent among children with physical health problems. However, when one only considers the descriptive statistics, it seems that our results contradict the proposition that overall HRQOL is equally compromised in children with mental and physical health problems [8]. This being said, the finding of lower total HRQOL scores among CSHCN with mental versus physical health problems was probably due to the severity of the main health problem being greater for the former group. Accordingly, multiple regression analyses indicated that reduced HRQOL was primarily associated with increased severity of the main health problem. This effect was also apparent for total HRQOL.

Despite the clear importance of the severity of the main health problems in predicting HRQOL, mental and physical health problems also contributed to the prediction of some HRQOL scores. For instance, the presence of mental health constraints was significantly associated with a poorer ‘school environment’ when self reports were considered, and tended towards poorer ‘school environment’ when parent reports were used. The effect on this subscale may be due to the composition of our sample, as the most frequently-reported mental health problems were attention deficits, learning difficulties and conduct problems, thus problems that share their strong impact upon school context. It seems that this impact was not entirely attributable to content similarities between the conceptualization of mental health problems and HRQOL items, because our results remained largely unchanged even when we controlled for item overlap (comparable to findings by Sawyer et al. [6]). Other psychosocial subscales (e.g., ‘psychological well-being’) were not reduced in CSHCN with mental health problems in multiple regression models. This might be because this subscale was less directly affected by the most frequently-represented mental health problems in our sample.

That the presence of a mental health problem was associated with enhanced parent-reported ‘physical well-being’, whereas the presence of a physical health problem predicted higher parent-reported ‘psychological well-being’ and ‘school environment’ may be due to compensatory and overly-positive ratings by parents in those HRQOL domains that are not directly related to the particular health constraint [23].

The status of medication use had no influence on HRQOL. This might be due to our inclusion of a variety of different health constraints in our study that were not treated uniformly. Hence, it is possible that the positive effects of some drugs on HRQOL were overlaid by the negative effects of others (e.g., drugs that have severe side effects).

The two included demographic characteristics also exerted an influence on some HRQOL scores in the multiple linear regression models. First and comparable with other studies (e.g., [24]), female gender was associated with reduced ‘physical well-being’ and ‘psychological well-being’, as well as with increased ‘social support & peers’ and ‘school environment’. Second, non-Swiss nationality had a significant negative impact upon self- and parent-reported ‘physical well-being’ as well as on self-reported ‘autonomy & parent relation’. This result (especially the former) can be explained by the compromised health status of the non-Swiss population (see, for instance [25]) that leads to reduced HRQOL. However, it also may be due to a more negative assessment of the same health status by non-Swiss relative to Swiss subjects.

Despite the strengths of our study, some limitations must be considered. Similar to other population-based studies (e.g., [10]), the most important limitation was that precise diagnostic information about the main health problem was missing (i.e., CSHCN were classified primarily based upon the parent-reported main health problem). However, population-based studies generally uncover similar results as those in which children are diagnosed through a specialist [10]. Furthermore the utilization of clinical samples may lead to biased results [6, 9]. Hence, population-based studies seem to be a worthwhile supplementation to studies with clinical samples [4]. It can be argued that the CSHCN who were not classifiable based upon their main health problem were classified according to answers to the fifth item of the CSHCN Screener. We cannot rule out that the need for treatment and/or counseling for emotional, developmental or behavioral problems was the consequence of a significant physical health problem. However, additional analyses indicated that CSHCN with mental health problems who were classified according to the parent-reported main health problem were similar, in terms of their HROQL, as CSHCN with mental health problems classified according to the fifth item of the CSHCN Screener. Hence, we assume that applying this second method was valid.

A second study limitation was that we had no information about co-morbid conditions, even though it is possible that these conditions might contribute to the prediction of HROQL.

Third, most of the independent variables (e.g. the health status) were based upon answers provided by the parents. Hence, that the explained variances in multiple linear regression models were larger when HRQOL was rated by parents than by children might be explained through shared-rater method variance.

A fourth limitation is that we used standardized HRQOL scores instead of the Rasch-scaled HRQOL scores proposed by the KIDSCREEN-27 developers. We decided against using Rasch-scaled HRQOL scores because they are difficult to interpret. To deal with this problem, the developers proposed recoding these values into T-scores, a transformation that is based upon norm values. However, only the German-speaking part of Switzerland was included to obtain these norm data, whereas we also drew subjects from the French- and Italian-speaking regions of Switzerland.

The present paper demonstrates the significant contribution of the severity of a child’s main health problem to predicting HRQOL, a contribution that has implications for the interpretation of the results of other studies. That is, HRQOL differences between children with mental and physical health problems could be more or less pronounced when the severity of health problems is taken into account. Besides the severity of the main health problem, additional variables were important in predicting HRQOL. One important finding was that the presence of a mental health problem predicted a poorer ‘school environment’, a HROQL domain that was most closely related to the most-frequently represented mental health problems (attention deficits, learning difficulties, and conduct problems). This finding seems not to be solely attributable to item overlap between mental health problems and HRQOL items, because our results remained much the same when we controlled for item overlap. Hence, HRQOL assessments are useful when attempting to gather wide-ranging information about CSHCN. This information, which goes beyond the clinical symptoms of mental health problems, can be used in many ways: 1) to expand knowledge about the impact of particular mental health constraints; 2) to improve clinical practices (e.g., by considering compromised HRQOL domains in therapy); and 3) to adapt public health practices (e.g., by considering a broad range of different health conditions and comparing their HRQOL constraints, so as to adequately plan services) [2].