Introduction

Over the last decades, technological advances in medicine have resulted in increased survival rates and a greater prevalence of chronic physical health conditions in childhood [1, 2]. Additionally, environmental (e.g., air pollution) and behavioral changes (e.g., nutrition and physical activity patterns) have contributed to an increased incidence of conditions such as asthma and obesity [3, 4]. The worldwide prevalence of chronic conditions in childhood is approximately 10%, although accurate rates are difficult to estimate because epidemiological studies often focus on specific diagnoses [5]. Within a non-categorical approach, chronic conditions in children are mainly characterized by (1) limitations in age-appropriate function, activities or social roles; (2) reliance on compensatory mechanisms such as medication, special diet, equipment or personal assistance; and (3) use/need for medical, psychological or educational services beyond routine care [6]. Emphasizing the similarities in functional limitations and their psychosocial consequences (recurrent hospitalizations, limited participation in school/social activities, etc.) across health conditions, rather than the idiosyncratic characteristics of specific diagnoses, enables inter-diagnostic comparisons and has been highly recommended for psychosocial research [7,8,9].

Studies comparing the health-related quality of life [HrQoL] of pediatric patients and physically healthy peers have yielded inconsistent findings regarding the magnitude of impairments and the most affected domains in pediatric patients. For instance, Varni et al. [2] reported significantly lower HrQoL across children with several disease clusters in comparison to healthy children for all domains of functioning assessed by the Pediatric Quality of Life Inventory (PedsQL). Specifically, medium to large effect sizes for overall HrQoL, physical health, psychosocial health and school functioning were found. In contrast, Grootenhuis et al. [10] reported only a few diagnosis-specific differences between pediatric patients and controls in motor functioning, autonomy and social functioning, but not in physical functioning and emotions, as assessed by the Dutch TNO-AZL Children’s Quality of Life questionnaire (TACQoL). These heterogeneous results might be explained by participants’ sociodemographic and clinical characteristics but also by the methodological challenges in pediatric HrQoL assessment, namely the selection of developmentally appropriate questionnaires, the HrQoL dimensions being assessed and the use of child- or parent-reports [11, 12].

To assess HrQoL in children/adolescents, the WHO [13] states that instruments should be child-centered, rely on subjective self-report (supplemented by proxy judgments if the child is too young/disabled to provide a self-report), be age-appropriate and cross-culturally comparable, include generic and specific modules, and emphasize the health-enhancing aspects of HrQoL. These requirements were thoroughly met by the KIDSCREEN project [14], which was simultaneously conducted in 13 European countries and followed the complex methodological approach proposed by the WHOQOL Group for international instrument development [15,16,17,18]. This project developed reliable, valid, age-appropriate and cross-culturally comparable measures in long (KIDSCREEN-52) and short forms (KIDSCREEN-27; KIDSCREEN-10 Index) to assess subjective health and well-being in healthy and chronically ill children/adolescents aged 8–18 years [19]. The KIDSCREEN measures have been translated and validated for more than 40 European and non-European countries and have been used in numerous clinical studies and large health surveys (e.g., the European SPARCLE project [20, 21]; the Health Behavior in School-aged Children (HBSC) study [22, 23]).

The main objectives of this meta-analytic review were (1) to estimate the mean differences (MD) in HrQoL, as assessed by the KIDSCREEN questionnaires, between 8- to 18-years-old children/adolescents with chronic physical health conditions and community/healthy controls; and (2) to identify possible causes of heterogeneity by examining differences between subgroups of studies according to diagnoses, informants (i.e., patient- vs. parent-reports), instrument versions (long vs. short-forms), and methodological quality ratings (i.e., low vs. average vs. high quality). By summarizing the results of case–control studies, this study might contribute to improving the current understanding of the consequences of pediatric chronic conditions for patients’ overall adaptation and identifying specific areas of functioning that should be targeted by medical treatments and/or psychosocial interventions.

Methods

Search strategy

To identify published literature that applied the KIDSCREEN questionnaires, a systematic search was conducted in three electronic databases: PubMed (US National Library of Medicine), PsycINFO, and the EBSCOhost Psychology & Behavioral Sciences Collection. The keyword “kidscreen” was searched in all fields because the instruments used to assess HrQoL are frequently not named in the title/abstract. The final search was conducted on February 14–15, 2018. Additionally, the reference lists of all eligible articles were screened to identify other potentially relevant articles.

Eligibility criteria

Eligibility criteria were defined for types of studies, participants, comparisons and outcome measures according to the PICOS approach [24]. Only empirical quantitative studies that were published in peer-reviewed journals and written in English were considered.

For participants, eligibility criteria included samples of 8- to 18-years-old children/adolescents with a chronic physical health condition. Only children/adolescents between 8 and 18 years of age were included because younger children might be unable to provide reliable and valid answers due to insufficient language/reading proficiency [25]. Moreover, both the child- and parent-report versions of the KIDSCREEN questionnaires were developed for 8- to 18-year-olds, and there is no evidence of their psychometric performance in younger children. Chronic health conditions in childhood were defined according to the four criteria proposed by Mokkink and colleagues: “A disease or condition is considered to be a chronic condition in childhood if: (1) it occurs in children aged 0 up to 18 years; (2) the diagnosis is based on medical scientific knowledge and can be established using reproducible and valid methods or instruments according to professional standards; (3) it is not (yet) curable or, for mental health conditions, if it is highly resistant to treatment and (4) it has been present for longer than 3 months or it will, very probably, last longer than 3 months, or it has occurred three times or more during the past year and will probably reoccur” [26] (p. 1444). The operationalization of these criteria was based on the ICD-10 classification [27]. Mental and behavior disorders (codes F00-F99) were excluded because of their distinctive etiology and diagnostic and therapeutic procedures and because greater HrQoL impairments have been described for mental health disorders compared to physical conditions [28, 29].

Regarding comparisons/controls, studies were required to report direct comparisons between pediatric patients and community/healthy controls. Community controls were defined as age-matched children/adolescents without a specific diagnosis but whose health status with regard to other physical and/or mental health conditions was unknown; conversely, healthy controls were considered when the study had clearly assessed and excluded children with a history of any chronic health conditions, mental or behavior disorders and/or developmental delays. Comparative studies using population norms as the reference group were also excluded because potential confounders could not be adjusted, while control groups could be specifically selected to differ from the targeted group only by the variables of interest and to prevent unforeseen variability (e.g., variability related to the geographic area or socioeconomic conditions) [30]. Moreover, normative samples may include both healthy children and pediatric patients in a proportion that resembles the prevalence of chronic health conditions in the population, and thus, they may partially overlap with the target group.

Regarding outcome measures, we considered the mean difference between cases and controls on the KIDSCREEN domains completed either by self- or parent-report. The KIDSCREEN-52 measures ten domains: Physical well-being (five items), Psychological well-being (six items), Moods and emotions (seven items), Self-perception (five items), Autonomy (five items), Parent relations and home life (six items), Financial resources (three items), Peers and social support (six items), School environment (six items), and Social acceptance (bullying) (three items). The KIDSCREEN-27 is derived from the 52-item version and assesses five domains: Physical well-being (five items), Psychological well-being (seven items retrieved from the Psychological well-being, Moods and emotions, and Self-perception domains), Autonomy and parent relations (seven items representing the Autonomy, Parent relations and home life, and Financial resources domains), Peers and social support (four items) and School environment (four items). The KIDSCREEN-10 Index is derived from the 27-item version and provides an index HrQoL score. All versions are scored on a five-point Likert scale, with higher scores indicating better HrQoL; the provision of T-scores with M = 50 and SD = 10 for the reference population enables comparability across instrument versions [19].

Study selection

Study selection was conducted in two stages. Initially, the first author screened the titles and abstracts of all retrieved records to identify articles with relevant research objectives and methods and to decide whether to obtain the full-text. This first screening was over-inclusive, and only obviously irrelevant and duplicate records were excluded. Subsequently, the full-texts were independently assessed for eligibility by the first and second authors. Inter-rater agreement on study selection was calculated with Cohen’s kappa coefficient, considering k < 0.00 as poor, k ≤ 0.20 as slight, k ≤ 0.40 as fair, k ≤ 0.60 as moderate, k ≤ 0.80 as substantial and k > 0.81 as almost perfect agreement [31]. Disagreements were resolved by discussion until consensus was reached.

Data extraction

A data collection form was developed for this review based on the Data Extraction Template for Cochrane Reviews [32]. Data were extracted by the first author and verified for accuracy by the second author. For each study, we extracted information on publication (authors, year); methods (study design, procedures, inclusion/exclusion criteria); participants (number of eligible/included participants, non-response rate, country, age, sex, clinical characteristics); outcomes (informants and instrument version used, reliability/validity of the KIDSCREEN questionnaires); and quantitative results for the pediatric and control groups (sample size [n], mean [M] and standard deviation [SD]). When summary data were not available, other descriptive statistics (e.g., medians and interquartile ranges, standard errors [SE], confidence intervals [CIs]) were converted into the desirable format [33, 34]. For studies reporting data separately for subgroups of participants (e.g., children and adolescents; boys and girls), data were gathered into a single sample size that combined M and SD values [33]. When MD could not be computed from the available data, additional information was requested from the corresponding author by email.

Methodological quality assessment

The methodological quality of the studies included in the systematic review was independently assessed by the first and second authors using an adapted version of the Newcastle–Ottawa Quality (NOQ) assessment scale [35]. The NOQ was developed using a Delphi process and subsequently tested on systematic reviews and further refined. Thus, it can be considered a valid, repeatable and simple tool to assess the quality of non-randomized case–control studies included in systematic reviews. Studies were awarded up to 11 points based on the selection of participants (definition and representativeness of cases, selection and definition of controls; maximum of four points), comparability between cases and controls (maximum of two points) and HrQoL assessment (the measures’ reliability/validity, informants, procedures, and response rate; maximum of five points). Studies awarded 0–4 points were classified as low quality, those with 5–6 points were average quality, and those with 7–11 points were high quality. Inter-rater agreement was calculated with intraclass correlation coefficients (ICCs) for quality scores and Cohen’s kappa coefficient for quality category; disagreements were resolved by discussion until consensus was reached.

Quantitative synthesis of results

Meta-analyses of continuous data were performed with Review Manager Version 5.3 [36] using the inverse-variance random-effects method to incorporate between-studies heterogeneity (i.e., covariates that have been related to HrQoL scores and that are likely to differ across studies). Because this statistical method assumes that the outcomes have a normal distribution, skewness was inspected in each study by checking whether the mean was smaller than 2SD for both cases and controls [37]. When T-scores were not directly reported in individual studies, M ± SD scores were transformed to Rasch person parameters according to the KIDSCREEN syntax for each dimension. Although the computation of a total score was not foreseen for the KIDSCREEN long versions, weighted M(SD) total scores were computed for subgroup analyses after adjusting for Rasch person parameters to enable comparability across instrument versions.

Once the results of all the individual studies were standardized, the MD between cases and controls and its 95% CI were computed for each outcome measure as the summary statistic for the estimate of effects. Separate meta-analyses were performed for each domain and for the HrQoL Index. The ten domains of the KIDSCREEN-52 were clustered to match the five-domain structure of the KIDSCREEN-27, although each domain/questionnaire version was treated as a subgroup. Considering the minimally important difference in HrQoL of a half SD [38] and as recommended in the KIDSCREEN Manual [19], an MD ≥ 5 was considered clinically significant.

Additional analyses

The τ2 statistic (squared estimated SD of underlying effects across studies) was used to describe between-studies variance, and the I2 index described the percent of variability in effect estimates due to heterogeneity [39, 40]. When considerable heterogeneity was observed (I2 > 50%) [40], differences between subgroups of studies were examined with χ2-tests (subgroup analyses by diagnostic category to explore the diversity of participants; subgroup analyses comparing instrument versions, informants and quality ratings to investigate methodological diversity).

Results

Study selection

The literature search identified 528 non-duplicated articles, of which 66 were selected for full-text eligibility assessment (Fig. 1). Forty articles were excluded for the following reasons: (1) the sample included children younger than 8 or older than 18 years and no separate data were reported for the 8–18 years age-group (n = 8); (2) absence of a sample of children/adolescents with chronic conditions defined according to the criteria proposed by Mokkink and colleagues [7] and operationalized by the ICD-10 classification [33] or the inclusion of patients with mental/behavior disorders classified under codes F00-F99 (n = 18); (3) absence of community/healthy controls or use of normative data as a reference group (n = 8); (4) no data on pediatric HrQoL as assessed by the KIDSCREEN questionnaires (n = 1); (5) no comparative data on HrQoL between pediatric and control groups (n = 4); and (6) insufficient/inconsistent data (n = 1).

Fig. 1
figure 1

Flowchart for selection of studies

Full size image

To avoid multiple-publication bias, two papers reporting data from the same sample were also excluded (priority was given to the article that reported data from the largest sample), and three papers reporting data from subsamples of the same study were extracted together. Additional quantitative data were requested from the authors for four studies, and summary statistics were obtained for three studies; the remaining study was excluded due to insufficient quantitative data. Ultimately, 21 different studies reported in 23 papers were included in the meta-analysis. Inter-rater agreement for the selection of papers was moderate (k = 0.56, p < .001).

Study characteristics

The 21 studies examining children/adolescents’ HrQoL (Table 1) included a total of 4952 pediatric patients aged 8–18 years (M = 11.92, SD = 2.77; 52.7% male) and 28,678 controls (mean age = 11.59, SD = 2.37; 49.4% male) from 15 European and two non-European countries. According to the ICD-10 [27], chronic conditions were broadly classified into (C00-C97) malignant neoplasms [41]; (D50-D89) diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism [42]; (E00-E90) endocrine, nutritional and metabolic diseases [43,44,45,46,47,48,49]; (G00-G99) diseases of the nervous system [44, 50, 51]; (H60-H95) diseases of the ear and mastoid process [52]; (I00-I99) diseases of the circulatory system [53]; (J00-J99) diseases of the respiratory system [44, 54]; (K00-K93) diseases of the digestive system [55,56,57]; (M00-M99) diseases of the musculoskeletal system and connective tissue [58, 59]; (N00-N99) diseases of the genitourinary system [60]; (Q00-Q99) congenital malformations, deformations and chromosomal abnormalities [61, 62]; and (R00-R99) symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified [63].

Table 1 Study design and sample characteristics reported in the 21 studies included in the systematic review
Full size table

Most studies had a cross-sectional design (n = 18, 85.7%), and the three studies with longitudinal designs [54, 57, 59] only assessed HrQoL at follow-ups. For HrQoL assessment, 16 studies (76.2%) used patient-reports [41,42,43,44,45,46,47,48,49, 53,54,55,56,57,58,59, 62, 63], one (4.8%) used parent-reports [50], and four (19.0%) adopted a multi-informant approach [51, 52, 60, 61]. Thus, self-reported data were retrieved from 20 studies and included 4852 pediatric patients and 28,578 controls; parent-reported data were retrieved from four studies and included 511 parents of pediatric patients and 433 parents of community/healthy children (parent data from one study [59] was excluded due to missing summary statistics for controls). The proxy-reports were mainly completed by the mothers (78.5%). For the patient-reports, the KIDSCREEN-52 was most commonly used (n = 11, 55%) [41,42,43, 49, 52, 56,57,58, 60,61,62], while four studies (19.0%) used the KIDSCREEN-27 [47, 48, 51, 53] and five (23.8%) used the KIDSCREEN-10 Index [44,45,46, 54, 55, 59, 63]. Regarding parent-reports, two studies (40.0%) used the KIDSCREEN-52 [52, 60], and three studies (60.0%) used the KIDSCREEN-27 [50, 51, 61].

Methodological quality

The percentage of studies awarded 0, 1 or 2 points according to the adapted NOQ assessment scale [35] is summarized in Fig. 2, and a detailed description of quality assessment is provided in Online Resource 1. Most studies (n = 15, 71.4%) were assessed as average quality; only two studies were assessed as low quality [53, 55], and four studies were assessed as high quality [50, 56, 57, 61]. The main reasons for lower quality scores were as follows: no description of controls’ health state (81.0%); no examination of the reliability/validity of the KIDSCREEN questionnaires in the study sample (81.0%); and response rates that were below 80%, not reported or differed significantly between respondents and non-respondents (76.2%). Inter-rater agreement was excellent for quality scores (ICC = .92, 95% CI .81/.97) and substantial for quality categories (k = 0.79, p < .001).

Fig. 2
figure 2

Quality assessment of the 21 studies included in the systematic review. a Case definition: 1 point for diagnosis of a chronic health condition established by a physician, based on medical records or on physiological indicators; b representativeness of cases:1 point for selection of a random sample of patients, all eligible patients in a defined healthcare/educational institution or consecutive series of patients over a defined period of time; c selection of controls: 1 point for selection of controls within the same community/geographic area as cases; d definition of controls: 1 point for selection of healthy controls with no history of chronic conditions; e comparability: 1 point if the study controlled for disease severity/clinical characteristics, and 1 additional point if the study controlled for sociodemographic confounders; f measures: 1 point for the ascertainment of good reliability/validity of the measures in the study’s sample; g informants: 1 point for the use of patient- or parent-reported measures or 2 points for inclusion of both informants; h procedures: 1 point for the use of the same procedures for assessing cases and controls; i Response rate: 1 point for response rate that was similar for cases and controls, higher than 80%, or non-significant differences between respondents and non-respondents

Full size image

Quantitative synthesis of results

Pooled estimates for HrQoL domains confirmed lower HrQoL in pediatric patients compared to healthy/community peers for Physical well-being (MD = − 4.84, 95% CI − 6.44/− 3.24 for self-reports, and MD = − 6.86, 95% CI − 10.42/− 3.29 for parent-reports; Fig. 3); self-reported Psychological well-being (MD = − 0.61, 95% CI − 1.16/− 0.06; Fig. 4); and Peers and social support (MD = − 1.29, 95% CI − 2.25/− 0.34 for self-reports, and MD = − 3.90, 95% CI − 5.28/− 2.52 for parent-reports; Fig. 5). Considering the threshold of a half SD [19, 38], the MD for parent-reported Physical well-being can be considered clinically significant, while the MD for self-reported Physical well-being, self-reported Psychological well-being and self- and parent-reported Peers and social support were statistically, but not clinically, significant. No significant differences were observed for parent-reported Psychological well-being, Autonomy and parent relation (Fig. 6), School environment (Fig. 7) and Social acceptance/bullying (Fig. 8), or for the HrQoL Index provided by the KIDSCREEN-10 (Fig. 9). Significant between-studies heterogeneity (I2 ≥ 50%) was found for all HrQoL domains, except self- and parent-reported Autonomy and parent relation.

Fig. 3
figure 3

Forest plots for meta-analysis of differences for (1) Physical well-being, (1.1) self-reported and (1.2) parent-reported, between pediatric patients and community/healthy controls

Full size image
Fig. 4
figure 4

Forest plots for meta-analysis of differences for (2) Psychological well-being, (2.1) self-reported and (2.2) parent-reported, between pediatric patients and community/healthy controls

Full size image
Fig. 5
figure 5

Forest plots for meta-analysis of differences for (4) Peers & social support, (4.1) self-reported and (4.2) parent-reported, between pediatric patients and community/healthy controls

Full size image
Fig. 6
figure 6

Forest plots for meta-analysis of differences for (3) Autonomy & parent relation, (3.1) self-reported and (3.2) parent-reported, between pediatric patients and community/healthy controls

Full size image
Fig. 7
figure 7

Forest plots for meta-analysis of differences for (5) School environment, (5.1) self-reported and (5.2) parent-reported, between pediatric patients and community/healthy controls

Full size image
Fig. 8
figure 8

Forest plots for meta-analysis of differences for (6) Social acceptance/bullying, (6.1) self-reported and (6.2) parent-reported, between pediatric patients and community/healthy controls

Full size image
Fig. 9
figure 9

Forest plots for meta-analysis of differences for (7) HrQoL Index, (7.1) self-reported, between pediatric patients and community/healthy controls

Full size image

Additional analyses

The high heterogeneity across studies can be explained by the patients’ diagnostic category, the KIDSCREEN version used and the informants (Table 2). Specifically, children/adolescents with malignant neoplasms; endocrine, nutritional or metabolic diseases; diseases of the nervous system; congenital anomalies; or chronic pain had significant HrQoL impairments, while patients with diseases of the respiratory, digestive, genitourinary, musculoskeletal or circulatory systems presented HrQoL similar to that of healthy peers. Further, studies that relied on the long versions of the KIDSCREEN were more likely to detect significant differences between pediatric patients and controls (with larger effect sizes and lower heterogeneity for studies using the KIDSCREEN-27) compared to studies using the KIDSCREEN-10 Index. Moreover, greater differences between cases and controls and lower heterogeneity were found for studies using parent-reports compared to those using patient-reported HrQoL. No significant differences were observed between subgroups of studies according to methodological quality ratings. The forest plots for subgroup analyses of HrQoL total scores by ICD-10 classification, instrument version, informants and methodological quality ratings are presented in Online Resources 2–5.

Table 2 Subgroup analysis for children/adolescents’ HrQoL total scores
Full size table

Discussion

This is the first meta-analytic review to gather comparative studies of HrQoL in children/adolescents with physical chronic conditions and community/healthy controls. Our main contribution was the ascertainment of the magnitude of HrQoL impairments and the identification of physical and social domains as the most affected in pediatric patients.

Regarding Physical well-being, a statistical, but not clinically, significant difference was found for self-reports, while for parent-reports, the MD between cases and controls reached clinical significance. A previous meta-analysis of HrQoL differences between healthy controls and children with specific diagnoses, such as central nervous system tumor survivors [64] and those with psoriasis [65], obesity [66], or asthma [67], reported significant HrQoL impairments, but most of these studies only considered global HrQoL or broad dimensions of physical and psychosocial functioning. The mean differences in the Peers and social support domain were also statistically significant for both self- and parent-reports. This result may reflect the inclusion of several studies with children/adolescents with obesity, which has been associated with great HrQoL impairments due to victimization by peers, weight-related teasing and social marginalization [68].

In addition, a statistically (but not clinically) significant difference was found in self-reported Psychological well-being. A previous overview of evidence-based research showed that children/adolescents with chronic conditions have a greater risk of psychosocial impairments, but few present clinically significant psychological symptoms [69]. A statistically significant difference was also found in the meta-analysis of studies using the parent-reported KIDSCREEN-27, which may reflect the inclusion of two of the three studies that included patients diagnosed with diseases of the nervous system [50, 51], which have been associated with decreased HrQoL and more psychological problems [70, 71]. The absence of significant differences between cases and controls in the Autonomy and parent relation, School environment and Social acceptance/bullying domains, together with considerable within-studies variability (i.e., a broad CI in most individual studies), suggest that other intrapersonal characteristics (e.g., age, sex, adherence to treatments) among patients may be associated with different levels of HrQoL impairments.

The subgroup analyses showed that diagnostic category, informants and instrument versions are clinical/methodological features that might contribute to explaining heterogeneity across studies. Regarding diagnostic categories, HrQoL impairments were greater for children with malignant neoplasms; endocrine, nutritional and metabolic diseases (e.g., obesity); diseases of the nervous system (epilepsy, cerebral palsy or muscular dystrophy); congenital malformations (congenital heart diseases or cleft lip/palate); and chronic pain. Analogous results have been found in previous studies comparing different chronic conditions [2, 44]. From a non-categorical perspective, impairments involving the nervous central system, the external visibility of the condition and the pain/discomfort inherent to the condition and/or treatments can be identified as potential risk factors with a detrimental impact on patients’ HrQoL and psychological adaptation. However, disease severity and treatment were not considered in our analyses; for instance, one study [59] only included children who had undergone surgery and attended a physiotherapy program more than 1 year postoperatively, which may explain why they reported enhanced HrQoL, even when their Legg-Calvé-Perthes disease had a severe impact on physical functioning.

Moreover, studies using the longer versions of KIDSCREEN were more likely to detect significant differences between cases and controls (with greater effect sizes for the KIDSCREEN-27), while the meta-analysis of studies using the KIDSCREEN-10 Index showed no significant differences (although the pooled pediatric sample included children/adolescents diagnosed with endocrine, nutritional and metabolic diseases and diseases of the nervous system, which have been consistently associated with significant HrQoL impairments [2, 44, 72]). These results suggest that, independent of diagnosis, profile measures accounting for the multidimensionality of the HrQoL construct might be more sensitive for detecting the most impaired domains of children’s HrQoL [25]. Finally, a major trend in the use of self-reports was observed, but studies that relied on parent-reports were more likely to identify significant HrQoL differences between cases and controls. This result is aligned with previous findings on parent–child agreement, which have consistently shown that parents are more prone to underrate the HrQoL of their children with chronic conditions [73,74,75,76].

Limitations

Some limitations should be considered at the levels of both the individual studies and the review. First, our literature search was restricted to papers published in peer-reviewed journals and written in English, and the “grey literature” (i.e., literature that has not been formally published in books or journal articles [33]) was not considered, which may have introduced publication bias. Second, the first screening of titles and abstracts was performed by only one author, although it was over-inclusive to decrease the chance of rejecting relevant records. In addition, the moderate inter-rater agreement for the selection of papers in the second screening may reflect the complexity of the eligibility criteria and should be considered a limitation. Third, potential publication bias was not verified in each meta-analysis because tests for detecting funnel plot asymmetry have low power when fewer than 10 studies are included and are prone to lead to false-positive test results in the presence of substantial between-study heterogeneity [33, 77, 78]. Fourth, although participant and methodological diversity were addressed using random-effects models and subgroup analyses, other potential factors explaining between-study heterogeneity (e.g., the health state of the controls, disease severity, treatment status) could not be examined across studies. Finally, the small number of studies precluded subgroup analyses for each HrQoL dimension and forced the use of total scores, even though the computation of total scores was not foreseen for the long versions of the KIDSCREEN. Although the focus on a single instrument ensured that the operationalization of the HrQoL construct was comparable across studies, the high correlations among instrument versions do not imply a similar ability to detect important changes/differences and might have introduced heterogeneity into the study results [79].

Conclusions and practical implications

This meta-analytic review identified some shortcomings in the existing literature and allowed the establishment of guidelines for future research and clinical practice. First, our review identified few studies meeting the eligibility criteria, although the comparison between healthy and chronically ill populations is one of the major applications of generic HrQoL assessment [25, 80]. Therefore, further comparative studies are still needed in the pediatric context [11], particularly studies investigating different diagnoses (e.g., rare diseases) and examining HrQoL changes in the course of child development. Second, our results contribute to a better understanding of the differential impact of specific clinical features (e.g., neurological compromise, external visibility, pain/discomfort) on physical, psychological, social, family and school functioning, which is essential for identifying and planning comprehensive interventions in pediatric healthcare. HrQoL assessment and psychosocial intervention should be routinely implemented in general pediatric healthcare, prioritizing oncology, endocrinology and neurology services [72]. Third, profile assessments are particularly relevant in guiding targeted interventions for children with chronic conditions. The KIDSCREEN-27 provides a health profile based on 27 items clustered into the most accepted HrQoL dimensions, thus preventing response burden [25]. Finally, a multi-informant approach is highly recommended, particularly in the clinical setting, to better understand patients’ and parents’ participation in clinical decision-making and disease management.