Child–parent agreement on health-related quality of life in children with newly diagnosed chronic health conditions: a longitudinal study

Abstract This study compared health-related quality of life scores between children newly diagnosed with chronic health conditions and their parents to identify potential domain discrepancies and quantified the magnitude of discrepancies during a six-month follow-up. Children aged ≥11 years and their parents completed the KIDSCREEN-27. Informant agreement was quantified using the intraclass correlation coefficient (ICC) and Bland and Altman’s limits of agreement. The proportion of child–parent dyads with a minimal clinically important difference (MCID) was computed. Children reported higher KIDSCREEN-27 scores for all subscales compared to parents. At baseline and at six months, child–parent agreement was generally low to moderate (ICC = .25–.67) and the proportion of MCID was high (36–79%). Over time, child–parent agreement improved on most domains. These exploratory findings suggest that the agreement between children newly diagnosed with chronic health conditions and their parents is low to moderate and meaningful discrepancies are common, though improve over time.


Introduction
Over the past 50 years, epidemiological studies have shown that the incidence of chronic health conditions in children is increasing (Perrin, Bloom, & Gortmaker, 2007;Public Health Agency of Canada, 2009). The peri-diagnostic period is often a distressing time for parents and children and represents the starting point for a complex relationship between a chronically ill child and their parents. Furthermore during this period, clinicians, children and parents are deciding optimal treatment and management regimens (Lowes & Lyne, 2000;Valentine & Lowes, 2008). Often, clinicians rely on reports of child health-related quality of life (HRQL) to facilitate patient-clinician communication, improve child/parent satisfaction, and to aid in clinical decision-making (Varni, Limbers, & Burwinkle, 2007).
Defined as the 'subjective and objective impact of dysfunction associated with an illness or injury, medical treatment, and health care policy' (Spieth & Harris, 1996), HRQL is a multidimensional patientreported outcome that clinicians and researchers have argued is a fundamental outcome for individuals, particularly children, with chronic health conditions when evaluating health care (Ingerski et al., 2010;Kaplan, 2001). Recognizing its importance, regulatory agencies have now mandated that assessments of ARTICLE HISTORY received 9 January 2017 accepted 16 february 2017 KEYWORDS adolescent; chronic disease; health-related quality of life; informant agreement; mental health; parent-proxy reports HRQL be identified as an outcome in clinical trials, health services research, and programme evaluations (FDA, 2006), and has become the impetus for the development of measures of HRQL that are valid and reliable .
Because reports of HRQL enrich understanding of disease processes and health outcomes, as well as influence health service use (Parsons, Fairclough, Wang, & Hinds, 2012;Upton, Lawford, & Eiser, 2008;Varni et al., 2007), potential discrepancies between child self-and parent-proxy reports of HRQL measures is an important issue, especially for children with chronic health conditions. As such, measures of HRQL that include both, a parent and child self-report version such as the KIDSCREEN, are highly valued. Developed by a consortium of researchers in Europe, the KIDSCREEN is a generic measure of HRQL that includes child and parent-proxy 52, 27 and 10-item versions (Ravens-Sieberer et al., 2014). Considerable effort has been devoted to ensuring the validity and reliability of the KIDSCREEN in a variety of populations (e.g. children with and without chronic health conditions) and findings with respect to its convergent/discriminant validity, internal consistency and test-retest reliability, and factor structure, appear robust , 2014Robitail et al., 2007).
While there is evidence suggesting that child-parent agreement on the KIDSCREEN for children with chronic health conditions is typically good (Ravens-Sieberer et al., 2014), gaps remain. First, much of the research addressing child-parent agreement on the KIDSCREEN has been cross-sectional and has not assessed the changes in child-parent agreement over time. Second, studies have not evaluated child-parent agreement on the KIDSCREEN in children with newly diagnosed chronic health conditions. Given the significance of the peri-diagnostic period in determining the outcomes associated with a chronic health condition (Valentine & Lowes, 2008), uncertainty with prognosis, and relevance of childparent agreement on HRQL measures for clinical decision-making, the objectives of this exploratory study were twofold: (1) assess informant agreement between children with newly diagnosed chronic health conditions and their parents to identify potential domain discrepancies in the KIDSCREEN at two separate time periods (i.e. at diagnosis and six months later), and (2) quantify the magnitude of such informant discrepancies.

Sample
Data were collected as part of a multisite prospective study of psychiatric and psychosocial outcomes in children newly diagnosed by a physician or nurse practitioner with a chronic health condition. Clinic nurses at two children's tertiary care centres in Ontario, Canada approached families about the study and provided them a letter describing the study which outlined what participation would entail. Inclusion criteria were a new case of asthma, diabetes, epilepsy, food allergy or juvenile arthritis in a child aged 6-16 years in whom diagnosis had been confirmed no longer than six months prior to recruitment, and a parent sufficiently fluent in English who was the child's primary caregiver for at least six months and would continue to be for the duration of the study. These analyses are restricted to child-parent dyads in which the child was aged ≥11 years (n = 28; 56% of the study sample), as this was the earliest age at which child self-reports were obtained. The study protocol received ethical approval from all relevant research ethics boards.

Data collection
Families interested in participating in the study consented for clinic nurses to send their contact information to study investigators who then followed-up with families by telephone to confirm eligibility, obtain oral consent from parents and children, and arrange for a convenient time to conduct a telephone interview to assess child psychiatric health. Parents and children also completed two mail questionnaires; one at baseline and one six months later, when a second telephone interview to assess psychiatric health was conducted. Parents and children also consented to have clinicians provide clinical information at the same measurement occasions.

Measures
Child HRQL was measured using the KIDSCREEN-27; a 27-item child self-and parent-proxy reported generic measure . It measures HRQL across five domains: Physical Wellbeing (five items; examines physical activity and energy), Psychological Well-being (seven items; examines emotional balance and life satisfaction), Autonomy & Parent Relations (seven items; examines family dynamics and age-appropriate freedoms), Social Support & Peers (four items; examines nature of peer relationships) and School Environment (four items; examines perception of cognition, learning, and feelings about school). Item responses are based on a five-point Likert scale and T scores for each domain are computed with a mean of 50 and standard deviation of 10, whereby higher scores indicate better HRQL. It has been found to be valid and reliable in children aged 8-18 years with and without chronic health conditions, demonstrating adequate psychometric properties Robitail et al., 2007). Internal consistency reliabilities for each domain from this study were good for both child (α > .75) and parent reports (α > .70).
Sociodemographic characteristics for children and parents were collected which included: child and parent age and sex, child immigrant status, parent relationship to child, parent marital status, parent education attainment and total household annual income. Questions were adopted from epidemiological studies conducted by Statistics Canada and had previously undergone appropriate validity and reliability testing (Statistics Canada, 2015).

Analysis
T scores, based on child and parent reports, were compared at baseline and at six months using paired t tests. Standardized effects (d) were also computed (Cohen, 1988). Agreement between child and parent reports for each domain of the KIDSCREEN-27 was calculated using the intraclass correlation coefficient (ICC) (Streiner & Norman, 2008), as well as the limits of agreement method described by Bland and Altman (Bland & Altman, 1986). To examine the extent to which child and parent reports resulted in substantial discrepancies, the proportion of child-parent dyads with KIDSCREEN-27 domain score differences that met or exceeded the minimal clinically important difference (MCID) were computed. Given that no MCID currently exists for the KIDSCREEN-27, we used half a standard deviation for each domain to approximate the MCID (Norman, Sloan, & Wyrwich, 2003). To assess agreement over time, child and parent agreement was classified according to the presence of the MCID at baseline and at six months. Based on the MCID, four classifications were created (Agree-Agree, Disagree-Disagree, Agree-Disagree and Disagree-Agree). Analyses were conducted using SAS 9.2 (SAS Institute Inc.). Table 1 shows the sample characteristics of the 28 child-parent dyads included in the study. The mean age of children was 13.1 (standard deviation 2.6) years and 54% were male. Primary diagnoses were equally spread across the five chronic health conditions and the mean illness duration was 1.6 (1.6) months. Parents had a mean age of 45.6 (5.4) years and 93% were female. Most parents were in partnered relationships (79%), had completed post-secondary education (75%), and reported annual household incomes of ≥$75,000 (75%). Baseline agreement, as calculated using the ICC was typically low across domains, ranging from .25 for the School Environment to .62 for the Physical Well-being domains. At six months, the ICC reached a low of .40 for the Autonomy & Parent Relations and a high of .67 for the Psychological Well-being domains. As compared to baseline, the ICCs for all domains except Physical Well-being increased. The greatest change was observed for the Psychological Well-being domain which increased from .32 at baseline to .67 at six months (Table 2). Bland-Altman plots showing the deviation from unbiased agreement between child and parent reports for each domain at baseline and at six months are illustrated in Figures 1 and 2. The proportion of child-parent dyads with differences that met or exceeded the cut-point for MCID was large. At baseline, 36% had a MCID in the Physical Well-being domain and 79% had a MCID in the Social Support & Peers domain. At six months, 39% had a MCID in the Psychological Well-being domain    Table 3. changes in child-parent agreement over time.

Comparison of child versus parent report at baseline and at six months
notes: child and parent agreement was classified as to whether the dyad had a McId or not at baseline and six months. for example, the 'agree-disagree' classification was based on children and parents who had less than the McId for the KIdscreen-27 at baseline, but had a McId or greater at six months.

Summary of findings
In this study of children aged 11-16 years newly diagnosed with chronic health conditions, findings showed that the agreement between child and parent reports of HRQL as measured by the KIDSCREEN-27, was generally low to moderate, and the proportion of score differences between informants that met or exceeded the MCID was high. Children in this study reported higher KIDSCREEN-27 scores across all domains at baseline and at six months compared to their parents -a finding consistent with a large body of research in the HRQL of children with chronic health conditions (Sattoe, van Staa, & Moll, 2012;Upton et al., 2008). Baseline differences between informants were largest, with effect sizes of medium, for the domains of Social Support & Peers, Autonomy & Parent Relations and Psychological Well-being. Physical Wellbeing was the exception which showed good agreement. This is similar to previous reports whereby a higher level of agreement was reported in more observable domains (i.e. Physical well-being) and a lower level of agreement was reported in less observable domains (i.e. Social Support & Peers, Autonomy & Parent Relations and Psychological Well-being) (Brunner et al., 2004;Eiser & Morse, 2001;Rajmil, López, López-Aguilà, & Alonso, 2013). However, whereas these previous reports indicated substantial agreement between children and parents, agreement between informants in the current study was considered poor according to established guidelines (Streiner & Norman, 2008). The discrepancies in findings between the current study and previous reports are likely attributable to differences in study methodology -whereas previous studies randomly sampled children from the general population in European countries, this study used consecutive sampling within tertiary care settings in Canada to sample children with chronic health conditions. Additional clinical studies are needed to make more definitive the extent to which differences in study findings are sample-dependent. Findings indicated that the child-parent agreement, with the exception of Physical Well-being, improved over time. Research suggests that children and parents differ with regard to their concerns and valuation of HRQL domains (De Civita et al., 2005). Possibly, at diagnosis, parents and children are both focused on physical health concerns resulting in high agreement on the Physical Well-being domain. However, as time progresses, the psychosocial consequences of having a chronic health condition become more apparent resulting in a higher child-parent agreement on these domains and a lower child-parent agreement for physical health. Furthermore, the peri-diagnostic period can be highly stressful for parents (Lowes & Lyne, 2000), negatively influencing their perceptions of their child's HRQL. However, as parents adapt to managing a child with a chronic health condition, overall child-parent agreement increases.
There are several potential explanations for our findings of informant discrepancy. First, as mentioned earlier, there is evidence that children and parents differ with regard to their concerns and valuation of HRQL domains (De Civita et al., 2005). Second, children with chronic health conditions may overestimate reports of their HRQL on account of the positive illusory bias -a compensation phenomenon observed when children rate their self-perceptions higher than more objective measures of performance or competence (Golden, 2007). Evidence of this self-protective mechanism has been observed in children with chronic health conditions (Diener & Milich, 1997;Ferro & Boyle, 2013;Heath & Glen, 2005). Third, parents, especially mothers, with symptoms of depression, may underestimate the HRQL of their children in a phenomenon known as depression distortion (Richters, 1992). Although, evidence has suggested depression distortion is common when assessing problem behaviour in children (De Los Reyes & Kazdin, 2005), there is less support for this type of informant bias in assessments of HRQL of children with chronic health conditions (Ferro, Avison, Campbell, & Speechley, 2010). Fourth, having a chronic health condition does not necessarily mean that a child is unsatisfied or views his/her life negatively, despite what other individuals might perceive. This disability paradox (Albrecht & Devlieger, 1999) can explain why higher HRQL scores were observed for child reports compared to parent reports.

Strengths and limitations
This exploratory study has a number of strengths including the inclusion of a variety of chronic health conditions, complete data on all eligible child-parent dyads, evaluation of child-parent agreement in a longitudinal design, and investigation of the extent to which the magnitude of informant discrepancies are clinically relevant. There is, however, one notable limitation to this study -small sample size. While this relatively small clinical sample study might restrict the generalizability of these exploratory findings, this limitation is tempered by the fact that the sociodemographic characteristics of our sample were similar to census data of the population residing within the catchment area of the tertiary care centres from which our sample was recruited (Statistics Canada, 2015) . Additionally, the KIDSCREEN-27 domain scores reported in our study were similar to those reported in population-based samples of children with chronic health conditions , 2014.

Conclusions
Our study provides preliminary evidence to suggest that the agreement between children with newly diagnosed chronic health conditions and their parents is low to moderate at diagnosis and tends to increase over time. This study further suggests that meaningful discrepancies are common. Replication of these findings is encouraged, as well as research aimed at understanding the factors associated with, as well as the mechanisms leading to child-parent discrepancies in reports of HRQL. Further quantification of the extent to which the magnitude of disagreement is clinically relevant is needed as this may have implications on the classification and estimation of changes in the disease course over time and treatment effects. The KIDSCREEN-27 remains a psychometrically robust measure of child HRQL. Clinicians and researchers should continue to consider both child and parent perspectives of HRQL in children with chronic health conditions, particularly in the period soon after diagnosis.