Evaluations of the well-being of children with juvenile idiopathic arthritis (JIA) typically rely on parents as proxy respondents. An assumption of several studies appears to be that mothers' and fathers' ratings are interchangeable, as reports do not always specify which parent completed the assessments nor, in repeated measures, whether they were completed by the same parent. The aim of this study was to examine the level of agreement between mothers' and fathers' ratings of their child's quality of life (QOL) and to identify possible predictors of disagreement.
Mothers and fathers (n = 82) of children with JIA completed ratings of their child's symptoms, QOL, and measures of their mood and beliefs about their child's illness and treatment. The number of active and limited joints and the physician's global assessment were also recorded.
Intraclass correlation coefficients between mothers' and fathers' ratings of physical and psychosocial QOL were high (0.824 and 0.755, respectively). However, calculation of difference scores revealed that 70.6% and 65.9%, respectively, were classified as discordant. Where parents differed, the direction of difference was not systematic. Discordance in parents' mood states and in their illness and treatment beliefs explained a small amount of the variance in discordance in QOL.
It should not be assumed that proxy ratings of a child's well-being can be generalized from one parent to the other. Studies that take repeated assessments should ensure that the same parent completes assessments at all time points. Other factors that may explain discordance between parents' ratings need to be explored.
Understanding parents' responses to their child's arthritis is crucial because their judgments of the well-being of their child are used, in part, to determine the relative success of therapy. Parents act as proxy respondents, particularly in the case of younger children. In juvenile idiopathic arthritis (JIA), 2 of the 6 core outcome variables that are used to define improvement in clinical trials (rating of overall well-being and the Childhood Health Assessment Questionnaire [C-HAQ]) (1) are often completed by parents. Parents' evaluations of their child's well-being are not only important in clinical assessment, but they also drive health care utilization (2).
Studies of proxy reporting in childhood chronic disease have usually focused on the level of agreement between the parent and child or between the parent and physician. Several studies have suggested poor to moderate correlations between parent and child ratings, with parents generally reporting the impact on the child to be greater than that reported by the children themselves (3, 4). In JIA, studies have compared parent and child ratings of pain and physical function (5–9), overall well-being (6, 8), and quality of life (QOL) (6, 8, 10–12). Agreement between parent and child has tended to be moderate, and there is some indication that the level of agreement varies with disease severity and the particular health domain being assessed.
Studies have also compared parents and physicians on their ratings of the child's pain (9), physical function (13, 14), and overall well-being (15, 16). Agreement between parent and physician ranged from 40–69% in these studies (13–16). Where the raters differed, there was no consistent finding across studies for either parent or physician to give poorer ratings.
An issue that has not had much attention is the assumption in many studies that it is not important whether a mother or father completes the ratings, because their responses are seen as interchangeable. In several of the studies mentioned above, it was not stated which of the parents provided the rating. This issue is particularly important in studies that require repeated parent ratings to be provided at more than one time point. It is striking that this information is not necessarily reported in clinical trial reports, which may specify only that a parent assessment was obtained, without specifying from which parent or whether it was the same parent each time (17). The assumption that mothers and fathers are interchangeable appears to have been made but this has rarely been tested (an exception being Garcia-Munitis et al, who compared patients' ratings with those of the patients' mothers, fathers, and physicians ).
Our study aimed to examine the relationship between mothers' and fathers' ratings of their child's QOL in JIA, and also to examine factors that might help explain the level of agreement or disagreement between them. Other studies have suggested that factors such as parents' illness beliefs and mood are likely to influence their assessments (8, 9). This hypothesis is consistent with Leventhal's theoretical model (18, 19), which proposes that people's cognitive and emotional responses to their illness influence how well they cope with it, which in turn has impact on health outcomes such as QOL. Our study aimed to examine the role of these variables when applied to a proxy measure. In this study, we compared mothers' and fathers' ratings of their child's QOL and examined whether the level of agreement between them was influenced by demographic characteristics, disease-related variables, agreement in parents' beliefs about their child's arthritis and its treatment, and parents' mood.
SUBJECTS AND METHODS
Participants were mothers and fathers of children with JIA, defined by International League of Associations for Rheumatology criteria (20), who were under the care of the rheumatology service at the Great Ormond Street Hospital for Children or the University College Hospital, London, UK. Children and parents were recruited as part of the Childhood Arthritis Response to Medication Study (CHARMS). All children were either currently taking methotrexate (MTX) or had been prescribed it in the past for ≥6 months. As part of the CHARMS, data were collected from parents about their child's response to treatment for JIA. In this report we present data only in cases where both parents completed the study questionnaires. The study had full ethical approval from the Institute of Child Health/Great Ormond Street Hospital Local Research Ethics Committee (reference 05/Q0508/95), and all participants gave full informed written consent. The study conforms to the principles outlined in the Declaration of Helsinki.
In addition to demographic information on the parents and child, the following measures were collected.
Clinician-assessed measures from the JIA core outcome variables (1), i.e., the number of active and limited joints and the physician's global assessment, were collected as indices of current JIA disease severity. JIA subtype, age of the patient at diagnosis, and disease duration were also recorded.
Parents' ratings of their child's QOL were assessed using the Pediatric Quality of Life Inventory (PedsQL) generic and rheumatology scales (10), both of which have parallel scales relating to children ages 2–4, 5–7, 8–12, and 13–18 years. The generic measure produces 2 composite scores of physical and psychosocial QOL. The rheumatology scale consists of 5 subscales: pain and hurt, daily activities, treatment, worry, and communication. The 2 latter subscales appear only in the scales relating to children age ≥5 years. Scores on all subscales of the generic and rheumatology scales are transformed to 0–100 scales, where a higher score signifies a better QOL. The PedsQL has been found to have good internal consistency and to distinguish between healthy children and those with rheumatic disease (10). Varni et al (21) have reported a clinically meaningful difference on the PedsQL generic scale of ±4.5 points.
Parents' ratings of their child's symptoms were assessed with 10-cm visual analog scales (VAS) for severity of pain, stiffness, and fatigue in the last week.
Parents' beliefs about their child's arthritis were assessed with the Revised Illness Perceptions Questionnaire (IPQ-R) (22), which was adapted to assess a proxy's beliefs about the patient's illness, rather than the patient's beliefs. The IPQ-R has been found to have good internal consistency, acceptable test–retest reliability, and to be able to discriminate between acute and chronic pain patients (20). The IPQ-R assesses beliefs in 9 domains, 7 of which are analyzed in this paper. Those not analyzed were identity: the symptoms that the person perceives to be related to the illness, and cause: beliefs about what caused the illness. The domains assessed in this study were 1) personal control: the parents' beliefs about their and their child's ability to control the illness, 2) treatment control: beliefs about the ability of treatment to control or cure the illness, 3) timeline acute/chronic: perception of the likely time course of the illness, 4) timeline cyclical: perception of the degree of unpredictability of the illness, 5) consequences: parents' perception of the impact of the illness both for themselves and their child, 6) coherence: how much parents understand the illness, and how much parents think their child understands the illness, and 7) emotional representation: the emotional responses generated by the illness in the parents themselves and in the child. Each subscale provides a score from 1–5, with a higher score representing a stronger belief.
Parents' beliefs about their child's treatment were assessed with the Treatment Representations Inventory (TRI) (23), which was adapted to assess a proxy's beliefs about a patient's treatment rather than the patient's beliefs. The TRI has been found to have good internal consistency, and to discriminate between patients undergoing different treatments (21). It assesses beliefs in 4 domains: 1) treatment value: beliefs about the positive effects of the treatment in controlling and arresting the progress of the illness, 2) concerns: beliefs about the emotional impact of the treatment (on the child) and parents' concerns about the treatment, 3) cure: beliefs about the ability of the treatment to resolve the illness and return their child to their normal life, and 4) decision satisfaction: parents' evaluation of the decision process for choosing their child's treatment. Each subscale provides a score from 1–5, with a higher score representing a stronger belief.
Parental mood was assessed with the Hospital Anxiety and Depression Scale (24). This scale provides separate scores for anxiety and depression, both ranging from 0–21, with higher scores indicative of greater depressed/anxious mood.
Levels of agreement between mothers' and fathers' assessments were calculated using intraclass correlation coefficients (ICCs). ICC values <0.40 reflect poor agreement, values 0.4 to <0.75 reflect moderate agreement, and values ≥0.75 reflect good agreement. Differences between mothers' and fathers' scores were also calculated for all measures to provide continuous difference scores between mothers and fathers on all variables. Agreement in QOL was also analyzed, using the Bland and Altman method (25).
To explore possible determinants of the level of discordance between parents in assessment of QOL, 2 multiple regression analyses were performed in which the dependent variables were parent difference scores on the PedsQL generic, physical, and psychosocial subscales. Potential predictor variables were child age and sex, disease duration, disease severity as indicated by the number of active and limited joints and the physician's global assessment VAS, parent age and education level, and difference scores on parent mood, illness, and treatment beliefs. The relationship between potential predictor variables and the QOL difference scores were examined initially by correlations (Pearson's correlation coefficients for continuous variables and Spearman's rho [ρ] for ordinal variables). To examine which variables accounted for the most variance in QOL difference scores, all significant variables (P < 0.01) identified from the univariate analyses were included in hierarchical multiple regressions using the enter method. The criterion level of P values less than 0.01 was used to ensure that the number of predictor variables did not exceed recommendations for power calculation in the multiple regression analysis (26).
The independent variables were entered into the regression in blocks in the following order: demographic variables, disease variables, illness and treatment beliefs, and mood. This order was used because it enabled the examination of whether psychosocial variables added to the explanation of discordance in parents' assessment of QOL after demographic and clinical variables had been taken into account.
Eighty-two parent dyads completed all assessments. The demographic and disease variables are shown in Table 1. All 82 children had been treated with MTX for their JIA for ≥6 months, and at the time of the assessment they were relatively well as indicated by the low disease activity scores such as active joint count and physician's global assessment (Table 1).
Values are the number (percentage) unless otherwise specified. IQR = interquartile range; JIA = juvenile idiopathic arthritis; RF = rheumatoid factor; VAS = visual analog scale; GCSE = General Certificate of Secondary Education; O-level = Ordinary level; A-level = Advanced level.
Age, mean ± SD years
8.3 ± 3.9
Time since diagnosis, median (IQR) years (n = 78)
Polyarticular, RF negative
Polyarticular, RF positive
Polyarticular, RF status unknown
Core outcome variables, median (range, IQR)
Number of active joints (n = 77)
0 (0–10, 0–1.5)
Number of limited joints (n = 77)
0 (0–10, 0–1.5)
Physician VAS (n = 65)
0 (0–5.8, 0–1)
Age, mean ± SD years
38 ± 6.3
Highest academic qualification
Missing data, no.
Age, mean ± SD years
40 ± 6.3
Highest academic qualification
Missing data, no.
Assessment of the degree of concordance between mothers and fathers.
Mothers' and fathers' ratings of their child's QOL are shown in Table 2. ICCs between mothers' and fathers' evaluations of generic QOL were good, with a slightly higher correlation in physical than in psychosocial QOL (Table 2). However, on the rheumatology scale, ICCs ranged from poor on the worry subscale to good on the pain subscale. For mothers' and fathers' ratings of their child's symptoms, ICCs were moderate for stiffness and fatigue and good for pain (Table 2).
Table 2. Mean ± SD scores and ICCs between mothers' and fathers' assessments of their child's QOL and symptoms*
Values are the mean ± SD unless otherwise indicated. ICC = intraclass correlation coefficient; QOL = quality of life.
Scored 0–100, with higher scores signifying better QOL, e.g., a higher score on pain signifies less-frequent pain-related problems.
Not completed for children age <5 years.
Measured on a 10-cm visual analog scale, where a higher score signifies more severe symptoms in the past week.
There was greater variation in the level of agreement between parents on their beliefs about their child's JIA and its treatment (Table 3). ICCs ranged from poor on beliefs about the level of personal control the child or parent had over the illness to moderate for beliefs about the ability of treatment to cure or control the illness.
Table 3. ICCs between mothers' and fathers' illness and treatment beliefs*
Values are the mean ± SD unless otherwise stated. All scores are on a scale of 1–5, with higher scores indicating stronger belief in the ability of the treatment to control the illness, more severe consequences, more severe emotional response, stronger belief in an unpredictable disease timeline, more coherence, belief in a longer disease duration, greater sense of personal control over the illness, stronger belief in the power of the treatment to cure the illness, greater satisfaction with treatment decisions, greater concerns about the treatment, and stronger belief in the value of the treatment. ICCs = intraclass correlation coefficiants; IPQ-R = Revised Illness Perceptions Questionnaire; TRI = Treatment Representations Inventory.
3.05 ± 0.46
3.06 ± 0.34
Consequences for parent
3.03 ± 0.80
3.00 ± 0.84
Consequences for child
3.12 ± 0.77
3.24 ± 0.68
Child's emotional representation
2.84 ± 0.70
2.87 ± 0.71
Parent's emotional representation
3.42 ± 0.78
3.19 ± 0.71
3.29 ± 0.92
3.05 ± 0.89
Coherence to parent
3.70 ± 0.92
3.62 ± 0.67
Coherence to child
2.98 ± 0.57
2.98 ± 0.61
3.03 ± 0.31
3.01 ± 0.32
Child's personal control
2.76 ± 0.44
2.80 ± 0.40
Parent's personal control
3.32 ± 0.68
3.21 ± 0.66
3.57 ± 0.65
3.55 ± 0.59
3.89 ± 0.44
3.77 ± 0.49
3.45 ± 0.70
3.40 ± 0.65
4.12 ± 0.44
4.11 ± 0.48
Mean ± SD scores for anxiety and depression were 7.43 ± 4.07 and 4.01 ± 3.47, respectively, for mothers and 6.35 ± 3.77 and 3.84 ± 3.02, respectively, for fathers. These are within the normal, nonclinical range for the scale. Paired t-tests found no significant differences between mothers and fathers on anxiety or depression (results not shown).
Although ICCs on the PedsQL generic scale were good, plotting of difference scores showed that there was nonetheless a relatively high degree of discordance between parents on these scales (Figure 1). A difference of ±4.5 points on the generic scale is considered clinically significant (21), and only 29.4% of mothers' and fathers' scores fell within this range on generic physical QOL, and 34.1% on psychosocial QOL (Figure 1). Although a clinically meaningful difference has not been reported for the rheumatology scale, we took the same criteria for this scale as are applied in the general scale to establish discordant judgments. The greatest agreement between mothers and fathers was on the daily activities subscale, and the poorest was on the worry subscale (Figure 1).
The direction of difference in the parents' assessments was not systematic (Figure 1). For example, the proportions for whom QOL was rated better by the mother were fairly similar to those for whom QOL was rated better by the father.
Factors that may help to explain the degree of concordance/discordance between mothers and fathers.
To determine whether the level of discordance in parents' assessments of their child's QOL was related to demographic or clinical variables, a series of univariate correlations was performed. These showed that the level of discordance was not related to the child's age or sex; JIA subtype; disease duration; or disease severity assessed by active joints, limited joints, and the physician's global assessment VAS.
To determine whether the level of discordance in parents' assessments of their child's QOL was related to differences in their age, education, mood, and perceptions of symptoms and beliefs regarding their child's JIA and its treatments, a further set of correlations was performed between the difference scores on these variables. The significant correlations between these difference scores are shown in Table 4. Greater discordance between parents in their ratings of their child's physical QOL was related to greater discordance in their assessment of pain and stiffness, their belief about the emotional impact of JIA on their child, and their level of depressed mood. So, for example, where one parent had a higher level of depressed mood than the other, that parent's rating of the child's physical QOL was also worse.
Table 4. Significant correlations (P < 0.01) of the differences between fathers and mothers in their evaluation of their child's QOL*
QOL = quality of life; VAS = visual analog scale; ns = not significant.
Belief about disease uncertainty
Belief about emotional impact on child
Parental depressed mood
Parental years education
The significant variables discussed above were entered into 2 multiple regression analyses, with discordance in physical QOL and psychosocial QOL as the dependent variables (Table 5). These variables explained 21% and 14% of the variance in physical and psychosocial QOL, respectively. Only discordance in pain rating and depressed mood remained significant predictors of discordance in physical QOL in the final equation. Discordance in mothers' and fathers' treatment concerns and in their beliefs about the cyclical nature of JIA were significant predictors of discordance in their ratings of psychosocial QOL.
This study has shown that the correlations found between mothers and fathers in their ratings of their child's QOL can mask high levels of discordant responses. This indicates that where parental proxy reports are used in JIA, any assumption that mothers' and fathers' ratings are interchangeable may not be correct. Although it was asserted by Jozefiak et al (27) that it is reasonable to generalize from mothers to parents, based on a study of parents' ratings of the QOL of healthy, school-age children, those findings appear to not generalize to parents of children with JIA.
Any assumption that parents would display a systematic difference in their ratings of QOL was not supported by the data in this study. Consequently, it is not possible to simply apply a correction factor to enable generalization from one parent to the other. It would seem that where possible, obtaining the ratings of both parents may be informative, but, as this will often be impractical, repeated assessments in studies examining changes in QOL should at least be completed by the same parent, and studies should report who has provided the proxy rating. It is important that the classification of discordance in the current study was based on a difference score in the PedsQL that is considered clinically meaningful (21). Therefore, the findings indicate that where repeated proxy assessments by parents in a trial are not provided by the same parent, it is possible that clinically important bias may enter study findings.
To our knowledge, only 1 previous study has compared in detail mothers' and fathers' ratings in JIA (9). That study compared mother, father, child, and physician ratings of the child's pain, and reported moderate levels of agreement about pain intensity between the child and the parents and poor levels of agreement between the child and the physician. Comparison of mothers and fathers found moderate to good agreement in ratings of their child's present pain, pain in the previous week, and C-HAQ scores, with ICCs of 0.73, 0.77, and 0.8, respectively. However, our study performed further analyses of parent responses, and we suggest that although some data indicate fairly good levels of agreement between parents, correlational analysis may mask underlying differences.
Studies have reported that proxy ratings tend to be in greater agreement with child reports for more observable phenomena, such as daily activities, than for cognitive and emotional attributes (3). The current study has provided some evidence to suggest that this is also the case between parent reports, in that the highest level of agreement was for daily activities and the lowest was for pain and worry (Figure 1). Such findings are probably to be expected, but it is helpful to be aware of where the discrepancies exist between raters because symptoms and emotional variables are important areas of the assessment of children with JIA.
Given the differences found between parents, our study aimed to identify variables that could help to explain the level of differences. Some studies that have compared parent and patient proxy reports have found higher levels of agreement at extreme ends of the scale but more disagreement midscale, i.e., raters are more likely to be in agreement if the child is very well or very unwell. This does not appear to be the case in the comparison of ratings between mothers and fathers, in which discordance was not influenced by the child's disease severity. Neither did demographic characteristics of the parents and children help to explain levels of discordance.
We also examined whether discordance between parents in their mood and beliefs about the illness and its treatment might help to explain discordance in their ratings of QOL. In the multivariate analysis of discordance between parents' ratings of physical QOL, discordance in pain rating and level of depressed mood were significant explanatory variables; however, discordance in illness and treatment beliefs did not contribute to the explanation of discordance in ratings of physical QOL. An association between mothers' ratings of their child's functioning and their own levels of depressed mood was found in another study of JIA (28). The child's depressive symptoms have also been found to be predictive of parent–child disagreement about the child's pain (7). The causal direction of the findings in the current study remains unknown, however. It is possible that a poorer perception of one's child's QOL increases the likelihood of depressed mood. On the other hand, a more depressed mood could lead to perceiving one's child's QOL in a more negative way. It is likely that both could be valid, but longitudinal analysis would be required to tease out the relationship.
The variables that helped explain discordance in psychosocial QOL were different from those that were significant in physical QOL. Discordance in parents' beliefs about the unpredictable nature of JIA (timeline cyclical) and concerns about treatment were significant variables in the multiple regression analysis of discordance in psychosocial QOL. Concerns about medication have been reported as one of the greatest stressors by mothers of children with JIA (28), and the unpredictable disease course has been associated with parental anxiety (29). It would appear that where these perceptions differ between parents, they are reflected in different ratings of the child's psychosocial QOL. Again, the causal direction is unknown, in that a poorer perception of the child's psychosocial QOL may result in greater concerns about their treatment and the unpredictable nature of the disease; however, the opposite causal route is also feasible.
The amount of variance in discordant ratings of both physical and psychosocial QOL explained by these variables was fairly small, indicating that other factors that were not assessed in this study must be important in explaining discordance between parents (Table 5). The findings suggest the need for further research to understand the underlying factors related to parental discordance, and one may speculate that factors such as family functioning, parental characteristics such as optimism, and parents' coping abilities may be important.
The current study's findings must be considered in the light of some limitations. The patients in our study had all been treated with MTX for ≥6 months and mostly had low disease severity at the time of assessment, shown by joint count and physician VAS scores, so the extent to which they are generalizable to the JIA population as a whole is unknown. Although the study examined only parent ratings and could have been enriched with the inclusion of child data, this would have excluded the parents of very young children who cannot complete self-report measures.
In conclusion, we suggest that studies of JIA that use parent proxy reporting to assess QOL and well-being in children need to detail which parent provides the reports and, where possible, to use the same parent, or both, throughout any study that uses repeated sampling of parents proxy ratings.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Newman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Kassoumeri, Wedderburn, Newman.
Acquisition of data. Mulligan, Etheridge, Kassoumeri, Wedderburn.
Analysis and interpretation of data. Mulligan, Kassoumeri, Newman.