Health of children with chronic arthritis: Relationship of different measures and the quality of parent proxy reporting

Authors


Abstract

Objective

To examine the strength of the association between different measures of health-related quality of life (HRQOL), disability, pain, and well-being in children with chronic arthritis. To evaluate whether HRQOL scores vary as a function of disability status beyond chance. To assess the quality of the parent proxy report for HRQOL as compared with disability, pain, and well-being.

Methods

Measures of HRQOL (visual analog scale [VAS] of health, Pediatric Quality of Life Inventory [PedsQL], Juvenile Arthritis Quality of Life Questionnaire (JAQQ), and modified standard gamble technique [SG]), disability (Childhood Health Assessment Questionnaire), VAS of pain, and VAS of well-being (VAS-well) were completed by the parents (n = 119) and patients ≥8 years (SG: ≥12 years).

Results

HRQOL was highest when measured by the SG, whose utilities were no more than weakly correlated with any of the other outcomes. The values of all other HRQOL measures were at least moderately correlated with each other and with the VAS-well. Irrespective of the measure used, disability was associated with significantly decreased HRQOL. There was fair to good agreement and moderate consistency of the HRQOL ratings (SG: fair consistency) between patients and parents with marked differences between health domains.

Conclusion

HRQOL measured by the PedsQL, JAQQ, and VAS are moderately to highly correlated with each other in children with chronic arthritis. The children's HRQOL significantly decreases with increasing disability. Despite more pronounced differences for some health domains, parents are moderate to good proxy reporters of HRQOL, disability, and well-being of children with chronic arthritis.

INTRODUCTION

There is a continuing need to evaluate the efficacy of medical interventions for pediatric patients, especially those with chronic disorders. In the past, for children with chronic arthritis, such evaluations were typically based on such traditional medical outcomes as growth, the number of affected joints, and the results of certain laboratory parameters. Although traditional outcomes are important, they do not adequately consider the child's and family members' perception of treatment benefits, nor do they capture the child's overall sense of health. Health-related quality of life (HRQOL) measurement provides additional information for judging the effects of therapy, may be helpful when comparing treatment alternatives, and useful when screening for children who have particular difficulties and therefore need remedial or counseling help (1). The revised International Classification of Functioning, Disability and Health was recently published by the World Health Organization in an effort to standardize the language and framework for the description of functioning, health, and health-related states, such as well-being (2). In this framework, the umbrella term functioning refers to all body functions, activities, and participation. Similarly, disability serves as an umbrella term for impairments, activity limitations, and participation restrictions. Health is the result of a patient's functioning and disability, whereas HRQOL constitutes the valuation of a certain level of health, e.g., functioning and disability. Conversely, health is only one aspect of an individual's overall wellbeing, along with her or his education, work, and social relationships (3).

Different approaches to measuring HRQOL have been developed. Available HRQOL measures can be classified as being generic measures that can be used to measure the HRQOL of both healthy children and those carrying different diagnoses. So-called disease-specific measures were developed to value health states of children with specific diseases. Because such measures are tailored toward the target diseases, they are thought to better capture changes in the health seen with these diseases (better responsiveness to change) than generic measures.

Preference-based HRQOL provides an individualized valuation of health states and can be measured using the standard gamble (SG) technique (4) and visual analog scales (VAS) (5). In contrast, non–preference-based measures provide HRQOL estimates based on the attitudes and values of the “average” patient (as determined by the developer of the scale). Not surprisingly, the various approaches to measuring health result in different values of HRQOL. Although the choice of a HRQOL tool for research purposes depends largely on the study question posed, it would be informative to know how the values of the different HRQOL tools compare with each other when used in patients with chronic arthritis. For example, such knowledge is required when comparing the results of studies using different HRQOL measures and for future meta-analyses.

Another challenge when measuring HRQOL, particularly in pediatrics, is that proxy reporting is often required. Using only parent proxy reports instead of patient self reports would fail to capture that parents and children may differ in their perception of health (6). Although patient self reporting of HRQOL has gained increasing acceptance in pediatrics (7, 8), parents remain the principle medical decision makers. Thus, there is a growing need to understand the relationship between parent proxy reporting and patient self reporting. Previous research suggests that the quality of the proxy reports in children with some chronic diseases may vary with patient age and depend on the health domain measured, but this has not been well examined across HRQOL measures for children with arthritis (1, 8, 9). The relationship between parent proxy reports and patient self reports of HRQOL is best clarified by directly comparing those ratings. When doing so, one has to consider at least 2 aspects of the relationship between parent and patient ratings. They are the agreement between parent and patient report, e.g., to what extent parents and patients coincide in their rating of HRQOL (consensus), and their consistency, e.g., whether parent and patient ratings, although they are not the same (no consensus), are at least importantly correlated with each other (10).

The purpose of the present study was to examine the strength of the association of HRQOL estimates measured by different types of tools for children with chronic arthritis. We were also interested in the strength of the association between HRQOL and disability, pain, or well-being, and whether HRQOL changes importantly as a function of the disability status. In addition, we wanted to compare the quality of the parent proxy report of HRQOL with that of disability, pain, and wellbeing. Based on previous reports in the literature, we hypothesized that HRQOL estimates are at least moderately associated with each other and that SG ratings would provide higher HRQOL values than other HRQOL measures. We also hypothesized that parents may be better proxy reporters when estimating the disability of their children than when rating pain and HRQOL of the patients (11–13).

PATIENTS AND METHODS

Patients.

Families of children with arthritis were recruited during routine clinic visits within a 3-month period. Eligible patients were between 1 and 18 years of age and had symptoms of chronic arthritis irrespective of a specific underlying diagnosis. To be included in the study, arthritis had to have been present for at least 3 months continuously. Excluded were families with children carrying the diagnoses of fibromyalgia, nonspecific myalgias or arthralgias, and those in which children had symptoms of <3 months' duration.

Families were recruited from the rheumatology clinics in a semiconsecutive fashion, depending on the availability of research personnel within a 3-month period. Less than 5% of the approached patients refused participation, mostly because of time conflicts. A standardized introduction to HRQOL was given to both parents and patients prior to the administration of the study instruments. All interviews were performed by 1 of 5 trained interviewers (MT, NB, AK, AJ, AB). Each family was interviewed twice at the time of clinic visits at least 4 weeks apart, but generally on 2 consecutive clinic visits. Parents and patients completed the questionnaires independent of each other while under direct supervision. The sequence in which the questionnaires were administered was altered between subjects to avoid sequence effects.

Outcome measures.

Measures completed for the study included different types of HRQOL measures and, for comparison and validity check, measures of well-being, disability, and such traditional outcomes as physician assessment of disease severity, pain, and the number of involved joints, e.g., those with active arthritis or limited range of motion. The primary caretakers and children age ≥12 years completed all study instruments. Children age 8–11 years completed all study questionnaires but were not asked to perform the SG because of the relative complexity of the procedure involved (14).

Disease severity.

The treating physician rated global disease severity on a VAS of 100-mm length anchored at either end by “very mild” and “very severe.” This scale has not been formally validated in the past.

Childhood Health Assessment Questionnaire (CHAQ).

The CHAQ consists of 2 components: disability and discomfort. Disability is assessed using 30 questions grouped in 8 domains covering major aspects of daily living over a 1-week period: dressing and grooming, arising, eating, walking, hygiene, reach, grip, and activities. Items are rated on a 4-point Likert scale (no difficulty, some difficulty, much difficulty, unable to do) with the option to mark “not applicable” if a child cannot be expected to perform a certain maneuver because of young age. Each domain contains at least 1 item that is developmentally appropriate for children according to their age. If aids or devices are used or assistance is required, the minimal score for the corresponding domain is 2. The disability index is calculated as the unweighted average of the 8 domain scores and yields a disability score between 0 (no disability) and 3 (most severe disability). A recent multicenter study (n = 6,644) supports that the CHAQ, when used in children with arthritis, has an important floor effect, often only fair to good internal consistency, with fair to poor test–retest reliability, especially with regard to the domain arising (15). Other studies, however, suggest excellent test–retest reliability of the CHAQ (16, 17). Discomfort is determined by the presence of pain over the preceding 1-week period, which is measured by a 100-mm VAS anchored at either end by no pain and very severe pain (VAS-pain).

Global rating of health and well-being.

Parents and patients were asked to rate the patient's well-being during the preceding 1 week using a 100-mm double-anchored linear analog scale presented with the sentence stem “My/my child's overall well-being is …” (13, 18, 19). The lower endpoint of the scale is marked with “extremely bad” and a sad face, whereas the upper endpoint of the scale is defined as “excellent” with a happy face (VAS-well). Similarly, preference-based global HRQOL is measured using the sentence stem “My/my child's health is …” (13, 18, 19). The lower endpoint of the scale is marked with “extremely bad” and a sad face, whereas the upper endpoint of the scale is defined as “excellent” with a happy face (VAS-health).

The Juvenile Arthritis Quality of Life Questionnaire (JAQQ).

The JAQQ is a non–preference-based, disease-specific measure of HRQOL (20, 21). The questionnaire consists of 74 items grouped into the following 4 domains: gross motor function, fine motor function, psychosocial function, and systemic symptoms. Patients are requested to consider the preceding 2 weeks. Each item is scored on a 6-point Likert scale (none of the time—never; hardly any time–10% of the time; some of the time–25% of the time; half of the time–50% of the time; most of the time–75% of the time; almost all of the time–90% of the time; all the time–always) with the option to answer “does not apply to me/my child” for items that are not expected from children because of their young age. Domain scores are calculated based on the unweighted average of the 5 highest scored items in the domain. The summary JAQQ score is derived from the unweighted average of the 4 domain scores (22). The JAQQ has excellent reliability, construct validity, and responsiveness to change when used in children with juvenile rheumatoid arthritis and other chronic arthritides (21, 22). Initial validation of the JAQQ (n = 40) suggests good agreement between parent proxy reports and patient self reports (23).

Pediatric Quality of Life Questionnaire Inventory (PedsQL).

The PedsQL comprises a parallel child self report and a parent proxy report with formats for various age ranges. Child self reports include ages 5–7 years (young child), 8–12 years (child), and 13–18 years (adolescent). Parent proxy reports include ages 2–4 years (toddler), 5–7 years (young child), 8–12 years (child), and 13–18 years (adolescent). Besides the PedsQL generic core scale version 4 (PedsQL-GC) (6), there are several disease-specific HRQOL modules, 1 of which is the PedsQL rheumatology module (PedsQL-RM).

The PedsQL-GC is a generic non–preference-based scale that encompasses the following 4 health domains for the preceding 4 weeks: physical functioning, emotional functioning, social functioning, and school functioning. Items are scored using a 5-point Likert scale (never, almost never, sometimes, often, always). From the sum of the raw scores of the 23 items a total health summary score ranging from 0 to 100 can be calculated, with higher scores indicating higher HRQOL. Reference scores for healthy children and children with rheumatic diseases are available (8, 24). Internal consistency and reliability for both parent and patient reporting, as well as construct validity and responsiveness, of the PedsQL-GC have been shown (6, 8, 24).

The PedsQL-RM is a non–preference-based, disease-specific scale designed to measure health dimensions that are particularly relevant for children with arthritis and has been proposed for use in combination with the PedsQL-GC (24). The PedsQL-RM is a 22-item scale that encompasses the following 5 different domains for the preceding 4 weeks: pain and hurt, daily activities, treatment, worry, and communication. Parent proxy report of the toddler age does not include a worry and communication domain. Items are scored on a 5-point Likert scale identical to that used for the PedsQL-GC. From the raw scores of the PedsQL-RM, a summary score of 0–100 was calculated, with higher scores indicating higher HRQOL. The PedsQL-RM has excellent internal consistency in children with chronic arthritis and its domain scores are moderately to highly correlated with PedsQL-GC scores (8, 24). Initial validation suggests moderate to high responsiveness of the PedsQL-RM and that the scale is, on average, more responsive to clinically important changes of children with chronic arthritis than the PedsQL-GC.

Standard gamble technique.

The SG is the classic method of measuring health state preferences (4, 25). This method is based on game theory and provides HRQOL values that are called utilities (range 0–1). Different from other HRQOL measures, the measurement of utilities requires the subject to take a certain risk. This is similar to many health care decisions, in which patient benefits are weighed against possible side effects of therapy. Using the classic form of the SG, patients are asked how much risk they would be willing to accept in a gamble that could provide them with “perfect health” (utility = 1) instead of her/his current health state against a certain risk of “instant, painless death” (utility = 0). The theory underlying the SG assumes that patients with severe symptoms with their disease will accept a higher risk of death compared with patients who are relatively well. The level of a patient's HRQOL (range 0–1) corresponds to the level of risk that is accepted by the patient when gambling for perfect health (utility = 1) against the odds of instant painless death (utility = 0). For adults with chronic arthritis, the test–retest reliability of the SG is generally better than that of the VAS (intraclass correlation coefficient [ICC] for VAS is 0.24–0.56 versus an ICC for SG of 0.43–0.79) (26, 27). Juniper et al reported ICC values of 0.55 for children ages 12 years and older (14).

Based on the findings and suggestions of previous studies in which SG utilities were extremely skewed (13, 27, 28), we performed the SG in a modified form. Modification of the classic SG has been used in the past for measuring patient preferences (29). Each participating parent was asked to imagine a “once-daily oral medication (white pill) without side effects that will instantly relieve the child from joint pains and provide the child immediately with normal physical function until the end of the child's natural life.” This health state was assigned the utility of 1. The parents was then asked to imagine another medication (black pill) that, when taken by her/his child once, “will lead to lifelong severe arthritis, requiring the child to use a wheelchair and receive help with the activities of daily living until the end of the child's natural life” (utility = 0). All parents were then shown a picture of a jar containing 100 white pills and all those who understood the concept of the SG chose to accept the certainty of cure rather than have their child remain in the present health state. Then a jar with 95% white pills and 5% black pills was shown. If they accepted, then the proportion of black pills was increased by 5% until the gamble was refused. At this point, the proportion of black pills was decreased in 1% increments until the point of indifference was reached. The indifference probability, that is, the point at which parents could not decide whether they would gamble or not, represents the utility that the parents placed on the health state of the child. To ensure that participants understood the task, they were asked to rate 2 practice paper cases of children with chronic arthritis (one mildly ill, the other moderately–severely ill). Parents were then asked to rate the health as they thought their own child with chronic arthritis would, and older patients (age ≥12 years) rated their own health. Utilities were measured by using the adapted form of a previously tested easy-to-use computer program (the GAMBLER) that was generously provided by Dr. Mark Eckman, Departments of Internal Medicine and Cost-effectiveness, University of Cincinnati.

Statistical analysis.

Data was analyzed using SAS 8e (SAS Inc., Cary, NC) and EXCEL 5.0 (Microsoft Inc., Redwood, WA). The HRQOL values derived from the JAQQ were rescaled to possible ranges of 0–1 to facilitate the comparison with the other HRQOL measures. This was done using the following formula:

HRQOL rescaled value = 1 − ([sum of the 5 highest scores of each of the 4 domains/4] − 5)/30.

The values of none of the measured outcomes were normally distributed (Kolmogorov-Smirnov test for normality; all P < 0.01). The scores of the CHAQ, disease severity, and pain were negatively skewed, whereas those of all HRQOL measures and well-being were positively skewed. Therefore, Spearman correlation coefficients (rs) were calculated for visit 1 and visit 2 separately to evaluate the relationship between the measured outcomes.

Disability categories were determined based on the CHAQ scores (parent proxy ratings) as previously suggested (19): CHAQ scores of 0 identified patients with no disability and those with CHAQ scores of 0–0.25 were categorized as having mild disability. Mild to moderate disability was defined by CHAQ scores of 0.25–1.25 and moderate disability by scores of 1.26–2.0. To examine the relationship between HRQOL and disability categories in more detail, nested parametric analysis of variance (ANOVA) was done using data from both interviews, as well as nonparametric ANOVA (Kruskal-Wallis statistic; P values are reported as Pkw) using data from visit 1 and visit 2 separately. The HRQOL of patients grouped into the different disability categories was tested for significant differences using both nonparametric and parametric ANOVA with post-hoc testing as previously suggested (30).

Agreement between parent proxy report and patient self report was determined using ICCs (31–34) based on the results of nested ANOVA, which considered that each family contributed up to 2 ratings (35). For the interpretation of ICC values, the following classification has been suggested: poor agreement: ICC <0.4; fair to good agreement: ICC ≥0.4–0.75; excellent agreement: ICC ≥0.75 (36).

Consistency between parent proxy report and patient self report was measured by Spearman's correlation coefficient. Data was assessed for visit 1 and visit 2 separately. Statistically significant correlations (P < 0.05) of patient self report with parent proxy report are considered as unrelated or nonconsistent for values of rs < 0.2. Weak consistency is defined by rs ≥ 0.2–0.4; moderate consistency by rs ≥ 0.4–0.6; strong–moderate consistency by rs ≥ 0.6–0.8; and strong consistency by rs ≥ 0.8 (37).

RESULTS

Patients and parents.

Families of children (n = 119; 91 girls and 28 boys) with chronic arthritis were studied. Most patients were diagnosed with juvenile rheumatoid arthritis (n = 102), but diagnoses also included Castleman syndrome (n = 1), juvenile dermatomyositis (n = 1), arthritis associated with inflammatory bowel disease (n = 1), juvenile psoriatic arthritis (n = 8), juvenile ankylosing spondylitis (n = 2), sarcoidosis (n = 1), systemic lupus erythematosus (n = 2), and mixed connective tissue disease (n = 1). The mean age of the enrolled children was 10.5 years (range 3–18 years; SD 4.3 years). The mean disease duration at the time of enrollment was 3.5 years (range 0.3–14.2 years). These families were recruited from the rheumatology clinics during routine visits after informed consent. Assent was obtained from children ages 8 years and older. The mean ± SD time between study visits was 3.5 ± 0.6 months. Twelve fathers, 106 mothers, and 1 grandmother provided the proxy ratings at the time of the first and second interview. Forty-three percent of the caretakers reported to have completed high school, 40% to have at least some college education, and 7% had a university degree.

Descriptive analysis of the outcome measures.

Because study evaluations were done during clinic visits, and in an effort to keep the duration of the study interviews shorter than 40 minutes, not all families were asked to complete all measures. Parent ratings of the VAS-health, the VAS-well, and the CHAQ (including VAS pain) were available from all parents (n = 119), but the SG (n = 58), the JAQQ (n = 58), the PedsQL-RM (n = 94), and the PedsQL-GC (n = 60) were completed by only some of the proxy reporters. The CHAQ (including VAS-pain), the VAS-health, the VAS-well, and the PedsQL-RM were completed by 87 children, the PedsQL-GC by 46 children, and the SG by 31 children age ≥12 years. Results of visit 1 were similar to those of visit 2 and are shown in Table 1. On a group level, HRQOL reported by the patients themselves was not significantly different from the proxy reports of HRQOL (P = not significant).

Table 1. Descriptive analysis*
 VisitCompleted bynPossible rangeObserved rangeMedianIQRMeanSD
  • *

    “All parents” refers to ratings from parents irrespective of whether a child report was available. “Parents with child” refers to ratings from families in which a child report and a parent report were available. IQR = interquartile range; VAS = visual analog scale; CHAQ = Childhood Health Assessment Questionnaire; JAQQ = Juvenile Arthritis Quality of Life Questionnaire; PedsQL-GC = Pediatric Quality of Life Inventory (PedsQL) generic core scale; PedsQL-RM = PedsQL rheumatology module.

Physician rating of disease severityVisit 1 1190–10.00–9.52.64.33.12.6
Visit 2 118 0–7.82.73.63.12.3
No. involved jointsVisit 1 119 0–46143.46.1
Visit 2 119 0–34244.59.8
VAS-painVisit 1Child870–10.00–10.01.12.92.22.4
  All parents119 0–9.01.74.52.52.5
  Parents with child87 0–9.01.14.62.62.6
 Visit 2Child87 0–10.01.33.62.22.6
  All parents119 0–9.91.24.62.62.7
  Parents with child87 0–8.11.94.62.62.6
CHAQVisit 1Child870–3.00–2.250.250.630.440.54
  All parents119 0–2.130.250.880.510.60
  Parents with child87 0–2.00.250.750.460.56
 Visit 2Child87 0–1.630.130.500.330.44
  All parents119 0–2.130.250.800.480.57
  Parents with child87 0–2.000.250.630.410.52
VAS-wellVisit 1Child870–10.02.0–10.08.03.07.42.0
  All parents119 2.0–10.08.02.07.81.7
  Parents with child87 2.0–10.08.02.07.71.8
 Visit 2Child87 3.1–10.08.22.37.91.8
  All parents119 2.1–10.08.22.07.91.6
  Parents with child87 2.2–10.08.31.98.01.5
VAS-healthVisit 1Child870–10.02.5–10.08.11.87.71.9
  All parents119 1.7–10.08.03.57.52.1
  Parents with child87 1.7–10.07.63.67.32.1
  Child87 4.0–10.08.12.37.81.9
  All parents119 3.2–10.08.32.67.91.8
  Parents with child87 3.0–10.08.11.97.81.9
Modified standard gamble Child310–1.00.50–1.000.940.130.890.13
Visit 1All parents58 0.64–1.000.960.150.910.11
  Parents with child31 0.64–0.990.940.190.890.11
 Visit 2Child31 0.65–1.000.950.100.930.10
  All parents59 0.70–1.000.980.050.950.07
  Parents with child31 0.70–1.000.980.050.950.08
JAQQVisit 1Child410–1.00.33–1.000.850.250.790.20
  All parents58 0.30–1.000.780.240.750.17
  Parents with child41 0.30–0.990.820.200.760.16
 Visit 1Child41 0.26–1.000.910.250.820.18
  All parents58 0.33–1.000.820.260.780.18
  Parents with child41 0.33–0.990.830.190.780.18
PedsQL-GCVisit 2Child460–1.00.18–0.990.810.280.780.17
  All parents60 0.37–1.000.850.280.790.19
  Parents with child46 0.27–1.000.790.190.770.13
 Visit 2Child46 0.26–1.000.870.200.830.17
  All parents59 0.49–1.000.870.240.820.15
  Parents with child46 0.26–1.000.800.240.800.18
PedsQL-RMVisit 1Child860–1.00.25–1.00.810.240.760.17
  All parents94 0.30–1.000.750.280.740.17
  Parents with child86 0.30–1.000.740.290.730.17
 Visit 2Child87 0.30–1.000.840.170.800.16
  All parents94 0.34–1.000.770.280.760.17
  Parents with child87 0.40–1.000.780.230.800.18

Strength of the association between the different types of HRQOL measures.

With the exemption of the SG utilities, the scores of all the other HRQOL measures (PedsQL-RM, PedsQL-GC, VAS-health, JAQQ) were moderately to highly correlated with each other. The SG utilities were weakly correlated with ratings of the VAS-health and the PedsQL-RM, but unrelated to HRQOL measured by the JAQQ and the PedsQL-GC (Table 2). In exploratory analysis, we examined whether certain health domains had a stronger correlation with SG utilities than the summary scores of the PedsQL-RM, the PedsQL-GC, and the JAQQ. Utilities moderately correlated with the pain and hurt domain of the PedsQL-RM (rs ≤ –0.50; P < 0.006). Parent proxy report of utilities was highly correlated with the systemic symptom domain of the JAQQ (rs = 0.62; P < 0.007), but not with any of the PedsQL-GC domains. SG utilities of children but not of parents were significantly correlated with patient age (rs = −0.46; P < 0.04).

Table 2. Spearman correlation coefficients of visit 1 data*
 nDisease severityNo. involved jointsPain (VAS-pain)CHAQWell-being (VAS-well)Global rating of health (VAS-health)Modified standard gambleJAQQPedsQL-GC
  • *

    For abbreviations, see Table 1.

  • P < 0.0001.

  • P < 0.05.

  • §

    P < 0.005.

No. involved joints1190.461.0       
Pain (VAS-pain)          
 Parent119−0.01−0.01       
 Child870.240.16       
CHAQ          
 Parent1190.110.240.28      
 Child870.01−0.030.31§      
Well-being (VAS-well)          
 Parent119−0.40−0.31§−0.36§−0.45     
 Child87−0.35§−0.09−0.53−0.23     
Global rating of health (VAS-health)          
 Parent119−0.28−0.16−0.65−0.520.77    
 Child87−0.34−0.07−0.57§−0.640.73    
Modified standard gamble          
 Parent58−0.06−0.27−0.31−0.250.290.29   
 Child31−0.33−0.25−0.15−0.190.180.32   
JAQQ          
 Parent58−0.17−0.16−0.54−0.650.590.570.17  
 Child410.02−0.03−0.45−0.740.360.660.47  
PedsQL-GC          
 Parent60−0.42§−0.020.12−0.220.640.530.020.73 
 Child46−0.24−0.20−0.36−0.320.44§0.660.280.78§ 
PedsQL-RM          
 Parent94−0.48−0.18−0.27−0.420.660.620.200.790.81
 Child86−0.46−0.18−0.60−0.470.450.60§0.520.760.80

Strength of the association between HRQOL and other outcomes.

HRQOL ratings were strongly correlated with patient well-being (VAS-well) for both parent proxy reports and patient self reports, with the exception of the SG utilities, which were weakly correlated with VAS-well ratings (Table 2). Similarly, HRQOL was weakly correlated with pain ratings (VAS-pain), with the exception of the JAQQ, whose scores were moderately correlated with the VAS-pain. The number of involved joints was only weakly correlated with SG utilities (parent and patient report), and it was basically unrelated to HRQOL measured by the PedsQL-RM, the PedsQL-GC, the JAQQ, and the VAS-health. Similarly, weak relationships were found between HRQOL and disease severity (Table 2). CHAQ scores were weakly correlated with the number of involved joints, pain, and disease severity.

Relationship between disability categories and HRQOL.

Disability as measured by the CHAQ was moderately correlated with the JAQQ, the PedsQL-RM, and the VAS-health, and weakly with the PedsQL-GC and the SG (Table 2).

We then evaluated whether HRQOL significantly differed depending on the disability category of the patients. Irrespective of whether parametric or nonparametric statistics were used, significant differences in HRQOL across all disability categories were found for the PedsQL-RM (F = 6.72, P < 0.0001, Pkw = 0.0047), the JAQQ (F = 20.44, P < 0.0001, Pkw = 0.0001), the VAS-health (F = 8.64, P < 0.0001; Pkw = 0.003), the VAS-well (F = 10.62, P < 0.0001, Pkw = 0.0004), and the VAS-pain (F = 5.36, P < 0.0004, Pkw = 0.044), but not for the SG, the PedsQL-GC, the number of involved joints, or disease severity. To further examine the differences between each of the disability categories, post-hoc testing was done and the results are summarized for the parent proxy report in Table 3. With the exception of the PedsQL-GC, the parent proxy ratings of all the other HRQOL measures significantly differed between patients without disability (CHAQ = 0) compared with patients with disability (CHAQ > 0; all P < 0.05). HRQOL as measured by the PedsQL-GC differed significantly only between patients without moderate disability (CHAQ ≤ 1.250) versus those with moderate disability (CHAQ > 1.250; P < 0.05). We expected that patients with higher disability would have lower SG utilities as compared with children without disability (Table 3). However, this was not the case, suggesting that SG utilities were not importantly influenced by patient disability. When patient self reporting was considered instead of parent proxy reporting, then similar relationships between disability categories and HRQOL were observed (Table 4).

Table 3. Parent proxy report of health-related quality of life and disability categories as measured by the Childhood Health Assessment Questionnaire
Disability level*nGlobal rating of health (VAS-health)Modified standard gamblePedsQL-GC§PedsQL-RMJAQQ#Well-being (VAS-well)**Pain (VAS-pain)††Disease severity‡‡No. involved joints§§
  • *

    Significance of differences between disability levels under consideration of multiple comparisons between groups using Tukey's post-hoc test. Analysis was repeated using Kruskal-Wallis statistics with no differences in significant comparisons (P ≤ 0.05).

  • Comparison: absent disability or mild disability versus mild/moderate disability or moderate disability; P < 0.005. VAS = visual analog scale.

  • None of the disability states is significantly different from each other.

  • §

    Comparison: absent disability or mild disability or mild/moderate disability versus moderate disability; P < 0.05. PedsQL-GC = Pediatric Quality of Life Inventory generic core scale.

  • Comparison: absent disability versus mild disability or mild/moderate disability; P < 0.007. Comparison: mild disability or mild/moderate disability versus moderate disability; P < 0.03. PedsQL-RM = Pediatric Quality of Life Inventory rheumatology module.

  • #

    Comparison: absent disability or mild disability versus mild/moderate disability; P < 0.0001. Comparison: mild/moderate disability versus moderate disability; P < 0.0001. JAQQ = Juvenile Arthritis Quality of Life Questionnaire.

  • **

    Comparison: absent disability versus mild disability or mild/moderate disability; P = 0.006. Comparison: mild disability or mild/moderate disability versus moderate disability; P < 0.0005.

  • ††

    Comparison: absent disability or mild disability or mild/moderate disability versus moderate disability; P < 0.001.

  • ‡‡

    None of the disability states is significantly different from each other.

  • §§

    Comparison: absent disability versus mild disability or mild/moderate disability or moderate disability; P < 0.001.

  • ¶¶

    Childhood Health Assessment Questionnaire (CHAQ) score = 0 using parent proxy report; median/mean CHAQ score (parent) = 0/0. IQR = Interquartile range.

  • ##

    <CHAQ score ≤ 0.25 using parent proxy report; median/mean CHAQ score (parent) = 0.25/0.20.

  • ***

    0.25 < CHAQ score ≤ 1.25 using parent proxy report; median/mean CHAQ score (parent) = 0.63/0.69.

  • †††

    1.25 < CHAQ score ≤ 2.00 using parent proxy report; median/mean CHAQ score (parent) = 1.63/1.65.

Absent¶¶45         
 Mean (SD) 8.71 (1.88)0.96 (0.09)0.81 (0.14)0.82 (0.12)0.88 (0.09)8.66 (1.07)1.40 (2.88)2.57 (2.01)2 (3.4)
 Median (IQR) 9.20 (1.42)0.95 (0.04)0.86 (0.27)0.82 (0.21)0.89 (0.11)8.60 (1.50)0.55 (1.85)2.40 (2.15)0 (3.0)
Mild##19         
 Mean (SD) 8.27 (1.19)0.88 (0.09)0.83 (0.15)0.75 (0.12)0.82 (0.09)7.57 (1.92)2.64 (1.92)3.51 (1.87)5 (7.0)
 Median (IQR) 8.70 (1.79)0.96 (0.23)0.87 (0.26)0.76 (0.23)0.81 (0.18)8.15 (2.00)1.85 (2.50)3.80 (3.90)2 (6.0)
Mild/moderate***37         
 Mean (SD) 7.20 (1.70)0.90 (0.08)0.83 (0.17)0.72 (0.11)0.73 (0.08)7.46 (1.58)3.09 (2.92)3.06 (1.89)5 (8.5)
 Median (IQR) 7.32 (3.10)0.95 (0.13)0.87 (0.19)0.75 (0.33)0.74 (0.22)7.95 (1.03)2.55 (4.20)2.5 (4.00)2 (5.0)
Moderate†††18         
 Mean (SD) 6.10 (1.74)0.92 (0.06)0.69 (0.18)0.62 (0.15)0.56 (0.09)6.37 (1.74)3.67 (2.52)3.74 (2.12)5 (5.1)
 Median (IQR) 6.80 (3.80)0.95 (0.07)0.69 (0.46)0.61 (0.19)0.58 (0.32)6.35 (3.52)4.00 (4.50)4.03 (4.17)2 (5.5)
Table 4. Patient self report of health-related quality of life and disability categories as measured by the Childhood Health Assessment Questionnaire
Disability levelnGlobal rating of health (VAS-health)Modified standard gamblePedsQL-GC§PedsQL-RMJAQQ#Well-being (VAS-well)**Pain (VAS-pain)††Disease severity‡‡No. involved joints§§
  • * See Table 3 for details on disability ratings.

  • Comparison: absent disability or mild disability versus mild/moderate disability; P < 0.0001. Comparison: mild/moderate disability versus moderate disability; P < 0.03. VAS = visual analog scale.

  • None of the disability states is significantly different from each other.

  • §

    Comparison: absent disability or mild disability or mild/moderate disability versus moderate disability; P < 0.05. PedsQL-GC = Pediatric Quality of Life Inventory generic core scale.

  • Comparison: absent disability versus mild disability or mild/moderate disability; P = not significant. Comparison: mild disability or mild/moderate disability versus moderate disability; P < 0.001. PedsQL-RM = Pediatric Quality of Life Inventory rheumatology module.

  • #

    Comparison: absent disability or mild disability versus mild/moderate disability; P < 0.0001. Comparison: mild/moderate disability versus moderate disability; P < 0.0001. JAQQ = Juvenile Arthritis Quality of Life Questionnaire.

  • **

    Comparison: absent disability versus mild disability; P = not significant. Comparison: mild disability versus mild/moderate disability; P < 0.001.

  • Comparison: mild/moderate disability versus moderate disability; P < 0.0001.

  • ††

    Comparison: absent disability or mild disability versus mild/moderate disability or moderate disability; P < 0.001.

  • ‡‡

    None of the disability states is significantly different from each other.

  • §§

    Comparison: absent disability versus mild disability or mild/moderate disability or moderate disability; P < 0.001.

Absent32         
 Mean (SD) 8.84 (1.49)0.92 (0.10)0.83 (0.13)0.84 (0.11)0.90 (0.12)8.11 (1.55)1.32 (0.35)2.89 (2.12)1.7 (5.7)
 Median (IQR) 9.00 (1.75)0.95 (0.09)0.88 (0.14)0.85 (0.07)0.94 (0.15)8.10 (1.45)1.00 (1.50)2.55 (3.75)0 (3.0)
Mild16         
 Mean (SD) 8.49 (1.08)0.93 (0.07)0.75 (0.11)0.80 (0.11)0.86 (0.08)7.93 (1.19)2.24 (0.41)3.13 (1.57)5.6 (4.5)
 Median (IQR) 8.65 (1.50)0.95 (0.08)0.79 (0.23)0.85 (0.20)0.90 (0.12)8.00 (1.75)1.70 (2.60)2.80 (1.46)3 (5.5)
Mild/moderate28         
 Mean (SD) 6.57 (1.51)0.89 (0.09)0.80 (0.15)0.78 (0.10)0.70 (0.11)7.24 (1.56)3.17 (0.39)3.04 (2.15)5.1 (5.7)
 Median (IQR) 6.87 (1.93)0.95 (0.17)0.83 (0.32)0.82 (0.22)0.74 (0.26)7.05 (3.44)2.85 (4.00)2.40 (4.41)2 (4.0)
Moderate11         
 Mean (SD) 4.75 (1.92)0.95 (0.15)0.68 (0.18)0.58 (0.14)0.42 (0.12)6.23 (2.03)3.87 (0.83)2.74 (2.65)0.5 (7.6)
 Median (IQR) 4.78 (0.33)0.94 (0.20)0.70 (0.17)0.57 (0.23)0.40 (0.13)5.24 (2.27)4.00 (2.80)2.80 (4.50)0 (0)

Quality of parent proxy report: agreement and consistency.

Irrespective of the HRQOL measure used, there was fair to good agreement between parent proxy report and patient self report of HRQOL (Table 5). Similarly, there was fair to good agreement between patient and parent ratings regarding disability (CHAQ) and well-being (VAS-well), but only poor agreement for the assessment of pain (VAS-pain).

Table 5. Agreement and consistency of parent proxy report with patient self report*
 Global rating of health (VAS-health)Modified standard gambleJAQQPedsQL-GCPedsQL-RMCHAQWell-being (VAS-well)Pain (VAS-pain)
  • *

    VAS = visual analog scale; JAQQ = Juvenile Arthritis Quality of Life Questionnaire; PedsQL-GC = Pediatric Quality of Life Inventory generic core scale; PedsQL-RM = Pediatric Quality of Life Inventory rheumatology module; CHAQ = Childhood Health Assessment Questionnaire.

  • Data from visit 1 only. Interpretation of statistically significant Spearman correlation coefficients: no consistency: rs < 0.2; weak consistency: 0.2 ≤ rs < 0.4; moderate consistency: 0.4 ≤ rs < 0.6; strong-moderate consistency: 0.6 ≤ rs < 0.8; and strong consistency: rs ≥ 0.8.

  • Data from visit 1 and visit 2. Interpretation of intraclass correlation coefficient (ICC) values: poor agreement ICC < 0.4; fair to good agreement 0.40 ≤ ICC < 0.75; excellent agreement ICC ≥ 0.75.

Spearman correlation coefficient0.600.540.650.460.670.540.550.30
P-value< 0.0001< 0.005< 0.0001< 0.005< 0.0001< 0.0001< 0.0001< 0.01
Intraclass correlation coefficients0.530.350.690.480.570.510.470.26

The consistency of HRQOL reporting between parents and patients was moderate when using the SG, the PedsQL-RM, and the PedsQL-GC and high for the JAQQ and the VAS-health. Parent proxy reports of pain (VAS-pain) were only weakly consistent with self reported pain, whereas there was moderate consistency when using the CHAQ and the VAS-well.

Age- and domain-specific differences in relationship of patient self report and parent proxy report.

Specific HRQOL domains are assessed by the PedsQL-GC, the PedsQL-RM, and the JAQQ. There was fair to good agreement and moderate consistency for all health domains with the exception of poor agreement and fair consistency for emotional functioning (PedsQL-GC: rs = 0.37, ICC = 0.36), worry (PedsQL-RM worry domain: rs = 0.31, ICC = 0.31), and communication domains (PedsQL-RM communication domain: rs = 0.37, ICC = 0.36). Although the impact of patient age on the agreement and consistency between patient and parent report for these outcomes was examined, no uniform relationship across all HRQOL measures was identified (Figure 1). Patient agreement and consistency appeared to significantly decrease with increasing patient age when using the SG and the VAS-health, whereas they increased when using the PedsQL-GC.

Figure 1.

Age-dependent agreement between parent proxy rating and patient self report of health-related quality of life. PedsQL-GCS = Pediatric Quality of Life Inventory generic core scale; PedsQL-RM = Pediatric Quality of Life Inventory rheumatology module; JAQQ = Juvenile Arthritis Quality of Life Questionnaire; VAS = visual analog scale.

DISCUSSION

Health is the result of a person's functioning and disability. HRQOL constitutes the valuation of a certain health state. Various approaches to measuring the HRQOL of children with chronic arthritis have been developed, but there is no direct comparison available between the estimates derived by different types of HRQOL measures. This study provides such comparison values and also supports that disability is associated with significantly decreased HRQOL in childhood arthritis. As for disability, parent ratings of patient HRQOL are generally in fair to good agreement with those of their children, but parents are less capable of judging pain, emotional functions, and psychosocial functioning of their children.

Despite the differences in measurement approaches and the fact that somewhat different constructs are being assessed, HRQOL measured by the JAQQ, the VAS-health, the PedsQL-GC, and the PedsQL-RM were surprisingly similar on a group level when patient self ratings are considered, whereas differences between parent ratings were somewhat more pronounced. In keeping with similar studies in adults, affected subjects (e.g., children) tended to rate their HRQOL somewhat higher compared with healthy subjects (e.g., parents) (38). However, these differences were very small and absent when using the SG. The mean scores of the PedsQL-RM and the PedsQL-GC measured in this study are similar to those previously reported in other cohorts of children with chronic arthritis (8). We confirm the previously reported very high correlation between the PedsQL-RM and the PedsQL-GC scores (8). Our data support that compared with HRQOL measured by the PedsQL-GC, PedsQL-RM scores are more strongly correlated with those of global measures of health, well-being, and disease severity; the JAQQ; and the SG. This suggests that HRQOL measured by the PedsQL-RM may adequately capture relevant HRQOL domains of children with chronic arthritis, making the suggested coadministration of the PedsQL-GC with the PedsQL-RM unnecessary, at least is some instances.

In initial studies that used the SG in patients with chronic arthritis, utility ratings appeared strongly influenced by risk aversion (13, 39). It has been hypothesized that risk aversion may dominate the utility measurement, making the valuation of the actual health state difficult (40). Therefore, we modified the SG, and the utilities derived by this modified SG were less skewed compared with traditional SG utilities, suggesting that there was less risk aversion. However, the modified SG utilities remained weakly correlated with HRQOL estimates using other measures. This finding is in keeping with previous studies in adults (40, 41) and children (14, 42) and is likely due to the fact that a different construct is being measured. Only the SG, but none of the other tested HRQOL tools, considers risk taking and also the burden of being in a certain health state for a prolonged period of time in the future. Thus a utility incorporates the value a certain health state both at the present time, as is it done by the other HRQOL measures, and its potential future impact. Exploratory analyses (data not shown) support that the SG is unlikely to have sufficient responsiveness to be used for the longitudinal assessment of day-to-day changes in the health state of children with chronic arthritis (response to treatments) (42).

The JAQQ and CHAQ scores of the studied families had similar correlations with pain and well-being as was reported in other cohorts (16, 21). However, we did not observe the previously noted moderate to strong correlations of the JAQQ or the CHAQ scores with the number of involved joints. As such, the number of involved joints was only weakly correlated with the CHAQ and the JAQQ. We hypothesize that this is because, with the help of the new biologic agents, the disease symptoms of the studied patients were better controlled (median number of involved joints = 1) than those of the previously examined cohorts.

Our data show that similar to adult arthritis patients, increasing disability is associated with significantly decreased HRQOL (43, 44). Although one would intuitively agree with this finding, we are not aware of any formal comparison of the relationship between disability categories and HRQOL in children with chronic arthritis. Thus, our results support previous suggestions that important changes in patient disability could be used for determining appropriate sample sizes for future trials of medications in childhood arthritis (19) because different levels of disability are associated with significant differences in HRQOL.

The quality of parent proxy reporting was examined in detail. Parents were generally acceptable proxy reporters of well-being, disability, and HRQOL, but their ratings differed more substantially when using the SG or estimating pain. The results of our study support previous results of high–moderate consistency (16, 17) and fair–good agreement (45) between parent and patient report when using the CHAQ. Patient and parent ratings were most closely related for the JAQQ, the PedsQL-RM, and the VAS-health. This confirms previous observations made for the JAQQ (23), the PedsQL questionnaires (8), the VAS-health (18), and the SG (13). Our results support previous observations that parents have special difficulties in judging the psychosocial and emotional health of their children with chronic arthritis (1, 8, 23).

Different from other researchers (14), and although the SG was only completed by children age 12 years and older, we did not find that parent/patient agreement increases with patient age when using the SG. We hypothesize that the younger children were still challenged by the difficulty of the task, thus providing utilities that were not necessarily governed by their health state. Among the outcomes measured, the ratings of pain (VAS-pain, hurt and pain domain of the PedsQL-RM) had the highest correlations with the SG utilities for both parents and patients, suggesting that SG utilities may be importantly influenced by the degree or perception of pain. Although differences in utilities between parents and patients may be due to differing attitudes toward risk taking (14), we hypothesize that the differences in pain ratings between parents and patients observed by us and others (11) contribute to the poorer agreement between parents and patients when using the SG. Further studies are required to assess the relationship of pain and utilities and the test–retest reliability of the modified SG in clinical practice.

A limitation of our study may be that not all currently available tools for children with chronic arthritis were tested. We chose among the tools that are commonly used for studies of chronic arthritis and whose measurement properties had been previously tested. We cannot exclude that there may be HRQOL tools that have better measurement qualities. However, given that the study was performed during routine clinic visits, the testing of additional tools would not have been feasible.

In addition, the disease severity scale used in this study has not been formally validated; however, this traditional outcome measure was included to serve as a comparison to HRQOL measures only. There are few studies using the SG in children (14, 46–49) in general and only 1 in those with arthritis (13). Because preference-based HRQOL measurement is still under development, the measurement properties of the SG when completed by children or by parent proxy reporters have not been well examined. The same is true for the modification of the SG used in this study. Thus, future research is required to examine whether the weak correlations of the SG with other HRQOL measures are due to, for example, poor test–retest reliability of the SG or the fact that somewhat different constructs are being assessed.

The measurement of HRQOL is important for children with chronic arthritis and enhances the basis upon which treatment decisions can be made. The choice of the HRQOL tool will be influenced by the type of health-related issue to be assessed for a certain medical decision or treatment. Research on the child's mental capacity to anticipate the effects of health decisions is necessary before patient perceptions of health can be used for important health decisions with confidence. Although parents don't completely agree in their perception of health with their child diagnosed with chronic arthritis, they are acceptable proxy reporters whose HRQOL reports can provide important health information that is not captured by traditional medical outcomes alone.

Acknowledgements

We thank Bin Huang, MSc, PhD for the review of the statistical methods of the manuscript.

Ancillary