To examine the degree to which nonmedical factors explain additional variance in parent proxy report and child self-report of health-related quality of life (HRQOL) among newly diagnosed children with juvenile idiopathic arthritis (JIA) after accounting for medical factors.
Parents (of children ages 2–16 years; n = 230) and patients (ages >5 years; n = 180) diagnosed within the previous 6 months completed surveys to assess medical (clinical parameters and functional status) and nonmedical (self-efficacy, coping, barriers to adherence, social support, parental distress, and access to care) factors and HRQOL (Pediatric Quality of Life Inventory Generic Core Scales). Physician-rated global assessment of disease activity, active joint count, and select laboratory variables (rheumatoid factor, antinuclear antibodies, and erythrocyte sedimentation rate) were recorded.
Nonmedical factors, including self-efficacy, coping with pain, barriers to adherence, social support, and parental distress, explained additional variance in HRQOL total, physical functioning, and psychosocial functioning scales (R2 increases of 6%, 1%, and 13% for parent proxy report and 16%, 7%, and 30% for self-report, respectively). Parental distress was uniquely associated with parent proxy-report HRQOL, while child self-efficacy and social support were uniquely associated with self-report HRQOL.
Nonmedical factors are associated with HRQOL in newly diagnosed patients with JIA after accounting for medical variables, particularly for psychosocial functioning.
Health-related quality of life (HRQOL), an individual's perception of their physical, psychosocial, and role functioning with respect to their health, is a key outcome in clinical care and clinical trials for children with juvenile idiopathic arthritis (JIA), the most common pediatric autoimmune disease affecting the musculoskeletal organ system. The importance of HRQOL as a primary outcome is echoed by the Food and Drug Administration ([1, 2]), the Centers for Disease Control and Prevention (), and the World Health Organization (). Implicit in the measurement of HRQOL as a JIA outcome is the notion that medical interventions, such as drug therapies, can affect not only clinical parameters (pain, joint count, and functional status), but also more distal outcomes, such as HRQOL.
Early research examined the etiologic role of psychosocial factors in juvenile arthritis (). In the mid-1980s, Varni and colleagues were among the first to document the effect of clinical and psychosocial factors on children's adjustment to juvenile arthritis and to put forth a conceptual framework linking both clinical and psychosocial factors to HRQOL ([6, 7]). Wilson and Cleary () further specified a causal pathway from physiologic variables to clinical status to functional outcomes and thus to HRQOL. Figure 1, which is based on these models, posits a causal cascade from biologic and physiologic variables to clinical status and functional status to HRQOL. Figure 1 also suggests a causal role, both direct and indirect, of nonmedical variables on HRQOL and, potentially, an expanded set of targets for clinical intervention.
Although much research has focused on understanding how biologic and physiologic variables affect clinical outcomes, far less is known about how clinical parameters are related to HRQOL in JIA patients and how characteristics of the child, family, or environment affect HRQOL in the context of ongoing treatment. Lack of such knowledge is an important problem. Researchers have documented worse HRQOL for children with JIA compared to healthy controls ([9-11]) and have found that, even with excellent disease control through the use of biologic medications, approximately 50% of children with polyarticular JIA continue to have lower HRQOL than healthy children (). This finding suggests that without a thorough understanding of the determinants of HRQOL and identification of potential modifiers, many children with JIA, even when treated with biologic agents and despite well-controlled arthritis, will continue to experience suboptimal HRQOL.
Given the growing understanding of HRQOL as an important outcome, the traditional focus of therapeutics on medical outcomes, and the relative lack of knowledge regarding how clinical parameters and therapeutic interventions affect HRQOL in children with JIA, there is a need to understand parent and child perceptions of HRQOL and their relationship to medical and nonmedical factors. We therefore examined the degree to which medical variables (biologic, physiologic, clinical, and physical function) and nonmedical characteristics of the child (coping, self-efficacy, social support, and adherence), family (parental distress and family climate), and environment (socioeconomic status and access to care) relate to HRQOL in newly diagnosed children with JIA. We hypothesized that both medical and nonmedical variables will be significantly associated with patient HRQOL.
Box 1. Significance & Innovations
This study examined the degree to which medical and nonmedical variables explain variance in health-related quality of life (HRQOL) in children newly diagnosed with juvenile idiopathic arthritis (JIA).
The results suggest that nonmedical variables, i.e., parent emotional distress, social support, and self-efficacy, are important explanatory variables in HRQOL for these children, particularly for psychosocial functioning.
These results can be used to develop clinical interventions to help improve HRQOL in JIA.
PATIENTS AND METHODS
In this study, we considered reports from children with JIA and their primary caregivers. Eligible index patients were male or female patients ages 2–16 years of any race or ethnicity who had been diagnosed within the last 6 months with any type of JIA as per the International League of Associations for Rheumatology criteria (), irrespective of the specific JIA category. Patients who carried the diagnosis of a medical condition that would otherwise severely impair their HRQOL (e.g., cerebral palsy, spina bifida, severe mental retardation, and fibromyalgia) were excluded from study participation. All participants gave assent or informed consent to participate in the study.
The study was a prospective, new-onset patient cohort study design. Two centers were involved in this study: Cincinnati Children's Hospital Rheumatology Clinic and the University of Louisville Pediatric Rheumatology Clinic. Participants were recruited from October 2008 through August 2012. During the study period, consecutive patients eligible for the study were approached as soon as a diagnosis of JIA was established. The study was reviewed and approved by the Institutional Review Boards of each of the 2 participating centers (Cincinnati Children's Hospital Medical Center and University of Louisville).
Measures: explanatory variables
Biologic and physiologic variables
Biologic variables included age, sex, race, and disease duration with JIA. Physiologic variables considered were rheumatoid factor (RF; positive/negative), antinuclear antibodies (ANAs; positive/negative), and erythrocyte sedimentation rate (normal range 0–10 mm/hour), and JIA category (systemic JIA; polyarticular RF-negative JIA, polyarticular RF-positive JIA, extended oligoarticular JIA, oligoarticular JIA, and psoriatic JIA; and undifferentiated JIA), active joint count, and number of joints with limited range of motion.
Clinical and functional status
Clinical variables included proxy- and self-reported patient pain, measured using a visual analog scale (anchors: 0 = no pain, 10 = very severe pain), and the Pediatric Quality of Life Inventory (PedsQL) Rheumatology Module ([11, 14, 15]) pain subscale (scaled 0–100, with higher values indicating less pain). We also measured overall physician global assessment of disease activity (anchors: 0 = inactive disease, 10 = very active disease). Physical function was measured using the PedsQL Rheumatology Module ([11, 15, 16]) functioning subscale (also scaled 0–100, with higher values indicating better functioning) and the revised version of the Childhood Health Assessment Questionnaire (C-HAQ) disability index (not including aides and help items) ([16, 17]). This index is calculated as the unweighted average of 30 questions in 8 domains covering major aspects of daily living over a 1-week period and yields a score between 0 (no disability) and 3 (most severe disability).
Nonmedical characteristics of the child
Nonmedical characteristics of the child were assessed by age-appropriate questionnaires completed by the child or, for younger children, by their parents.
Coping was measured using the Waldron/Varni Pediatric Pain Coping Inventory (PPCI) (). The PPCI is a 41-item measure used to identify strategies children use to cope with pain. Respondents are asked to rate, on a 3-point response scale, how frequently coping takes the form of 1) cognitive self-instruction, 2) seek social support, 3) strive to rest and be alone, 4) cognitive refocusing, and 5) problem solving/self-efficacy. Higher scores indicate more frequent use of the particular coping strategy. The PPCI was completed by children ages ≥5 years and by parents.
Self-efficacy was measured using the Children's Arthritis Self-Efficacy Scale (CASE) (). The CASE is an 11-item scale that measures children's self-efficacy with respect to managing symptoms, emotional consequences, and activities, each with score ranges from 0–4, with higher scores indicating more self-efficacy. Although validated for children ages 7–17 years, we used it with children ages 5–17 years.
Social support was measured using the Harter Social Support Scale for Children (SSSC) (). The 21-item SSSC was designed to assess social support from parents, classmates, teachers, and friends, with each score ranging from 1–4, where higher values indicate stronger social support. The SSSC has been shown to moderate the effect of stress (daily hassles) on depression in children with rheumatic disease (). It was completed by participants ages 5–16 years.
Adherence was measured via the Medication Adherence Self-Report Inventory (MASRI). The 5-item MASRI is validated for pediatric rheumatic disease () and other chronic disease (). Parents of children ages 5–11 years and patients ages 12–16 years completed the MASRI. In addition, we used the PedsQL Rheumatology Module () treatment problems subscale (range 0–100, with higher values indicating fewer barriers) to measure barriers to adherence.
Nonmedical characteristics of the family
The primary caregiver reported on nonmedical characteristics of the family. Parental distress was measured using the Symptom Checklist-10 (SCL-10), a 10-item self-report measure of psychological symptoms rated on a 5-point Likert scale (). The SCL-10 ranges between 0 and 4, with higher values indicating higher stress. Family climate was assessed via the Family Environment Scale (FES). The FES () consists of 90 items rated as “true” or “false” to assess family climate. In this study, the Family Relationship Index subscore was used due to its extensive use in previous research (). The FES yields a standardized T score centered at 50 with an SD of 10.
Nonmedical characteristics of the socioeconomic and financial environment
To measure socioeconomic status, the mother's education level, zip code, and insurance type (private or publicly financed) were used as proxies. Financial access to care was measured by parents' report of whether their child had health insurance and parents' report of the degree to which cost was a problem (i.e., a big problem, a small problem, or not a problem) (). Realized access was measured through parents' reports of foregone care (), i.e., any time when the child should get medical care, but did not.
Response variable: HRQOL
HRQOL was measured using the PedsQL Generic Core Scales (). The PedsQL consists of parallel forms for children (ages 5–18 years) to report on their own HRQOL and for parents of children ages 2–18 years to report on their child's HRQOL. We measured the total scale, physical functioning subscale, and psychosocial functioning subscale scores. Scale scores ranged from 0–100, with higher scores indicating better HRQOL.
Upon consent, the participants ages 5–16 years completed the CASE, SSSC, PPCI, PedsQL, and PedsQL Rheumatology Module. Participants ages 12–16 years also completed the MASRI survey. Children unable to complete the survey by themselves were read the surveys by the research coordinator or a parent. Accompanying parents or guardians completed the PPCI, PedsQL, PedsQL Rheumatology Module, FES, SCL-10, C-HAQ, and the nonmedical characteristics for children ages 12–16 years and completed the MASRI for children ages 2–11 years. The biologic, physiologic, and clinical function data from the same visit were obtained from the clinical record.
The goal of the study was to examine the effect of nonmedical factors, after accounting for medical factors and functional status, on HRQOL in newly diagnosed children with JIA. The primary outcome is the PedsQL total score, and the secondary outcomes are the PedsQL physical and psychosocial subscale scores. Examination of the distributions of the study outcomes showed reasonable fit with a normal distribution assumption. We ran parallel linear regression analyses for child self-report and parent proxy-report HRQOL. In the first stage, multivariate linear regression analyses were performed for each set of explanatory variables in the conceptual framework (Figure 1) separately (i.e., biologic, physiologic, clinical, functional status, child, family, and environmental variables). From each of these sets of explanatory variables, only those that remained statistically significant were selected and entered into the next stage of analyses. The rationale for this first stage of analyses is that it allows us to identify the strongest independent variables to represent a set of potentially correlated variables within the same domain. At the second stage, the variables selected from the first-stage analyses are entered into a linear regression analysis. The biologic variables of age, sex, race, and age at JIA disease onset are always included in linear regression modeling, regardless of their statistical significance. Stepwise selection method was used to select the best subset of the variables that maximize the R2, and only the variables remaining significant at 0.05 levels are retained in the final model. Finally, the type I squared partial correlations are assessed for each domain of variables remaining in the final model, following the order of causal chain as specified in the conceptual model, i.e., biologic and physiologic domain followed by clinical and functional status, then followed by nonmedical characteristics. The analyses for the parent proxy-report PedsQL were performed both with and without parents of children ages <5 years. This was done to examine the sensitivity of the study results to the inclusion and exclusion of the younger children, and to be able to compare the results between parent proxy report and self-report for children ages ≥5 years.
Three hundred sixteen children screened eligible for the study, 262 families were approached, and 230 participated, yielding a 73% participation rate and an 88% response rate. Of the 230 families, 180 included children old enough to complete the surveys. The biologic and demographic characteristics of the sample (n = 230) are shown in Table 1. Patient age ranged from 2–16 years and was distributed fairly consistently. The patients were predominantly girls (69.1%) and white (92.6%). Table 1 also displays the JIA subtype and physiologic variables. The polyarticular RF-negative and oligoarticular subtypes were the most common, and most patients in the study were ANA (67.8%) and RF (92.6%) negative. Baseline clinical, functional status, HRQOL, individual, and family characteristics are shown in Table 2.
Table 1. Biologic, demographic, JIA subtype, and physiologic characteristics of the sample*
Values are the number (percentage) unless indicated otherwise. JIA = juvenile idiopathic arthritis; RF = rheumatoid factor; ESR = erythrocyte sedimentation rate; ROM = range of motion.
Age, mean ± SD years
9.42 ± 4.49
Age at onset of symptoms, mean ± SD years
8.43 ± 4.41
Age group, years
Disease duration, median (Q1, Q3) days since diagnosis
57 (35, 105)
Days between symptoms and diagnosis, median (Q1, Q3)
126 (65, 328)
Polyarticular RF positive
Polyarticular RF negative
Less than college
College or more
Both private and public insurance
Problem with costs of care
ESR, median (Q1, Q3) mm/hour
11.5 (7.0, 22.5)
Active joint count, median (Q1, Q3)
2 (0, 5)
No. of joints with limited ROM, median (Q1, Q3)
1 (0, 4)
Table 2. Clinical, functional status, health-related quality of life, individual, and family characteristics of the sample*
Values are the mean ± SD. VAS = visual analog scale; PedsQL = Pediatric Quality of Life Inventory; C-HAQ = Childhood Health Assessment Questionnaire; CASE = Children's Arthritis Self-Efficacy Scale; PPCI = Pediatric Pain Coping Inventory; MASRI = Medication Adherence Self-Report Inventory; SSSC = Social Support Scale for Children; SLC-10 = Symptom Checklist-10; FES = Family Environment Scale. † Proxy report is by a parent except as noted.
aSelf-report is by patients ages ≥5 years.
bMASRI is by self-report for patients ages ≥12 years and by proxy report for patients ages <12 years. Only item E is used.
Table 3 shows the regression coefficients for the variables that significantly predicted parent proxy-report and child self-report HRQOL: Generic Core total scale, physical subscale, and psychosocial subscale. Pain, activity limitations, coping with pain, barriers to adherence, and parental emotional distress predicted parent proxy-report HRQOL. Pain, functional status, self-efficacy, coping with pain, barriers to adherence, and social support predicted patient self-report HRQOL.
Table 3. Regression beta coefficients for parameters predicting health-related quality of life, adjusted for age, sex, race, age at onset, and juvenile idiopathic arthritis subtype*
Each column represents separate models for the corresponding outcomes specified by the column header. SR = self-report; PedsQL = Pediatric Quality of Life Inventory; PR = proxy report; C-HAQ = Childhood Health Assessment Questionnaire.
Table 4 shows the percentage of variance (R2) accounted for by each set of explanatory variables after accounting for the domain of variables entered earlier, where the ordering of domains is based on the conceptual model (Figure 1), predicting parent proxy-report and child self-report HRQOL: Generic Core total scale, physical subscale, and psychosocial subscale. These models include biologic variables and JIA subtype, as well as the variables from the domain-specific analyses that remained significant in the final model. Collectively, biologic and physiologic variables, clinical and functional status, and nonmedical characteristics explained 70% and 75% of variance in parent proxy-report and child self-report HRQOL. After adjusting for biologic and physiologic domains, clinical and functional status explained 44% of proxy-report and 35% of self-report total score for HRQOL. After adjusting for biologic and physiologic variables and clinical and functional status, nonmedical characteristics explained only 7% of proxy-report total HRQOL, but 30% of self-report total HRQOL. This is reflected in the relatively larger proportion of variance explained by nonmedical characteristics for self-report (30%) versus proxy-report (13%) psychosocial HRQOL. The same results hold true when excluding children ages <5 years.
Table 4. Total R2 from the final model and the type I square partial correlation corresponding to each domain*
Results from the final model are shown in Table 3.
aVariables are entered in sequential order according to the theoretical framework shown in Figure 1. The square partial correlation for each domain is adjusted for the preceding domain of variables.
Biologic and physiologic
Clinical and functional status
We examined medical (biologic, physiologic, clinical, and functional status) and nonmedical (individual, family, and environmental) explanatory variables of HRQOL in a cohort of children newly diagnosed with JIA. Consistent with earlier conceptual frameworks and our hypotheses, the results suggest that nonmedical factors are significant explanatory variables of HRQOL and that parent proxy-report and child self-report HRQOL explanatory variables overlap incompletely.
Our first hypothesis was that nonmedical factors are significant explanatory variables for HRQOL. We tested this by stepwise inclusion of sets of potential explanatory variables in regression models to predict HRQOL. Nonmedical factors did explain additional variance in overall HRQOL after accounting for medical variables, more so for self-reported HRQOL (an additional 30% of explained variance) than for parent proxy-reported HRQOL (an additional 7%). In the subdomain of HRQOL related to psychosocial functioning, the size and difference were even more striking: nonmedical variables accounted for an additional 30% of self-report variance and 13% of proxy-report variance in psychosocial functioning. In contrast, when all sets of potential explanatory variables were entered, no biologic or physiologic variables explained significant variance, nor did physician ratings of global disease assessment. Explaining the majority (e.g., 70–75%) of variance in a construct such as HRQOL is quite unusual and corresponds to a very large effect size ().
Our second hypothesis was that the set of explanatory variables significant for parent proxy-report and child self-report HRQOL would overlap incompletely. We found support for this hypothesis as well. For the PedsQL total scale score, for example, reports of pain and functioning explained both proxy and self-report, but proxy-report HRQOL was explained additionally by parent emotional distress, while self-report HRQOL was predicted strongly by classmate and parent social support, as well as self-efficacy.
Together, these results complement emerging literature on the relationship between clinical factors and HRQOL in children with JIA. Seid et al () showed significant variation in HRQOL despite excellent symptom control, and Haverman et al () showed that perceived difficulty with treatment regimen and missed school were additional predictors of HRQOL. This study adds to the literature by demonstrating that psychosocial factors such as parental distress and children's perception of social support are strong predictors of HRQOL in newly diagnosed patients. Of note, the point change estimate for the PedsQL total scale score associated with a 1-point difference in classmate (5.2) and parent (4.4) social support and with parent emotional distress (−4.3) is similar to the minimum important difference of the PedsQL total scale (∼4.5) (). This suggests that changes in social support and parent emotional distress are related to changes in HRQOL that are not only statistically significant, but also clinically important.
This study has some limitations. We sampled patients from only 2 clinical centers, limiting our ability to generalize. This is a cross-sectional analysis of baseline data from a cohort of new-onset patients. The results are applicable to the early stage of the new-onset patients. It is not clear whether the results will hold the same at later stages of the disease course. The small variance contributed by nonmedical variables after accounting for medical variables could suggest a mediated pathway, as shown in our conceptual model. However, these data are cross-sectional and the casual direction of the effect is unclear. The fact that parent-report explanatory variables and self-report explanatory variables were significantly related to proxy-report and self-report HRQOL, respectively, raises the possibility of a rater effect, i.e., that the relationship found is due to the same person completing the surveys rather than to whether the constructs are related. In addition, we used simple linear regression and did not account either for higher-order regression or for mediating or moderating variables. Future research should sample additional care centers, examine longitudinal effects, and explore the role of mediating and moderating variables.
Nevertheless, this study improves our understanding of possible determinants of HRQOL in children with JIA and raises the possibility of additional intervention targets for clinicians to consider.
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 published. Dr. Seid 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. Seid, Brunner, Lovell.
Acquisition of data. Seid, Niehaus, Brunner, Lovell.
Analysis and interpretation of data. Seid, Huang, Niehaus, Lovell.