|Home | About | Journals | Submit | Contact Us | Français|
Self-versus proxy- perspectives may produce different results that are important for clinical decision-making and for assessing outcomes in research studies. We examined differences in child versus parent-report of the child's health-related quality of life (HRQOL) in a large prospective, community-based study of newly diagnosed childhood epilepsy that included children with epilepsy (case) and sibling controls.
HRQOL was assessed 8–9 years after initial diagnosis of epilepsy in a subset of 143 case-control matched pairs using the Child Health Questionnaire (CHQ), a generic HRQOL measure with child (CHQ-CF87) and parent (CHQ-PF50) versions.
There were no significant differences between self-reported case and sibling control HRQOL scores on 9 of 11 scales or 2 global-items. However, parent ratings were significantly better (higher HRQOL) for sibling controls compared to epilepsy cases on 10 of 12 scales, global behavior and general health items, and the physical and psychosocial summary scores (p≤0.05). Parent-child agreement was low for cases and controls (kappa 0.27–0.33) for three single-item questions with the same wording on parent and child versions. Parent ratings of the case’s HRQOL were often significantly associated with 5-year remission status and current anti-epileptic drug use, but the case’s self-reported HRQOL scores were not. In contrast, current pharmacoresistance was often associated with the child and parent ratings of the child’s HRQOL.
Children with epilepsy report HRQOL that is comparable to that of sibling controls while parents rate children with epilepsy as having lower HRQOL than sibling controls. Measuring outcomes in studies of this population should incorporate both perspectives.
Epilepsy is one of the most common chronic neurological diseases, affecting approximately 1% of the population . Approximately one-half of epilepsy in the population first occurs during childhood. Children with epilepsy face many challenges including but not limited to seizures; they often have learning, cognitive and school difficulties, medication side effects, social stigma, psychiatric comorbidity, and behavioral problems [2–12]. Given these challenges it is important to measure health-related quality of life (HRQOL) in children with epilepsy. HRQOL assessments in children, however, are complex; while ideally elicited directly, proxy ratings are often used to assess HRQOL in young children. Children, however, may have different perspectives from that of their adult proxies. Identification of such potential differences is important for assessing outcomes in research studies and for developing comprehensive treatment programs for children and families with epilepsy.
While a number of studies have evaluated HRQOL in children with epilepsy, the source of report for HRQOL often varies [13–18]. Modi et al. reported that children with a single seizure, or with newly diagnosed untreated epilepsy, had significantly worse Pediatric Quality of Life Inventory (PedsQL) scores compared to general population norms . Sabaz et al. reported that children with “symptomatic” epilepsy have worse HRQOL than those with idiopathic epilepsy using the Child Behavior Checklist (CBCL), the Child Health Questionnaire (CHQ), the Quality of life in Childhood Epilepsy (QOLCE), and the Child Seizure Profile (CSP) . In both of these studies, HRQOL assessments were elicited by parent-report only. Austin et al. showed that youth with epilepsy had worse HRQOL than youth with asthma; psychological, social, and school domains of QOL were measured using five different instruments with child, parent, or teacher report . HRQOL for children with epilepsy has also been assessed by child-report directly; studies by Devinsky et al. and Benavente-Aguilar et al. assessed self-report of HRQOL in adolescents with epilepsy using the QOLIE-AD48 [15, 16].
Few studies, however, have compared child and parent report of HRQOL in children with epilepsy. Verhey et al. compared parent and child-report of HRQOL using the CHEQOL-25, an epilepsy-targeted HRQOL measure . Disagreement was observed between child and parent reports for subjective domains of HRQOL (secrecy and present worries); parent-report of HRQOL was lower, or worse, on all domains compared to child-report, with the exception of one domain, present worries. Similarly, in a sample of children with first-time seizures, Hamiwka et al. found that compared to normative data, parent-report of HRQOL was lower on 6 out of 11 scales and 1 summary score on the Child Health Questionnaire (CHQ), while children reported lower HRQOL on only the change in health item . Additionally, in a sample of adolescents with epilepsy, Huberty et al. compared parent, child, and teacher behavioral ratings using the Child Behavior Checklist (CBCL) and found that mother’s evaluations of their child’s behavioral problems were significantly higher (or worse) than the youth’s evaluations on 8 of 11 subscales . Finally, although parent-child agreement was high (r≥0.5) in a sample of epilepsy patients using the German version of the PedsQL, the mean responses of child-report of HRQOL were higher, or better, compared to parent-report on all scales with the exception of a school functioning domain . These studies provide some preliminary support for the hypothesis that children and parents report HRQOL differently in children with epilepsy.
Understanding whether child versus parent perspectives produce different results is important for clinical decision-making and for planning research studies that focus on these outcomes, particularly in children with epilepsy. Furthermore, as children with epilepsy, or other chronic diseases, become adults, recognition of these potential differences may be important for identifying those with special health needs requiring transition support . We examined parent versus child-report of HRQOL in a group of children with epilepsy and sibling controls to parent-report of HRQOL in these children at long-term follow-up in a prospective, community-based cohort study of newly diagnosed epilepsy.
The Connecticut Study of Epilepsy is a prospective, community-based cohort study that recruited 613 children with newly-diagnosed epilepsy between 1993 and 1997 with ongoing prospective follow-up; details of methods, recruitment, and follow-up procedures have been published previously . Initial enrollment inclusion criteria included: 1) initial diagnosis of epilepsy by participating physicians (or diagnosis by a non-participating physician with referral to a participating physician within 3 months of diagnosis) during the recruitment period (January 1993 –December 1997), and 2) first of two or more unprovoked seizures occurred between 1 month (28 days) through 15 years of age.
Eight to nine years after initial enrollment (March 2002-February 2006), 311 subjects from the original cohort and family members of those subjects participated in a comprehensive reassessment protocol; siblings without a neurological disability, within 3 years of the patient’s age, and less than 18 years of age were also enrolled as controls . The protocol included an assessment of HRQOL reported by the child with epilepsy (case) and a sibling control, and a parent’s assessment of both the case and the sibling control’s HRQOL.
Institutional Review Board approval was obtained at all sites (G07-07–094—UCLA). Written informed consent was obtained from the parent and written assent from the child, at the time of initial enrollment and at the 8 to 9 year reassessment.
Clinical and demographic data of cases and parents were obtained via a structured in-person interview (clinic or subject’s home) with the parent or guardian interviewed by trained research associates during initial enrollment . This interview included details of the child’s unprovoked or provoked seizures, medical, and developmental history. All baseline medical records were reviewed and information including neurological examination, presence of mental retardation, neuroimaging findings, specific underlying etiology, history of prior provoked seizures, presence of other medical conditions, and EEG findings were extracted and coded on an ongoing basis. Families were called every three to four months by trained research associates and asked whether their child had experienced any interval seizures and information regarding the dates, types of seizures, and circumstances of any reported seizure, in addition to any changes in anti-epileptic medications were ascertained. Patients and families were asked to keep seizure logs. At the 8–9 year follow-up comprehensive reassessment, clinical and demographic data of cases, parents, and sibling controls were obtained via a structured interview in-person (in clinic or subject’s home), or via phone.
HRQOL was assessed using the Child Health Questionnaire (CHQ), a self-administered, generic HRQOL measure with child (CHQ-CF87) and parent (CHQ-PF50) versions . The CHQ-CF87 includes 87 items used to create 11 scales (physical function, role/social limitations–physical, role/social limitations–behavioral, role/social limitations–emotional, mental health, self-esteem, behavior, bodily pain/discomfort, general health perceptions, family activities), and 2 global items (global behavior and global general health). The CHQ-PF50 includes 50 items used to construct 12 scales (physical function, role/social limitations–physical, role/social limitations–emotional/behavioral, mental health, self-esteem, behavior, bodily pain/discomfort, general health perceptions, family activities, parent impact on time, parent emotional impact), and 2 global items (global behavior and global general health); two summary scores (physical and a psychosocial summary scores) can also be generated from the CHQ-PF50.
Raw scores for each scale are transformed to a 0–100 scale with 0 representing worst health and 100 indicating best health. Physical and psychosocial summary scores (CHQ-PF50 only) are transformed to T-scores (mean = 50, SD = 10) calculated against a reference general U.S. population. The CHQ-CF87 and the CHQ-PF50 contain similar but not identical items, making direct score comparisons problematic except for three single-items that have similar wording.
To compare construct validity across case and parent versions, we examined associations of the case and parent-reported HRQOL measures with whether or not the child had been seizure-free for at least 5-years at the time of the follow-up reassessment; whether or not the child was currently taking anti-epileptic drugs (AEDs); whether the child’s epilepsy was associated with a known structural or metabolic underlying cause, or an electroclinical syndrome considered to be an epileptic encephalopathy, designated here as “complicated epilepsy” ; and whether or not the child’s epilepsy was pharmacoresistant and the child was not seizure free for the last 5 years. Pharmacoresistance was defined as the failure of two different appropriate AEDs to bring seizures under complete control when used as prescribed and pushed to the maximum tolerated levels . This definition is consistent with that proposed by the Task Force of the International League Against Epilepsy (ILAE) Commission on Therapeutic Strategies .
Sociodemographic and clinical characteristics of cases and controls and parents were compared using bivariate statistics (matched t-tests, Wilcoxon rank sum, chi-square). We estimated internal consistency reliability (Cronbach’s alpha) for each scale of the CHQ-CF87 and the CHQ-PF50 for cases and controls.
Because direct comparisons between child and parent CHQ versions were problematic, we compared self-reported HRQOL in cases versus sibling controls on the CHQ-CF87 and then parent-reported HRQOL in cases versus sibling controls on the CHQ-PF50. Only subjects for whom HRQOL was evaluated for both the case and the sibling control, and for whom HRQOL of the case and control were rated by the same parent (n=143) were included in this matched analysis. Mean differences between cases and controls (child-report and parent-report) were calculated. We compared case versus control self-reported HRQOL means for the 11 scales and 2 global items of the CHQ-CF87 using matched, or paired t-tests. Similar analyses were performed to compare parent-report of case versus control HRQOL on the 12 scales, 2 global items, and 2 summary scores of the CHQ-PF50. A sensitivity analysis was performed excluding the 29 complicated cases, generally associated with more severe neurological impairment. To compare HRQOL responses to a reference healthy standard population, parent-report of case and control mean HRQOL responses (CHQ-PF50) were compared to age-adjusted and gender-adjusted norms, using t-tests . (Child-reported norms from a representative U.S. population of healthy children are not available, thereby precluding a similar analysis for the CHQ-CF87).
To assess the level of parent-child agreement, we calculated weighted kappas between case and parent-report of HRQOL between parent and child on the single-items that had the same wording (bodily pain/discomfort, global general health, and global behavior) .
We used multitrait-multimethod matrix (MTMM) to assess the construct validity for case and control HRQOL by comparing child (CHQ-CF87) and parent (CHQ-PF50) reports for the nine scales for which the child and parent versions have common constructs or domains [30, 31]. Correlations between different methods (CHQ-CF87 versus CHQ-PF50) should be higher if they are assessing the same HRQOL domain compared to when they are assessing different domains.
We assessed construct validity of the child and parent versions of CHQ for epilepsy by evaluating child and parent-report of HRQOL for cases with more severe disease compared to those with less severe disease as reflected by 5-year remission status, current AED use, having complicated epilepsy or not, and whether or not the epilepsy met criteria for pharmacoresistant and the child had not been seizure free for the prior five years. Only study participants whom the HRQOL was evaluated for both the case and the parent were included in this analysis (n=279). We evaluated differences in mean HRQOL scores for cases that were seizure free versus not seizure free, taking AEDs versus not AEDs, complicated versus not complicated, and were currently pharmacoresistant versus not pharmacoresistant using t-tests for both the child and parent-report.
Analyses were performed using Stata (9.2), SAS (9.1), and the MTMM.exe program . We used an a priori p-value of p ≤ 0.05 for statistical significance.
Sociodemographic and clinical characteristics for children, sibling controls, and parents are shown in Table 1. The CHQ was completed for both the case and the parent, but with no sibling control in 279 dyads. For 143 cases, the CHQ was also completed by a sibling control and the same parent rated HRQOL in these 143 case-control pairs. There was no significant difference in the parent-child testing dates between cases and controls (p=0.59).
Cronbach’s alphas were high, ranging from 0.75 to 0.90 for CHQ-CF87 subscales (Table 2) and from 0.74 to 0.97 for CHQ-PF50 subscales (Table 3). Because the CHQ-CF87 has been recommended for ages 10 and over, we analyzed internal consistency reliability for the subset of children age 9 and under (cases n=18; controls n=28); Cronbach’s alpha for all scales except one exceeded 0.80; Cronbach’s alpha was 0.69 for one scale, general health perceptions.
There were no differences between self-reported case and control HRQOL scores on 9 of 11 scales and 2 global items of the CHQ-CF87 (Table 2, Figure 1a). Cases reported worse HRQOL than controls on 2 scales: physical function (p=0.01) and role limitations-physical (p=0.04). In contrast, parent–report of HRQOL for controls was significantly better than cases on 10 of 12 scales, 2 global items, and both summary scores of the CHQ-PF50 (all p≤0.05) (Table 3, Figure 1b). Excluding 29 complicated cases from our analyses, findings were similar except that child and sibling control self-reports of HRQOL were now no different on all 11 CHQ-CF87 scales and parent-report of case and sibling control HRQOL was still significantly better on 10 of 12 scales, both summary scores, and only 1 global item (global general health).
Compared to age-adjusted norms, parent-report of case HRQOL was significantly lower for only 3 of 12 scales (p≤0.01), although parent-report of control HRQOL was significantly higher for 7 scales and both summary scores (p≤0.01) with the exception of one scale, family activities, which was significantly lower (p≤0.05) (Table 3). Similar results were found for parent-report of case and control HRQOL compared to gender-adjusted norms (results not shown).
Parent-child agreement was low for the single-items with the same wording between the CHQ-CF87 and CHQ-PF50. Weighted kappas ranged from 0.27 to 0.33 for case-parent and 0.27 to 0.28 for control-parent. MTMM analysis showed relatively poor agreement between parent and child on common domains. The average correlation between parent and child when reporting on the same HRQOL domains was a little higher for cases (r=0.374) than controls (r=0.334), but the average correlation was also higher when they were reporting on different domains (r=0.285 vs. r=0.246). Evidence for construct validity was low; only 82 out of 288 (case) and 98 out of 288 (control) t-tests of the different pairs of MTMM discriminant validity correlations were significant in the hypothesized direction.
In contrast to child self-report, where HRQOL was only infrequently associated with 5-year remission status, AED use, and having complicated epilepsy, parent-report of case HRQOL was often significantly associated with CHQ-PF50 scales (Table 4). For example, parent-reported HRQOL for seizure-free (n=178) cases was significantly better compared to non-seizure free (n=101) cases in 7 scales, 2 global items, and both summary scores (p≤0.03). In contrast, both parents’ and childs’ ratings of the child’s HRQOL were significantly associated with pharmacoresistance status for about half the HRQOL scales on the child and on the parent versions (Table 4).
The measurement of HRQOL is more complex in children compared to adults; the assessment of child HRQOL often relies on a parent or guardian report. Children, however, may have different perspectives from that of their adult proxies. We found significant differences in child self-report versus parent-report of HRQOL for children with epilepsy compared to sibling controls. Overall, children with epilepsy report HRQOL that is comparable to that of healthy sibling controls while parents rated children with epilepsy as having lower, or worse, HRQOL than sibling controls broadly across multiple scales, with no predominance of one domain. These results suggest children and parents may not necessarily share similar views about the overall impact of epilepsy.
There are several potential explanations for our findings. First, the concerns of a child with epilepsy may differ from that of a parent. For example, Arunkumar et al. reported that out of 26 identified concerns about living with epilepsy, children identified 5 domains that were not identified by parents (hatred of seizures, dislike of hospital visits, dating, social embarrassment, and headache) . In this regard children and parents may draw on different values and perspectives from which to evaluate HRQOL. Children and parents, however, may not be aware of these different perspectives . Second, it is likely that parent-report of HRQOL for children with epilepsy is influenced by parental anxiety that could lead to a lower, or worse, parent-report of HRQOL compared to the child-report. Parental anxiety may stem from the knowledge that children with epilepsy face many challenges including but not limited to seizures, including behavior, psychiatric, learning and school performance problems [2–11], in addition to poor long-term outcomes with respect to employment, educational attainment, socioeconomic status, marital status and quality of life [34–37]. In fact, parents are highly aware of the behavioral, emotional, and cognitive difficulties faced by their children with epilepsy . Parental anxiety was found to be significantly associated with decreased quality of life for children with epilepsy in a study by Williams et al. in which parent ratings of quality of life were measured using the Impact of Childhood Illness Scale . Lastly, having a chronic disease, or disability, does not necessarily mean that a person is unsatisfied with their life, despite what others may think—a term denoted the disability paradox . Such a paradox could explain our findings that parent-report of HRQOL was low compared to child-report. In this regard, parental perception of HRQOL may be distorted due to having a “sick” child; such distortion could not only lead to an underestimate of the HRQOL of the child with epilepsy, but also, it may result in an overestimate of the HRQOL of a “healthy” sibling—a phenomenon that could explain our findings that parents viewed sibling controls as having better HRQOL than general population controls.
Our findings are consistent with prior research, which has suggested that parents of children with health conditions tend to underestimate the child’s HRQOL . In fact, parent-report of lower, or worse, HRQOL compared to child-report has been observed in a number of different disease states including cancer, asthma, diabetes, inflammatory bowel disease, and attention deficit disorder [12, 22, 42–45]. Few studies, however, have evaluated parent-child differences in the report of HRQOL in children with epilepsy [19, 20, 22]. Verhey et al. found that parents rated children’s HRQOL lower compared to the child’s report, as measured by an epilepsy-specific measure, the CHEQOL-25 . Only a small proportion of children in this sample, however, were seizure free (11.5% seizure free for 12 months). In contrast, our sample uniquely identifies these parent-child differences in a sample of children with predominantly well-controlled epilepsy (> 60% were seizure free for 5-years). Furthermore, although Hamiwka et al. reported child-parent differences in a sample of first-time seizure patients, this study was limited by the fact that comparisons were made to different parent and child normative samples; the CHQ-CF87 norms were based on a Australian normative sample while the CHQ-PF50 norms were based on a US normative sample . Additionally, Felder-Puig et al. found that parent-report of HRQOL was lower compared to the child-report of HRQOL using a German version of the generic PedsQL . However, the epilepsy sample in this study was small (n=41) and no clinical information was provided regarding epilepsy disease severity (seizure frequency or remission status). Finally, in a sample of adolescents with epilepsy, Huberty et al. found that adults (parents and teachers) tended to agree with each other than with the youths, with youths reporting fewer problems than adults . Although analogous to our findings, the Hubert study focused on analyzing different informant perspectives of behavior, while we analyzed differences in perspectives on health-related quality of life.
There are several limitations to our study. First, because the CHQ is a generic instrument, it may not be sensitive to the nuances of epilepsy. While published data are available evaluating the use of the CHQ-PF50 in a sample of children who underwent resective epilepsy surgery, no such data are available for the CHQ-CF87 . As such, the differences that we found may reflect our finding of better construct validity for epilepsy disease severity in the CHQ-PF50 compared to the CHQ-CF87, perhaps suggesting that parent-reports are more valid than child self-reports in this population. Second, the CHQ-CF87 and the CHQ-PF50 have varied item wording and scale length making direct comparisons of agreement across the measures problematic; we did, however, find low parent-child agreement on the three items with the same wording between parent and child versions . Furthermore, our comparison of child- and parent-report of HRQOL with that of a sibling control, however, allows for indirect comparisons between the child and parent versions. Third, although the CHQ-CF87 is primarily intended for administration in children ages 10–18; we included children under age 10 (18% cases and 28% sibling controls in the matched case-control sample). Internal consistency reliability for the subsample, however, was high (Cronbach’s alpha ≥ 0.80 for all but one scale). Fourth, although we noted statistically significant differences on multiple domains of HRQOL in the parent-report of case compared to sibling control—only differences in 7 of the scales (role/social limitations-physical, role/social limitations-emotion/behavior, behavior, general health perceptions, family activities, parent impact on time, parent emotional impact) and the 1 global general health item may represent clinically meaningful differences (>5 points) . Furthermore, we did not correct for multiple comparisons, although we used an a priori p-value of p ≤ 0.05 for statistical significance; however, we note that the outcomes (dependent variable) are correlated and a strict adjustment would be too conservative. Lastly, the generalizability of our findings may be limited, not only because we were only able to compare the CHQ-PF50 to normative data because there is no U.S. based normative data for the CHQ-CF87, but also because our sample population was relatively homogeneous; subjects were recruited from only one U.S. state (Connecticut), and were predominantly Caucasian (>80%), and highly educated (43% of parents had college education or higher). The predominance of subjects with relatively high socioeconomic status in our sample, however, may explain our finding that parent-report of HRQOL for the sibling control was higher compared to normative data.
This study uniquely showed differences in parent and child-report of HRQOL in a prospective, community-based sample of children with predominantly well-controlled epilepsy compared to sibling controls. Because parent ratings are often used to assess HRQOL in children it is important to understand that children and parents may have different perspectives, both of which need to be taken into account when assessing HRQOL. Patient-reported outcomes, including HRQOL, are increasingly being recognized as an important outcome measure in research studies, particularly in clinical anti-epileptic medication trials and prospective studies evaluating surgical outcomes. Furthermore, the evaluation of HRQOL is essential for developing disease management programs for children and families with epilepsy that not only provide comprehensive treatment and education, but also help to facilitate dialogue between parents and children given the different perspectives that each may have. Understanding the potential different, yet valuable, perspectives provided by parents and children is integral to these evaluations and also for developing appropriate support services targeted for children as they transition into adulthood. As such, future studies are needed to further validate our findings and to elucidate the factors that may drive these potential differences.
Funding support: Supported by grant from National Institutes of Health, NINDS R01-NS31146 (PI-Dr. Berg). Dr. Baca was supported by the Robert Wood Johnson Foundation Clinical Scholars Program (Grant 59982). Dr. Hays was also supported in part by the UCLA Resource Center for Minority Aging Research/Center for Health Improvement in Minority Elderly (RCMAR/CHIME), NIH/NIA Grant Award Number P30AG021684, and the UCLA/Drew Project EXPORT, NCMHD, 2P20MD000182.