You are on page 1of 14

PedsQL™ 4.0: Reliability and Validity of the Pediatric Quality of Life Inventory™ Version 4.

0
Generic Core Scales in Healthy and Patient Populations
Author(s): James W. Varni, Michael Seid and Paul S. Kurtin
Source: Medical Care, Vol. 39, No. 8 (Aug., 2001), pp. 800-812
Published by: Lippincott Williams & Wilkins
Stable URL: http://www.jstor.org/stable/3767969
Accessed: 16-02-2016 06:30 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Lippincott Williams & Wilkins is collaborating with JSTOR to digitize, preserve and extend access to Medical Care.

http://www.jstor.org

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
MEDICAL CARE
Volume39, Number8, pp 800-812
02001 LippincottWilliams& Wilkins,Inc.

4.0: Reliabilityand Validityof the PediatricQualityof


PedsQLTM
Version4.0 GenericCore Scales in Healthy
LifeInventoryTM
and PatientPopulations

SEID, PHD,* ANDPAULS. KURTIN,


JAMESW. VARNI,PHD,*t MICHAEL MD*

BACKGROUND. The PedsQL (Pediatric Quality ent report), Physical Health Summary Score
of Life Inventory) (Children's Hospital and (a = 0.80 child, 0.88 parent), and Psychosocial
Health Center, San Diego, California) is a Health Summary Score (a = 0.83 child, 0.86
modular instrument for measuring health- parent) were acceptable for group compari-
related quality of life (HRQOL) in children sons. Validity was demonstrated using the
and adolescents ages 2 to 18. The PedsQL 4.0
known-groups method, correlations with indi-
Generic Core Scales are multidimensional cators of morbidity and illness burden, and
child self-report and parent proxy-report scales factor analysis. The PedsQL distinguished be-
developed as the generic core measure to be tween healthy children and pediatric patients
integrated with the PedsQL Disease-Specific with acute or chronic health conditions, was
Modules. The PedsQL 4.0 Generic Core Scales
related to indicators of morbidity and illness
consist of 23 items applicable for healthy
burden, and displayed a factor-derived solu-
school and community populations, as well as
tion largely consistent with the a priori
pediatric populations with acute and chronic
health conditions. conceptually-derived scales.
METHODS. The 4 PedsQL 4.0 Generic Core
CONCLUSION. The results demonstrate the re-
Scales (Physical, Emotional, Social, School) liability and validity of the PedsQL 4.0 Generic
were administered to 963 children and 1,629 Core Scales. The PedsQL 4.0 Generic Core
Scales may be applicable in clinical trials,
parents (1,677 subjects accrued overall) re-
cruited from pediatric health care settings. research, clinical practice, school health set-
Item-level and scale-level measurement prop- tings, and community populations.
erties were computed. Key words: Health-related quality of life;
RESULTS. Internal consistency reliability for pediatrics; health; children; adolescents. (Med
the Total Scale Score (a = 0.88 child, 0.90 par- Care 2001;39:800-812)

In clinical trials,1'2population health,3 clinical minimum of the physical, mental, and social
improvement,4,5and among purchasers of health health dimensions delineated by the World Health
care services,6 health-related quality of life Organization (WHO).7
(HRQOL) has emerged as an essential health Pediatric HRQOL measurement instruments
outcome. HRQOL instruments must consist at the must be sensitive to cognitive development and

*From Center for Child Health Outcomes, Children's Address correspondence and reprint requests to:
Hospital and Health Center, San Diego, California. James W. Varni, PhD, Professor and Senior Scientist,
Center for Child Health Outcomes, Children's Hospital
tFrom the Department of Psychiatry, University of and Health Center, 3020 Children's Way, San Diego, CA
California, School of Medicine, San Diego, California. 92123. E-mail: jvami@chsd.org. The PedsQL is available
at http://www.pedsql.org.
This research was supported by intramural grants
from the Children's Hospital and Health Center, San Received February 8, 2000; initial review April 17,
Diego. 2000; accepted March 6, 2001.

800

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
Vol.39, No. 8 PEDSQL4.0

include child self-report and parent proxy- Methods


report.Imperfectconcordancehas been consis-
tently noted in adult HRQOLresearchbetween Subjects and Settings
patients'self-reports and ratings by health care
providersand significantothers.8Imperfectcon- Subjects were children ages 5 to 18 years
cordancehas been documented in the HRQOL (n = 963) and parents of childrenages 2 to 18
assessment of children with asthma,9cystic fi- years (n = 1,629),with 1,677participantsaccrued
brosis,10chronicheadache," limb deficiencies,12 overall.For 915 childrenages 5 to 18 years,both
arthritis,13cancer,14-16as well as healthy chil- child-self report and parent proxy-reportwere
dren.17In pediatric health assessment, agree- available.To maximizethe heterogeneityof the
ment has been found to be lower for internal- sample,subjectswererecruitedfromthreetypesof
sites. Subjectswere childrenpresentingat pedia-
izing problems (eg, depression, pain) than for tricians'offices for scheduled well-child checks
externalizingproblems (eg, hyperactivity,walk-
(n = 98, 5.9%),childrenat one of four hospital
ing).1317The presence of imperfectconcordance
specialty clinics-orthopedics (n = 127, 7.6%),
suggests a critical need in pediatric HRQOL
measurement for reliable and valid child self- cardiology (n = 156, 9.3%), rheumatology
(n = 95, 5.7%),and diabetes(n = 49, 2.9%)-and
report instruments for the broadest age range children who had been seen as inpatients or
possible. outpatientsat the hospitalor its outpatientcom-
Although the relativemerit of generic versus munity clinics at least 3 months previously
disease-specific instrumentsis a matter of em- (n = 1,141,68.0%)(11 children,0.7%had missing
piricalinquiry,18there may be advantagesto an data).
integrated approach.19'20 Disease-specific mod- Two modes of administrationwere used-in
ules may enhance measurement sensitivity for person (n = 439, 26.2%) and by telephone
health domains germane to a chronic health (n = 1,227,73.2%).Forthe in-personmode,ques-
condition (eg, fatigue in cancer), whereas a tionnaireswere self-administeredfor parentsand
generic instrument enables comparisonsacross for children ages 8 to 18 and interviewer-
groups and bench-marking with healthy administeredfor childrenages 5 to 7. The mea-
populations. sures were administeredin two languages-En-
The PedsQL (PediatricQuality of Life Inven- glish (n = 1328, 79.2%) and Spanish (n = 349,
tory) MeasurementModel (Children'sHospital 20.8%).
and Health Center,San Diego, California)inte- Forall formscombined,the averageage of the
grates generic core scales and disease-specific 815 boys (48.6%)and 830 girls (49.5%;32, 1.9%
modules into one measurement system. The missing)was 9.3years(SD = 4.37)with a rangeof
PedsQL4.0 builds on programmaticinstrument 2.0 to 18.8years.Forchildself-report,the average
development researchduring the past 15 years, age of the 643 boys (48.1%)and 664 girls (49.7%;
beginning with the measurement of pain and 30, 2.2% missing) was 10.78 years (SD = 3.61)
functionalstatus.21,22 The PedsQL1.0,20derived with a range of 5.0 to 18.8. The sample was
from a cancerdatabase,15,16,23 was designed as a heterogeneouswith respectto race/ethnicity, with
612 (36.5%)White non-Hispanic,667 (39.8%)
generic instrument to be utilized noncategori-
Hispanic, 117 (7.0%) Black non-Hispanic, 49
cally across pediatricpopulations. The PedsQL
2.0 and 3.0 included additional constructs and (2.9%)Asian/PacificIslander,18 (1.1%)American
Indian or AlaskanNative, 99 (5.9%)Other,and
items, a more sensitive scaling range, and a 115 (6.9%) missing. With respect to insurance
broaderage range.The PedsQL4.0 has resulted
from this iterativeprocess, and was designed to type, 31% of the sample had commercialinsur-
ance,with 56%coveredby Medicaid,2.8%Other,
measure the core health dimensions delineated 0.7%self-pay,and 9.5%missing.The samplewas
by WHO,7including role (school) functioning. also diversewith respectto socioeconomicstatus,
This study presents the measurementproper-
using mother'seducationas a proxy.In the sam-
ties of the PedsQL 4.0 Generic Core Scales, ple, 7.1%of mothershad no more than an ele-
reporting on their reliabilityand validity in a mentaryschool education,15.8%had some sec-
diverse sample of healthy childrenand pediatric ondaryeducationbut did not graduatefromhigh
patientswith acute or chronichealth conditions. school, 14.8%were high school graduates,25.8%

801

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
VARNI ET AL MEDICALCARE

had some posthigh school education, 12.2% were scale.21'24Parent proxy-report also includes the
college graduates, 4.4% had a graduate or profes- toddler age range (ages 2-4), which does not
sional degree, with 19.8% missing data. include a self-report form given developmental
The sample included chronically ill children, limitations on self-report for children younger
acutely ill children, and healthy children. The than 5 years of age,24'25and includes only 3 items
chronically ill children were those whose parents for the school functioning scale.
reported the presence of a chronic health condi- Items are reverse-scored and linearly trans-
tion (n = 683, 40.7%); acutely ill children were formed to a 0 to 100 scale (0 = 100,1 = 75, 2 = 50,
those whose parents did not report the presence of 3 = 25, 4 = 0), so that higher scores indicate better
a chronic health condition, but who were assessed HRQOL. Scale Scores are computed as the sum of
at one of the specialty clinics (n = 207, 12.3%); the items divided by the number of items an-
and healthy children were those assessed either in swered (this accounts for missing data). If more
physicians' offices during well-child checks or by than 50% of the items in the scale are missing, the
telephone and whose parents did not report the Scale Score is not computed.26For this study, more
presence of a chronic health condition (n = 730, than 99% of child and parent respondents were
43.5%; 57 (3.4.%) missing). The response rate for included in the Scale Score analyses. The Physical
subject accrual is unknown because information Health Summary Score (8 items) is the same as the
on nonparticipants was not collected. Physical Functioning Subscale. To create the Psy-
chosocial Health Summary Score (15 items), the
mean is computed as the sum of the items divided
Measures by the number of items answered in the Emo-
tional, Social, and School Functioning Subscales.
PedsQL 4.0 (Pediatric Quality of Life
InventoryVersion 4.0)
PedsQL Family InformationForm
The 23-item PedsQL 4.0 Generic Core Scales
encompass: 1) Physical Functioning (8 items), 2) The PedsQL Family Information Form, com-
Emotional Functioning (5 items), 3) Social Func- pleted by parents, contains demographic informa-
tioning (5 items), and 4) School Functioning (5 tion on the child and parents. It also asks for
items), and were developed through focus groups information on the number of days during the past
and cognitive interviews.20 30 days that the child needed care or missed
The Generic Core Scales are comprised of par- school because of health, the number of days the
allel child self-report and parent proxy-report for- parent missed work because of the child's health,
mats. Child self-report includes ages 5 to 7,8 to 12, and the impact of the child's health on the parent's
and 13 to 18 years. Parent proxy-report includes daily work routine and ability to concentrate at
ages 2 to 4 (toddler), 5 to 7 (young child), 8 to 12 work.
(child), and 13 to 18 (adolescent), and assesses
parent's perceptions of their child's HRQOL. The
items for each of the forms are essentially identi- Procedure
cal, differing in developmentally appropriate lan-
guage, or first or third person tense. The instruc- Inclusion criteria were: child ages 2 to 18 and
tions ask how much of a problem each item has parent and child English- or Spanish-speaking. At
been during the past 1 month. A 5-point response pediatricians' offices, an exclusion criterion was:
scale is utilized across child self-report for ages 8 presenting complaint of acute symptomatology. At
to 18 and parent proxy-report (0 = never a prob- pediatrician offices and specialty clinics, the sam-
lem; 1 = almost never a problem; 2 = sometimes ple was a convenience sample based on consecu-
a problem; 3 = often a problem; 4 = almost al- tive patients. Subjects were identified through
ways a problem). To further increase the ease of clinic appointment schedules. At these sites, par-
use for the young child self-report (ages 5-7), the ents of possible study participants were informed
response scale is reworded and simplified to a of the study after checking in for their appoint-
3-point scale (0 = not at all a problem; 2 = some- ment, but before being seen by their health care
times a problem; 4 = a lot of a problem), with each provider. Written parental informed consent and
response choice anchored to a happy to sad faces child assent were obtained. Parents and children

802

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
Vol. 39, No. 8 PEDSQL 4.0

TABLE1. Scale Descriptives and Internal Consistency Reliability for PedsQL 4.0 Child Self-Report and
Parent Proxy-Report
InternalConsistency
ScaleDescriptives Reliability
Percentage Percentage
Numberof Floor Ceiling Total
Scale Items N Mean SD Ill/Healthy Ill/Healthy Sample Range
Self-Report Ages 5-18
Totalscore 23 960 79.62 15.26 0.0/0.0 1.9/7.2 0.88 0.85-0.90
Physicalhealth 8 959 80.19 19.30 0.0/0.0 13.1/25.8 0.80 0.77-0.85
Psychosocialhealth 15 958 79.37 15.70 0.0/0.0 5.2/12.0 0.83 0.78-0.85
Emotionalfunctioning 5 958 78.10 20.66 0.3/0.8 22.4/29.8 0.73 0.70-0.78
Socialfunctioning 5 958 84.09 18.50 0.0/0.0 33.2/47.1 0.71 0.67-0.74
Schoolfunctioning 5 933 75.87 19.71 0.3/0.5 13.0/23.1 0.68 0.59-0.75
Proxy-Report Ages 2-18
Totalscore 23 1622 80.87 16.73 0.2/0.0 4.1/10.3 0.90 0.89-0.90
Physicalhealth 8 1613 81.38 23.18 2.3/0.0 18.5/39.6 0.88 0.85-0.89
Psychosocialhealth 15 1621 80.58 16.52 0.2/0.0 5.6/13.8 0.86 0.80-0.86
Emotionalfunctioning 5 1622 77.95 20.67 1.4/0.1 19.5/29.5 0.77 0.69-0.80
Socialfunctioning 5 1615 85.38 19.17 0.5/0.0 34.4/58.1 0.75 0.68-0.78
Schoolfunctioning 5 1417 77.80 22.00 1.7/0.3 15.5/34.5 0.76 0.57-0.78
Note: PercentageFloor/Ceiling= the percentageof scoresat the extremesof the scalingrange.

completed the PedsQL separately. One parent Statistical Analysis


(80% mothers; 13% fathers; 7% other) completed
the proxy-reportversion. A research assistant (RA) Item-Level Analysis
was available to answer questions regarding the
parent self-administered instruments. A RA ad- Feasibility was determined from the percentage
ministered the PedsQL for the young child (ages of missing values for each item and the distribu-
5-7), and was available to assist the self- tion of item responses. Multitrait scaling analysis
administered instrument for the child (ages 8-12) was conducted to determine the extent to which
and adolescent (ages 13-18) after the instructions individual items correlated with the hypothesized
had been given and clarified. scale construct rather than with other scales.27
For the telephone sample, a random sample of Multitrait scaling analyses were summarized via
children ages 2 to 18 seen as inpatients or outpa- tests of individual item scaling success, defined as
tients at Children's Hospital and Health Center the number of times an item correlated higher
was generated from a list of all discharges between with its hypothesized scale construct rather than
April 1, 1998 and June 30, 1998. Children with a with another scale by >2 standard errors,28which
discharge status of expired, whose payer was from provides an approximation of scaling success. The
the victim/witness fund, and whose parents had percentage of item scaling successes relative to the
requested their phone number and address be total number of item scaling tests was calculated
for each scale.28
kept private were excluded. RAs telephoned par-
ents of children on this list, explained the study,
and obtained verbal parental informed consent
and child assent. The RA verbally administered the Scale-Level Analysis
PedsQL individually to the parents and their chil-
dren. If the child was not home at the time of the Range of measurement was based on the per-
initial call, the RA arranged for a call at another centage of scores at the extremes of the scaling
time. The Institutional Review Board approved this range, that is, the maximum possible score (ceiling
research protocol at Children's Hospital and effect) and the minimum possible score (floor
Health Center, San Diego. effect).28

803

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
VARNI ET AL MEDICAL
CARE

TABLE
2. One-Way ANOVA Comparing Chronically Ill, Acutely Ill, and Healthy Children:
Child Report
Scale N Mean SD df F P
Totalscore 2,913 15.05 0.001
Chronicallyill* 367 77.19 15.53
Acutelyill* 148 78.70 14.03
Healthy 401 83.00 14.79
Physicalhealth 2,912 14.16 0.001
Chronicallyill* 366 77.36 20.36
Acutelyill* 149 78.88 19.10
Healthy 400 84.41 17.26
Psychosocialhealth 2,911 11.42 0.001
Chronicallyill* 367 77.10 15.84
Acutelyill 148 78.68 14.66
Healthy 399 82.38 15.51
Emotionalfunctioning 2,911 4.84 0.001
Chronicallyill* 366 76.40 21.48
Acutelyill 148 77.33 20.04
Healthy 400 80.86 19.64
Socialfunctioning 2,911 10.16 0.001
Chronicallyill* 367 81.60 20.24
Acutelyill* 148 82.83 16.66
Healthy 399 87.42 17.18
Schoolfunctioning 2,888 6.50 0.001
Chronicallyill* 362 73.43 19.57
Acutelyill 143 75.68 18.04
Healthy 386 78.63 20.53
Note: * = different from healthy children at P <0.05.

Scale internal consistency reliability was deter- vealed no relation between health status and
mined by calculating Cronbach's coefficient al- gender. There was a significant association be-
pha.29Scales with reliabilities of 0.70 or greater are tween age group and health status (X26 = 16.8,
recommended for comparing patient groups, P = 0.01). Adolescents were over represented in
whereas a reliability criterion of 0.90 is recom- the chronically ill group (standardized residu-
mended for analyzing individual patient scale al = 2.4) and underrepresented in the healthy
scores.30,31 group (standardized residual = -2.1). There was a
Construct validity was determined utilizing the significant association between ethnicity and ill-
known-groups method. The known-groups ness group (X210= 67.4, P = 0.001). Whites were
method compares scale scores across groups over represented in the chronically ill group (stan-
known to differ in the health construct being dardized residual = 2.1) and underrepresented in
investigated. In this study, groups differing in the healthy group (standardized residual = -3.6).
known health status (healthy vs. acute or chronic Hispanics were underrepresented in the chroni-
health conditions) were computed,3233 using one- cally ill group (standardized residual = -2.0) and
way analysis of variance (ANOVA). We hypothe- over represented in the healthy group (standard-
sized that healthy children would report higher ized residual = 3.8).
PedsQL scores than pediatric patients with acute Given these associations, 2-Way ANOVA'swere
and chronic health conditions. We examined po- performed to explore interactions. No interactions
tential confounding between health status and were found for self-report. Although interactions
age, gender, and ethnicity. Cross-tabulations re- between health status and age and ethnicity were

804

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
Vol. 39, No. 8 PEDSQL 4.0

TABLE
3. One-Way ANOVA Comparing Chronically Ill, Acutely Ill, and Healthy Children:
Parent Report
Scale N Mean SD df F P
Total score 2, 1575 128.39 0.001
Chronically ill 662 74.22 18.40
Acutely ill 199 80.42 15.26
Healthy 717 87.61 12.33
Physical health 2, 1566 91.72 0.001
Chronically ill 653 73.28 27.02
Acutely ill 199 81.81 20.46
Healthy 717 89.32 16.35
Psychosocial health 2, 1574 98.95 0.001
Chronically ill 661 74.80 18.16
Acutely ill 199 79.56 15.51
Healthy 717 86.58 12.79
Emotional functioning 2, 1575 38.90 0.001
Chronically ill 661 73.05 23.27
Acutely ill 199 78.82 18.00
Healthy 718 82.64 17.54
Social functioning 2, 1568 72.55 0.001
Chronically ill 657 79.77 21.91
Acutely ill 198 83.58 18.29
Healthy 716 91.56 14.20
School functioning 2, 1376 73.25 0.001
Chronically ill 601 71.08 23.99
Acutely ill 167 74.74 20.95
Healthy 611 85.47 17.61
Note: For each scale, each subgroup significantly different from every other subgroup P <0.05.

found for parent proxy-report,these accounted for PedsQL items. MTMM analysis was conducted
approximately 1% of the variance (eta using MTMM.EXE software (Berkeley Heights,
squared = 0.008 for health status*age; eta New Jersey).34,35 MTMMassumes that heterotrait-
squared = 0.013 for health status*race). Given the monomethod correlations (eg, correlations among
large sample size and the small eta squared, we Subscales within self-report and proxy-report)
did not pursue these 2-Way interactions. should be lower than monotrait-heteromethod
Construct validity was further assessed through correlations (eg, concordance between self-report
PedsQL correlations with indicators of morbidity and proxy-report for the same Subscale). Correla-
and illness burden and examination of the PedsQL tions are designated as small (0.10-0.29), medium
factor structure. It was hypothesized that higher (0.30-0.49), and large (>0.50).36 Given shared
PedsQL scores would be associated with fewer method variance31 and that the PedsQL items
days during the past month in which the child were developed to measure an integrated multidi-
needed care for a physical or mental health con- mensional construct (pediatric HRQOL), it was
dition, fewer days missed from school for children expected that heterotrait-monomethod correla-
and work for parents, and less impact on work tions among the Subscales would be medium to
routine and concentration for parents who worked large (r >0.40).37 Parent/child concordance for the
outside the home. same Subscale was expected to demonstrate me-
PedsQL factor structure was examined via dium to large effect sizes, but not so large that
multitrait-multimethod (MTMM) analysis of the child and parent reports would be redundant.
PedsQL Subscales and factor analysis of the Based on previous literature,1317it was anticipated

805

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
VARNI ET AL MEDICALCARE

4.
TABLE Pearson Correlations Between Indicators of Morbidity and Child Self-Report and Parent
Proxy-Report
Work
WorkRoutine Concentration
Scale CareNeeded SchoolMissed WorkMissed Impact Impact
Self-Report r n r n r n r n r n
Total -0.24 865 -0.22 598 NS 216 -0.23 209 -0.28 209
Physical -0.27 864 -0.21 596 -0.20 216 -0.28 209 -0.31 209
Psychosocial -0.18 863 -0.20 598 NS 216 NS 209 -0.22 209
Emotional -0.13 863 -0.13 597 NS 216 -0.16 209 -0.25 209
Social -0.13 863 -0.11 598 NS 216 NS 209 NS 209
School -0.17 841 -0.24 591 -0.21 211 NS 206 -0.19 206
Proxy-Report
Total -0.38 1522 -0.29 816 -0.30 279 -0.44 277 -0.50 277
Physical -0.38 1515 -0.26 811 -0.30 279 -0.31 277 -0.38 277
Psychosocial -0.31 1521 -0.25 816 -0.23 278 -0.43 276 -0.48 276
Emotional -0.19 1521 -0.12 817 -0.20 280 -0.40 278 -0.44 278
Social -0.27 1518 -0.13 812 -0.24 280 -0.34 278 -0.34 278
School -0.28 1324 -0.33 786 -0.23 268 -0.27 268 -0.35 268
Note: All displayedcorrelationcoefficients(r) are significantat P <0.01.

that the Physical Functioning Subscale would vs. telephone) and language (English vs. Spanish)
demonstrate the largest concordance. Heterotrait- by examining the percent missing data, floor and
heteromethod concordance was expected to be ceiling effects, and scale interal consistency
small. Principal components analysis with oblique across mode of administration and language. No
rotation was performed on the 23 items to test the differences were found across mode of adminis-
PedsQL underlying dimensions.38 tration or language. Thurstone scaling analysis
Statistical analyses were conducted using SPSS was conducted to examine response choice equiv-
for Windows.39 Response equivalence was as- alence between the English and Spanish
sessed across mode of administration (in-person forms.40'41Eight bilingual respondents were pre-

5. Multitrait-Multimethod IntercorrelationsBetween and Among PedsQL 4.0 Subscales


TABLE

Self-Report Proxy-Report
Scale Emotional Social School Physical Emotional Social School

Self-Report
PhysicalFunctioning 0.45 0.48 0.48 0.50 0.25 0.33 0.24
EmotionalFunctioning 0.42 0.47 0.19 0.36 0.26 0.25
SocialFunctioning 0.45 0.20 0.19 0.37 0.19
SchoolFunctioning 0.21 0.24 0.17 0.41
Proxy-Report
PhysicalFunctioning 0.43 0.49 0.43
EmotionalFunctioning 0.47 0.49
SocialFunctioning 0.49
=
Notes: N 871
All correlationsare significantat P <0.001.
Effectsizes are designatedas small (0.10),medium(0.30),and large(0.50).
Multitrait-monomethod correlationsare underlined;multitrait-
correlationsare in bold; monotrait-multimethod
multimethodcorrelationsare italicized.

806

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
Vol. 39, No. 8 PEDSQL 4.0

TABLE6. PedsQL 4.0 Factor Loadings for Child Self-Report and Parent Proxy-Report*

Scale/Item Factor1 Factor2 Factor3 Factor4 Factor5

PhysicalFunctioning
Hardto walk more than one block -0.049 -0.715 0.003 0.118 0.049
0.830 -0.106 0.036 0.018 -0.015
Hardto run 0.100 -0.773 -0.060 0.049 0.118
0.836 -0.175 0.072 0.035 -0.005
Hardto do sportsor exercises 0.094 -0.779 0.009 -0.003 0.063
0.797 -0.091 0.100 0.064 0.009
Hardto lift somethingheavy 0.135 -0.456 0.058 0.161 0.045
0.746 0.032 0.102 0.039 -0.080
Hardto take bath or shower -0.063 -0.574 0.006 -0.140 -0.238
0.719 0.224 -0.107 -0.082 -0.016
Hardto do choresaroundhouse -0.039 -0.646 0.005 -0.105 -0.208
0.737 0.193 -0.019 -0.069 -0.016
Hurtor ache 0.223 -0.253 0.046 0.204 0.025
0.310 -0.148 0.326 0.291 -0.021
Low energy 0.385 -0.257 0.030 0.142 -0.025
0.274 -0.027 0.374 0.304 -0.038
EmotionalFunctioning
Feel afraidor scared 0.782 -0.034 0.114 -0.175 0.052
0.042 0.000 0.721 -0.069 0.044
Feel sad or blue 0.776 0.102 0.065 0.012 -0.015
-0.016 0.049 0.766 -0.023 0.077
Feel angry 0.657 0.007 -0.040 -0.020 -0.105
-0.044 0.090 0.664 -0.025 0.103
Troublesleeping 0.371 -0.129 0.046 0.128 -0.179
0.042 0.081 0.539 0.155 -0.013
Worryaboutwhat will happen 0.705 0.002 -0.072 0.040 0.015
-0.034 0.035 0.778 -0.049 -0.015
SocialFunctioning
Troublegettingalong w/peers 0.014 0.038 0.679 -0.077 -0.175
0.035 0.252 0.089 -0.156 0.604
Otherkids not wantingto be -0.041 0.024 0.813 -0.111 -0.054
friend
-0.036 0.022 0.062 -0.016 0.836
Teased 0.093 0.102 0.747 0.053 -0.012
-0.061 -0.096 0.098 0.078 0.810
Doing things otherpeers do 0.065 -0.308 0.417 0.183 0.169
0.484 0.024 -0.075 0.126 0.405
Hardto keep up when playwith -0.025 -0.350 0.423 0.242 0.206
others
0.547 0.149 -0.127 0.115 0.374
SchoolFunctioning
Hardto concentrate 0.119 -0.056 0.075 0.041 -0.710
0.025 0.809 0.094 0.008 0.058
(Continued)

807

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
VARNI ET AL MEDICAL
CARE

TABLE
6. (Continued)
Scale/Item Factor1 Factor2 Factor3 Factor4 Factor5

Forgetthings 0.214 0.003 0.093 0.135 -0.522


0.031 0.734 0.129 0.077 -0.026
Troublekeepingup with -0.046 -0.119 0.144 0.204 -0.678
schoolwork
0.019 0.773 0.044 0.151 0.051
Miss school - not well -0.002 -0.002 -0.050 0.800 -0.143
-0.040 0.113 -0.044 0.901 -0.050
Miss school - doctorappointment -0.031 0.059 -0.008 0.848 -0.068
-0.026 0.071 -0.021 0.855 0.073
Eigenvalues 6.22 1.93 1.46 1.24 1.11
7.35 2.70 1.80 1.35 1.05
PercentVariance 27.02 8.37 6.35 5.40 4.82
31.94 11.72 7.82 5.86 4.54

*Reportedfor ages 5 yearsand older.


TotalVarianceExplainedfor ChildSelf-Report:52%;for ParentProxy-Report: 62%.
Bold = highestfactorloadingfor each item.
Note: In each cell, Child Self-ReportLoadingsare shown above and ParentProxy-
-Reportloadingsare shown
below in italics.

sented with two 100-cm lines anchored by the missing item responses was 1.54% and 1.95%,
extreme response choices ("Never" and "Almost respectively.
Always"for English, "Nunca"and "Casi Siempre"
for Spanish). Respondents were asked to make a
mark on the line to indicate where they thought Item Response Distributions
the intermediate English and Spanish response
choices fell. The Spanish intermediate response A full range for each item was demonstrated,
choices fell within 5-cm of the English response with item distributions tending to be skewed
choices, indicating that the response ratings across toward higher HRQOL.
language were equivalent. Therefore, responses
were pooled across mode of administration and
language. Responses were also pooled across the
Item Descriptives
age ranges for both self-report and proxy-report.
For all 23 items, item means for self-report
ranged from 70.8 to 92.5 with 13 of 23 items falling
Results within a 10-point range. Item means for proxy-
Data tables with item-level missing values, re- report ranged from 65.9 to 88.4, with 16 of 23
items falling within a 10-point range. Standard
sponse distributions, means, and SDs, item-scale deviations ranged from 25.6 to 35.3 for self-report
correlations, and scaling tests, are available from items and 23.3 to 35.2 for proxy-report items. For
the first author.
the 15 items in the Psychosocial Health Summary
Score, item means ranged from 71.1 to 86.3 for
self-report and 65.9 to 86.8 for proxy-report. Stan-
Feasibility: Missing Item Responses dard deviations ranged from 25.7 to 31.6 for
self-report and 24.1 to 34.2 for proxy-report. Sim-
To assess the feasibility of administration, the ple linear averaging is possible when the means
percentage of missing values was calculated. For and standard deviations of items within each scale
self-report and proxy-report, the percentage of or summary score are roughly equivalent.28

808

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
Vol. 39, No. 8 PEDSQL 4.0

Item-InternalConsistency demonstrate small to medium correlations with


the number of days the child needed care in the
Item-scale correlations demonstrated that most
past 30 days, and with the number of days missed
items (19/23) for self-report and all items for from school. Proxy-report shows medium correla-
proxy-report met or exceeded the 0.40 standard tions with the number of days missed from work.
for item-internal consistency corrected for item
Self-report shows small to medium, and parent
overlap.30 proxy-report shows medium to large, correlations
with parent-report of the extent to which their
child's HRQOL interfered with their daily routine
Item Scaling Tests and ability to concentrate at work.
The MTMM matrixis shown in Table5. Consis-
For self-report and proxy-report,scaling success
tent with hypotheses, self-report and proxy-report
rates ranged from 87% to 100%.
heterotrait-monomethodcorrelationsare in the me-
dium to large effect size range, as are parent/child
concordance for the same Subscale (monotrait-
Range of Measurement heteromethod). Heterotrait-heteromethod correla-
Table 1 presents the percentages of scores at the tions were small. Subscale successes were 2/24 for
floor and ceiling. There were no floor effects for monomethod and 23/24 for heteromethod. The
healthy or ill children, as no scale had more than complete MTMMresults are availablefrom the first
2.3% of respondents scoring the minimum. Ceil- author.
The results of the factor analysis for self-report
ing effects existed in some cases. These ranged
from minimal (eg, 7.2% and 10.3% of healthy and proxy-report are shown in Table 6. An eigen-
value cutoff of 1.0 resulted in a five factor solution
respondents, respectively, for the self- and proxy-
for self-report and proxy-report, accounting for
report Total Scale Score) to moderate (eg, 47.1% to
58.1% of healthy respondents, respectively for the 52% and 62% of the variance, respectively. School
self- and proxy-report Social Functioning Sub- Functioning items split into two different factors,
but otherwise, the factors that emerged are con-
scale). The ceiling effects were in the expected
sistent with the a priori hypothesized factor struc-
direction, with healthy children and their parents
ture, with few exceptions.
reporting more ceiling effects than children with
health conditions.

Discussion
Internal Consistency Reliability
The PedsQL 4.0 internal consistency reliabilities
Internal consistency reliabilityalpha coefficients generally exceeded the standard of 0.70 for group
are presented in Table 1. Most self-report scales comparisons. Across the ages, the Total Scale
and proxy-report scales approached or exceeded Score for self-report and proxy-report approached
the minimum reliability standard of 0.70. an alpha of 0.90, recommended for individual
patient analysis,30 making the Total Scale Score
suitable as a summary score for the primary anal-
Construct Validity ysis of HRQOL outcomes in clinical trials and
other group comparisons. The Physical Health and
One-Way ANOVA's comparing chronically ill, Psychosocial Health Summary Scores are recom-
acutely ill, and healthy children for all scales are mended for secondary analyses. The School Func-
displayed in Tables 2 and 3. For self-report and tioning Subscales for proxy-report for ages 2 to 4
proxy-report, scales demonstrated differences and self-report for ages 5 to 7 were the only two
among the three groups. The hypothesis was Subscales that did not approach or exceed 0.70.
confirmed that healthy children would manifest The Emotional, Social, and School Functioning
higher scores than acutely or chronically ill Subscales may be utilized to examine specific
children. domains of functioning, with the caveat that until
Table 4 shows the correlations between the further testing is conducted, these Subscales
PedsQL scales and indicators of morbidity and should be used for descriptive or exploratoryanal-
burden of illness. Self-report and proxy-report yses. The means and standard deviations of the

809

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
VARNI ET AL MEDICALCARE

items within scales were similar, supporting the The items selected for the PedsQL reflect
scaling assumption of equivalent item means and those that are of universal concern across age
variance. groups. Attempts were made to keep wording,
The PedsQL performed as hypothesized utiliz- and thus the content, as similar as possible
ing the known-groups method. The PedsQL dif- across parallel forms, while being sensitive to
ferentiated HRQOL between healthy children and developmental differences in cognitive ability.
those with acute or chronic health conditions, and This consistency facilitates the evaluation of
was correlated with measures of morbidity and differences in HRQOL across and between age
illness burden. The MTMM analyses tested con- groups, as well as the tracking of HRQOL
vergent and discriminant validity across methods. longitudinally. The PedsQL 4.0 is the only ge-
The heterotrait-monomethod analyses are consis- neric pediatric HRQOL measurement instru-
tent with the conceptualization of the PedsQL as ment that we are aware of to span ages 2 to 18
for self-report and proxy-report while maintain-
measuring an integrated multidimensional con-
struct. However, further work is necessary to sup- ing item and scale construct consistency.
The present findings have several potential lim-
port the construct validity of the PedsQL scales as
itations. Test-retest reliability and responsiveness
measuring distinctly unique dimensions of
are not reported. Information on nonparticipants
HRQOL. The results of the factor analysis in
was not available,which may limit generalizability.
general support the hypothesized factor structure
of the PedsQL. When a two-factor solution was The prevalence of Medicaid coverage may be
associated with lower HRQOL, potentially atten-
forced, 19 of the 23 items loaded as hypothesized
on the a priori Physical Health and Psychosocial uating ceiling effects. Parents reported health con-
dition and illness burden information. Objective
Health Summary Scores, for both self-report and
measures of health condition and illness burden
proxy-report. Total variance explained was 35.4% will strengthen the validation process. Item-
for self-report and 45.6% for proxy-report. Be-
cause this finding may be sample-specific, further response theory provides another analytic ap-
proach for investigation.
testing with other samples will be conducted The PedsQL is undergoing further field-testing
before reconfiguring the conceptually-derived
nationally and interationally, including examin-
Summary Scores.
The PedsQL Measurement Model emphasizes ing the usefulness of the Disease-Specific Modules
in pediatric chronic health conditions. The PedsQL
the child's perceptions. The items chosen for
Measurement Model represents a conceptual
inclusion were derived from the measurement framework for a measurement instrument that
properties of the self-report scales, whereas the must be at once disease-specific and also reflective
proxy-report scales were constructed to directly of broader generic concerns.43,44
parallel the self-report items. The imperfect
concordance observed between self- and proxy-
reports supports the need to measure the per- References
spectives of child and parent in evaluating pe- 1. Hlatky MA, Rogers WJ, Johnstone I, et al.
diatric HRQOL. Although self-report is
Medical care costs and quality of life after randomization
considered the standard for measuring per- to coronary angioplasty or coronary bypass surgery.
ceived HRQOL, it is the parent's perception of N Engl J Med 1997;336:92-99.
their child's HRQOL that influences health care 2. Spilker B. Quality of life and pharmacoeconom-
utilization.12 This perspective may help explain ics in clinical trials. Philadelphia, PA: Lippincott-Raven;
in part the pattern of higher correlations be- 1996.
tween proxy-report with the number of days 3. Gold M, Franks P, Erickson P. Assessing the
missed from school and work, care needed, and health of the nation: The predictive validity of a
work impact. The use of proxy-report to esti- preference-based measure and self-rated health. Med
mate HRQOL may be necessary when the child Care 1996;34:163-177.
is either unable or unwilling to complete the 4. Peskin SR. Applications of QOL measurements:
HRQOL measure. Nevertheless, proxy-reports A managed care perspective. Oncology 1995;9:S127-
should be conducted with the knowledge that a S128.
proxy rater's estimate may be insufficiently 5. Ganz PA. Impact of quality of life outcomes on
accurate .42 clinical practice. Oncology 1995;9:S61-S65.

810

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
Vol. 39, No. 8 PEDSQL 4.0

6. Seid M, Sadler BL, Peddecord KM, et al. 20. Vari JW, Seid M, Rode CA. The PedsQL:
Accountability: Protectingthe well-being of America's Measurementmodel for the PediatricQualityof Life
childrenand those who carefor them. Alexandria,VA: Inventory.Med Care1999;37:126-139.
NationalAssociationof Children'sHospitalsandRelated 21. Vari JW, Thompson KL, Hanson V. The
Institutions;1997. Vami/Thompson PediatricPainQuestionnaire:I. Chronic
7. WorldHealth Organization.Constitutionof the musculoskeletalpain in juvenile rheumatoidarthritis.
WorldHealth Organizationbasic document.Geneva, Pain1987;28:27-38.
Switzerland:WorldHealthOrganization; 1948. 22. Varni JW, Wilcox KT, Hanson V, et al.
8. Sprangers MAG, Aaronson NK. The role of Chronicmusculoskeletalpain and functionalstatus in
healthcareprovidersand significantothersin evaluating juvenilerheumatoidarthritis:An empiricalmodel.Pain
the qualityof life of patients with chronicdisease:A 1988;32:1-7.
review.J ClinEpidemiol1992;45:743-760. 23. Vari JW, Rode CA, Seid M, et al. The Pedi-
9. Guyatt GH, Juniper EF, Griffith LE, et al. atricCancerQualityof LifeInventory-32(PCQL-32):II.
Childrenand adult perceptionsof childhoodasthma. Feasibilityand range of measurement.J Behav Med
Pediatrics1997;99:165-168. 1999;22:397-406.
10. Czyzewski DI, MariottoMJ,BartholomewK, 24. Vari JW, Waldron SA, Gragg RA, et al.
et al. Measurementof qualityof well beingin a childand Developmentof the Waldron/Vami PediatricPainCoping
adolescent cystic fibrosis population. Med Care Inventory.Pain1996;67:141-150.
1994;32:965-972. 25. Thompson KL, Vami JW. A developmental
11. LangeveldJH, Koot HM, Loonen MCB,et al. cognitive-biobehavioral approachto pediatricpain as-
sessment.Pain1986;25:282-296.
A qualityof life instrumentfor adolescentswith chronic
headache.Cephalalgia1996;16:183-196. 26. Fairclough DL, Cella DF. FunctionalAssess-
ment of CancerTherapy(FACT-G): Non-response to
12. Vari JW, Setoguchi Y. Screeningfor behav-
individualquestions.QualLifeRes 1996;5:321-329.
ioral and emotionalproblemsin childrenand adoles-
centswith congenitalor acquiredlimbdeficiencies.Am J 27. Hays RD, Anderson R, Revicki D. Psycho-
Dis Child1992;146:103-107. metricconsiderationsin evaluatinghealth-relatedqual-
ity of life measures.QualLifeRes 1993;2:441-449.
13. Vari JW, Rapoff MA, Waldron SA, et al.
Chronicpain and emotional distress in childrenand 28. McHorney CA, Ware JE, Lu JFR, et al. The
adolescents.J DevelopBehavPeds 1996;17:154-161. MOS 36-item short-formhealth survey (SF-36):III.
Testsof data quality,scalingassumptions,and reliabil-
14. Vari JW, Katz ER, Colegrove R, et al. Ad-
ity across diverse patient groups. Med Care
justment of children with newly diagnosed cancer: 1994;32:40-66.
Cross-informant variance. J Psychosocial Oncol
29. CronbachLJ.Coefficientalphaandthe internal
1995;13:23-38. structureof tests. Psychometrika1951;16:297-334.
15. VarniJW,KatzER,Seid M, et al. ThePediatric
30. Nunnally JC, Bernstein IR. Psychometricthe-
CancerQualityof LifeInventory-32(PCQL-32):I. Reli-
ory3rd ed. New York,NY:McGraw-Hill; 1994.
abilityand validity.Cancer1998;82:1184-1196.
31. Pedhazur EJ, Schmelkin LP. Measurement,
16. VarniJW,KatzER,Seid M, et al. ThePediatric
CancerQualityof Life Inventory(PCQL):I. Instrument design,and analysis:An integratedapproach.Hillsdale,
NJ: Erlbaum; 1991.
development,descriptivestatistics,and cross-informant
variance.J BehavMed 1998;21:179-204. 32. McHorney CA, Ware JE, Raczek AE. The
MOS36-itemshort-formhealthsurvey(SF-36):II.Psy-
17. Achenbach TM, McConaughy SH, Howell chometricand clinical tests of validity in measuring
CT. Child/adolescentbehavioraland emotionalprob-
physical and mental health constructs. Med Care
lems: Implicationsof cross-informantcorrelationsfor 1993;31:247-263.
situationalspecificity.PsycholBull1987;101:213-232.
33. McHorneyCA, WareJE,RogersW, et al. The
18. Patrick DL, Deyo RA. Genericand disease-
validityand relativeprecisionof MOS short- and long-
specificmeasuresin assessinghealthstatusand quality formhealthstatusscalesand DartmouthCOOPcharts:
of life. Med Care1989;27:S217-S233. Resultsfrom the MedicalOutcomesStudy.Med Care
19. SprangersMAG, Cull A, BjordalK, et al. The 1992;30:MS253-MS265.
EuropeanOrganizationfor Researchand Treatmentof 34. Hayashi T, Hays RD. A microcomputerpro-
Cancerapproachto qualityof life assessment:Guide- gram for analyzing multitrait-multimethodmatrics.
lines for developingquestionnairemodules. Qual Life Behav Res Methods Instruments Comput 1987;
Res 1993;2:287-295. 19:345-348.

811

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions
VARNI ET AL MEDICALCARE

35. Hays RD, Hayashi T. Beyondinternalconsis- inary results from the IQOLA Project International
tencyreliability:Rationaleand user'sguideforMultitrait Quality of Life Assessment. Soc Sci Med
Analysis Programon the microcomputer.Behav Res 1995;41:1359-1366.
MethodsInstrumentsComput1990;22:167-175. 41. Keller SD, Ware JE, Gandek B. Testingthe
36. Cohen J. Statisticalpower analysisfor the be- equivalenceof translationsof widely used response
havioralsciences,2nd ed. Hillsdale,NJ:Erlbaum;1988. choicelabels:Resultsfromthe IQOLAProjectInterna-
37. Smith KW, Avis NE, Assmann SF. Distin- tional Quality of Life Assessment. J Clin Epidemiol
guishing between qualityof life and health status in 1998;51:933-944.
qualityof life research:A meta-analysis.Qual Life Res 42. BlazebyJM, WilliamsMH, AldersonD, et al.
1999;8:447-459. Observervariationin assessmentof qualityof lifein patients
38. Floyd FJ,Widaman KF. Factoranalysisin the withoesophageal cancer.BrJ Surg1995;82:1200-1203.
developmentand refinementof clinicalassessmentin- 43. Guyatt G, Feeny D, Patrick D. Issues of
struments.PsychologicalAssess 1995;7:286-299.
quality-of-lifemeasurementin clinicaltrials.Controlled
39. SPSS.SPSS8.0 forWindows.Chicago,IL:SPSS; ClinTrials1991;12:81S-90S.
1998. 44. Wallander JL, Vami JW. Effectsof pediatric
40. Bullinger M. German translation and psy- chronicphysicaldisorderson child and familyadjust-
chometrictesting of the SF-36 Health Survey:Prelim- ment.J ChildPsycholPsychiat1998;39:29-46.

812

This content downloaded from 165.190.89.176 on Tue, 16 Feb 2016 06:30:58 UTC
All use subject to JSTOR Terms and Conditions

You might also like