Participants were prospectively enrolled after presenting to a large, urban, children’s hospital ED for treatment of minor injury (defined as a single injury, occurring within 24 hours of presentation, in patients discharged to home after ED care). We included children and adolescents aged 2 to 18 years. Exclusion criteria included non–English-speaking patients or caretakers, trauma team activation, no parent in the ED, or suspected child maltreatment. The study was approved by the hospital’s institutional review board.

Enrollment occurred on randomly selected study days (18 h/d, 7 d/wk) during the 12-month period from June 1, 2008, through May 31, 2009. Informed consent and assent were obtained before enrollment.

Demographic and injury data were abstracted from the ED record at the time of the visit, and baseline HRQOL was assessed (child and parent reports).

Short-term clinical outcomes were collected at telephone follow-up interviews with the parent and with the child (if he or she was 8 years or older) at 1 week (attempts on days 6–9) and 2 weeks (attempts on days 13–18) after the ED visit. Age-appropriate HRQOL measures (child report for children and adolescents aged 5–18 years and parent report for children and adolescents aged 2–18 years) were also administered at the telephone follow-up.

The PedsQL is a multidimensional measure initially developed and validated for use in children with chronic disease.^34–36 The instrument uses parallel child (self) and parent (proxy) reports for children and adolescents aged 5 to 7, 8 to 12, and 13 to 18 years and parent report only for children aged 2 to 4 years. It takes 5 to 10 minutes to complete and has been validated for self-completion, face-to-face interview, and telephone administration. The total score (23 items) includes 2 subscales (physical health and psychosocial summaries) and is made up of items in the physical, emotional, social, and school functioning domains. Individual items are reverse scored and linearly transformed to a scale of 0 to 100, with higher numbers indicating higher HRQOL. The acute version uses a 7-day recall period instead of the 30-day period for the standard PedsQL. The acute version was validated in an outpatient subspecialty clinic setting^35,38 and in an acute exacerbation of a chronic disease.²³ A single study in pediatric ED patients assessed only the parent (proxy) reports.³³

For all families, the PedsQL was administered at the ED visit and again at the telephone follow-up. At the ED visit, the 7-day recall was specified as the 7 days before the injury, and these findings constitute the baseline PedsQL scores. Follow-up PedsQL scores were from the first successful follow-up telephone contact (at 2 weeks if 1-week attempts were unsuccessful). The PedsQL was administered according to the terms of the use agreement between the authors and distributors (http://www.mapi-trust.org).

Short-term clinical outcomes, specified as resulting from the minor injury, were collected by telephone interviews with the parent and with children and adolescents 8 years or older. They included days of pain/discomfort after the ED visit, days to return to baseline activities/routines for the child and family, and days of daycare/school/work or regularly scheduled activities missed by the child or the parent. To evaluate construct validity, clinical outcomes were considered individually, first continuously and then categorized as poor (lasting for ≥1 week) or good (<1 week). The 1-week cutoff was chosen a priori on the basis of previous studies of short-term outcome^32,39 and to correspond with the recall periods of the follow-up telephone interviews.

For known-group validity testing, an a priori composite grouping of clinical outcome was defined. As in previous studies, patients were dichotomized as having a composite good or poor outcome. A composite poor outcome was assigned if any 1 or more of the following was reported by the parent or the child: 7 or more days of pain, 7 or more days of abnormal patient or family activity, or 5 or more days of daycare, school, or work missed by the child or the parent. All other outcomes were assigned to the composite good outcome group.

Demographic characteristics for the study sample and patients lost to follow-up were compared using χ² test for categorical variables and unpaired, 2-tailed t tests or Mann-Whitney tests for continuous variables (based on data distribution). We calculated PedsQL scores according to the developer’s instructions. Treatment of missing items followed the developer’s protocol; scores were computed as the sum of items divided by the number answered. When more than half the items from any scale or subscale were missing, the score was not computed.³⁵ Data normality for reliability and validity analyses was assessed with the Kolmogorov-Smirnov test. Group comparisons were made using adjusted marginal means derived from univariate generalized linear models, with age and sex as covariates. Data analyses were conducted using commercially available software (SPSS, version 17.0; SPSS, Inc, Chicago, Illinois).

Feasibility was assessed by the success of telephone follow-up, by floor and ceiling effects (the percentage of scores at the bottom and top of the scale, respectively), and by calculating values for the percentage of missing items in the parent and patient report forms.

Instrument reliability reflects whether information is measured in a reproducible fashion, and internal consistency reliability looks at this reproducibility between instrument items. Internal consistency reliability was assessed using the Cronbach α, with values of at least 0.70 considered adequate for comparisons of groups and at least 0.90 for comparing individual patient scores.⁴⁰ Parent-child concordance was examined using intraclass correlation coefficients, with values of no more than 0.40 indicating poor to fair agreement; 0.41 to 0.60, moderate agreement; 0.61 to 0.80, good agreement; and 0.81 to 1.00, excellent agreement.⁴¹ We hypothesized that the internal consistency reliability in our setting would be similar to prior evaluations of the PedsQL, exceeding the standard for comparing groups for all subscales.^34,35,37,42 Moderate parent-child agreement was expected, as reported in previous PedsQL studies.^25,43

Test validity is assessed through evidence that the tool is measuring what is intended.⁴⁴ To evaluate construct validity, linear regression was used to compute the change in mean PedsQL scores for each day that the individual clinical outcomes persisted after the ED visit (ie, days with pain or days of disrupted activities). We hypothesized that HRQOL (PedsQL scores) should decrease as days of clinical symptoms increased.

Known group comparisons were used to demonstrate discriminative validity of the PedsQL in our setting.⁴⁵ Patients were grouped by injury characteristics or by clinical outcomes that were expected to represent different levels of HRQOL at follow-up. After adjusting for age and sex, mean PedsQL scores were calculated with 95% confidence intervals of group differences. For patients grouped by injury location and injury type, we hypothesized that the groups should not have significantly different PedsQL scores at baseline but show expected significant differences (statistically and clinically) at follow-up as in prior studies of long-term injury follow-up.^46–48 We expected that fractures would have poorer HRQOL at follow-up than soft-tissue injuries and that lower extremity injuries would have a greater negative effect on HRQOL than upper extremity injuries. For the groups dichotomized by their clinical outcomes as good and poor (as defined for individual and composite clinical outcomes), we expected to find an association between short-term clinical outcomes and follow-up PedsQL scores, with significantly higher mean scores for children with our definitions of good clinical outcomes than for those with poor clinical outcomes.

Responsiveness of a scale is the ability to measure change over time, commonly understood as true change after adjusting for the “noise” of chance or inherent variation of scale scores.^44,45,49 For this study, we assessed responsiveness by examining the change in PedsQL scores in relation to changes in clinical status during our follow-up period. We assessed distribution-based indicators of change, including the effect size (ES) and estimates of the minimal important difference (MID). Effect size was calculated as the mean change in score divided by the standard deviation of the score in the population at baseline.^32,50,51 In previous PedsQL studies, the MID of a score change of 4 has been proposed.³⁴ However, it has been recommended that several estimates of the MID should be calculated and used to establish a range for the MID.⁵² We calculated previously suggested distribution-based approximations of the MID, including the sample’s standard error of measurement (SEM), and one-half of the baseline standard deviation.^50,52–54 We estimated the SEM by the product of the standard deviation of the scale score and the square root of 1 minus the internal consistency reliability coefficient for the scale score.⁴³

A total of 404 children and adolescents aged 2 to 18 years and their parents were enrolled. Telephone follow-up was successful in 334 of these families (82.7%). Patient characteristics are detailed in Table 1. Differences between patients lost to follow-up (70 of 334) and those successfully contacted were not statistically significant except for injury location. The median number of days between baseline and follow-up PedsQL was 9 (interquartile range [IQR], 8–15). The mean PedsQL scores at baseline were nearly identical to those reported for healthy populations.^35,55

The percentages of all missing values were small. For parents with children aged 2 to 4 years, 111 of 2184 items (5.1%) were missing, but all missing items were from the school/daycare subscale for children not in daycare or school. For parents with children aged 5 to 18 years, 31 of 9015 items (0.3%) items were missing. For children aged 5 to 18 years, 34 of 6739 items (0.5%) were missing. Baseline PedsQL forms were easy to administer during the ED visit. Floor and ceiling rates of the baseline and follow-up PedsQL scores are detailed in Table 1 and are very similar for child reports to general and healthy populations in prior studies.^35,55 Parent reports had higher ceiling rates than these populations only at baseline in our study, which may reflect a contrast effect of the recent injury of their child.

Good reliability of the PedsQL in our ED setting was reflected by high Cronbach α values. All parent and child scales were internally consistent (α range, 0.73–0.93). The α values for total scores, in all but child reports for children aged 5 to 7 years (α = 0.84), approached or exceeded 0.90. Parent-child concordance showed moderate agreement (interclass correlation coefficient for the total score, 0.52; physical summary score, 0.48; and psychosocial summary score, 0.49).

Clinical outcomes reported by parents and children were considered separately. First, the relationships between follow-up PedsQL scores and individual clinical outcomes (days of pain and missed or disrupted activities) were evaluated (Table 2). We found the expected overall inverse relationship: for each day that the clinical symptoms persisted, there were consistent decreases in mean HRQOL scores. Table 3 gives known group comparisons of baseline and follow-up PedsQL scores for different injury characteristics, including injury type (soft-tissue vs fracture) and injury location (upper vs lower extremity). After adjusting for age and sex, the compared groups were not different at baseline, but HRQOL at follow-up for lower extremity injuries was clinically and statistically significantly poorer than for upper extremity injuries and for fractures than for soft-tissue injuries. Effect sizes, measuring score change over time adjusted for normal score variations, were moderate to large in the physical subscales, with ES for lower extremity injury consistently larger than ES for upper extremity injury and fracture ES larger than soft-tissue injury ES.

Tables 4, 5, and 6 use clinical outcomes categorized as good and poor for known group comparisons. In Tables 4 and 5, if an individual short-term outcome had resolved by 7 days (good outcome), children had significantly better follow-up HRQOL scores than if it persisted for 1 week or more (poor outcome). These differences in follow-up PedsQL scores were particularly large, ranging from 1 to 2 SDs of the baseline score in all but a few psychosocial subscale comparisons, and were statistically significant for all child- and parent-reported outcomes.

Table 6 details mean PedsQL total and subscale summary scores for patients grouped by our composite definition of good and poor outcomes. Differences between outcome groups were not significant at baseline but became significant at follow-up. We found large differences at follow-up in total scores (parent and child reports, 13.7 and 13.3, respectively) and in the physical summary score (22.4 and 25.0, respectively), with smaller but still clinically and statistically significant differences in summary scores between good and poor outcome groups in the psychosocial subscale (8.9 and 7.1, respectively).

Responsiveness was evaluated in several ways. In validity testing, PedsQL scores reflected decreased HRQOL in groups with increasing severity of clinical symptoms, and poor outcome groups had significant negative changes in score during the follow-up period. Effect sizes were moderate to large for fractures (total and physical summary ES, 0.56–1.87) and lower extremity injuries (0.80–2.44) and were substantially smaller for soft-tissue injuries (Table 3). Effect sizes for the full sample of all injuries were small for total score (0.26 for the parent report and 0.30 for the child report), moderate to large for the physical summary score (0.62 and 1.10, respectively), and minimal or insignificant in the psychosocial summary score (0.06 and 0.14, respectively). For our entire sample, the SEM ranged from 4.0 to 6.5 for the parent- and child-reported total and subscales, and the half-SD index ranged from 6 to 9 (data not shown).