The linkage between pediatric quality of life and health conditions: establishing clinically meaningful cutoff scores for the PedsQL

Abstract

Objective

To link pediatric health-related quality of life (HRQOL) and health conditions by establishing clinically meaningful cutoff scores for a HRQOL instrument, the PedsQL.

Methods

We conducted telephone interviews with 1745 parents whose children were between 2–18 years old and enrolled in the Florida KidCare program and Children’s Medical Services Network in 2006. Two anchors, the Children with Special Health Care Needs (CSHCN) Screener and the Clinical Risk Groups (CRGs), were used to identify children with special health care needs or chronic conditions. We established cutoff scores for the PedsQL’s physical, emotional, social, school, and total functioning using the areas under the curves (AUCs) to determine the discriminative property of the PedsQL referring to the anchors.

Results

The discriminative property of the PedsQL was superior, especially in total functioning (AUC > 0.7), between children with special health care needs (based on the CSHCN Screener) and with moderate and major chronic conditions (based on the CRGs) as compared to healthy children. For children < 8 years, the recommended cutoff scores for using total functioning to identify CSHCN were 83, 79 for moderate and 77 for major chronic conditions. For children ≥ 8 years, the cutoff scores were 78, 76 and 70, respectively.

Conclusions

Pediatric HRQOL varied with health conditions. Establishing cutoff scores for the PedsQL’s total functioning is a valid and convenient means to potentially identify children with special health care needs or chronic conditions. The cutoff sores can help clinicians to conduct further in-depth clinical assessments.

INTRODUCTION

There is a growing interest in measuring health-related quality of life (HRQOL) in addition to biomedical indicators to assess pediatric health outcomes [1]. HRQOL assessments can help identify unexpected functional disability, monitor disease progression, and improve physician-patient communication [2]. However, pediatricians’ practical use of HRQOL information in clinical decision-making is limited. It is estimated that < 25% of pediatric clinicians use outcome measures including HRQOL in their clinical practice [3], despite the American Academy of Pediatrics (AAP) recommendations that psychosocial assessments are performed at every well-child check [4, 5]. A key barrier to using these measures is that clinicians lack a sense of the linkage between HRQOL measures and health conditions. They also desire the guidance to help interpret HRQOL scores in a clinically meaningful way [6–8].

Establishing cutoff scores for HRQOL measures is a potentially useful approach for increasing the use of HRQOL information in clinical practice [9]. By having a cutoff score available, a clinician can use HRQOL measures as an aid to suggest whether further in-depth, targeted clinical assessments are necessary. In adult HRQOL measures, cutoff scores have been established for various instruments, such as the Barthel Index that measures physical disability [10] and the Beck Depression Inventory that measures depressive symptoms [11]. To our knowledge, only one study has reported cutoff scores of a pediatric HRQOL instrument [12]. This study used one standard deviation below the population mean to establish the cutoff scores of the PedsQL and found that cutoff scores were close to the mean scores for children with major chronic conditions. For example, the cutoff score for total functioning (65.4) [12] was similar to mean scores of children who were currently on treatment for cancer (66.7 [13]), asthma (68.8 [14]), rheumatic condition (68.7 [14]), and end stage renal disease (69.6 [14]). However, the “one standard approach” used in this previous study to establish cutoff scores for clinical applications is questionable because it is sample dependent and its validity as it relates to clinical measures is unclear. Additionally, this previous study assumed that the cutoff scores of the PedsQL were universal across all age groups [15]. This present study addresses the gap in knowledge by establishing cutoff scores using clinical relevant measures and reporting the scores by different age groups.

Successful establishment of cutoff scores for HRQOL measures relies on the use of a “gold standard” to capture a patient’s genuine health status. Unfortunately, a universally acceptable gold standard for assessing pediatric health status does not exist [16]. Alternatively, the use of “anchors” can be considered in lieu of a gold standard if the anchors adequately indicate disease status, are clinically interpretable, and show acceptable measurement properties [17].

Two types of measures can serve as anchors to establish cutoff scores for pediatric HRQOL measures: 1) diagnostic information grouped into health status categories by software programs such as the Clinical Risk Groups (CRGs) and 2) parent-reported screening tools that asses the child’s health such as the Children with Special Health Care Needs (CSHCN) Screener [18]. Software programs like the CRGs use group diagnoses assigned at the time of a health care visit and collected from claim and encounter data, making it efficient to classify children into mutually exclusive categories indicative of healthy, acute or chronic conditions [19]. In contrast, the CSHCN Screener relies on parental report of a child’s functional ability (including physical, psychological and social functioning) and health service utilization to classify children who are at risk for or who have special health care needs [18]. These two approaches are complementary and each can be useful for research and programmatic evaluations.

The purpose of this study was to investigate the associations between pediatric HRQOL using the PedsQL 4.0 and their health conditions. We specifically established clinically meaningful cutoff scores of the PedsQL for children who are 2–18 years old and potentially have special health care needs or chronic conditions. The PedsQL is a widely used generic instrument for measuring pediatric HRQOL and covers the domains of physical, emotional, social and school functioning. Evidence suggests that the PedsQL has acceptable psychometric properties and is suitable for use with healthy children and children with acute or chronic conditions [20, 21]. We used the CSHCN Screener and the CRGs as anchors to establish cutoff scores for the total and individual domains of the PedsQL.

METHODS

Study population

This is a cross-sectional study using data collected from two sources, with the child as the unit of analysis: 1) the 2006 annual evaluation of the Florida KidCare Program, which is composed of Medicaid and the Title XXI State Children’s Health Insurance Program (SCHIP), and 2) the 2006 satisfaction survey of the Florida Children’s Medical Services Network (CMSN), which is the State Title V CSHCN Program for children with special health care needs. All children in this sample were also enrolled in Medicaid. This pediatric population under our investigation is important to assess because they are at a great risk of poor outcomes due to poor socioeconomic circumstances and a high prevalence of chronic and life-limiting conditions [22, 23]. This population is also racially and ethnically diverse [24].

Data collection

As part of a statewide annual evaluation, we conducted telephone surveys using a sample of parents between 9/2006 and 12/2006 for the KidCare evaluation and between 12/2006 and 03/2007 for the CMSN survey. Parents whose children were enrolled in the KidCare or the CMSN for six months or longer were randomly selected to participate. Parents of the children were first sent an introductory letter explaining the purpose of the surveys. For those parents who agreed to participate, we sent informed consent and set up separate dates and time for interviews. The telephone surveys are conducted through the UF Bureau of Economic and Business Research (BEBR) using the Sawtooth WinCATI System. Multiple callbacks (at maximum of 10 times) were performed if phone numbers were busy or not answered.

The response rate was 50% for those in the KidCare evaluation and 49% for those in the CMSN survey, which is similar to other studies conducted with families that are publically insured [25, 26]. Responders and non-responders did not differ significantly on parental age, gender, or educational background (p > 0.05). The study sample consisted of 2269 subjects who completed the surveys (1642 from the KidCare and 627 from the CMSN). In the present study, we used the PedsQL scoring instructions to impute item scores of a domain for subjects who had less than 50% of items missing by using a mean score of the completed items for that domain. If more than 50% of the items in the domain were missing, the domain score was not calculated. We recognize that this approach is based on an assumption – missing at random. Our recent HRQOL imputation study, however, suggests that it is difficult to justify the appropriate use of imputation methods unless sufficient auxiliary variables are available [27]. After calculating the domain score for each subject we further deleted those subjects who had missing scores in any domain because we wanted to estimate scores of total functioning using scores of individual domains (physical, emotional, social, and school functioning). In total, 524 subjects were further excluded (95 who were missing the entire HRQOL section and 429 who were missing one or more PedsQL domain scores), leaving 1745 subjects for the final analyses (1290 from the KidCare and 455 from the CMSN). Our ad hoc analyses suggest that those who were missing HRQOL data did not differ significantly from those completing that section on the characteristics of parental age, gender, and educational background (p > 0.05).

Health-related quality of life: The PedsQL

We used the parent proxy version of the PedsQL 4.0 to assess pediatric HRQOL. The PedsQL provides four validated age-specific versions, 2–4, 5–7, 8–12 and 13–18 years, each with minor modifications in wording based on the children’s ages [20, 21]. The PedsQL consists of 23 questions which cover four domains: physical, emotional, social, and school functioning. A domain-specific score is calculated from the corresponding questions, ranging from 0 (worst HRQOL) to 100 (best HRQOL), which can be combined for a total functioning score.

The anchors: CSHCN Screener

We used the CSHCN Screener to assess whether children had a special health care need based on parental reports. The CSHCN Screener, by design, identifies children who currently use health services or those who may need services above what is normally expected for children [18]. The Screener asks whether a child: 1) needs or uses prescription medicines prescribed by a doctor; 2) has above-routine need for medical, mental health, or educational services; 3) is limited or prevented in any way in his or her ability to do things that most children of the same age can do; 4) needs or uses specialized therapies such as speech, occupational, and physical therapies; and 5) needs or receives treatment or counseling for an emotional, behavioral, or developmental problem.

If respondents answer “yes” to any of the five questions, they are then asked up to two follow-up questions to determine whether the consequence is attributable to a medical, behavioral, or other health condition lasting or expected to last at least 12 months. Only those who provide positive responses to one or more question sequences and each of the associated follow-up questions are classified as having a special health care need [18]. Schmidt et al. reported that among children with chronic conditions (e.g., asthma, arthritis, dermatitis, epilepsy, cystic fibrosis, and cerebral palsy), 80% of them were classified as having a special health care need using the CSHCN Screener [28]. Additionally, Bethell et al. demonstrated that compared to children without special health care needs, CSHCN were more likely to report fair/poor vs. excel/good health status (odds ratio [OR] = 6.9; p < 0.001) and more likely to visit their doctors vs. no visits (OR = 6.7; p < 0.001) [18].

The anchors: CRGs

The CRGs is an International Classification of Diseases, 9^th Revision (ICD-9-CM)-based system that uses claim and encounter data to classify children into mutually exclusive categories: 1) healthy, 2) significant acute conditions, 3) minor chronic conditions (e.g., attention deficit/hyperactivity disorder), 4) moderate chronic conditions (e.g., asthma), or 5) major chronic conditions (e.g., cystic fibrosis) [19]. The CRGs classification for the children represents their health status during the one year time period immediately preceding the telephone survey. Florida’s Agency for Health Care Administration (AHCA) provided health care claims and encounter data to generate the CRGs. We excluded children classified as having significant acute conditions from the analyses because the available case number is small (Table 1). Rolnick et al. found that the 66% to 73% of children identified as having a chronic condition by the CRGs were confirmed by through medical record reviews [29]. The use of the CRGs also yields great performance in the prediction of medical expenditures, with a R² = 42.8% [30].

Table 1.

Characteristics of study sample (N=1745)

	Mean (SD) or %
Age in years (SD)	10.3 (4.6)
Age in years (%)
2–4	11.4
5–7	22.8
8–12	29.9
13–18	35.9
Gender (%)
Male	53.0
Female	47.0
Race (%)
White	37.9
Black	28.5
Hispanic	27.1
Other	6.6
Children with special health care needs (%)
Yes	53.8
No	46.3
Clinical risk groups (%)
Healthy	45.5
Significant acute †	6.0
Mild chronic	10.2
Moderate chronic	23.5
Major chronic	14.9

^†Children with significant acute conditions were excluded from the analyses

Analyses

Score differences in HRQOL by the anchors

We estimated the differences in HRQOL scores among children with different health conditions using the CSHCN Screener and the CRGs. Specifically, for the CSHCN Screener, we compared HRQOL scores among children who have special health care needs to those who do not (reference group). For the CRGs, we compared HRQOL scores for children who have mild, moderate or major chronic condition to those who are healthy (reference group). We estimated an effect size to demonstrate the magnitude of score differences, which is the absolute discrepancy in HRQOL scores between a specific group (e.g., major chronic condition) and the referent (e.g., healthy) divided by the standard deviation (SD) of the score in the referent. We defined an effect size as ignorable (< 0.2), small (0.2–0.49), moderate (0.5–0.79) or large (≥0.8) [31]. If the effect size is moderate to large, this suggests that the use of the CSHCN Screener and/or the CRGs to establish cutoff scores is meaningful and valid.

Cutoff scores of the PedsQL

We used the Receiver Operating Characteristic (ROC) approach [32] to determine the extent to which the HRQOL scores from the PedsQL can optimally predict a health care need as indicated by the CSHCN Screener and chronic conditions as defined by the CRGs. We used the areas under the ROC curves (AUCs) to measure discriminative property of the PedsQL by the CSHCN Screener and the CRGs. We considered the area under the ROC curve as acceptable if it is between 0.7 and 0.8 and excellent if > 0.8 [33]. In turn, we generated different cutoff scores for the PedsQL based on their sensitivities and specificities associated with discriminating children who have special health care needs or chronic conditions [34, 35]. We established the optimal cutoff scores for individual HRQOL domains based on the equivalent sensitivities and specificities. Sensitivity was defined as the percentage of children with special health care needs who were identified by the PedsQL given a cutoff score. Specificity was defined as the percentage of children without special health care needs who were not identified by the PedsQL given a cutoff score. We chose this balance because we want to not only recommend the identification of as many children who have special health care needs or chronic conditions as possible, but also discriminate against false positives for children who do not have special care needs or chronic conditions.

For the CSHCN Screener, we established a single cutoff score for each HRQOL domain by comparing children with special health care needs to those without special health care needs. For the CRGs, we established three cutoff scores for each domain by comparing children with minor, moderate, or major chronic conditions to healthy children.

Because age, gender and race/ethnicity may confound the association of PedsQL scores with anchors (the CSHCN and the CRGs) [12, 15], we tested any associations and report the cutoffs by significant stratifications [36]. We conducted the ROC and cutoff score analyses using SAS 9.1.3 [37] and other analyses using STATA 9.0 [38]. This study received approval of human subjects research from the Institutional Review Board at the University of Florida.

RESULTS

Characteristics of study sample

Table 1 shows the characteristics of the 1745 children analyzed in this study. The CSHCN Screener identified half (54%) of the children with special health care needs, and the CRGs classified 54% as non-healthy: 6% had significant acute conditions, 10% had minor chronic, 24% had moderate chronic, and 15% had major chronic conditions. The rate of CSHCN reported in this study was higher than the national average (13–20%) [39] because our samples were collected from poor socioeconomic circumstances who may have more impaired health status. In addition, some of our samples were collected from the Children’s Medical Services Network (CMSN), which is the State Title V CSHCN Program for children with special health care needs.

We found that the classification of children’s health status by the CSHCN Screener was consistent with the CRGs. In fact, by using the CRGs, 99% of children with major chronic conditions, 92% with moderate chronic conditions and 80% with minor chronic conditions were identified with special health needs. In contrast, only 32% of children who were classified as healthy using the CRGs were identified with special health care needs using the CSHCN Screener.

In bivariate analyses, we found that age was significantly associated with the PedsQL scores and health conditions classified by the CSHCN and the CRGs (p < 0.05); race/ethnicity and gender did not (p > 0.05). Specifically, children who were older and severer in health conditions classified by the CSHCN or the CRGs had more impaired HRQOL compared to their counterparts. With further testing, we found that the magnitude in differences for the PedsQL scores associated with anchors between ages 2–4 and 5–7 years and between ages 8–12 and 13–18 years, respectively, were not significant. This allows for collapsed reporting of ages 2–7 and 8–18 years.

Score differences in HRQOL: the CSHCN Screener

Table 2 shows the score differences between the classifications of the CSHCN Screener. Compared to children without special health care needs, HRQOL scores were significantly lower among children with special needs across all domains (p < 0.001). Of note, the differences in effect size were greater for each functioning (except school functioning) among children ≥ 8 years compared to children < 8 years. In the age group < 8 years, the effect sizes were large in the domains of total, physical, social and school functioning, but moderate in the domains of emotional functioning. In the age group ≥8 years, the effect sizes were large across all domains.

Table 2.

Mean score differences in the PedsQL among the CSHCN groups

Domain	Age < 8 years				Age ≥ 8 years
	Mean	SD	Difference	Effect size †	Mean	SD	Difference	Effect size
Total
No CSHCN ‡	87.6	11.2	Reference		85.4	11.6	Reference
CSHCN	72.5	17.6	15.1***	1.35	66.1	18.7	19.3***	1.66
Physical
No CSHCN	91.7	11.4	Reference		90.5	12.8	Reference
CSHCN	76.1	22.9	15.6***	1.37	70.0	26.6	20.9***	1.64
Emotional
No CSHCN	82.4	15.6	Reference		82.2	16.0	Reference
CSHCN	72.3	20.0	10.1***	0.65	67.6	21.9	14.6***	0.91
Social
No CSHCN	88.5	15.9	Reference		86.8	16.5	Reference
CSHCN	72.3	23.0	16.2***	1.02	64.7	25.3	22.2***	1.35
School
No CSHCN	84.5	16.3	Reference		78.9	17.9	Reference
CSHCN	66.4	22.0	18.1***	1.11	60.5	21.5	18.4***	1.03

^†Magnitude of effect size: ignorable (<0.2), small (0.2–0.49), moderate (0.5–0.79), large (≥0.8);

^‡CSHCN: Children with special health care needs;

^***p<0.001

Score differences in HRQOL: the CRGs

Table 3 shows the score differences between the groups defined by the CRGs. Compared to healthy children, those with major chronic conditions demonstrated the largest discrepancies in the HRQOL domain scores, followed by those with moderate and then minor chronic conditions. In the age group ≥ 8 years, the effect sizes were large for children with major or moderate chronic conditions, and moderate for children with minor chronic conditions across all domains. The effect sizes were largest for children with major chronic conditions in the domains of physical and total functioning, which were 2.12 and 1.70, respectively. In the group of children < 8 years, the effect sizes were at least moderate for children with major or moderate chronic conditions. However, the effect sizes for children with minor chronic conditions were small in the total, social and school functioning, but ignorable in the remaining domains.

Table 3.

Mean score differences in the PedsQL between the CRG groups

Domain	Age < 8 years				Age ≥ 8 years
	Mean	SD	Difference	Effect size ‡	Mean	SD	Difference	Effect size‡
Total
Healthy	83.7	15.2	Reference		82.0	15.1	Reference
Minor †	80.2	12.8	3.5	0.23	70.3	17.9	11.7***	0.77
Moderate †	73.3	15.4	10.5***	0.69	65.4	19.6	16.6***	1.10
Major †	63.9	20.8	19.8***	1.30	56.4	19.3	25.6***	1.70
Physical
Healthy	89.5	15.5	Reference		87.0	17.0	Reference
Minor	87.5	11.9	2.0	0.13	78.6	20.9	8.4**	0.49
Moderate	77.0	19.1	12.5***	0.81	69.6	26.1	17.4***	1.02
Major	62.2	29.9	27.4***	1.77	51.1	31.3	36.0***	2.12
Emotional
Healthy	77.5	18.3	Reference		80.6	17.4	Reference
Minor	76.8	18.0	0.7	0.04	67.6	24.2	13.1***	0.75
Moderate	75.6	17.8	1.9	0.10	66.4	23.9	14.2***	0.82
Major	67.6	22.5	9.9**	0.54	65.7	22.2	15.0***	0.86
Social
Healthy	82.9	21.3	Reference		81.2	22.0	Reference
Minor	78.8	19.3	4.2	0.20	68.9	25.1	12.3***	0.56
Moderate	71.0	21.7	12.0**	0.56	62.7	26.8	18.5***	0.84
Major	70.5	24.4	12.4**	0.58	58.3	25.1	22.9***	1.04
School
Healthy	80.9	19.2	Reference		76.2	19.6	Reference
Minor	73.4	19.8	7.5	0.39	61.3	21.6	14.9***	0.76
Moderate	66.5	22.6	14.4***	0.75	60.3	21.6	15.9***	0.81
Major	54.8	22.5	26.1***	1.36	53.6	23.7	22.5***	1.15

^†Minor, moderate, and major chronic conditions;

^‡Magnitude of effect size: ignorable (<0.2), small (0.2–0.49), moderate (0.5–0.79), large (≥0.8);

^**p<0.01,

^***p<0.001

The moderate to strong effect sizes by the PedsQL for the CSHCN and the CRG groups (especially moderate and major chronic conditions) suggests a high level of discrimination to use both anchors to establish cutoff scores.

Area under the ROC curve and cutoff scores of the PedsQL

Table 4 shows the areas under the ROC curves (AUCs) and the associated cutoff scores using the CSHCN Screener. The AUCs were acceptable (> 0.7) for all domains of the PedsQL, except for the emotional functioning among children < 8 years. AUCs were larger for the total functioning than the individual functioning domains. Likewise, AUCs were larger for older children (≥ 8 years) than younger children (< 8 years) across all domains (Table 4 and Figure 1). This yielded optimal cutoff scores that were consistently lower for older children compared to younger children.

Table 4.

AUC, optimal cutoff scores and associated sensitivity and specificity for individual PedsQL domains: CSHCN as an anchor

Domain	Age < 8 years				Age ≥ 8 years
	AUC †‡	Sensitivity	Specificity	Cutoff	AUC	Sensitivity	Specificity	Cutoff
Total	0.76	0.71	0.73	83	0.81	0.74	0.74	78
Physical	0.72	0.67	0.65	91	0.75	0.71	0.69	88
Emotional	0.65	0.56	0.70	75	0.70	0.65	0.70	75
Social	0.71	0.67	0.65	85	0.77	0.73	0.66	80
School	0.74	0.68	0.70	75	0.74	0.71	0.65	70

^†Magnitude of AUC: acceptable (0.7–0.8) and excellent (> 0.8);

^‡Reference group: children without a special health care need.

Figure 1.

ROC curves associated with CSHCN (left panel) and CRGs (right panel) for PedsQL’s the total functioning

Table 5 shows the AUCs and the associated cutoff scores using the CRGs. Compared to the CSHCN Screener, the AUCs using the CRGs were smaller yet the values of the AUCs were large and at the acceptable level in the domains of total and physical functioning. Using healthy children as the referent, the AUCs for children with major chronic conditions were largest, followed by those with moderate or minor chronic conditions (Table 5 and Figure 1). This pattern was observed for all domains and across both age groups.

Table 5.

AUC, optimal cutoff scores and associated sensitivity and specificity for individual PedsQL domains: CRG as an anchor

Domain	Age < 8 years				Age ≥ 8 years
	AUC †‡	Sensitivity	Specificity	Cutoff	AUC	Sensitivity	Specificity	Cutoff
Total
Minor #	0.60	0.60	0.60	82	0.70	0.65	0.65	78
Moderate #	0.70	0.69	0.68	79	0.75	0.71	0.69	76
Major #	0.78	0.74	0.73	77	0.85	0.79	0.80	70
Physical
Minor	0.60	0.55	0.64	91	0.63	0.62	0.62	88
Moderate	0.71	0.69	0.69	88	0.71	0.65	0.70	84
Major	0.80	0.74	0.75	84	0.83	0.76	0.78	78
Emotional
Minor	0.52	0.45	0.62	75	0.66	0.61	0.64	75
Moderate	0.54	0.51	0.62	75	0.68	0.60	0.64	75
Major	0.64	0.65	0.62	75	0.70	0.63	0.71	70
Social
Minor	0.60	0.55	0.57	85	0.65	0.57	0.63	75
Moderate	0.66	0.69	0.63	80	0.70	0.65	0.63	75
Major	0.65	0.67	0.63	80	0.76	0.68	0.70	70
School
Minor	0.61	0.60	0.63	75	0.69	0.65	0.58	70
Moderate	0.68	0.64	0.63	75	0.71	0.66	0.70	65
Major	0.81	0.72	0.70	67	0.76	0.70	0.70	65

^†Magnitude of AUC: acceptable (0.7–0.8) and excellent (> 0.8);

^‡Reference group: healthy children;

^#Minor, moderate, and major chronic conditions

The optimal cutoff scores were highest for children with minor chronic conditions, followed by those with moderate chronic conditions, and then those with the major chronic conditions. In addition, the cutoff scores were higher among younger (< 8 years) than older children (≥ 8 years). Of note, the cutoff scores associated with the CSHCN Screener and the CRG’s minor chronic conditions were equivalent (Table 4 and 5).

Because AUCs associated with the CSHCN Screener and the CRGs’ moderate and major chronic conditions for the total functioning domain were acceptable (> 0.7) and superior to individual domains in both age groups, we therefore recommend using the total functioning of the PedsQL to identify CSHCN and those with chronic conditions. For children < 8 years, the recommended cutoff scores for the total functioning to identify potential CSHCN, moderate and major chronic conditions were 83, 79 and 77, respectively. For children ≥ 8 years, the recommended cutoff scores were 78, 76 and 70, respectively.

DISCUSSION

In this study, we investigated the associations of pediatric HRQOL as measured by the PedsQL with health conditions as measured by the CSHCN Screener and the CRGs. We also established clinically meaningful cutoff scores for the PedsQL. Using the CSHCN Screener and the CRGs as anchors, we found that the discriminative property of the PedsQL, as indicated by effect size and area under the ROC curve, was acceptable for both age groups, particularly in total functioning for children with special health care needs as identified by the CSHCN Screener and for those with moderate and major chronic conditions as measured by the CRGs. The cutoff scores varied by age group, with younger children (< 8 years) having higher scores than older children (8 ≥ years).

Compared to a previous study of cutoff scores, which was based on one standard deviation below the population mean [12], the cutoff scores reported in this present study were slightly higher. The cutoff scores established in both studies, however, were not based on the same age categories (our study was based on children < 8 years and ≥ 8 years, whereas a previous study focused on children across the entire age 2–18 years). For total functioning, the cutoff scores derived from the present study were 82 for age < 8 years and 78 for age ≥ 8 years, which were higher than the previous study (cutoff = 65) [12]. For physical functioning, the cutoff scores derived from the present study were 84 for age < 8 years and 78 for age ≥ 8 years, which were also higher than the previous study (cutoff = 63) [12]. Importantly, we found that the cutoff scores established in the present study were bounded by mean scores of children who were enrolled in one large scale study (N=3,652) and had a variety of major chronic conditions (e.g., cancer, rheumatic condition, cerebral palsy, asthma, renal disease and so on) [15]. For example, the cutoff score of total functioning corresponding to CRG major chronic conditions for children < 8 years old was 77, which was bounded by mean scores of 70–79 for children with a variety of major chronic conditions in a previous study [15]. The cutoff score corresponding to CRG major conditions for children ≥ 8 years was 70, which was bounded by mean scores of 69–73 [15]. In contrast, the cutoff score of total functioning based on one standard deviation below the population mean (65.4) [12] was below the lower boundary of mean scores. Therefore, given the strong linkage between the cutoff scores and clinical measures, the present study supports the validity and clinical meaning of using this cutoffs system in clinical practice.

Pediatric HRQOL measures, despite numerous studies showing acceptable psychometric properties and clinical feasibility, are not frequently used in clinical settings outside of clinical research [1]. Although this study attempts to translate this research tool into a clinically useful one, a major problem in establishing cutoff scores is the lack of a gold standard for comparisons. Guyatt et al., highlighted that the anchors selected must be measured independently, interpretable, and at least moderately correlated with the target HRQOL instruments [40]. Revicki et al., suggested that to facilitate the interpretation of HRQOL scores, the selection of anchors should be based on clinical endpoints (e.g., laboratory measures and clinical ratings), patient-reported health outcomes, or some combination of both [17]. For pediatric populations, the use of other anchors such as behavioral observations and school performance derived from self-reports or expert ratings (e.g., teacher) may also be useful for establishing cutoff scores for HRQOL measurements. We believe that the CSHCN Screener and the CRGs adequately fulfill clinical criteria and provide cutoffs with statistically acceptable qualities for assessing children’s total functioning.

Our use of the CSHCN Screener and the CRGs provides complementary information. We are not surprised that they generated different cutoff scores because, by design, these tools capture different concepts of health and are not entirely interchangeable. One study reported that about 70% of children who were identified as having a special health care need using the CSHCN Screener were also classified as having a chronic condition by the CRGs, whereas the remaining 30% were not [18]. Because the CRGs are based on diagnoses assigned at the time of a health care encounter, children with chronic conditions who have not used health care during the classification period may be missed [41]. In addition, children with chronic conditions who were seen for preventive care or the treatment of minor conditions, and did not have their underlying chronic condition recorded, would not be identified as having a chronic condition by the CRGs. In contrast, the CSHCN Screener is not dependent on information recorded during a health care encounter, but covers broad health issues (including limitations in functioning), reliance on prescription medication, and elevated service use. The CSHCN Screener also focuses on children who are at risk for developing a special health care need [42]. We think that the use of these contrasting approaches will provide a necessary variation to establish cutoff scores that are valid and reasonable.

The cutoff scores provide some interesting variations. We found that different sets of cutoff scores are necessary for different ages. We also found that the discriminative property of the PedsQL for the CSHCN Screener and the CRGs was greater in older children (≥ 8) than younger children (< 8 years). This could be interpreted that the PedsQL may be more sensitive in identifying CSHCN and those with chronic conditions for older children given the simple fact that older children may have more chronic conditions than younger children [43]. This dual set of age cutoffs also echoes the cognitive development theory suggesting that around age 7–8 years children demonstrate more concrete thoughts about health status than their counterpart, allowing for different reporting of HRQOL [44, 45]. Alternatively, this finding may suggest that parents are using different internal standards to judge their child’s HRQOL depending on the child’s age and/or developmental stage, which is known as a measurement non-invariance [46]. Further studies are needed to explore this specific issue.

The provision of clinically meaningful cutoff scores of the PedsQL can help identify children who are potentially at risk of having special health care needs or severe health conditions. Establishing cutoff score of HRQOL explicitly expands the initial utility of HRQOL instruments, which is providing information of physical, emotional, social and school functioning. Clinicians in community or primary care settings can use the PedsQL as a screening tool to assist in making clinical assessments in community or primary care settings, when the complex diagnostic information and software programs such as the CRGs are not available. If the rating in the PedsQL is lower than an established cutoff score, then the child would likely benefit from more sophisticated diagnostic assessments or referral to other health care specialists [6, 47]. However, there are barriers to implementing such a program, including financial and time constraints. These barriers have been identified in other areas of screening in primary care pediatrics [8, 48]. More research is needed to demonstrate how to best implement periodic screening using the PedsQL, as well as other measures to screen for psychosocial problems to facilitate meeting the AAP requirements for such screening [4, 5].

Some potential limitations merit attention. First, this study is restricted to children who were enrolled in Florida KidCare. This population is at or below 200% of the Federal Poverty Level. Parents with lower incomes may perceive their children’s HRQOL differently than higher income families. Therefore, the generalizability of our findings is limited because cutoff scores are expected to vary within the context of different populations. Second, we did not cross-validate our data. The use of a cross-validation method is to iteratively split the data into training and testing samples for establishing the cutoff scores and evaluating their performance (sensitivity, specificity, and areas under the curves). The strength is that the training and testing samples would be independent of each other, and hence the findings would not be overly optimistic. Further studies are also needed to cross-validate model performance of the present study by using new samples which are comprised of children of different social economic backgrounds and different health conditions. Third, we only established cutoff scores for the PedsQL using the CSHCN Screener and the CRGs. If different anchors (e.g., the Questionnaire for Identifying Children with Chronic Conditions – a parent report measure [49] or the Ambulatory Care Groups – that uses diagnostic information [50]) were used, the cutoff scores likely would change. Further studies are encouraged using diverse populations and anchors to focus on a small range of cutoffs or a single value by pooling the findings of several analyses [51]. Fourth, parental mental and functioning status may influence the establishment of cutoff scores. Previous studies suggest that parents’ self-reported depression and functioning status may confound their reports of children’s HRQOL [52, 53]. This may lead to estimations of different cutoff scores, especially if these factors are not equally distributed among disease groups (e.g., healthy, minor, moderate and major groups in the CRGs). However, in the present study we did not collect these variables and are unable to verify these confounding effects. Finally, the use of parents’ ratings of their child’s HRQOL may differ from the child’s or adolescent’s own rating. Further studies in older children are needed to determine whether the derived cutoff scores are valid for the PedsQL when completed by children themselves.

In summary, we found that pediatric HRQOL was significantly associated with health conditions. Establishing cutoff scores for the PedsQL’s total functioning provides a valid and convenient mean to help clinicians identify children potentially with special health care needs or chronic conditions in community and primary care settings.