Comparison of a Cardiac Specific Versus a General Instrument to Assess Health-Related Quality of Life 4-5 Years After Cardiac Surgery in Early Infancy

Abstract

Background

We aimed to compare health-related quality of life (HRQL) using a disease-specific and a generic instrument in children after complex cardiac surgery in early infancy.

Methods

A prospectively followed inception cohort of children having complex cardiac surgery at age ≤6 weeks had HRQL assessed at age 4-5 years using Pediatric Quality of Life Inventory (PedsQL) 3.0 Cardiac Module. Scores were compared with the PedsQL Generic Core Scales in the same population. Predictors of Cardiac Module dimensions were determined by univariate and multivariate linear regression, with P ≤ 0.05 considered statistically significant.

Results

Among 425 patients who had the PedsQL 3.0 Cardiac Module, the HRQL mean (standard deviation) scale scores were as follows: Heart Problems and Treatment, 78.7 (19.0); Treatment II, 94.5 (15.2); Perceived Physical Appearance, 93.6 (13.9); Treatment Anxiety, 71.1 (29.5); Cognitive Problems, 67.3 (22.7); and Communication, 67.5 (31.2). Correlations between Cardiac Module and Generic Core Scales were as follows: Heart Problems and Treatment × Physical Functioning 0.57; Perceived Physical Appearance × Psychosocial Summary 0.21; Cognitive Problems × School Functioning 0.48; Treatment Anxiety × Physical Functioning 0.38; and Communication × School Functioning 0.30. Thirteen predictor variables were independently associated with at least 1 dimension of the Cardiac Module, and these only partially overlapped with independent predictors of Generic Core Scale summary scores.

Conclusions

The Cardiac Module measured a different HRQL construct than the Generic Core Scales, reflected by low correlations among scores and different independent predictors in the same population of children. Future studies should determine the relation of these scores with longer-term outcomes including school achievement, mental health, and well-being.

Mortality rates in children with congenital heart disease (CHD) have decreased significantly in the last few decades.1, 2, 3 Attention now has changed from mortality alone to also considering morbidity in surviving children. According to the study published by Fink et al.⁴ that described a Core Outcome Set for Pediatric Critical Care, aiming to reduce the heterogeneity in pediatric critical care medicine outcome publications, research should focus on 5 domains: cognitive, emotional, physical, family, and overall health. The latter includes health-related quality of life (HRQL).

The instruments used to measure HRQL can be divided into generic or specific tools. Generic instruments can be used in a wide range of diseases and in healthy states, and allow comparisons of HRQL among different diseases and with healthy populations. On the other hand, specific instruments are developed to evaluate a more specific population, with items more focused on common problems related to specific symptoms or conditions. One of the main reasons to use specific instruments is that it is believed that they are more responsive to changes in that specific disease condition than the generic instruments.⁵

One of the most used generic instruments to assess HRQL in children is the Pediatric Quality of Life Inventory (PedsQL), a family of instruments with a generic scale, and with the option of multiple disease-specific instruments.⁶ In 2003, Uzark et al.⁷ published on the use of the PedsQL 3.0 Cardiac Module, the specific instrument of the PedsQL used for children with heart disease. Since then, the instrument has been translated into multiple languages and applied in different pediatric populations with cardiac diseases.8, 9, 10, 11, 12

Although HRQL is an important outcome after pediatric critical illness, there is a gap in the literature on PedsQL 3.0 Cardiac Module outcomes in the most severely affected children with critical CHD, those who have complex cardiac surgery (CCS) early in infancy. These outcomes, their validity, and their acute care predictors are important to determine in this critically ill CHD population to work toward improving outcomes over time. In this study, we aimed to address this gap by using the PedsQL 3.0 Cardiac Module to (1) describe the HRQL at age 4-5 years in children having had CCS at ≤6 weeks of age, (2) assess the construct validity of the PedsQL 3.0 Cardiac Module (by determining correlation with PedsQL 4.0 Generic Core Scales), and (3) determine predictors of each domain of the PedsQL 3.0 Cardiac Module (and whether these predictors differ from those for PedsQL 4.0 Generic Core Scales). We assessed the construct validity of the PedsQL 3.0 Cardiac Module similarly to the original validation study by Uzark et al.,⁷ hypothesizing correlations between scales in terms of direction and magnitude, and that predictors of PedsQL 3.0 Cardiac Module domain outcomes would be similar to those of the PedsQL 4.0 Generic Core Scale summary outcomes. If the correlations between and/or the predictors of the 2 HRQL measures are different, discussion of how these tools may be measuring different aspects within the theoretical concept of HRQL will be required.

Methods

This was a prospective inception-cohort study including patients who had CCS at ≤6 weeks of age at the referral Stollery Children’s Hospital, Edmonton, Alberta, from the years 2000 to 2014. Patients were registered with the Complex Pediatric Therapies Follow-up Program in Western Canada, and survivors were followed to kindergarten age (4-5 years of age) with HRQL outcomes determined using the parent-proxy-report PedsQL 3.0 Cardiac Module. We previously reported HRQL outcomes at 4-5 years of age using the parent-proxy-report PedsQL 4.0 Generic Core Scales in the same patients, allowing for paired comparison of the 2 HRQL instruments.¹³

This registry and follow-up program includes multidisciplinary teams that follow children having had invasive complex therapies at the Stollery Children’s Hospital in a pediatric intensive care unit, including open cardiac surgery performed at age ≤6 weeks of life.^14,15 Briefly, families were approached during intensive care unit admission for consent, and by discharge, the prespecified perioperative variables were collected. The follow-up of these patients occurred in Edmonton or at the original sites from where patients were referred for their surgeries. On the follow-up visit at 4-5 years of age, the parent-proxy-report PedsQL questionnaires were completed by the parents/caregivers.¹⁴

Study variables

The Complex Pediatric Therapies Follow-up Program prospectively collects multiple previously agreed-upon variables in the perioperative period as well as during the follow-up of these patients, which will be used to describe the population and to assess for variables independently associated with long-term outcomes, including HRQL. Supplemental Table S1 shows the variables that were collected (and were included in the univariate analyses). Supplemental Table S2 shows the variables that had multicollinearity for each outcome, with only one of the collinear variables included in the multiple regressions.

HRQL instrument

As described by Amodeo et al.,¹⁶ the PedsQL 3.0 Cardiac Module comprises 23-25 items, depending on the version, distributed into 6 dimensions: 7 items for Cardiac Symptoms or Heart Problems and Treatment, 3 items for Adherence to Treatment or Treatment II, 3 items for Perceived Physical Appearance, 4 items for Treatment Anxiety, 3-5 items for Cognitive Problems, and 3 items for Communication. The proxy report version for children aged 2-4 years had only 3 items for Cognitive Problems. All other versions, including for preschool children (5-7 years of age), both for patients or parent proxy report, have 5 items for the Cognitive Problems dimension. The items are reverse scored, where the higher the score, on a 0-100 scale, the better the HRQL.

The PedsQL 4.0 Generic Core Scale has the same age versions as the Cardiac Module and is also reverse scored, with higher scores indicating better HRQL. There are 4 dimensions to the PedsQL 4.0 Generic Core Scale: Physical Functioning (8 items), Emotional Functioning (5 items), Social Functioning (5 items), and School Functioning (5 items). Different from the Cardiac Module, it also has 2 summary scores—Physical Health Summary Score (using the same items of the Physical Functioning dimension), and the Psychosocial Health Summary Score, which is created with the average of the other 3 dimension scores—and a total score, composed of the average of all 23 items.¹⁷

All HRQL outcomes were using parent proxy report as the children were too young to do patient-report questionnaires (few were ≥5 years, with age at assessment mean [standard deviation] of 54.9 [4.65] months). Both measures were completed by the same parent at the same time, just before the clinical assessment.

Statistical analysis

Categorical variables are reported as numbers and proportions (percentages). Continuous variables are reported as means and standard deviations or as medians and interquartile ranges as appropriate. We compared our PedsQL 3.0 Cardiac Module and PedsQL 4.0 Generic Core Scale results with previously published raw scores using independent samples t tests.7, 8, 9^,18

To evaluate the construct validity of the PedsQL 3.0 Cardiac Module in our population, correlations between the PedsQL 3.0 Cardiac Module domain scores and the PedsQL 4.0 Generic Core Scales summary scores were guided by previous work examining construct validity by Uzark et al. in 2003,⁷ and subsequently by others.8, 9, 10 The correlations examined were as follows: Heart Problems and Treatments vs Physical Summary Score, Perceived Physical Appearance vs Psychosocial Summary Score, Cognitive Problems vs School Functioning Score; Treatment Anxiety vs Psychosocial Summary Score, and Communication vs School Functioning Score. These domain and summary score comparisons were considered clinically reasonable, as they measured what we hypothesized to be similar constructs, and we therefore expected to observe clinically important correlations. The domain and summary scores on each PedsQL instrument asked different questions, making the hypothesis of clinically important correlations that would indicate measurement of highly similar constructs, as suggested by others,7, 8, 9, 10 questionable; nevertheless, we aimed to test this hypothesis in our prospective cohort. We used Pearson correlation coefficients to determine these associations, with P < 0.05 considered statistically significant. We considered a correlation r ≥ 0.5 as clinically important (a large effect size) and a correlation of r = 0.3-0.49 as potentially clinically important (a moderate effect size). Lower correlations were considered as nonclinically important.

To determine the association of demographic and perioperative variables with Cardiac Module dimension score outcomes at 4-5 years of age, we first performed univariate linear regression analyses, conducted for each of the domain scores as the main outcome variable. A purposeful variable selection method was used to identify predictors that would be included in multiple linear regression models. Our purposeful selection process began by considering all univariate variables that occurred in ≥5% of the survivor population and that had a P value of ≤0.05 for multiple linear regression analysis, after checking for multicollinearity. Variables were removed from the multiple linear regression model when they were nonstatistically significant (P value >0.05) and a nonconfounder (defined as a <20% change in remaining parameters when compared with the full model). Age at first surgery and socioeconomic status (SES, determined using the Bishen Index)¹⁹ were forced into the final models, as we had hypothesized that these were clinically relevant variables based on previous research on outcomes in this population.²⁰ Finally, we fit a multiple linear regression model with the statistically significant variables, confounding variables, and clinically relevant variables. Missing data could not be assumed to be missing at random, and therefore, no missing data imputation was used. Statistical analysis was performed on R software, using version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria). Results are presented as effect sizes (the variable coefficient) with 95% confidence intervals and with 2-sided P values. Statistical significance in multiple regression was considered at P values ≤0.05.

We made the following hypotheses regarding these statistical tests: (1) there would be large effect sizes in the correlations between the Cardiac Module dimensions and the PedsQL Generic summary scores, which would indicate that these scores measured similar constructs (and lower correlations we considered would indicate that the scores measured different constructs); and (2) the independent predictors of Cardiac Module dimension scores and the PedsQL Generic summary scores would be similar, which would indicate that these scores measured similar constructs (and having different independent predictors we considered would indicate that the scores measured different constructs). These hypotheses were based on the assumption that the 2 instruments have a similar purpose and are intended to validly measure the same construct (convergent validity).²¹ The Cardiac Module was used assuming that it provided a similar but more detailed assessment of the construct of HRQL in children specifically with CHD.

Results

Study cohort

The study population was composed of 425 patients having had CCS during early infancy from 2000 to 2014 and surviving to follow-up age of 4-5 years (see Fig. 1). Demographic and perioperative variables recorded for the patients were previously described and are shown in Supplemental Table S1.¹³ For the patients who responded to the PedsQL 4.0 Generic Core Scale, approximately 90%-94% also responded to dimensions of the PedsQL Cardiac Module. The dimension that had the lowest completion rate was the Adherence to Treatment (Treatment II) domain of the Cardiac Module, because most parents of the patients who were not taking cardiac medications at the time of follow-up did not respond to this module.

Figure 1.

Patient flow chart showing the included cohort of children having outcomes on the 2 health-related quality of life instruments. GCS, Generic Core Scales.

The patients were mostly born at term (90%) and of male sex at birth (63%). The most common CHD was transposition of the great arteries (32%). Single ventricle palliation was the CCS in 31% of patients. Extracorporeal membrane oxygenation was used in less than 7% of patients, and heart transplant in 3% of patients.

The scores on the PedsQL Cardiac Module are shown in Table 1, and were compared with previously published cohorts of children with CHD using the independent samples t test. Our patient population had higher scores on Treatment II and Perceived Physical Appearance dimensions than on the other dimension scores. The score for Perceived Physical Appearance was significantly higher than that reported by Uzark et al.,⁷ Berkes et al.,⁸ and Gonzales-Gil et al.,⁹ whereas scores for Cognitive Problems and Communication were significantly lower than those reported by Uzark et al.⁷ and Berkes et al.⁸ Our cohort’s Cardiac Module scores were similar to those obtained by Ruggiero et al.¹⁸

Table 1.

PedsQL 3.0 Cardiac Module results with comparison with previously published cohorts

PedsQL 3.0 Cardiac Module domain		Heart Problems and Treatment	Treatment II	Perceived Physical Appearance	Treatment Anxiety	Cognitive Problems	Communication
Current study	N	425	211	422	421	419	405
	Mean	78.7	94.5	93.6	71.1	67.3	67.5
	SD	18.98	15.22	13.85	29.45	22.71	31.22
Uzark et al. 2003 (proxy report)7	N	343	–	336	338	338	273
	Mean	79.6	–	82.9	71.2	71.2	74.6
	SD	17.62	–	22.96	28.84	25.73	26.76
	P	0.497	–	<0.001	0.963	0.029	0.002
Berkes et al. 2010 (proxy report)8	N	252	95	223	250	237	241
	Mean	76.4	93.7	82.8	69.8	73.4	74.5
	SD	17.46	15.75	23.00	26.73	21.03	28.31
	P	0.109	0.678	<0.001	0.558	0.001	0.004
Gonzales-Gil et al. 2012 (proxy report)9	N	174	–	173	174	172	165
	Mean	77.7	–	76.0	68.8	69.5	73.8
	SD	18.0	–	26.9	32.1	27.9	28.0
	P	0.544		<0.001	0.416	0.360	0.010
Ruggiero et al. 2018 (proxy report)18	N	39	39	39	39	39	39
	Mean	72.1	97.1	90.1	66.8	67.4	71.5
	SD	20.3	9.5	15.9	35.4	24.5	18.2
	P	0.057	0.163	0.191	0.466	0.981	0.230

Dashes indicate data not reported.

Comparison with the current study conducted using the independent samples t test.

Uzark et al.⁷ studied a population from a cardiology clinic with ages from 2 to 18 years; the numbers presented are the proxy report for their study. Berkes et al.⁸ studied a population of parents and patients aged 2-18 years attending a cardiology outpatient unit in Hungary; the numbers presented are the proxy report for their patients. Gonzales-Gil et al.⁹ studied “children between 2 years and 18 years of age diagnosed with severe congenital heart disease, at least 1 year after corrective or palliative surgical repair conducted within the pediatric cardiology and surgery service” in Spain; the numbers presented are the proxy report for their patients. Ruggiero et al.¹⁸ studied a population of children aged 5-12 years with congenital heart disease submitted to cardiac surgery before 1 year of age; the numbers presented here are those referring to proxy report on patients aged 5-7 years.

“–” These components of the Cardiac Module were not reported in the reference study.

PedsQL, Pediatric Quality of Life Inventory; SD, standard deviation.

Table 2 shows the PedsQL 4.0 Generic Core Scale scores for our population and the same studies cited above. With this instrument, our scores were similar to those found by Uzark et al.⁷ and Gonzales-Gil et al.,⁹ whereas Berkes et al.⁸ and Ruggiero et al.¹⁸ reported lower scores than in our cohort (except for social functioning and school functioning).

Table 2.

PedsQL 4.0 Generic Core Scales results with comparison with previously published cohorts

PedsQL 4.0 Generic Core Scales domain		Total score	Physical summary	Psychosocial summary	Emotional functioning	Social functioning	School functioning
Current study	N	425	425	375	425	425	375
	Mean	80.0	84.1	77.5	75.5	81.2	75.8
	SD	15.93	18.27	16.46	18.39	19.91	21.74
Uzark et al. 2003 (proxy report)7	N	344	344	343	343	343	296
	Mean	79.4	83.1	77.4	74.7	82.5	73.1
	SD	16.50	18.73	17.27	20.45	20.11	20.35
	P	0.611	0.457	0.937	0.573	0.371	0.098
Berkes et al. 2010 (proxy report)8	N	212	212	212	212	212	183
	Mean	76.0	77.7	75.1	68.5	82.1	74.6
	SD	15.30	18.73	15.49	18.06	19.68	18.62
	P	0.002	<0.001	0.079	<0.001	0.588	0.500
Gonzales-Gil et al. 2012 (proxy report)9	N	168	168	168	168	168	168
	Mean	78.5	81.3	76.9	74.5	82.8	73.2
	SD	16.3	19.5	16.0	17.5	19.6	21.2
	P	0.310	0.110	0.689	0.537	0.373	0.191
Ruggiero et al. 2018 (proxy report)18	N	39	39	39	39	39	39
	Mean	72.8	76.5	70.0	64.3	85.5	69.9
	SD	16.8	22.8	20.2	22.1	15.3	29.4
	P	0.014	0.050	0.030	<0.001	0.109	0.230

Comparison with the current study conducted using the independent samples t test.

Uzark et al.⁷ studied a population from a cardiology clinic with ages from 2 to 18 years; the numbers presented are the proxy report for their study. Berkes et al.⁸ studied a population of parents and patients aged 2-18 years attending a cardiology outpatient unit in Hungary; the numbers presented are the proxy report for their patients. Gonzales-Gil et al.⁹ studied “children between 2 years and 18 years of age diagnosed with severe congenital heart disease, at least 1 year after corrective or palliative surgical repair conducted within the pediatric cardiology and surgery service” in Spain; the numbers presented are the proxy report for their patients. Ruggiero et al.¹⁸ studied a population of children aged 5-12 years with congenital heart disease submitted to cardiac surgery before 1 year of age; the numbers presented here are those referring to proxy report on patients aged 5-7 years.

PedsQL, Pediatric Quality of Life Inventory; SD, standard deviation.

Construct validity: correlations between PedsQL Cardiac Module and PedsQL Generic Core Scales

The correlations between Cardiac Module dimensions and Generic Core Scale summary scores are shown in Table 3, with comparison to other published CHD cohorts. The correlation of Heart Problems and Treatment with Physical Functioning had a large effect size, although our correlation was smaller than those found by Uzark et al.,⁷ Berkes et al.,⁸ Gonzales-Gil et al.,⁹ and Capitello et al.¹⁰ The correlation of Cognitive Problems with School Functioning and of Treatment Anxiety with Physical Functioning had moderate effect sizes. Again, the former was lower than that found by the authors mentioned above. The latter was similar to that found by the authors mentioned above, except for Capitello et al.,¹⁰ who found a much higher correlation. The correlation of Perceived Physical Appearance with Psychosocial Summary had a small effect size, which was similar to that reported by Gonzales-Gil et al.,⁹ but smaller than that reported by Uzark et al.⁷ and Berkes at al.⁸ The correlation of Communication with School Functioning had a borderline moderate effect size, again smaller than that reported by Berkes et al.⁸ and much smaller than that reported by Capitello et al.¹⁰

Table 3.

Correlation between PedsQL 3.0 Cardiac Module and PedsQL Generic Core Scales (Pearson correlation coefficient)

PedsQL 3.0 Cardiac Module dimension	PedsQL Generic Core Scale dimension	Correlation coefficient (r)
		Our population, Canada (with P value)	Uzark et al. 2003 (proxy report), USA7	Berkes et al. 2010 (proxy report), Hungary8	Gonzales-Gil et al. 2012 (proxy report), Spain9	Capitello et al. 2021, Italy10
Heart Problems and Treatment	Physical Functioning	0.57 (<0.001)	0.713	0.66	0.685	0.922
Perceived Physical Appearance	Psychosocial Summary	0.21 (<0.001)	0.493	0.38	0.227	–
Cognitive Problems	School Functioning	0.48 (<0.001)	0.776	0.60	0.728	0.739
Treatment Anxiety	Physical Functioning	0.38 (<0.001)	0.382	0.34	0.337	0.752
Communication	School Functioning	0.30 (<0.001)		0.37		0.55

Uzark et al.⁷ studied a population from a cardiology clinic with ages from 2 to 18 years; the numbers presented are the proxy report for their patients. Berkes et al.⁸ studied a population of parents and patients aged 2-18 years attending a cardiology outpatient unit in Hungary; the numbers presented are the proxy report for their patients. Gonzales-Gil et al.⁹ studied “children between 2 years and 18 years of age diagnosed with severe congenital heart disease, at least 1 year after corrective or palliative surgical repair conducted within the pediatric cardiology and surgery service” in Spain; the numbers presented are the proxy report for their patients. Capitello et al.¹⁰ studied “patients aged 2-18 with congenital or acquired heart disease attending an outpatient cardiology unit” in Italy; the numbers presented are from the parent report on 5- to 7-year-old patients.

PedsQL, Pediatric Quality of Life Inventory.

Univariate and multiple linear regression analyses

Table 4 gives the univariate (all variables with P ≤ 0.05 and the clinically significant variables prespecified to be forced into multiple regressions) and the final multiple linear regression analyses performed for each of the PedsQL Cardiac Module dimensions in our population.

Table 4.

Univariate and multiple linear regressions for predictors of PedsQL 3.0 Cardiac Module dimension scores among survivors of complex cardiac surgery performed in early infancy

Score	Univariate linear regression		Multiple linear regression
	Effect size (95% CI)	P value	Effect size (95% CI)	P value
Heart Problems and Treatment
Single ventricle palliation	–17.97 (–21.50, –14.44)	<0.001	–14.79 (–18.55, –11.03)	<0.001
Female sex	3.84 (0.11, 7.57)	0.044	3.73 (0.46, 7.00)	0.025
Number of hospital admissions for noncardiac reasons after discharge from first surgery	–1.96 (–2.65, –1.27)	<0.001	–1.14 (–1.77, –0.51)	<0.001
Dialysis at any time up to the 4- to 5-year visit	–13.63 (–18.79, –8.47)	<0.001	–5.42 (–10.31, –0.52)	0.030
Highest lactate on first surgery postoperative days 6-10 (mmol/L)	–3.40 (–4.83, –1.96)	<0.001	–1.82 (–3.16, –0.48)	0.008
Age at first surgery (d)	–0.04 (–0.26, 0.18)	0.727	–0.12 (–0.33, 0.09)	0.266
Socioeconomic status (Blishen)	0.06 (–0.07, 0.20)	0.340	0.09 (–0.03, 0.20)	0.127
Highest modified inotrope score on postoperative days 1-5	–0.34 (–0.51, –0.17)	<0.001
Need to return to CPB during the first surgery	–8.72 (–14.74, –2.71)	0.005
Gestational age at birth (completed weeks)	1.34 (0.34, 2.34)	0.009
Weight at first surgery (kg)	3.75 (0.76, 6.74)	0.014
Antenatal diagnosis	–8.48 (–12.02, –4.94)	<0.001
Lowest arterial pH on first surgery postoperative day 1	25.20 (1.13, 49.27)	0.04
Treatment II
Dialysis at any time up to 4- to 5-year visit	–7.80 (–12.93, –2.68)	0.003	–7.56 (–12.47, –2.64)	0.003
Number of hospital admissions for noncardiac reasons after discharge from first surgery	–1.19 (–1.84, –0.54)	<0.001	–1.09 (–1.72, –0.45)	0.001
Age at first surgery (d)	0.29 (0.04, 0.55)	0.026	0.26 (0.01, 0.51)	0.040
Socioeconomic status (Blishen)	–0.05 (–0.21, 0.10)	0.51	–0.03 (–0.18, 0.11)	0.640
Perceived Physical Appearance
Reopening sternum on first postoperative day of first surgery	–8.87 (–14.33, –3.42)	0.001	–6.20 (–11.98, –0.42)	0.036
Socioeconomic status (Blishen)	0.11 (0.01, 0.20)	0.033	0.10 (0.01, 0.20)	0.032
Age at first surgery (d)	–0.04 (–0.20, 0.13)	0.650	–0.09 (–0.25, 0.08)	0.310
Highest lactate on first surgery postoperative days 1-5 (mmol/L)	–0.68 (–1.16, –0.21)	0.005
Extracorporeal membrane oxygenation at any time up to 4- to 5-year visit	–6.38 (–11.59, –1.17)	0.017
Dialysis at any time up to 4- to 5-year visit	–5.39 (–9.24, –1.54)	0.006
Treatment Anxiety
Single ventricle palliation	–13.73 (–19.68, –7.77)	<0.001	–12.63 (–18.63, –6.64)	<0.001
Preterm (gestational age <37 completed weeks)	–11.82 (–21.28, –2.36)	0.014	–10.10 (–19.58, –0.62)	0.037
Age at first surgery (d)	0.01 (–0.34, 0.36)	0.967	–0.01 (–0.36, 0.33)	0.936
Socioeconomic status (Blishen)	0.02 (–0.19, 0.23)	0.837	0.03 (–0.17, 0.23)	0.792
Sepsis at any time up to 4- to 5-year visit	–8.86 (–16.95, –0.77)	0.032
Lowest arterial pH on first surgery postoperative days 2-5	82.81 (39.32, 126.30)	<0.001
Cognitive Problems
Known chromosomal abnormality	–10.83 (–18.19, –3.48)	0.004	–9.43 (–16.83, –2.04)	0.013
Cardiopulmonary bypass time at first surgery (min)	–0.07 (–0.12, –0.03)	0.001	–0.06 (–0.11, –0.01)	0.028
Age at first surgery (d)	–0.33 (–0.59, –0.06)	0.017	–0.38 (–0.65, –0.11)	0.006
Socioeconomic status (Blishen)	0.11 (–0.06, 0.27)	0.206	0.08 (–0.08, 0.24)	0.340
Highest modified inotrope score on first surgery postoperative days 1-5	–0.28 (–0.49, –0.06)	0.011
Lowest arterial pO2 on first surgery postoperative days 1-5	0.18 (0.04, 0.32)	0.011
Risk Adjustment for Congenital Heart Surgery (RACHS 1) score 4-6	–5.45 (–10.15, –0.75)	0.023
Sepsis at any time up to 4- to 5-year visit	–7.53 (–13.72, –1.34)	0.017
Dialysis at any time up to 4- to 5-year visit	–8.04 (–14.41, –1.68)	0.013
Total cardiopulmonary bypass time up to 4- to 5-year visit (min)	–12.93 (–21.90, –3.97)	0.005
Sternum left open from operating room at first surgery	–4.77 (–9.14, –0.40)	0.033
Highest lactate on fist surgery postoperative days 1-5 (mmol/L)	–0.61 (–1.40, 0.18)	0.132
Lowest arterial pH on first surgery postoperative days 2-5	57.86 (24.04, 91.67)	0.001
Highest lactate on first surgery postoperative days 6-10 (mmol/L)	–3.06 (–4.80, –1.31)	0.001
Total number of days with sternum open at first surgery	–0.99 (–1.56, –0.42)	0.001
Cardiopulmonary resuscitation on first admission (before or after first surgery)	–9.85 (–19.20, –0.51)	0.039
Communication
Total hospital days up to 4- to 5-year visit	–0.16 (–0.27, –0.06)	0.002	–0.15 (–0.23, –0.06)	0.001
Highest creatinine on first surgery postoperative days 6-10 (μmol/L)	0.12 (0.00, 0.24)	0.048	0.18 (0.05, 0.31)	0.008
Age at first surgery (d)	–0.22 (–0.60, 0.16)	0.260	–0.23 (–0.61, 0.15)	0.232
Socioeconomic status (Blishen)	0.13 (–0.10, 0.36)	0.254	0.11 (–0.11, 0.34)	0.332
Preterm (birth gestational age <37 completed weeks)	10.72 (0.43, 21.02)	0.041
Total number of days with sternum open at first surgery	–0.85 (–1.65, –0.05)	0.038

CI, confidence interval; CPB, cardiopulmonary bypass; PedsQL, Pediatric Quality of Life Inventory.

In total, we found 13 different variables that were statistically significant independent predictors of at least 1 dimension in the multiple linear regression analyses. There were 5 significant variables associated with Heart Problems and Treatment. Single ventricle palliation had a large effect size, as on average, these patients had a score lower by 14.8 points. The other significant variables were number of hospital admissions that were for noncardiac reasons after discharge from first surgery, female sex, dialysis at any time up to the 4- to 5-year visit, and the highest lactate on first surgery postoperative days 6-10.

For Treatment II, there were 3 significant variables: number of hospital admissions that were for noncardiac reasons after discharge from first surgery, dialysis at any time up to the 4- to 5-year visit, and age in days at first surgery. For the Perceived Physical Appearance, Treatment Anxiety, and Communication dimensions, there were 2 variables that were significant for each dimension: reopening sternum on the first postoperative day of first surgery and SES; single ventricle palliation and preterm birth; and total hospital days up to the 4- to 5-year visit and the highest creatinine level on postoperative days 6-10 of first surgery, respectively. For the Cognitive Problems dimension, there were 3 significant variables: known chromosomal abnormality, cardiopulmonary bypass time at first surgery, and age in days at first surgery.

Discussion

We followed 425 survivors of CCS performed in early infancy to determine HRQL using the PedsQL 3.0 Cardiac Module at age 4.5 years. The main findings included the following. First, the construct validity of the Cardiac Module according to our prespecified criteria was limited, as correlations between PedsQL Cardiac Module and Generic Core scores was large only for Heart Problems and Treatment (with Physical Functioning), only moderate for Cognitive Problems and Treatment Anxiety (with School Functioning and Physical Functioning), borderline moderate for Communication (with School Functioning), and nonclinically important for Perceived Physical Appearance (with Psychosocial Summary). Second, independent predictors were found for lower scores on each dimension of the Cardiac Module, including single ventricle palliation, days of age at first surgery, severity of acute illness variables (dialysis, highest lactate, highest creatinine, and reopening of sternum), and severity of ongoing illness variables (number of hospital admissions that were for noncardiac reasons after discharge from first surgery and total hospital days to age 4.5 years). However, only about half of the independent predictors of the Cardiac Module dimensions overlapped with those of Generic Core Scale summary scores (as further discussed below), again suggesting that these 2 instruments measured different constructs.

This study had many strengths relative to the literature that had been published so far.²² First, our study was unique in determining HRQL using the Cardiac Module in a relatively large prospective inception cohort of patients assessed at a uniform age (of 4.5 years). Second, unlike previous studies, we examined a cohort of patients who were at uniquely high risk for adverse outcomes, that is, those with CHD severe enough to require CCS at ≤6 weeks of age. These patients were high risk for 2 reasons: the severity of their CHD and the intervention (CCS) required, and that they experienced this severe illness in early infancy, the period of most vulnerability to brain injury. The brain doubles in size in the first year of life, and all major brain networks are wired before the age of 2 years, making injury during these early years a period of unique vulnerability.²³ Third, predictors of adverse HRQL using the Cardiac Module in such a high-risk cohort have not been previously reported in the literature. Very importantly, in exploring predictors, we aimed to avoid bias that previous studies were susceptible to. By prospectively prespecifying the cohort, well-defined potential demographic and acute-care predictor variables, outcomes, and statistical analysis plan (including which variables to force into multiple regressions), we avoided excessive flexibility in design, definitions, outcomes, and analytic modes that contribute to bias from so-called data dredging, multiple testing, and post hoc selection of variables.²⁴ As discussed below, the risk factors identified have implications for patient counseling and management, and future research directions. Fourth, we report insights beyond those in our previous study in this same population that described HRQL using the PedsQL Generic Core Scales, determined that this was lower than in the healthy normative pediatric population, and described predictors of the Generic Core Scales. The current study described HRQL using the Cardiac Module and compared this with the few reports of other cardiac populations finding differences that were likely related to the unique severity of CHD in our inception cohort. Moreover, construct validity of the Cardiac Module was examined not only using correlations between domains on the 2 HRQL instruments, but also in a novel way by comparing independent predictors between the 2 PedsQL instruments, emphasizing the degree to which different constructs are being measured.

The PedsQL 3.0 Cardiac Module is a disease-specific instrument, such that there is no standard healthy population to compare with our results. Altin et al.¹² have applied this instrument to a “control” population formed by a matching number of healthy children. Similarly, Spanaki et al.²⁵ measured the HRQL using the Cardiac Module in children with hypertrophic cardiomyopathy and compared results with a “control” population of children who went for screening but were negative. In both studies, the “control population” had statistically significantly higher HRQL scores than the cardiac patient population. This suggests that the Cardiac Module can differentiate the HRQL in children with heart disease from healthy children, and therefore that the instrument is measuring the theoretical construct intended.

Comparing the results in our population with results in other studies is difficult, as study populations differed in age and severity of disease. As shown in Table 1, the Cardiac Module dimension scores often had statistically significant differences between our population results and those of previously published cohorts.7, 8, 9^,18 Our population of children had higher scores on Perceived Physical Appearance and lower scores on Cognitive Problems and Communication than those reported by Uzark et al.⁷ and Berkes et al.⁸ Similarly, our population of children had scores that were higher on Perceived Physical Appearance and lower on Communication than those reported by Gonzales-Gil et al.⁹ On the other hand, our population of children had scores that were quite similar to those reported by Ruggiero et al.¹⁸ Table 2 shows comparison of the PedsQL 4.0 Generic Core Scale scores from our population to the same studies above, with findings that were different from comparisons using the Cardiac Module. Interestingly, our population of children had scores that were not statistically different from those reported by Uzark et al.⁷ and Gonzales-Gil et al.⁹ However, both Berkes et al.⁸ and Ruggiero et al.¹⁸ reported lower scores than in our population of children, except on social functioning and school functioning, suggesting a worse HRQL in their population, a finding that was not captured by the Cardiac Module comparisons. For example, compared with the Berkes et al.⁸ population, our population of children had worse scores in 2 dimensions, similar scores in 3, and only 1 dimension with a higher score (Perceived Physical Appearance).

These differences in HRQL results may be due to differences in the study populations. Uzark et al.⁷ in the United States and Berkes et al.⁸ in Hungary studied a population of children at a cardiology clinic with varying severity of heart diseases, some having had no cardiac surgery, at ages 2-18 years. Gonzales-Gil et al.⁹ in Spain studied children 2-18 years old with severe CHD at least 1 year after surgical repair. Ruggiero et al.¹⁸ studied a population of children with CHD at ages 5-12 years after having had cardiac surgery before 1 year of age, with the figures we provided in Tables 1 and 2 referring to the proxy report from 5- to 7-year-old patients, characteristics more similar to our study population, and potentially explaining the more similar scores on the Cardiac Module. For example, it may not be surprising that younger children in our study and the study by Ruggiero et al.,¹⁸ who have not yet entered grade school, were less concerned with their sternotomy scar and physical appearance than older children in the other studies.

We found statistically significant correlations between the PedsQL Cardiac Module and the PedsQL Generic Core Scales for the dimensions as tested by Uzark et al. and others.7, 8, 9, 10 However, the correlation values in our population were usually smaller than those found by others (Table 3).7, 8, 9, 10 We found a large effect size correlation only for Heart Problems and Treatment with Physical Functioning (r = 0.57). The other correlations were of only moderate or low effect size. Overall, we did not confirm our hypothesis that there would be strong correlations between the 2 HRQL instruments, suggesting the 2 instruments measured different constructs within the concept of HRQL. For example, even the highest correlation of r = 0.57 suggested that Heart Problems and Treatment scores explained only r² = 32% of the variance in Physical Functioning score on the PedsQL Generic Core Scales.

In the present study, the multiple linear regression analyses showed some predictors statistically significantly independently associated with the Cardiac Module HRQL dimensions. Single ventricle palliation, age at first surgery, dialysis at any time, and number of hospital admissions that were for noncardiac reasons after initial hospital discharge were each associated with 2 Cardiac Module dimensions. Variables each associated with 1 Cardiac Module dimension were female sex, highest lactate on postoperative days 6-10, reopening of the sternum on the first postoperative day after first surgery, SES, preterm birth, known chromosomal abnormality, cardiopulmonary bypass time at first surgery, total hospital days, and highest creatinine on postoperative days 6-10. Interestingly, from the 13 predictors cited above, 6 were also statistically significant in our previous report on the PedsQL Generic Core Scales in this same cardiac population.¹³ These 6 predictors were single ventricle palliation, number of hospital admissions that were for noncardiac reasons after initial hospital discharge, total hospital days, known chromosomal abnormality, age at first surgery, and SES. The other 7 statistically significant predictors of HRQL outcomes on the Cardiac Module were not statistically significant in our previous study of the PedsQL Generic Core Scales. In addition, 2 of the 8 predictors that were statistically significant in our previous study of the PedsQL Generic Core Scales (ECMO at any time up to the 4- to 5-year visit, and total days on mechanical ventilation before first surgery) were not statistically significant predictors in the present study of Cardiac Module outcomes. We found different independent predictors of the Cardiac Module domains and Generic Core Scale summary scores, even for the dimensions where we found the highest correlation (Heart Problems and Treatment × Physical Functioning), where the first had 5 statistically significant predictors, and the second had 4, but only 2 were the same for both scales. For the correlation between Perceived Physical Appearance × Psychosocial Summary (the lowest correlation found with r = 0.21), the first had 2 predictors and the second had 4 predictors, and none of these were the same for both scales. Therefore, we did not confirm our hypothesis that the independent predictors of the 2 HRQL instruments would be the same. Again, these results suggested that the 2 instruments were measuring different HRQL constructs.

We found that the 2 PedsQL HRQL instruments measured overlapping constructs, but they were not quite the same. HRQL is not an objective construct, like height or weight, that can be measured directly. In fact, there is disagreement on the definition of HRQL and how it should be measured.26, 27, 28 Most of the scales used to measure HRQL are based on psychometrics, which, according to the American Psychology Association Dictionary of Psychology, is “the branch of psychology concerned with the quantification and measurement of mental attributes, behavior, performance, and the like, as well as with the design, analysis, and improvement of the tests, questionnaires, and other instruments used in such measurement,”²⁹ and the classical test theory, as proposed by Spearman at the beginning of 20th century, where any measure is composed of the true score plus an error.³⁰ The problem with this approach is that often the construct being measured is not well defined. As the PedsQL Generic Core Scale and the PedsQL 3.0 Cardiac Module have different (though related) question items, it may not be surprising that they were not measuring the exact same construct.

For example, the Physical Functioning of the Generic Core Scale asks about usual activities such as the ability to walk, run, self-care, perform chores, and pain and energy. On the other hand, the Heart Problems and Treatment asks about common concerns of patients with heart disease, such as dyspnea, tachycardia, cyanosis, and chest pain. Both questionnaires are related to difficulties in performing physical activities, but with a very different approach. Of course, the disease-specific instrument uses questions that are important to patients with cardiac disease, but the instruments are not asking the same questions.

The independent associations we identified can direct future clinical care and research in novel ways. Nonmodifiable predictor variables, including single ventricle palliation and known chromosomal abnormality, may indicate cohorts to prioritize in preoperative parental counseling and for follow-up and access to early intervention aimed to help the child achieve their full potential. Potentially modifiable predictor variables may indicate areas to focus on in patient management and quality improvement. For example, total hospital days and number of hospital admissions that were for noncardiac reasons are novel findings that may indicate a group of patients to prioritize for in-hospital developmental care, aimed to prevent medical trauma and empower parents to provide a safe stable nurturing relationship, that is, “to address the key environmental conditions necessary for optimal early childhood development and hence facilitate children's ability to thrive.”³¹ Age in days at first surgery may be amenable to intervention if delay is due to reasons such as lack of operating room time or intensive care beds. Severity of illness variables (eg, dialysis, highest creatinine, reopening of sternum, and highest lactate) may be amenable to a focus on low cardiac output syndrome management and identification of residual cardiac lesions that require early intervention.^32,33 Future outcomes research, including before-after study designs, may iteratively determine whether these interventions improve subsequent outcomes in future cohorts and at older ages (eg, self-report at school-age).

This study had some limitations. First, we reported HRQL evaluation at only 1 time point (age 4-5 years), which did not allow assessment of responsivity of the instrument over time in our population, a main reason to use a disease-specific instrument. Second, the HRQL scores were reported in proxy by children’s parents/guardians, which may not be ideal; however, this was the feasible method to determine HRQL at the young age of 4-5 years in our population studied. Third, this was a single referral-center study, which may limit generalizability. Fourth, as in any observational study, findings may not be causal associations, as we cannot rule out unmeasured confounders. Fifth, we did not have the data needed to perform psychometric methods such as Item-Response Theory or Rasch analysis, which is a direction for future research.

Strengths of this study included the large number of patients evaluated at a similar age (of 4-5 years) in a prospective inception cohort, the largest study using the PedsQL 3.0 Cardiac Module in patients with critical CHD of which we are aware. The cohort included only children who had CCS in early infancy (before 6 weeks of age), reflecting the most severe spectrum of CHD, and assessed these children at kindergarten age when they were entering the school system. The Cardiac Module and Generic Core Scales were collected at the same time in the same patients, which allowed for comparisons to be made.

Conclusion

The PedsQL 3.0 Cardiac Module dimensions had statistically significant correlations with the PedsQL 4.0 Generic Core Scale scores. However, each instrument may be measuring different HRQL constructs, as the magnitude of the correlations was not strong, and the statistically significant independent predictors of scores in the same population were often different. Further studies are needed to better define those constructs and their relevance for longer-term patient-important outcomes such as school achievement, mental health, and individual well-being.³⁴

Ethics Statement

Ethics approval was obtained from the Health Research Ethics Board at the University of Alberta (ID Pro00001030; last renewal August 2025), as well as at each follow-up site.

Patient Consent

The authors confirm that patient consent forms were obtained for this article. All parents or guardians signed a written informed consent for registration (which included consent for data collection and long-term follow-up). All the study procedures followed were in accordance with ethical standards of the University of Alberta Health Research Ethics Board, and with the Helsinki Declaration of 1975, as most recently amended.

Funding Sources

The CPTFP, a multidisciplinary western Canadian collaboration, has been supported over years by funding contributions from Alberta Health, Stollery Children’s Hospital, and Women and Children’s Health Research Institute (https://www.wchri.org/about-wchri/thepower-of-partnership/), referral-site follow-up clinics (Saskatoon and Regina, Saskatchewan; Winnipeg, Manitoba; Calgary, Alberta; Vancouver, British Columbia), and Glenrose Rehabilitation Hospital Research Trust, with ongoing funding by Glenrose Rehabilitation Hospital. Funding agencies had no role in the design and conduct of the study; collection, management, analysis, and interpretation of data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclosures

The authors have no conflicts of interest to disclose.