Type 1 diabetes mellitus (T1DM) is one of the most common chronic childhood illnesses, affecting approximately 1 in every 400-600 children and adolescents [1]. Treatment includes a multifaceted regimen with daily subcutaneous insulin injections (or insulin infusion), blood glucose monitoring, carbohydrate counting, dietary plan, and physical activity [2]. Due to its complexity, T1DM management can be overwhelming even for the most competent patient [3]. Children might feel different from their peers because of their need for disruptive self-care activities [4] and the impact of T1DM on psychological aspects [5]. Health-related quality of life (HRQoL) captures the individual’s health perception, adding the patients’ perspective into T1DM clinical monitoring and research.

The lack of correspondence and continuity between HRQoL instruments designed for children and those designed for adults hinders the analysis of HRQoL changes using a life course approach. To solve this gap, in 2006, the EuroQol Group developed the EQ-5D-Youth (EQ-5D-Y) as a version of the EQ-5D for adults. The EQ-5D-Y was designed by adapting the EQ-5D dimensions to the requirements of HRQoL measurement in children and adolescents aged 8 years or older [6,7]. The main rationale of the new version was to enable young individuals to self-report their health [6,8]. A EQ-5D-Y proxy version was also developed [9] for children below 8 years of age. Given its generic and econometric nature, it allows comparison between different populations and conditions, and cost-utility analysis for economic evaluation. Another econometric instrument, the Health Utility Index [10], has a self-administered version for adults and adolescents and a proxy version for children (5-12 years of age), but its administration burden is substantially greater.

The EQ-5D-Y is a very short (2-3 minutes) and simple instrument to fill out. There are several studies assessing the metric properties of the EQ-5D-Y in the general population [11,12] or in school environments [13-16]. A Swedish study [12] suggested that the EQ-5D-Y was comprehensible, acceptable, and feasible for self-completion. A multicountry study supported its feasibility, reliability, and validity [11]. Furthermore, a Spanish study [16] compared the paper and Web-based versions of EQ-5D-Y, showing acceptable levels of agreement.

Only a few studies have assessed the metric properties of EQ-5D-Y in children with specific chronic conditions. A multicenter study among patients with cystic fibrosis [17] concluded that the EQ-5D-Y can be considered a valid instrument, which reflects the differences in health according to the progression of this life-long chronic disease. A comparison between the EQ-5D-Y and the Pediatric Asthma Quality of Life Questionnaire [18] supported EQ-5D-Y’s convergent validity for asthmatic children and adolescents. In addition, a study in youth with chronic kidney disease provided evidence of the EQ-5D-Y validity, as it discriminated among groups that differed with regard to the disease-related clinical burden [19]. Finally, we have only identified one descriptive study of HRQoL in children with T1DM [20] measured with the EQ-5D-Y, but without assessing its metric properties. Our aim was to evaluate acceptability, validity, reliability and responsiveness of the EQ-5D-Y administered online in children and adolescents with T1DM.

Data was obtained from a clinical trial on children and adolescents with T1DM, designed to evaluate the impact of routine use of HRQoL assessment in clinical practice by using the KIDSCREEN-27 as the primary outcome. Previous publications of this trial reported that the baseline HRQoL of children and adolescents with T1DM [21] was similar to that of the European population of the same age, but with slightly worse physical well-being, and that routine assessment and face-to-face patient-physician discussion of HRQoL results improved after a year of follow-up, especially psychological well-being and school environment [22].

To the best of our knowledge, this is the first study evaluating metric properties of the new EQ-5D-Y in T1DM children and adolescents. We found this generic preference-based instrument to be acceptable for children and adolescents with T1DM and easy to administer online, but with a high ceiling effect (50.7% of participants reported no problem in any dimension). The EQ-5D-Y showed good validity, considering the results obtained in the known groups’ analysis based on general health, acute diabetic decompensations, mental health, and family function, and those obtained from the multitrait, multimethod matrix with KIDSCREEN. Test-retest reproducibility was high, indicating good reliability, and the moderate-large change observed between the first and fourth visits among the selected subsample of “improvement” supports its responsiveness.

Of the 136 participants in the T1DM study, all responded to the EQ-5D-Y at the first visit and 119 (87.5%) responded at the fourth visit, which is similar to the response rates reported by studies carried out in the clinical (96% [18]) and school (77% [16]) settings. It is important to remark that the participants answering the EQ-5D-Y entirely completed the instrument, including the five dimensions, and the EQ-VAS. These rates of completion support the acceptability of the EQ-5D-Y among children and adolescents with T1DM when administered online. Taking into account the quick expansion of online recreational habits among the new generations born in the digital era, acceptability could be even higher with the incorporation of other devices such as apps for smartphones or tablets. However, completion rates in routine clinical practice could be lower than under clinical trial conditions, which indicates the need for closer monitorization.

Ceiling effect in the sample of children and adolescents with T1DM was high for the EQ-5D-Y summary score (50.7%) and for the three out of its five dimensions, which exceeded 85% of participants reporting no problem (“mobility,” “looking after myself,” and “doing usual activities”). A European multicountry study of EQ-5D-Y metric properties [11] reported a similar ceiling effect for these dimensions, which was up to 80% in children with chronic conditions. Half of the children and adolescents with no problem in any dimension (value=0 in EQ-5D-Y summary score) could be considered to show a really high ceiling effect, taking into account the established recommendations of 15% for HRQoL scores [42]. However, this is a general standard, as there are none specific for children. This high ceiling effect could be considered coherent with the relative low impact of diabetes on HRQoL for most children and adolescents, partially explained by their capacity of adaptation to the demanding tasks related to care [43]. The main disadvantage of this high ceiling effect is that it avoids the detection of improvement in children and adolescents who reported the highest level of health, which could affect the instrument’s responsiveness.

On the other hand, our study showed a substantial proportion of children and adolescents with T1DM reporting problems in the dimensions of “having pain or discomfort” (n=47, 34.6%) and “feeling worried/sad/unhappy” (n=43, 28.7%). These results are consistent with a previous study [44] showing this latter EQ-5D-Y dimension as the most affected in children and adolescents with T1DM. They are also consistent with studies using the 21-item Beck Depression Inventory and the Child Health Questionnaire PF-50 [45,46], which highlighted the impact of T1DM on the psychological dimension, especially due to the lack of autonomy, worrying for future chronic complications and the self-care routines that these children and adolescents have to experience.

As hypothesized, the EQ-5D-Y presented a good discrimination capacity between groups defined by the general health question and SDQ score in the summary score and in three dimensions: “doing usual activities,” “having pain/discomfort,” and “feeling worried/sad/unhappy.” These dimensions presented the highest correlations with general and mental health in previous studies of children and adolescents with T1DM [41] or chronic conditions [10]. The magnitude of the difference between groups defined by these constructs was larger when HRQoL was measured with the EQ-5D-Y summary score (effect size for general health and mental health groups=0.7 and 1.5, respectively) than with the KIDSCREEN-10 index (effect size for general health and mental health groups=0.6 and 0.9, respectively).

It is important to highlight the problems reported in the “looking after myself” dimension among those participants who had previously experienced acute diabetic decompensations, which is consistent with previous studies comparing good and bad metabolic control [37,47]. These participants showed differences in the physical well-being and health perception dimensions, measured with KINDL-R [37] and the Diabetes Quality of Life questionnaire [42]. A priori hypotheses on the relationship between family function and HRQoL were based on studies that reported an increased risk for depressive symptoms in families with a disadvantaged socioeconomic status [39,48]. Our results, showing problems of “having pain and discomfort” more frequently and poorer HRQoL with KIDSCREEN-10 index in children and adolescents with dysfunctional families, are in line with the social component of health [39,48].

Correlations between EQ-5D-Y and KIDSCREEN showed the expected pattern for convergent and divergent validity, which were very similar to those obtained in the abovementioned European multicountry study [11]. The dimensions reflecting the emotional aspects of EQ-5D-Y (“feeling worried/sad/unhappy”) and KIDSCREEN-27 (“psychological well-being”), in addition to the EQ-5D-Y summary score and EQ-VAS, presented a substantial correlation with the KIDSCREEN-10 index, which provides a summary of the HRQoL. The EQ-5D-Y “mobility” dimension failed to display convergent validity with the KIDSCREEN dimension “physical wellbeing,” similar to a previous study [11] (r=0.10-0.25). Considering the content of the instruments, it can be argued that KIDSCREEN is more focused on the energy level and less focused on the physical functioning, which is the case of the EQ-5D-Y.

Our study shows good reproducibility of the questionnaire according to the established standards [7] for reliability coefficients: 0.7 and 0.9 for group and individual measurements, respectively. The ICC of 0.83 and 0.74 obtained for the EQ-5D-Y summary score and EQ-VAS allows recommending its use. Nine months between the first and the fourth visit is too long a period to evaluate reproducibility, since HRQoL could change considerably during this time. This long period explains the low number of children and adolescents in the stable subsample (n=18). Previous EQ-5D-Y studies used the kappa coefficient to evaluate test-retest reporting from good to fair agreement, according to dimensions [11,15].

The EQ-5D-Y detected differences in the selected subsample of children and adolescents with T1DM classified as experiencing “improvement” between the first and the fourth visit, showing the capacity of this instrument to detect change. It is important to highlight that no other studies testing the responsiveness of the EQ-5D-Y have been found. A systematic review assessing metric properties of diabetes-specific HRQoL instruments also highlighted that the lack of testing for responsiveness is a major shortcoming [49]. The EQ-5D-Y summary score indicates a moderate-large improvement over time, a magnitude which can be considered reasonable when accounting for the length of the follow-up of participants (9 months), the nature of the intervention, and the course of the T1DM.

The T1DM clinical trial provides a comprehensive and complete database of EQ-5D-Y and disease-related variables to allow assessment of metric properties including validity, reliability, and responsiveness. However, some limitations need to be addressed. First, the assessment of the acceptability is not complete: Other indicators such as administration time needed to complete the questionnaire and patient views about the instrument should be included in further studies. Second, since the EuroQol group has not developed the value set for the EQ-5D-Y yet [23], the unweighted summary score was calculated as was done in previous studies [24-26]. However, this summary score did not allow cost-utility analysis. Third, the EQ-5D-Y does not cover social dimension, a key aspect of children’s HRQoL, but it presents an acceptable capacity of discrimination between functional and dysfunctional families in our study. There are other generic pediatric instruments that cover social dimensions, such as KIDSCREEN [50] (“autonomy and relationships with parents,” “social support,” “relationship with friends,” and “school environment”), PedsQol [51] (“social functioning” and “school functioning”), or the Child Health Questionnaire [52] (“role/social emotional and behavioral functioning”). Fourth, this is a secondary analysis of a study designed for purposes other than the evaluation of the metric properties of EQ-5D-Y. For example, as commented above, the 9-month period between test and retest is too long for measuring reproducibility, and no large change is expected with the intervention applied. Furthermore, results on stability should be interpreted with caution due to the small sample size. Finally, regarding external validity, the results of this study should not be generalized to other chronic conditions.

Despite the limitations discussed above, our results provide considerable evidence supporting the appropriate metric properties of the EQ-5D-Y in patients with T1DM. In conclusion, these findings suggest that the EQ-5D-Y administered online is an acceptable, valid, reliable, and responsive instrument for evaluating HRQoL in children and adolescents with this chronic condition. Since it is a preference-based health status measure, it will allow calculating quality-adjusted life-years (combining both the quantity and quality of life) for economic evaluations, once social preferences specifically for children become available. Given its short and easy administration, the EQ-5D-Y is a practical instrument to implement in primary care to routine monitoring. It is a promising instrument to compare the efficiency of different programs or treatment strategies, helping prioritize and invest at different levels.