Except for families who were not able to speak or read Norwegian, all 8-19 year old children or adolescents with type 1 diabetes scheduled for follow-up at three pediatric departments in eastern Norway between October 1^st, 2009 through February 28^th 2010 and one of their parents were invited by mail to participate in the study before a scheduled consultation.

Whether the cohort in the present study was representative for Norwegian children and adolescents with diabetes was assessed by comparing demographic and clinical characteristics with that of the Norwegian Childhood Diabetes Registry, which is a population based, nationwide registry covering all pediatric departments in Norway. In 2010, 95% of all children and adolescents with diabetes treated by pediatricians were included in the registry [²²].

The children and adolescents and their parents gave written consent according to Norwegian requirements. The study was approved by the Regional committee on medical research ethics.

Results are presented as means with one standard deviation (SD) or as rates (percentages). Floor and ceiling effects are reported in numbers of patients with HRQL scores of 0 (floor) and 100 (ceiling). A percentage above 25 was characterized as high.

Internal consistency refers to the degree to which the different items in a scale measure the same construct. For the DISABKIDS questionnaires reliability was assessed by tests of internal consistency of each of the subscales and the overall sum score. Cronbach's alpha coefficients above 0.70 are generally viewed as acceptable when instruments are used for group level analyses [³³,³⁴]. With short scales as in DCGM-37 and DDM-10 it is often more appropriate to report mean inter item correlations. Upper and lower limits of mean inter item correlations are a matter of discussion. Some authors claim that values between 0.2 and 0.4 are optimal [³⁵], while others argue that a mean inter item correlation consistently above 0.70, may indicate redundancy [³⁶]. In the present article, we consider mean inter item correlations between 0.2 and 0.7 as satisfactory.

Convergent and divergent validity of the DISABKIDS questionnaires DCGM-37 and DDM-10 were assessed with reference to the generic questionnaires CHQ- CF87 and CHQ-PF50, respectively, using Pearson correlation adjusted for age and gender. A correlation coefficient (r) above 0.5 between measures of construct related to each other was considered as high and a coefficient between 0.3 and 0.5 as moderate convergence, while measures were not considered to be related if correlation coefficients were below 0.3 [³⁴].

DISABKIDS' discriminant validity in relation to age, gender, duration of diabetes, mode of insulin delivery and metabolic control (i.e. levels of HbA1c) was assessed using multiple regression analysis.

Paired sample t-tests were used to assess associations between scores obtained by the children and their parents.

Significance was defined as p < 0.05. SPSS version 18.0 (SPSS IBM, NY, USA) was used for analyses.

For the children's questionnaires the internal consistency of the DCGM-37, calculated as Cronbach's alpha, varied between 0.55 and 0.92 (Table 2.). Cronbach's alpha was above 0.70 for all subscales except for "Physical ability" and "Social inclusion". Mean inter item correlations were above 0.20 and below 0.50 for all subscales except for the "Physical ability" subscale which was 0.19. For the DDM-10 Cronbach's alpha was 0.79 for both scales, and mean inter-item correlations were 0.41-0.49.

For the parent's questionnaires, Cronbach's alpha varied between 0.74 and 0.89 for the DCGM-37 and was 0.83 for both scales on the DDM-10. Mean inter item correlations varied from 0.32 ("Physical ability") to 0.55 ("Inner strengths").

Correlation coefficients between the DCGM-37 and DDM-10 scales and the "Mental health" subscale in CHQ-CF87 were in the range 0.54-0.81 (Table 3). Correlation coefficients were in the range 0.65-0.81 between the DCGM-37 total score and six of the twelve subscales in the CHQ-CF87, and in the range 0.49-0.67 between "Role emotional" in CHQ-87 and the DCGM-37 "Treatment" scale and the DDM-10 ("Impact" and "Treatment" scales).

Correlation coefficients between the DCGM-37 scales and the CHQ-CF87 scales "Physical function" "Role behavioral", "Global behavior" and family related dimensions were low.

The generic as well as the diabetes-specific module of DISABKIDS discriminated between age groups and levels of HbA1c (Table 4). Higher age and increasing HbA1c were associated with lower HRQOL scale scores.

Boys tended to score higher than girls, and children using insulin pump higher than those on multi-injections on total sum-scores and scores on most subscales. With respect to treatment modality, the differences in mean score between insulin pump and multi-injection users were between 3.5-4.0 on "Social equality "and "Physical treatment" subscales on DCGM-37 as well as for both subscales in the DDM-10 (results not shown).

There were no differences in scores with regard to duration of diabetes. Twelve (12%) of the participants had diabetes duration of less than one year. This group scored slightly higher on all subscales as well as on the HRQOL total score, but these differences were not statistically significant.

Generally, children and adolescents tended to give higher scores than their parents. This was true for all subscales in DCGM-37 and for the "Impact" scale in DDM-10. Significant mean (SD) differences were found for the DCGM-37 total score (78 ± 11.0 vs. 76 ± 11.1, p = 0.03), the subscales "Inner strength" (77 ± 15.8 vs. 72 ± 13.1, p < 0.01) and "Social inclusion" (80 ± 11.7 vs. 74 ± 15,0 p < 0.001). In the DDM-10 "Treatment" scale parents tended to score higher than their children (66 ± 17.1 vs. 63 ± 20.4 p = 0.14).

The major strength of the study was that the DISABKIDS questionnaires were applied to a wide age-range of children and adolescents (8-19 years old). Also, a broad perspective was taken by the collection of both self-reported and parent data. Furthermore, the results of the DISABKIDS were compared with those obtained with a HRQOL instrument (CHQ) which is well validated in Norwegian populations. Major weaknesses were the limited sample size and low participation rate. The reasons why 48% of the total population did not participate were multifactorial. The main reason reported by the nurses at the participative centers was that not all families were approached during periods with large work load in the clinic. However, this happened at random and we are confident that it did not introduce a bias.

Compared to the children registered in the national diabetes registry the study cohort had a somewhat lower mean HbA1c and higher proportion of insulin pumps users, but we still suggest that the results are representative of the national cohort for the following reasons: The difference in mean HbA1c was probably too small to be of major significance, and the difference in proportion of pump users was likely due to differences in treatment traditions. In Norway, pumps are, in practice, available without extra charge for all children, and the proportion choosing pumps is mainly a result of how familiar and confident the medical staff is with this modality. The clinics participating in the study have an active approach to encourage the use of pump, and the proportion of pump users in the study group was similar to the proportion among all the patients followed in the clinics.

Due to the relatively small sample size we did not perform factor analyses to assess the factor structure of the DCGM-37 or DDM-10 instruments, which is advocated when the instrument is applied to a larger population. With regard to reliability the study may be criticised for not applying test- retest reliability scores. However, this was not included because of concern for the patients and logistic challenges.

It is suggested that the CHQ-CF87 can be read to children less than ten years of age. The medical staff, after piloting, reported that some children between 8-9 years had insufficient reading skills to respond reliably on the DISABKIDS. Use of the instruments might therefore have a limitation in the youngest population if the questions are not read to them.

Most of the sub-scales showed Cronbach's alpha coefficients above 0.7, which are in agreement with the European field study [¹⁷]. The "Physical ability" scale had a low Cronbach's alpha compared to what was reported in the European field study where the patient populations consisted of seven different chronic conditions [¹⁷]. The "Physical ability" items in the DCGM-37 range from questions about difficulties with moving and running to questions on how life is ruled by the condition. Young persons with diabetes rarely experience physical complications due to the disease and, therefore, usually have no physical limitations. However, they experience significant practical and often emotional challenges due to repeated blood glucose measurements and administration of insulin, fear of hypoglycaemia, ketoacidosis and long term complications on a daily basis. Therefore, a low Cronbach's alpha was not unexpected. The Cronbach's alpha was also low for the "Social inclusion" subscale. In line with earlier reports, we suggest that the demands of adhering to treatment may create a feeling of separation from peers [²⁰]. Also, differences in the experiences between users of pump and multi-injection treatments may create an incoherent scoring, explaining the lower alpha on this scale.

In general, few items in a scale, such as six items in the DISABKIDS subscales, may make Cronbach's alpha calculations vulnerable to variations between items, and mean inter-item (MII) correlation has been suggested as an alternative analysis of consistency [³⁵]. This method modifies the findings in our study, as only the "Physical ability" subscale had a MII correlation below the lower acceptable limit of 0.2 underscoring that this subscale may not be informative for young people with diabetes. Furthermore, no scales in the two questionnaires had a MII above 0.50, strongly suggesting that the items in the scales were not redundant.

On the parents' reports Cronbach's alpha was above 0.7 on all subscales, and consistently higher than those of their children. Still, the pattern was the same as for their children in that the same subscales "Physical ability" and "Social inclusion" had the lowest Cronbach's alpha. These findings are in accordance with the European study [²⁴].

The heading in both the child's and parent's form "About your typical day" may not be explicit enough for young people to connect it to "Physical ability". Parents, however, may have interpretive abilities that cause them to answer more consistently.

Considering the content of the DISABKIDS questionnaires, some scales were expected to correlate better than others with the CHQ-CF87 scales. The pattern of associations between the subscales of the two instruments largely supported the validity of the DISABKIDS instruments. The European validation procedure used only a few of the questions from CHQ-CF87 (personal communication, John Chaplin, 2009). As far as we know a comparison with complete CHQ questionnaires has not been done earlier when examining the validity of DISABKIDS.

Six of the subscales from CHQ-CF87 showed high correlation with DCGM -37 total score. This is as expected since these six subscales, "Role emotional", "Bodily pain", "Behavior", "Mental health", "Self esteem" and "General health" all have items that are similar to those in the DCGM-37 Questionnaire. On the other hand, the DCGM-37 does not have any sub-scales mirroring the "Change in health", "Family activity" or "Family cohesion". It is therefore appropriate that these scales in CHQ-CF87 showed low correlation with the subscale scores in DCGM-37. This feature may have clinical and scientific implication for children with diabetes. Other instruments than DISABKIDS may therefore be more applicable if the primary goal is to measure family related factors [³⁷,³⁸]. Furthermore, clinicians and researchers need to be aware that the DISABKIDS instruments seem to have their strength in measuring the mental and emotional aspects rather than detecting physical health and family related aspects of HRQOL.

The different subscales of the two instruments demonstrated correlations that may seem surprisingly high or low with reference to how they are named. However, when considering the content of the respective items in the DCGM-37 and CHQ-CF87, associations were mainly as expected. For example, the names of the subscales "Physical ability" in the DCGM-37 and "Physical function" in the CHQ-CF87 give the impression that they measure similar functions. However, the "Physical function" subscale in CHQ-CF87 measures limitation in nine specific physical activities due to health problems, while DCGM -37's "Physical ability" scale is constructed differently, e.g.-items like "Is your life ruled by you condition" and "Does it bother you that you have to explain to others what you can and can't do?" measure emotional reactions to living with impairments, while the CHQ-CF87 to a larger extent addresses physical limitations. The "Physical ability" subscale in DCGM-37 correlated best with "Mental health" in CHQ-CF87, suggesting that the six questions measure a wider construct than physical abilities alone.

"Inner strength" (DCGM-37) correlated well with "Mental health" (CHQ-CF87) as would be expected due to the construction of single items in the two scales. The same was true for "Mental independence" in DISABKIDS vs. "Mental health" and "Self esteem" subscales in CHQ-CF87. The three constructs "Mental health", "Self esteem" and "General health" in CHQ-CF87 have 40 items covering most of the 37 item DISABKIDS questionnaire. The DCGM-37 therefore seems to be well suited for measuring the mental and emotional burden of a disease like diabetes.

The two DDM-10 subscales "Impact" and "Treatment" and correlated well with the "Role emotional" subscale on the CHQ-CF87. The two DDM-10 subscales actually measure the emotional consequences of having a chronic illness, quite similar to what "Role emotional" measures. The "Treatment scale" in DCGM-37 and the "Role emotional" scale were similarly correlated. These findings suggest that the DISABKIDS instruments are suitable for an early detection of mental and emotional worries with possible negative influence on self-management. The fact that these instruments are available as computer programs and can be completed by the patient and automatically scored prior to consultation, make them particularly attractive in clinical practice [³⁹].

The DISABKIDS generic instrument discriminated between patients based on age and metabolic control. It is likely that the scores on both of these variables have clinical significance as well [³⁴]. Older children scored lower than the younger, a finding that is consistent with other studies on HRQOL using different instruments [³¹,⁴⁰]. The effect of age may indicate a higher level of stress during puberty and late adolescence because of greater responsibility for own disease management and better cognitive ability to understand possible consequences of the disease, while parents, by taking more responsibility, tend to relieve the younger children from psychosocial burden [⁴⁰].

The finding of higher HRQOL score with better metabolic control is also in accordance with earlier studies [⁴¹]. The fact that the DISABKIDS instruments are able to identify these differences is important, and of interest for both clinical work and further studies.

The reason why we did not find significant differences in scores related to gender, type of treatment or duration of diabetes may be due to lack of statistical power because of limited sample size. The trends were, however, similar to what has been found as statistical significant differences in larger studies [⁴¹].

When evaluating differences in HRQOL studies, it is equally important to evaluate the statistical findings and numeric differences in scores in relation to clinical importance. We believe that the statistically significant findings in our study are clinically significant as well. Generally, it has been suggested that on a scale of 0-100 a change of 5-10 points is clinically significant for an individual [³⁴]. The numeric differences between those using insulin pump and multi-injections on the subscales "Impact" and "Treatment" on DDM-10 and on "Social equality" on the DCGM-37 may be clinically important, although they were not statistically significant in this limited study. The "Social equality" subscale consists of questions regarding external stigma. It has been reported that pump users have the ability to "hide" their disease better than those in need of other equipment for insulin delivery [⁴²]. The feelings of less impact of disease and smaller problems related to treatment might therefore contribute to the pump users scoring higher on this subscale.

In agreement with earlier HRQOL studies we did not exclude diabetes patients with onset of disease less than one year prior to the study[⁴⁰]. Patients with diabetes less than one year had non-significantly better HRQOL scores than those with longer duration. The lack of substantial difference may partly be due to the fact that most of those with duration less than one year had duration more than 6 months.

The DISABKIDS generic and diabetes specific modules showed differences between the children's and parent's score. The same findings, with parents tending to perceive their children's HRQOL lower than the children themselves, have been reported previously, for DISABKIDS as well as for other HRQOL instruments, and for other chronic diseases [²⁰,⁴³,⁴⁴]. These findings may be due to changes in conceptualization of health related quality of life over the course of the disease trajectory towards better acceptance with duration of the disease [³¹,⁴⁵]. It is also notable that the reverse was found on the DDM-10 "Impact scale", i.e. that disease was felt to have less impact by the parents than by their children or adolescents. The reason for this finding is not obvious, but may at least partly be due to the high proportion of pump users and parents believing that treatment with pumps implies less impact of the disease.

All subscales are scored from 0 to 100 with higher scores indicating higher self-perceived HRQOL.

^a Pearsons's correlation coefficient adjusted for age and gender

*p < 0.005. **p < 0.01. ***p < 0.001

^aAdjusted for gender, duration of diabetes and use of insulin pump vs. multi-injections using multiple linear regression analyses

*Effect per year of age, **Effect per % increment of HbA1c.

All subscales are scored from 0 to 100 with higher scores indicating higher self-perceived HRQOL.