Cognitive interviews were conducted with patients from two pediatric cardiology centers and one pediatric outpatient diabetes unit in Sweden. Patients were eligible for these cognitive interviews if they were aged 14–19 years and if they had a scheduled appointment in the participating units in the week that the interviews were held (November-December 2015). Participants were recruited consecutively and comprised 6 patients with CHD (5 girls, 1 boy) and 3 with DM1 (1 girl, 2 boys), with an age range of 14–19 years.

The participants were asked to fill out the scale together with one of the researchers (ELB, CSL, AB) and to provide feedback on the comprehensibility of the items (probing), response options available and possible suggestions on how to improve the instrument. Interviewers documented the participants’ comments in their field notes. This process helped to identify structural problems in the GYPES and issues with interpretability.

Data from all interviews were discussed by the interviewers and items were rephrased if the participants stated difficulties in understanding the meaning of an item and/or considered it to be repetitive.

As part of the STEPSTONES project (Swedish Transition Effects Project Supporting Teenagers with chrONic mEdical conditionS), a cross-sectional study was conducted on young persons with CHD in Sweden. The main aim of this cross-sectional questionnaire study was to describe transitional care outcomes such as the level of empowerment, transition readiness, knowledge, health behavior, perceived health status, illness perception and quality of life in adolescents with CHD, and to investigate its correlates. Eligible participants were selected from the Swedish Registry of Congenital Heart Disease (SWEDCON) [15]. Patients were included if they had a confirmed diagnosis of CHD, defined as “a gross structural abnormality of the heart or the intrathoracic great vessels that is actually or potentially of functional significance” [16]; if they were aged 14–18 years at the time of the study; and if they were receiving follow-up care at one of the four pediatric centers of CHD in Sweden that participated in the study. Patients were excluded if they had a cognitive and/or physical limitation that did not allow them to answer the questionnaire; had undergone heart transplantation; or had not provided consent to participate.

In all, 593 patients met the inclusion criteria and were sent a set of questionnaires. In total, 202 patients completed and returned the questionnaires, corresponding with a response rate of 34.1%. Demographic characteristics of this sample are presented in Table 1. The mean age of the sample was 15.7 years; 55% were boys. At the time of the study, the majority of participants were in junior high school (57.4%) (Table 1). Participants and non-participants did not differ on sex and disease complexity. There was a significant difference (p<0.05) between the participants (15.7±1.1) and non-participants’ age (15.5±1.1). However, this difference was considered not clinically relevant since the Cohen’s d was 0.18.

Each participant received a package containing information about the study, an informed consent, a set of questionnaires and a pre-addressed envelope. They were asked to return completed questionnaires and the signed informed consent in the pre-addressed envelope. To minimize the non-response rate, a modified Dillman procedure was used [17]. Three weeks after the documents were sent, non-responders received a personalized reminder letter. After five weeks, persistent non-responders received a second reminder letter and a new set of questionnaires. Finally, if participants had not returned the questionnaires after seven weeks, they were contacted by telephone. On this occasion they were asked whether they had received the questionnaires, if their address was correct and whether they wanted to participate in the study. Data collection ran from January 25^th till August 31^st 2016.

In Study II, the psychometric evaluation of the GYPES was based on content validity, factorial validity, internal consistency and responsiveness. For content validity, the proportion of missing values and invalid scores was calculated because these parameters indicate how intelligible an item is [18].

Factorial validity was evaluated through confirmatory factor analysis (CFA) to assess the hypothesized factor structure of the scale, i.e. the five dimensions of empowerment. A chi-square index (χ² index) as small as possible, comparative fit index (CFI) value >0.90, a root mean square error of approximation (RMSEA) and standardized root mean square residual (SRMR) <0.08 were used to assess an adequate model fit [19, 20]. After testing a first-order five-factor model, we tested a second-order model with five first-order factors and a second-order global empowerment factor.

Reliability was assessed by calculating the Cronbach’s alpha coefficient as a measure for internal consistency. Coefficients were calculated for the overall scale and for each dimension [21].

Floor and ceiling effects were calculated in order to assess the scale’s potential responsiveness. Floor and ceiling effects were considered present if more than 15% of the participants achieved the lowest or highest score [22, 23].

Statistical analyses were performed with Mplus version 7 software (Muthèn & Muthèn, Los Angeles, CA) for CFA. IBM SPSS Statistics for Windows version 22 was used for other analyses (Armonk, NY: IBM Corp.).

As part of the “Better Transition in Diabetes project”, a cross-sectional study was conducted on young persons with DM1 in the Netherlands. The main aim of the cross sectional study was to describe experiences with transitional care and outcomes, such as level of empowerment, transfer experiences and coping with DM1 in daily life. Eligible participants were members from the Dutch Diabetes Association (DVN) and Stichting ééndiabetes (a Dutch foundation for young persons with DMI), all aged 12–25 years.

A total of 273 individuals participated in the study. Respondents’ characteristics are shown in Table 1. Three quarters of the patients were female (78.0%). The mean age was 19.9 years and most of the participants were in senior high school (44.3%) or at college or university (40.3%).

Both patient organizations disseminated the invitation for this survey through their networks and posted a call on their Facebook pages. DVN also sent an email to its members and a reminder after two weeks. Ééndiabetes posted a news article about the project on their website. Respondents voluntarily filled out the GYPES using a web-based system. All questionnaires were submitted anonymously and no identifying data was collected from the participants. To encourage participation, five gift vouchers worth €50 were raffled among the respondents. Data collection ran from mid-June to mid-September 2016.

The same statistical approach has been used as in Study II.

Based on this feedback, we modified some responses of the 5-point Likert scale in the Swedish version in order to increase interpretability of the options.

One item from the “enabling others” dimension was reworded. After these changes, the GYPES resulted in a 15-item instrument measured in a 5-point Likert scale (strongly disagree, disagree, neither agree nor disagree, agree and strongly agree) scale that was later used in Study II.

The proportion of missing values ranged from 0.5–3.4% (Table 2). Items from the “shared decision-making” and “enabling others” dimensions had the highest missing values (Table 2).

The aim of the CFA was to test a five-factor structure, in accordance with the five dimensions of empowerment. Based on the fit indices (dƒ: 80; χ²: 154.948, p<0.0001; CFI: 0.916; RMSEA: 0.068; SRMR: 0.069), the model had an adequate fit (Table 3). Factor loadings are shown in Table 2. The factor loadings of the items varied between 0.475 and 0.892, apart from the identity factor, which had one item with a low factor loading (0.185, p = 0.014). If this item was deleted and a new model with the five-factor structure with only two items in the “identity” dimension was run, an improved model fit was observed (see Table 3). We also evaluated the second-order factor model but the model did not converge properly, which may be due to including only two items in the “identity” dimension.

Cronbach’s alpha coefficients for all factors were above 0.6, apart from identity, which had a value of 0.521 (Table 4). The value for the overall scale was of 0.819, reflecting an adequate level of internal consistency. The mean total empowerment score was 54.5 ± 10.5 on a scale from 15 to 75.

None of the patients had the lowest possible score and only 1.5% of the participants had the highest possible score, indicating there were no floor or ceiling effects.

We have modified some items of the version of the GYPES used in Study II on statistical and substantive grounds. For statistical reasons, item 9 was rephrased because it turned out to be a problematic item in the “identity” dimension. For substantive reasons, we rephrased items 7 (identity), item 11 (shared decision-making) and item 14 (enabling others) to further improve consistency and increase understanding.

The revised version of GYPES was translated from English into Dutch following a forward-backward translation process [24]. Two translators (one bilingual in Dutch and English) translated the instrument independently from each other and presented the results to the Dutch team. The translation was tested for face validity with two mothers and a 16-year old adolescent with epilepsy. Consensus was reached after discussing both versions. This process resulted in the final version of the GYPES that was evaluated in Study III (Additional File 1: Final version of the GYPES).

No missing values were found for the GYPES items, suggesting that the wording of the items was intelligible.

CFA was also used to examine if the five-factor structure fitted with the observed variables. Fit indices generally indicated an adequate model, although CFI did not reach the threshold of 0.90 (dƒ: 80; χ²: 235.375, p<0.0001; CFI: 0.897; RMSEA: 0.084; SRMR: 0.059) (Table 3). In order to improve model fit based on the modification indices we evaluated a second model allowing one error correlation between factors (item 6 with 15). This was the largest error correlation and even when these items belong to different dimensions, it can be theoretically expected that individuals who are more actively involved in their care, feel more capable of sharing their experiences and coping techniques with other individuals [25]. This showed an adequate fit across all indices (dƒ: 79; χ2: 201.950, p<0.0001; CFI: 0.919; RMSEA: 0.076; SRMR: 0.057) with factor loadings that ranged from 0.385 to 0.941 (Table 3). The latent factor correlations of this model are shown in Fig 1.

Moreover, a third model assessing the second-order factor model showed an adequate fit as well (dƒ: 84; χ²: 222.788, p<0.0001; CFI: 0.908; RMSEA: 0.078; SRMR: 0.061). The items’ factor loadings for this model had values from 0.379 to 0.938. The factor loadings of each first-order factor in relation to the global empowerment factor had values exceeding 0.660 (knowledge and understanding: 0.729; shared decision-making: 0.660; personal control: 0.950; identity: 0.777; and enabling others: 0.617). Results from this model support the calculation of subscale scores, as well as an overall empowerment score.

The final version of the GYPES showed to be internally consistent with a Cronbach’s alpha of 0.858. Alpha values were acceptable for all the subscales, with values above 0.6 (Table 4). The mean empowerment score in this sample was 58.9±7.9 on a scale from 15 to 75.

None of the participants had the lowest scoring and only 2.2% of the participants had the highest possible score. Thus, the GYPES is not subject to floor or ceiling effects.