The Inception COhort of Newly diagnosed patients with JIA (ICON), a, prospective observational cohort study, was used to evaluate the PARQ and CARQ. ICON aimed to observe patients with recent-onset of JIA according to the ILAR classification criteria [1] for at least 10 years. Eleven of the largest pediatric rheumatology centers in Germany recruited patients for this project from 2010 onwards and documented clinical characteristics and treatments by a standardized physician questionnaire every 3 months during the first year of observation and every 6 months in the subsequent years until the end of 2021. More details on the ICON cohort were described by Sengler et al. 2015 [21].

The PARQ and CARQ, global assessments of adherence (GAA) to drug treatment, exercises and wearing splints, and other questionnaires were completed by parents and patients (if age ≥ 8 years) 48 months after enrolment in ICON.

Between 2010 and 2012, 951 children and adolescents with JIA were enrolled in the ICON study. By the survey date (4-year follow-up) of PARQ and CARQ, there were n = 176 drop-outs (18.5%). Of the resulting 775 available patients, n = 243 did not participate in the 4-year follow-up but did participate in subsequent visits and thus remained under further observation. Thus, a total of 532 patients and their parents were available to complete PARQ and CARQ as part of ICON.

Sociodemographic and clinical information such as age, sex, parents’ country of origin, disease onset, date of diagnosis, JIA category and global assessment of disease activity (measured by numerical rating scale, NRS 0–10) were provided by the treating pediatric rheumatologist. Patient-reported outcomes included the assessment of overall well-being and pain using NRS. Functional ability of the patients was assessed by the German version of the C-HAQ [22]. Health-related quality of life was reported along with the German versions of the Pediatric Quality of Life Inventory (PedsQL) generic core scales and rheumatology module scales (range 0–100, best = 100). The PedsQL rheumatology module assesses treatment-related problems with 7 single items [23].

Disease activity was measured by the clinical Juvenile Arthritis Disease Activity Score-10 (cJADAS-10), which includes the physician's global assessment of disease activity, the patient's global assessment of general well-being and the number of active joints (up to a maximum of 10). The 2014 JADAS-cutoff values for oligo- and polyarthritis were applied dependent on the cumulative joint involvement [24, 25].

The socioeconomic status of the patients was calculated using an established German multidimensional aggregated index [26]. As the parental work status is not assessed in the ICON study, the calculation of this index was modified to be based only on the parental educational level (including school education and vocational training) and the net household income. The study by Listing et al. includes more details [27]. The lower and upper quintiles of the sum of the education and income scores (6.55, 12.1) were used as cut-off points to define low, moderate, and high SES.

The German versions of the PARQ and the CARQ were fully translated from its original Canadian- English version and cross-culturally adapted according to international guidelines [28, 29]. Two forward- and backward translations each were conducted by an expert committee consisting of a health scientist, three pediatric rheumatologists whose mother tongue was the target language (German) as well as an American-English native speaker with no medical background. Each translator performed a separate translation before a consensus version was implemented.

Families who completed the 4-year follow-up between December 2015 and December 2016 were sent the PARQ and CARQ (for children ≥ 8) again about 2 weeks later for retesting in complete.

The PARQ and CARQ were distributed to parents and patients, respectively, enrolled in ICON.

The CARQ, which is a simplified version of the PARQ, addressing children ≥ 9 years, has been pretested in both English and French [19, 20]. In our study, children ≥ 8 years were asked to fill out the CARQ as the ICON questionnaires are distributed age-group wise and the corresponding age group includes children from 8 years onwards.

The PARQ and CARQ are composed of three sections, each of which commences with a question including the phrase “in the past 3 months”, which is intended to emphasize the time since the last prescription of treatment.

Background section identifies who is primarily responsible for different treatment components: taking medication, doing exercises, and wearing splints.

Patients and methods section measures the child’s ability related to i) general level of difficulty in following treatment, ii) frequency of following treatment and iii) negative reactions in response to treatment which is evaluated with a 100 mm Visual Analogue Scales (VAS, 100 = “very hard” for item i and “always” for items ii) and iii). Responses to items i) and iii) are reversed so that higher scores indicate fewer difficulties and fewer negative reactions, respectively. An overall score for the child's abilities (Child Ability Total Score = CATS, 0—100) can be derived by averaging the answers to each of the three items. Furthermore, it is asked whether errors were made in medication behavior using a yes/no format: (1) ever forget to take medicine; (2) being careless at times about taking medicine; (3) when feeling better sometimes stopped taking medicine; and (4) when feeling worse when taking medicine, sometimes stopped taking it. Positive responses are summed, with total scores ranging from 0 (no errors) to 4.

Results section asks about caregiver´s and child`s perception regarding the helpfulness of each treatment component (e.g., medication, exercises, and splints) using a 100-mm horizontal VAS (100 = “very helpful”). Caregivers are also asked about the therapies they most preferred. The final question (only PARQ) inquires about potential barriers that families have encountered with regard to treatment.

Global Adherence Assessment (GAA) by parents and patients ≥ 13 years of age was assessed by asking how frequently i) medications, ii) exercises, and iii) splints were taken, performed, or worn, respectively. Responses were documented on a self-constructed 5-point Likert scale for each measure. Response options ranged from "almost always," "frequently," "sometimes," "rarely," to "not at all" and included "no therapy required."

Standard descriptive statistics were used to describe the distribution of the single PARQ and CARQ items. The rate of missing values was calculated for each PARQ and CARQ item The basic population for the investigation of missing values was always the patients who had been prescribed the respective form of therapy.

Construct validity was investigated based on the PARQ and CARQ by assessing convergent and discriminant validity.

Within the assessment of convergent validity, scales measuring the same construct were correlated with each other: For each treatment component (medications, exercises, splints), the Spearman rank correlations were calculated for each of the VAS "frequency of adherence" and the CATS in relation to the GAA. In addition to that, we assessed polychoric correlations between the yes/no format question on errors in medication behavior: “when feeling worse when taking medicine, sometimes stopped taking it” and the PedsQL treatment subscale single item “medicine makes child feel sick”. For the assessment of exercises, we looked at Spearman rank correlations between the VAS iii) “negative reactions in response to treatment” and the PedsQL treatment subscale single item “physical therapy or daily exercises hurt”.

Within the assessment of discriminant validity, our underlying hypotheses were that some parameters are associated with better adherence: a younger age (since the parents can "control" more), a higher disease activity as well as more pain (since the burden of suffering is higher and thus also the expectation of an improvement of the state of health through e.g. medication), a higher SES (because therapies may also be carried out that cost the patients or their parents extra money), as well as a shorter duration of therapy with methotrexate (because with longer therapy the rate of patients with MTX aversion increases and thus presumably more patients no longer want to take MTX).

Different groups were formed based on clinical and sociodemographic parameters and correlated with the CATS for each corresponding treatment type. The groups included in our analyses were age (\(\le 12\mathrm. >12 \mathrm)\), sex (male vs. female), socioeconomic status (low, moderate, high), treatment with methotrexate (MTX, < 2 years vs. ≥ 2 years), pain (NRS = 0 vs. NRS > 0, overall well-being (NRS = 0 vs. NRS > 0), functional limitations (CHAQ = 0 vs. C-HAQ > 0) as well as disease activity (cJADAS ≤ 1 vs. cJADAS > 1). T-tests (Welch test) were performed to evaluate the ability of the PARQ and CARQ for their discrimination between selected patient groups with a hypothesized difference in adherence. To evaluate concordance of these questionnaires, comparisons of the CATS PARQ with the CATS CARQ were performed using a paired t-test.

Test–retest performance was carried out by the kappa coefficient and intraclass correlation coefficient (ICC) for categorical and continuously distributed variables, respectively. ICC values greater than 0.90 were considered excellent, between 0.75 and 0.90 good; whereas values between 0.50 and 0.75 or less than 0.50 were considered moderate and poor agreement, respectively [30]. As suggested by Landis and Koch [31] we regarded kappa coefficients below 0.21 as indicative of slight agreement, values from 0.21 to 0.40 as speaking for fair agreement, values between 0.41 and 0.60 as moderate, from 0.61 to 0.8 substantial and kappa values > 0.81 indicating almost perfect agreement.