Measuring social participation in children with chronic health conditions: validation and reference values of the child and adolescent scale of participation (CASP) in the German context

This study took place in three different settings resulting in two clinical and one population-based samples: i) an outpatient paediatric neurology centre (clinical sample 1; SPZ Frankfurt/Main, Federal State of Hesse, Germany), ii) an inpatient paediatric rehabilitation clinic (clinical sample 2; Klinik Viktoriastift Bad Kreuznach, Federal State of Rhineland-Palatinate, Germany) and iii) within the city limits of Mainz and the rural district of Mainz-Bingen (population-based sample; Federal State of Rhineland-Palatinate, Germany). This approach was chosen to i) cover a broad range of chronic conditions, living environments, and health care services and ii) test the CASP’s ability to differentiate between children with and without chronic health conditions of various severities.

The clinical sample 1 consisted of children with predominantly chronic neuro-paediatric conditions (e.g. cerebral palsy, epilepsy, neuro-genetic conditions) and was recruited via convenience sampling at regularly scheduled appointments at the outpatient centre between November 2015 and April 2016. The clinical sample 2 was a convenience sample of children and adolescents with a broad spectrum of different chronic health conditions such as diabetes, obesity, and attention-deficit/hyperactivity-disorder (ADHD), who were recruited at the inpatient clinic between March and June 2016. The population-based sample consisted of participants of an ongoing prospective cohort study of children officially registered for school entry in autumn 2014 [14]. Recruitment for the current study was performed at the end of first grade in June 2015.

Inclusion criteria for the clinical samples were: 3–11 years of age, a certified medical diagnosis, and sufficient German language skills of parents and children. Inclusion criterion for the population-based sample was active participation at the time of recruitment in the underlying cohort study. Written informed consent was obtained in parents and in children of more than 8 years of age. The study was approved by the Medical Ethics Review Board of the Medical Faculty Mannheim at Heidelberg University (2015-550 N-MA), the Ethics Committee of the Federal Physician Chamber of Hesse (MC180/2015) and the Ethics Committee of the Regional Medical Association of Rhineland-Palatinate (837.544.13 [9229]).

A prospective validation study was performed. After instrument translation, validity and reliability was assessed in using cross-sectional (validity) and longitudinal (reliability) study designs. Data were largely obtained by questionnaires, which were either handed out in person (e.g. at regular appointments in the clinical samples) or mailed to parents in the population-based sample. Parents took about 10–15 min to fill in the questionnaires and were asked to hand back the questionnaires after the appointment or send it back to the study centre using prepared stamped envelopes. Beyond this survey, detailed information on chronic health conditions was retrieved either from the local clinical information system (clinical samples) or from the preschool examination database (population-based sample) provided by the regional Department of Public Health (County Government Mainz-Bingen).

Social participation was measured using a German version of the CASP [11]. The CASP is a 20-items 4-scales parent-report instrument covering attendance and involvement in the areas “home”, “community”, “school”, and “living activities”. The instrument measures the extent to which children of 3 to 18 years of age participate in activities in comparison to healthy children of the same age. Due to its shortness, however, the CASP does not differentiate between different aspects of participation such as attendance and involvement. Responses are rated on a 4-point scale ranging from 4 “age expected” “(i.e. full participation)” and 3 “somewhat limited” to 2 “very limited” and 1 “unable to participate”. An additional fifth response category of “not applicable” is provided for all items. In addition, the instrument offers three open-ended questions that can be used for individualized family-centred care planning. The original CASP was shown to have high reliability (Intra-class coefficient = 0.94; Cronbach’s α > 0.96) and usefulness to discriminate between children with and without disabilities [11, 15]. Scores are summed up and calculated for the total scale and 4 subscales [16]. The scores of the scales range from 0 to 100%, higher scores thereby indicate more favourable participation.

The original English version was translated and back-translated by two independent translators as requested by international standards [17]. Based on comparisons of translated versions, a final version was consented after discussion among translators and an additional expert in the field (HP). This version was then applied to seven parents and nine experts in the field to assess cultural and content appropriateness. Based on the feedback, further cultural and language adaptations were undertaken to increase acceptance in German language settings. Cultural adaptions encompassed primarily a change of illustrative examples for activities as well as the use of simplified vocabulary to make the items easier to understand for parents with migrant background. For example, leisure-time physical (German: “Kinderturnen”) and musical education (German: “Musikschule”) were added as important additional after-school activities.

The identification of chronically ill children was based on the following data available: i) ICD-10 codes from the clinical information systems (clinical samples), ii) parent-reported doctor diagnoses from the preschool examination database (population-based sample), and iii) information on special health care needs as assessed in the preschool examination of Rhineland-Palatine (all three samples).

As there was no international definition for the identification and severity rating of chronic health conditions, we developed an algorithm based on the criteria published by Perrin et al. [18] and the ICF principals. Ratings were performed independently by two medical experts (FDB (author, pediatrician), CS (speech therapist)) and reviewed by a third one (HP (author, pediatrician, pediatric neurologist)). The following chronic health conditions were considered: combined developmental problems, obesity, ADHD, behavioral and conduct problems, cerebral palsy, epilepsy, autism, genetic syndromes, intellectual disability, and former cancer.

First, presence of one of the above chronic health conditions was determined by reviewing the ICD-10 codes and parent-reported doctor diagnoses. Second, based on ICD-10 codes and doctor diagnoses, the health condition was rated as severe, if long-lasting, irreversible structural and functional deficits were present or could be expected. Otherwise, the condition was rated as mild. Third, needs and uptake of prescribed medication for a medical problem were evaluated. Depending on the presence of such a medication, the medical condition could be upgraded (e.g. from no to mild or mild to severe). In a fourth step, needs and uptake of physical, occupational, or speech therapy were evaluated. Depending on the presence of such a treatment, children were upgraded from “without chronic condition” to mild health condition, to also include conditions not yet recognized as diagnosis. Last, needs and uptake of social or educational support or mental health services were evaluated. Again, depending on the support, the children were upgraded from “without chronic condition” to mild health condition. The identification and rating steps of this algorithm are illustrated in Fig. 1. This algorithm allowed the combination of information of the condition itself with information the condition’s consequences and impact and the actual health care use.

Health-related quality of life (HrQoL) was measured using the KINDL^R parent-proxy instrument [19]. The KINDL^R is a 24-items 6-scales parent-reported instrument covering the domains “physical wellbeing”, “emotional wellbeing”, “self-esteem”, “family”, “peers”, and “pre-school/school”. Responses are rated on a five-point scale ranging from “never” and “rarely” to “sometimes”, “frequently” and “always”. Comparable to the CASP, a total scale (scores range from 0 to 100) and subscales are available. Results for validity and reliability of the instrument (Chronbach’s α = 0.85 for the total scale) have been published elsewhere [20].

Although the KINDL^R was developed as a measure of HrQoL, only 36% of its items were found to be congruent with the WHO HrQoL definition [8]. In contrast, more than 50% of the items were congruent with one of the ICF-CY domains [8]. We, hence, expected some positive relationships between parent-reported CASP and KINDL measures, although the concepts of HrQoL and participation are theoretically quite different.

Age, sex, and migrant background of the child as well as parental education was obtained by additional questionnaire items. Parental education was defined based upon the highest parental level of education achieved, according to international standard classification of education (ISCED; high = ISCED level 4–5, medium = ISCED level 3, and low = ISCED levels 0–2). Migrant background was operationalized based on the recommendation by the German National Statistics Bureau [21], which accounts for parental nationality and country of birth.

Descriptive statistics (n and percentage, mean and standard deviation) were used to describe the distribution of socio-demographic and clinical factors among samples and disease severity groups (Tables 1 and 2).

Content validity was assessed using a face-validity approach. Eight experts (2 paediatricians, 2 speech therapists, 2 physiotherapists, and 2 occupational therapists) rated comprehensibility, importance, and conceptual comprehensiveness of the instrument on a 5-point rating scale ranging from “not at all” to “very much”.

Exploratory factor analysis (EFA) using a promax rotation was performed to analyse the underlying structure of the CASP items. Due to missing values (mostly because parents chose the “not applicable” option), an imputation technique with an expectation-maximization (EM) algorithm was applied [22]. This method was found to provide a better basis for EFA than other more traditional methods (e.g. complete case analysis or mean imputation [23]). To specify nominal sample size, we use the column-wise minimum, which is defined as the number of complete cases for the variables with the most missing values [22]. The Kaiser criterion (eigenvalues > 1) was applied to define the number of factors. Finally, internal consistency of the CASP was analysed using Cronbach’s alpha for both, the total scale and all subscales.

In absence of a German reference standard for measuring participation, relationships between CASP and KINDL^R measures were investigated. Therefore, a single questionnaire item covering parental perceived overall social participation ranging from 1 “unable to participate” to 4 “age expected” was introduced and Pearson’s correlation coefficients between CASP scales, KINDL^R scales, and perceived overall participation were calculated.

Test-retest reliability was assessed by repeated administrations of the instrument in children from the clinical sample 1 with a time lapse of 2 to 4 weeks between measurements. Therefore, intra-class correlation (ICC) coefficients were calculated for the CASP total scale and all subscales.

CASP reference values were calculated based on the population-based sample and contrasted to the results of the clinical samples. The association between disease severity and participation was analysed in using linear regression analysis with the CASP total scale as the depending variable and disease severity, age, sex, and parental education as independent variables. The ability of the CASP to classify children according to their disease severity was investigated in using receiver operating characteristic (ROC) curves and area under this curve (AUC). Results for the CASP were contrasted to results for the KINDL. All statistical analyses were performed using STATA 13.1 [24].

In total, 327 children were enrolled into the present study (clinical sample 1: 84 children, clinical sample 2: 28 children, population-based sample: 215 children). Demographic characteristics stratified by sample are given in Table 1. In total, 103 children (31.5%) were classified as without chronic health condition, while 174 children (53.2%) were categorized as mild chronic health condition and 50 children (15.3%) as having a severe chronic health condition. Demographic characteristics stratified by disease severity are given in Table 2.

Here: Tables 1 and 2.

Comprehensibility of the final German CASP items (see Additional file 1) was rated to be 3.8 (i.e. 76% of a maximum of 5) for subscale “home”, 4 (80%) for subscale “community”, 4.7 (93%) for subscale “school”, and 4.6 (91%) for subscale “living activities”. Importance ratings were as follows: 4.7 (i.e. 93% of a maximum of 5), 4.6 (91%), 4.7 (93%) and 4.4 (89%) for all subscales, respectively. Finally, conceptual comprehensiveness was rated 4.4 (88% of a maximum of 5), 4.1 (82%), 4.2 (84%), and 4.6 (91%) for all the subscales, respectively. These results corresponded to good to excellent content validity.

Results for the EFA are provided in Table 3. The column-wise minimum was n = 290. The Kaiser criterion implied an underlying one-factor structure. This factor contributed to 88.7% of the variance explained. Cronbach’s alpha for both, the total scale (α = 0.98) as well as for the subscales A-D (A: α = 0.94, B: α = 0.92, C: α = 0.93, and D: α = 0.93) indicated excellent internal consistency.

We found a moderate positive correlation between the CASP and the KINDL^R total scales in the total sample (n = 314, r = 0.45, p-value< 0.001). The correlation was highest in children with mild conditions (n = 169, r = 0.50, p-value< 0.001), lower in children with severe conditions (n = 43, r = 0.35, p-value = 0.022) and weak in children without chronic health conditions (n = 102, r = 0.19, p-value = 0.062). There was an excellent correlation between the CASP total scale and the results for the single questionnaire item covering parental perceived overall social participation (n = 308, r = 0.86, p-value< 0.001).

Based on the population-based sample of 215 children, mean CASP total score was 98.2 (±5.1), with a high left skewness (− 8.0) and kurtosis (82.6), and the following percentiles (5th: 92.1, 10th: 95, 25th: 98.7, median: 100). Children with mild chronic health conditions had distinctly higher CASP values (mostly> 80) than children with severe chronic health conditions (between 50 and 80: see Fig. 2). In line with Fig. 2, ROC analyses demonstrated that children with mild compared to severe chronic health conditions differed much more in CASP (AUC = .92), than in KINDL (AUC = .75, p-value for difference < 0.001, Fig. 3).

Compared to children without chronic health conditions, children with mild conditions had 3.4 points (±1.2; p-value< 0.01) and children with severe conditions had 28.7 points (±1.9; p-value< 0.001) lower values regarding the CASP total scale (Table 4). The differences across groups remained after adjusting for age, sex, and parental education.

This study has several strengths and limitations. It is the first study that provides validation data on a comprehensive instrument to measure social participation across a broad age range in German language. For this study, we intentionally sampled participants from three different populations to be able to draw conclusions on the generic value of the CASP. This sampling strategy should increase generalizability to a broad spectrum of settings and diseases in childhood. However, we acknowledge that the convenience sampling impeded a non-responder analysis, thereby limiting overall generalizability.

Another limitation is the age range in our sample. While Bedell [11] assessed validity in a sample of children aged 3–22 years, the age range in this study was only 3–11 years. Therefore, further validation steps of the German CASP in adolescents are definitively needed. On the other hand, social participation in adolescents should be assessed rather by self-reports than by parent-reports. Therefore, it is acceptable that a parent-report instrument is validated only up to an age range of 12 years.

Although we tried to cover a broad range of social and ethnic parental backgrounds in our sample, sufficient German language skills was an important inclusion criterion for study enrolment, which may hamper generalizability of results, in particular, to parents with migrant background and insufficient language skills. In fact, our study was not large enough to demonstrate acceptance and comprehensibility of the instrument (or the concept of social participation at all) in various ethnic groups. This should be acknowledged when the CASP is applied to parents with migrant background.

Similar to Bedell [11], diagnoses within the different disease groups were quite diverse. Thus, particularities and differences between certain conditions were not accounted for. Given that there is no international definition for the identification and severity rating of chronic health conditions, we developed a new algorithm based on the criteria published by Perrin et al. [16] and the ICF principals. However, this attempt had some limitations in acknowledging health care use, but nevertheless seemed to be useful to examine differences in participation across the condition severity spectrum. Further investigations in children with certain diseases or groups of similar diseases might be needed to assess the specific differences in participation. For example, Hwang et al. separated mental from physical disability in their study on the Chinese CASP version [26]).

Both, the participation and the HrQoL results were based on parental reports. For HrQoL, there is evidence for only limited overlap between self-reports and proxy-reports in children [27]. In analogy, self-reports of participation may be more accurate compared to parent-reports in older children and adolescents, which should be addressed in future validation and outcome studies. On the other hand, a positive relationship between participation and HrQoL can be expected to some extent, if the reports are not independent from each other. However, well-validated self-report participation instruments are not yet available for children and we explicitly chose the CASP to be able to cover a broad age range, which is one advantage of this instrument.

At last, we had to cope with a substantial high number of missing data within the CASP due to items that were “not applicable” as predefined for the original CASP. Parents may have misinterpreted the term `age-expected´, especially for the CASP items concerning `home and community living activities´. To simulate everyday health care practice, data were not imputed for most of the analyses, leading to a data loss of up to 33% for regression and correlation analyses. Given that subscales are only valid, if less than 40% of items are missing, we decided to impute missing values for the factor analysis. This could be one reason why we were not able to confirm the 3-factor structure of the original CASP. In the CASP Administration and Scoring Guideline [16], it is recommended “… to not include the non-applicable item in the scoring or to first take the average of the specific subsection and use this as the score for the non-applicable item”. However, studies investigating the effect of various missing imputation techniques on CASP results are missing. In summary, the problem of imputing missing data for the CASP is not satisfactory solved and future validation studies may focus on this specific topic.