Transdiagnostic Validation of the German Benet Finding Scale (BFSC) for Youth With Chronic Conditions

We examined the psychometric properties of the 10-item Benet Finding Scale (BFSC) in a transdiagnostic sample of German youth facing chronic conditions (N = 304; 12 – 21 years). Exploratory factor analysis with a rst subsample revealed a one-dimensional factor structure. Conrmatory factor analysis with a second subsample veried the one-dimensionality with an acceptable t. The BFSC exhibited acceptable internal consistency (α = .87 – .88). Benet nding (BF) was positively correlated with age, disease severity, optimism, self-esteem, self-ecacy, sense of coherence, and support seeking. There were no correlations with avoidance, wishful thinking, emotional reaction, and health-related quality of life. Sex differences in BF were not consistent across subsamples. BF was negatively associated with social status. The BFSC is a psychometrically sound and transdiagnostic instrument to assess BF in youth and may facilitate further research on positive adaptation processes in response to chronic conditions.


Introduction
Stress and coping research is shifting from focusing exclusively negative effects of chronic conditions (CC) to an emphasis on ways in which these conditions promote positive life changes (Park et al., 2009). Bene t nding (BF), de ned as individual differences in perceiving positive life changes resulting from adversity and negative life stressors (Helgeson et al., 2006;Park, 2009), herein, emerged as a key construct and gained increasing attention in the context of CC (Algoe & Stanton, 2009). Positive life changes may manifest themselves in domains including intrapersonal bene ts (e.g., feeling stronger and wiser), interpersonal bene ts (e.g., feeling closer with friends and family), and changes in priorities and goals (e.g., reordering goals and emphasis of enjoyment in life) (Tedeschi & Calhoun, 2004). There is rst meta-analytic evidence that BF in response to several health stressors is associated with lower levels of depression and global distress as well as more positive well-being (Helgeson et al., 2006). While BF was studied among adults with various CC (Helgeson et al., 2006;Park, 2009), studies among youth are lacking (Meyerson et al., 2011).  Tran et al., 2011). However, only one measure for BF was psychometrically evaluated for children and children and adolescents with cancer (Phipps et al., 2007). The Bene t Finding Scale for Children (BFSC) was adapted by pediatric clinicians from scales used among adult patients with cancer (Phipps et al., 2007). Conducting a principal component analysis (PCA), the authors identi ed a single component, which accounted for 41% of the variance, and showed that the BFSC had an adequate internal consistency. Further studies on children and adolescents with cancer supported the reliability and construct validity of the BFSC (e.g., Maurice-Stam et al., 2011).
However, it is crucial to ensure the measure provides appropriate psychometric properties, when it is introduced to new populations (Loehlin, 2004), namely youth facing different CC. To the best of our knowledge, this is the rst study validating a BF measure for a transdiagnostic sample of youth facing different CC, simultaneously providing the rst age-appropriate, German version. The study aimed at examining the factor structure of the BFSC, using both exploratory factor analysis (EFA) and con rmatory factor analysis (CFA). Moreover, we examined the scale's construct validity by focusing on associations with positive intra-and interpersonal resources and coping strategies. Convergent constructs were selected based on previous reported correlates of BF, such as optimism, self-esteem, self-e cacy, empathy, acceptance, social support, and support seeking (Cassidy et al., 2014;Helgeson et al., 2006;Phipps et al., 2007). Discriminant constructs were chosen based on theoretical considerations. We hypothesized BF to be unrelated to measures of negative emotional reactions and passive coping strategies (Compas et al., 2012), such as cognitive avoidance, wishful thinking, and distancing oneself from the CC. Finally, we tested the BFSC against a measure of health-related quality of life (hrQoL), as an independent criterion (concurrent validity).

Translation process
The translation process followed the WHO guidelines (World Health Organization, 2018). With authorization of the authors of the BFSC (Phipps et al., 2007), two psychologists independently translated the BFSC into German. In an expert panel, discrepancies between both versions were discussed and a pre-nal version provided. This version was then back translated by a bilingual person. Finally, a pilot group of youth (N = 5) proved the items for understanding.

Procedure
Data were collected between June 2018 and August 2019 by an online questionnaire. The sample was recruited via social networks (e.g., Facebook), via various self-help forums, rehabilitation facilities, and outpatient clinics in Germany. Inclusion criteria were as follows: participants' age between 12 and 21 years, informed consent, the presence of CC con rmed by the Children with Special Health Care Needs Screener (Bethell et al., 2002), and the completion of the entire questionnaire. Participants received gift coupons (10 Euros) as incentives.

Measures
Bene t nding BF was assessed with the German translation of the BFSC (Phipps et al., 2007). Responses were recorded on a 5-point Likert scale ranging from "not at all true for me" to "very true for me".

Social support
The Berlin Social Support Scales (BSSS; Schulz & Schwarzer, 2003) were used to assess perceived social support and support seeking on a 4-point Likert scale ranging from "strongly agree" to "strongly disagree". The internal consistencies in the present (original) study were α = .93 (α = .85) for perceived support and α = .87 (α = .81) for support seeking.
Health-related quality of life The 12-item short form for the DISABKIDS chronic generic module (DCGM-12) was applied to assess general subjective hrQoL in children and adolescents with CC (Schmidt et al., 2006). The items cover mental, social, and physical hrQoL. Responses were recorded on a 5-point Likert scale ranging from "never" to "always". As two items are referring to pharmacological treatment and as some participants (18.1%) in our sample had no prescribed medication, we calculated total scores for a 10-item version, too.

Disease history
In addition, subjective disease severity and the age at diagnosis were assessed with single items ("I perceive my illness as severe" -5-point Likert scale ranging from "not at all true for me" to "very true for me"; "How old were you when your illness was diagnosed by a doctor?").

Data analysis
The main analyses were conducted using R (R Core Team, 2019). A two-step analytic procedure, consisting of an EFA followed by a CFA, was performed to test the factor structure (Worthington & Whittaker, 2006). First, the total dataset was split into subsamples for EFA (n = 100) and CFA (n = 204) via random sampling in IBM SPSS version 27.0. The respective sample sizes ful lled the subject to item ratio of 10:1 and were therefore considered to be su cient, given the level of the reported factor loadings < .50 (Tabachnick & Fidell, 2013) and recommendations from simulations studies (e.g., Mundfrom et al., 2005). The factor structure of the BFSC was assessed in the rst subsample (n = 100) using Ordinary Least Squared extraction (OLS). OLS is known to provide results similar to Maximum Likelihood (ML) and is considered as more robust to non-normal distributed data (Osborne, 2014). A quartimax rotation was used, as we expected a single, orthogonal factor (Osborne, 2014). Factor loadings were interpreted as follows (Tabachnick & Fidell, 2013): .71 and above excellent, .63 -.70 very good, .55 -.62 good, .33 -.45 fair, and .32 or lower poor.
Data from the second subsample (n = 204) were subjected to CFA using lavaan (Rosseel, 2012). Hypothesized modelling was based on the results of the EFA in the rst subsample, as well as the expected one-dimensional factor structure. The CFA was performed with ML estimation with robust (Huber-White) standard errors and a scaled test statistic that is (asymptotically) equal to the Yuan-Bentler test statistic (Muthén & Muthén, 2017). Because χ 2 test is sensitive to sample sizes, three indices were used to assess the model t. An acceptable model t was indicated by using the cut-off values of these indices: comparative t index (CFI) of ≥ 0.90, root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR) of ≤ 0.08 were considered as acceptable (Hu & Bentler, 1999).
As a measure of internal consistency, Cronbach's α was calculated. In both subsamples, we examined sex differences and correlations with age, social status, disease severity, and time since diagnosis (age minus age at diagnosis). Effect sizes were calculated and interpreted by applying Cohen's guidelines (Cohen, 1977): d = 0.20 -0.50 small, d = 0.50 -0.80 medium, d ≥ 0.80 large effect sizes. Convergent as well as discriminant validity was examined via Pearson correlations with respective variables (r >.10 small, r >.30 medium, r >.50 large effect size; Cohen, 1992). As the level of missing data in the EFA subsample was very low (<1%), missing data were imputed using multiple imputation via fully conditional speci cation implemented by the MICE algorithm (van Buuren & Groothuis-Oudshoorn, 2011). Multiple imputation is a robust missing data handling procedure that requires the least stringent assumptions about missing data mechanism compared to other traditional data handling methods (Enders, 2010).

Acceptance of BFSC
Nearly all participants (98.7%) missed no items on the BFSC. In total, the BFSC showed 0.2% missing data points, indicating a very low level of missing data. Little's Test was not signi cant, χ2(33) = 13.72, p = .99, suggesting that the missing data pattern was Missing Completely at Random.

EFA
Means and standard deviations for all BFSC items are presented in Table 1. The data was suitable for EFA based on item distribution, average correlation to other items, and item-total correlation (Clark & Watson, 1995). Bartlett's test of sphericity indicated correlation adequacy, χ 2 (45) = 391.85, p < .001, and the Kaiser-Meyer-Olkin (KMO) measure indicated sampling adequacy, MSA = 0.87. The parallel analysis (Horn, 1965) as well as the scree plot examination (Cattell, 1966) recommended extraction of one factor. The results of the OLS EFA indicated that only a single factor should be extracted (λ = 4.29), explaining 42.9% of the total variance. As can be seen in Table 1, most items had good-to-excellent factor loadings except for item 4. Table 1 Items of the Bene t Finding Scale, descriptive statistics, and item-factor loadings in the rst sample (n = 100).

CFA
Based on EFA results, we examined the t of the hypothesized one-factor solution using CFA in the second subsample. The standardized estimates of factor loadings for the best-tting model were predominantly good-to-excellent (see Figure 1). Item 4 showed a fair factor loading.

Discussion
The purpose of this study was to provide a German version of the BFSC ( Our results are consistent with this literature: Using EFA, we found that all ten items of the German BFSC loaded onto the same latent dimension. Furthermore, using CFA in a second subsample, we were able to con rm that this one-dimensional model had an adequate t following modi cation. Although the overall pattern of loadings was meaningful, item 4 showed only fair factor loadings, which, however, was in accordance with previous validation studies. To ensure comparability with the original study, we did not exclude this item from further analyses. In addition, the results of our study uphold the internal consistency and construct validity of the BFSC. The BFSC showed positive correlations with a wide range of convergent constructs, while there were no signi cant correlations with discriminant constructs, including avoidance, wishful thinking, distance, and emotional reaction. However, it should be acknowledged that the associations between BF and acceptance, social support and distance were not consistent across subsamples. Replicating the ndings of the original study (Phipps et al., 2007), the BFSC was not signi cantly related to hrQoL. This highlights the notion that positive experiences (e.g., "Having had my illness has helped me to deal better with my problems") do not simply imply an absence of negative experiences (e.g., "Does your condition get you down"), but that both represent rather independent and co-occurring dimensions. Future studies should consider alternative criterions for validation by including measures of positive well-being and satisfaction with life.
While previous studies reported no sex differences between females and males ( Overall, the present study had several strengths, namely the very good data quality, and the su cient sample size. Our study covered a broad age range and a wide range of underlying chronic diseases enhancing the generalizability of our results. It should be further stressed that a methodological sound approach with an EFA-to-CFA strategy was applied, thereby overcoming the limitations of previous studies using a PCA, which is inappropriate for the identi cation of latent constructs and factor structure of a set of variables (Widaman, 1993). By focusing on intra-and interpersonal resources and coping strategies, our study provides initial evidence for potentially relevant starting points for diagnostic comparisons as well as transdiagnostic programs promoting BF in youth with different CC.
Several limitations must be acknowledged, though. First, the recruitment strategy may have resulted in a selection bias towards generally lower levels of distress, as youth with higher levels of distress might be less likely to participate in online surveys. Second, the cross-sectional design of our study precluded the assessment of test-retest reliability or stability of BF over time. To further strengthen the psychometric basis for the BFSC, studies with adequately-sized samples are needed to verify whether BFSC scores are invariant across group membership (e.g., sex group and diagnostic group) and measurement occasion (Putnick & Bornstein, 2016). Finally, future studies should examine whether bene t nding predicts positive adaptive outcomes, not only directly, but incrementally over and above established constructs, such as emotion regulation (e.g., positive reappraisal), to further ensure the validity of BF. Despite these limitations, the available evidence con rmed the one-dimensional factor structure of the BFSC also in German. This is important as it will facilitate comparison across cultures and diagnoses in future work. The BFSC is an economic, psychometric sound and transdiagnostic measure that accounts for positive life changes of youths' responses to CC. Its application in future research will help to get a more comprehensive picture of the psychosocial consequences of CC.

Declarations
Ethics approval and consent to participate.
The study was conducted in accordance with the principles of Good Clinical Practice, the Declaration of Helsinki (https://www.wma.net/wpcontent/uploads/2016/11/DoH-Oct2008.pdf), and current ethical standards. Informed consent was obtained from each participant. Depending on the age of the participant, informed consent from the legal representative or guardian was also required. The central Ethics Committee of Potsdam University approved the study (date 02/02/18, request number 52/2017).

Consent for publication.
Not applicable.
Availability of data and material.
Fully anonymized data will be available from the corresponding author on reasonable request.
Competing interests. Path diagram and estimates for the one-dimensional model of the Bene t Finding Scale. The large oval is the latent construct, with the rectangles representing measured variables, and the small arrow with numbers representing the residual variables (variances). The path factor loadings are standardized with signi cance levels were determined by critical ratios (all p < .001).