Validation and Psychometric Properties of the German Version of the Evidence Based Practice Attitudes Scale (EBPAS-36D)

Background: The implementation of evidence-based practice (EBP) in mental health care confers many benets to patients, and research into factors facilitating the implementation of EBP is needed. As an important factor affecting the implementation of EBP, service providers’ attitudes towards EBP emerged. The Evidence-Based Practice Attitude Scale (EBPAS-36) is an instrument with good psychometric characteristics that measures positive and ambivalent attitudes towards EBP. However, a German version is missing. The present study therefore aims to provide a validated German translation of the EBPAS-36. Methods: The scale was translated and back-translated as recommended by standard procedures. To evaluate the psychometric properties of the translated version, German psychotherapists (N = 599) were recruited to participate in an online survey. They provided demographic and professional information, completed the EBPAS-36, the Implementation Climate Scale (ICS) and the Intention Scale for Providers (ISP). Standard item and reliability analyses and exploratory (EFA) and conrmatory factor analyses (CFA) in two subsamples (random split) were conducted and correlations with demographic and professional characteristics calculated. Results: The item analyses showed a mean item diculty of p i = .64, a mean inter-item correlation of r = .18, and a mean item-total correlation of r itc = .40. The internal consistency was very good for the total score (α = .89) and ranged from adequate to very good for the subscales (.65 – .89), indicating high reliability. The original factor structure showed an acceptable model t (RMSEA = .064 (90% CI = .059–.068); SRMR = .0922; AIC = 1400.77), conrming the 12-factor structure of the EBPAS-36. However, a second-order factor structure derived by the EFA had an even better model t (RMSEA = .057 (90% CI = .052–.062); SRMR = .0822; AIC = 1274.56). When the EBPAS-36 was entered in a hierarchical regression model with the criterion Intention to use EBP, the EBPAS-36 contributed signicantly to the prediction (Change in R 2 = .28, p < .001) over and above gender, age and working in science. Conclusions: The present study conrms good psychometric properties and validity of a German version of the EBPAS-36 in a sample of psychotherapists. to participate in the free into rae for a voucher of 500 Euros for a wellness-weekend or a popular events and ticket agent. of the potential full study information and required to provide consent before they were able to proceed with the survey. the last survey, asked whether they wanted to receive information about the study results or to take part in the rae. If they chose one or both of options, they linked to a separate page where they could register their e-mail addresses independently of their survey

Accordingly, high educational support and supervision have been linked to more positive attitudes towards the adoption of EBP (40,29). The association between the organizational climate for the implementation of EBP and providers' attitudes towards EBP was partially mediated by the organizational support for EBP use (22).
The Evidence-Based Practice Attitude Scale (EBPAS) (20) is an instrument that has been identi ed as a psychometrically strong measure assessing positive as well as ambivalent attitudes towards EBP (20,30,18). In line with suggestions put forward in the literature (41,18), it is based on mental health dissemination and implementation theories and has been developed in collaboration with service providers and researchers (20,21,42). The original 15-item version showed strong psychometric properties and validity in various settings and samples from the US and several European countries (30,34,36,33). In an effort to incorporate additional relevant dimensions that add to the understanding of provider attitudes towards adopting EBP and might increase the utility of the scale, it was expanded by 35 items (EBPAS-50) (21). However, the resulting 50-item version was considered too long, causing irritation by respondents, risking response biases, missing answers and potential for reduced validity (43). Thus, a shorter and more user-friendly 36-item version was developed, retaining the 12 domains of the longer version (43). On the one hand, the domains assess positive attitudes towards EBP: the intuitive Appeal of EBP, the willingness to adopt EBP given the Requirements to do so, providers' Openness to new practices and manualized interventions, the Fit of EBP with values and needs of providers and clients, and providers' perceptions of an increased Job Security provided by learning EBP, of the Organizational Support for learning EBP and of receiving Feedback (21). On the other hand, the domains assess ambivalent attitudes towards EBP: the Divergence between research-based interventions and current practice, the Limitations of EBP due to not addressing client needs, negative perceptions of Monitoring by supervisors, the perceived Balance of clinical skills and science in therapy, and the Burden of learning EBP (21). This scale has shown good psychometric properties and cross-cultural validity in US and Norwegian samples (43).
As argued by Kien et al. (44), German instruments assessing implementation science constructs are scarce and psychometric properties rarely reported. This stands in contrast with the growing relevance and increasing efforts of implementation research in German-speaking countries (45)(46)(47). Over the last decades, psychotherapy in Germany has experienced a signi cant professionalization and focus on EBP. Since 1999, a law ('Psychotherapeutengesetz') regulates the practice of psychotherapy, de ning that only state approved practitioners may offer treatment (48). The license can either be granted to medical doctors with a psychiatric or psychosomatic specialization or to psychologists (and to people with a degree in social work and education science for child and adolescent psychotherapy) with a training as psychotherapists (49,50,48). Training for the license is provided at state recognized institutions in one of the therapy orientations that are scienti cally accepted, i.e. CBT, psychodynamic psychotherapy, psychoanalytic therapy and systemic therapy. In September 2019, the German parliament approved an adapted law that aims to further align the postgraduate training for psychotherapy to the structure of medical education (49). Learning about psychotherapists' attitudes toward EBP in the German health care may inform about their readiness to implement EBP and identify strategies that need to be targeted in implementation efforts. In light of the new 2019 law, such identi cation is timely, as strategies and implementation measures could then directly be incorporated in the new psychotherapy master program. To the best of our knowledge, two independent German translations of EBPAS-15 exist (51,52), but no translation of the EBPAS-36 is available. Therefore, the present study aims to present a German translation of the EBPAS-36 and evaluate its psychometric properties.

Ethics
The online survey study was approved by the Internal Review Board of the University of Marburg (approval number: 2019-58k). Participants received study information and provided informed consent before they were able to access the survey. Data were collected anonymously. All raw data were stored securely at the Department of Clinical Child and Adolescent Psychology at Philipps University in Marburg, Germany.

Participants
Eligible participants were licensed psychotherapists and psychiatrists for adults, children and adolescents as well as psychotherapists and psychiatrists enrolled in postgraduate training to obtain such a license. No exclusion criteria were applied. The link to the online survey was clicked 2.417 times. Overall, 913 participants continued after informed consent. Of these, 863 met the inclusion criteria (i.e., profession). A total of 261 participants were excluded due to dropout before completion of the EBPAS-36, two due to implausible answers (for example being 99 years old), one due to conspicuous response patterns in EBPAS-36 (e.g., straight-lining despite reverse coded items). Of the remaining 599 participants, 502 were female (83.8 %) and their age ranged from 23 to 82 years (M = 36.62, SD = 11.26). Roughly half of the sample (56.26 %) stated being in postgraduate training to become psychotherapists or psychiatrists. 42.7 % of the German psychotherapists have ever or are currently working in science. Further information on profession are presented in Table 1. Recruitment and data collection. All data were collected online via an openly accessible online survey, using the scienti c survey platform SoSci Survey (www.soscisurvey.de). The survey was online and accessible from 14th November 2019 to 27th April 2020. The link to the online survey was widely distributed via e-mail lists of professional psychotherapy organizations that all licensed psychotherapists are members of, universities, training institutes, and psychiatric in-and outpatient institutions as well as Facebook groups of psychotherapists and psychiatrists. As an incentive to participate in the study, participants were free to enter into a ra e for a voucher of 500 Euros for a wellness-weekend or a popular events and ticket agent. On the rst page of the survey, potential participants received full study information and were required to provide consent before they were able to proceed with the survey. On the last page of the survey, participants were asked whether they wanted to receive information about the study results or to take part in the ra e. If they chose one or both of these options, they were linked to a separate page where they could register their e-mail addresses independently of their survey answers.

Measures
Demographics and information on training and profession. Participants were asked for their age, gender and nationality. Additionally, they completed questions concerning their education and occupation, e.g. their university degree, license status, therapy orientation, and current occupation. Implementation Climate Scale (ICS). The ICS is an 18-item instrument measuring the implementation climate in organizations and work groups (53). The original English version was translated into German by the rst author (KS) and back-translated by the bilingual English-speaking senior author (AB). In order to adapt the scale for psychotherapists in private practice, a parallel version was constructed that captures the implementation climate in the health system.
Respondents are asked to rate their agreement with statements describing how the respondents perceive the climate in the institution they work at with regard to the implementation of evidence-based interventions. A 5-point Likert scale ranging from 0 ('not at all') to 4 ('to a very great extent') was used. Six subscales can be calculated: Focus on EBP, Educational Support for EBP, Recognition for EBP, Rewards for EBP, Selection for EBP, and Selection for Openness towards EBP. Means of the subscales are computed to create a total scale. In the present study, the internal consistency for the ICS regarding organizations was Cronbach's α = .91 for the total scale and between α = .77 (Selection for Openness) and α = .92 (Focus on EBP) for the subscales. For the ICS with respect to the health system, Cronbach's α was = .90 for the total scale and between α = .77 (Rewards for EBP) and α = .88 (Educational Support for EBP) for the subscales. (A separate manuscript for this measure is in preparation).
Intention Scale for Providers (ISP). The ISP is a 70-item instrument assessing individual behavioral intentions for EBP use (54)  Global assessments. As a subjective self-assessment measure, participants were asked to rate their interest in EBP on visual analogue scales for nine questions (see supplemental material 3). A total score was computed (α = .84). At the end of the survey, participants were asked to rate the honesty of their responses ('How honestly did you answer the questions of this study?') and their self-reported tendency towards social desirability when answering the survey ('Did social desirability play a role in the survey?') on visual analogue scales.

Statistical analysis
All statistical analyses were performed using IBM SPSS 26 for Windows (Chicago, IL, USA). For the con rmatory factor analysis (CFA), SPSS AMOS version 26.0.0 was used. P values < .05 were set as thresholds for statistical signi cance in all analyses. For the EBPAS-36D, means were computed if there was a maximum of one missing item per scale. Otherwise, respondents would have been excluded from analyses. For item analyses, item di culties, corrected itemwhole correlations and Cronbach's alpha if item is deleted were calculated. To obtain internal reliability coe cients of the scales and subscales, Cronbach's alpha was calculated. Values above .70 are regarded as acceptable, higher than .80 as good, higher than .90 as excellent. To assess the factorial validity of EBPAS-36D, the total sample was randomly divided into two samples: With the rst one, we conducted an exploratory factor analysis (EFA), followed by a con rmatory one (CFA) with the other half. Differences between both samples regarding age, gender distribution and the EBPAS-36D total scale and subscales were examined with independent t-tests. The suitability of data for EFA was assessed with the Kaiser-Meyer-Oklin (KMO) sample adequacy measure (56,57) and Bartlett's test (58). To determine the number of components for the EFA, Horn's parallel analysis and Velicer's MAP test were conducted using the SPSS programs available online (https://people.ok.ubc.ca/brioconn/nfactors/nfactors.html) and the results compared (59). Since parallel analysis of principal factor analysis tends to over-extract factors (60), parallel analysis of principal component analysis was conducted with raw data permutation and 1000 datasets. The EFA was conducted using principal axis factoring analysis with promax correlated factors rotation method. Subsequently, a CFA was conducted to test and compare the original 12-factor structure of EBPAS-36 against a second-order factor structure derived by the EFA, merging the EFA components 4 (Constraints by the institution), 5 (Monitoring) and 6 (Burden) into one second-order factor, and another second-order factor solution that was proposed by Rye  (61), skewness < 2, excess < 7), an increased χ 2 value was expected and the Bollen-Stine bootstrap procedure (1000 samples) was performed. The chi-square test statistic was conducted as model t index. Since this statistic is dependent on sample size and might lead to rejection of plausible models with increasing sample sizes (62), the χ 2 /df ratio was computed. Smaller values indicate better model t, with values of 3 being regarded as acceptable and vales of 2 or smaller being regarded as good (62). Additionally, the root mean square error of approximation (RMSEA), the standardized root mean squared residual (SRMR), the comparative t index (CFI) and the parsimony-adjusted comparative t index (PCFI) were reported as t indices. RMSEA values less than .06, SRMR values less than .08, and CFI values greater than .95 indicate good model ts (63). The Akaike Information Criterion (AIC) was computed to compare the models. Smaller scores indicate a better model t. To assess the convergent validity of EBPAS-36D, Pearson correlation coe cients were calculated between the EBPAS-36D total scale and the direct and indirect scales of attitudes (A-D, A-ID) of the ISP as well as the interest in EBP score. According to Cohen (64), r = .50 indicates high correlations, r = .30 medium correlations, r = .10 low correlations. To test whether the EBPAS-36D total scale is an incremental predictor of the direct scale of behavioral intention to use EBP of the ISP (BI-D), a hierarchical linear regression analysis with the method ENTER was conducted. Gender and age (block 1), having ever worked in science (block 2), and the EBPAS-36D total scale (block 3) were successively included in the regression model to assess incremental improvements of model t. Mean differences across gender and professional groups on the EBPAS-36D were assessed with independent t-tests. Pearson coe cients were calculated to assess correlations between age as well as demographic/professional variables and the EBPAS-36D. Lastly, Pearson correlations between ICS and EBPAS-36D were assessed.

Item analysis
Item di culties of EBPAS-36D ranged between p i = .21 (item 28) and p i = .93 (item 13) with a mean di culty of p i = .64. The mean inter-item correlation was r = .18. The item-total correlations of the individual items with the total scale ranged from r itc = .07 (item 26) to r itc = .62 (item 2) with a mean item-total correlation of r itc = .40. Eight items showed item-total correlations under .30 (see Table 2). Considering the subscales, the correlations of the individual items with their subscales ranged from r itc = .44 (item 5) to r itc = .87 (item 9 and 29).

Reliability
The

Subscale Correlations
The correlation coe cients between the EBPAS-36D total scale and the 12 subscales are presented in Table 3. The highest correlation was between the total scale and the Openness subscale (r = .689). On subscale level, high correlations were between the Appeal and Fit subscales (r = .609), the Divergence and Limitations subscales (r = .550), the Openness and Divergence subscales (r = − .531), the Appeal and Openness subscales (r = .514) and the Job Security and Organization Support subscales (r = .547).   Table 4 shows the rotated factor matrix. Eleven items loaded on factor 1 and explained 24.33% of the variance (factor loadings from .357 to .844). Ten items loaded on factor 2 and explained 9.   Figs. 1, 2 and 3, respectively. Although model t of the original factor structure was adequate, both second-order models showed even better model ts (see Table 5). For all three models, all regression weights were signi cant.  Regression analysis. The EBPAS-36D total scale was included in a hierarchical regression model to predict the Behavioral Intention Scale of the ISP as a third block, subsequent to the predictors gender and age (block 1) and ever or currently working in science (block 2). The inclusion improved the model t (Change in R 2 = .28, F = 267.32, p < .001) and the signi cant regression coe cient of the EBPAS-36D total scale (ß = 2.13; t = 16.35; p < .001) indicated incremental prediction beyond the previous predictors (see Tables 6 and 7).  Notes. n = 543. Dependent variable: Behavioral Intention Scale of Intention Scale for Providers.
The EBPAS-36D total scale was associated with the total score of the ICS on organizations and work groups (r = .432, p < .001, n = 408), but not with the total scale of the ICS on the health system (r = .138, p = .080) that was exclusively completed by licensed psychotherapists and psychiatrists working in private practices (n = 161). The self-rated honesty when answering the survey showed a small correlation with the EBPAS-36D total scale (r = .146, p = .001, n = 556).

Discussion
The present study is the rst to present a German version of the EBPAS-36 and investigates its validity and psychometric properties. In a sample of psychotherapists and psychiatrists, the original factor structure was con rmed and the EBPAS-36D demonstrated good item properties, internal consistency and convergent validity.
Regarding the psychometric properties of the individual items, most item di culties were in the medium range. In the context of attitude measurements, a high item di culty translates into low endorsement of the item. Medium di culty is desirable as it is optimal to differentiate between respondents with different attitudes. Items of the subscales Fit, Feedback and Appeal received high approval (subscale means > 3.2), whereas items of the subscales Burden, Job Security and Divergence were less strongly endorsed (subscale means < 1. The internal consistency of the EBPAS-36D total scale obtained is good and comparable to those found for the US and Norwegian versions. The internal consistencies of the subscales ranged from acceptable to good, with the subscales Appeal, Divergence, Fit and Balance demonstrating the lowest internal consistencies, as seen in the Norwegian sample (43). In consideration of the extreme brevity of the subscales (3 items), the overall reliability of the EBPAS-36D subscales can be considered high.
The CFA con rmed the 12-factor structure of the original EBPAS-36 by demonstrating adequate model t. Nonetheless, two second-order factor structures, one derived by EFA in the present sample and one that was proposed by Rye et al. (37), showed even better model ts. Therefore, second-order models might map the actual underlying construct of attitudes towards EBP even better than the original factor structure. The four second-order constructs we found are: Positive alignment with EBP, consisting of the Openness, Appeal, Fit and Feedback subscales, Reservations towards EBP, consisting of the Divergence, Limitations and Balance subscales, Institutional Endorsement, consisting of the Job Security and Organizational Support subscales and Constraints by the institution, consisting of the Requirements, Monitoring and Burden subscales.
As expected, the EBPAS-36D showed high correlations with two other scales assessing attitudes towards the use of EBP, the direct and indirect measurement scales of attitudes of the ISP. This con rms the convergent validity of the scale. Accordingly, a high correlation was found between the EBPAS-36D and a global assessment of participant's interest in EBP, a nine-item subjective self-rating of one's interest in research on psychotherapy and clinical psychology, clinical guidelines, EBP and treatment manuals with high face validity. Moreover, the EBPAS-36D allowed incremental prediction of the intention to use EBP as assessed with the Behavioral Intention Scale of the ISP beyond gender, age and working in science.
Responders of the 50-item version of the scale commented on being annoyed and fatigued by answering the items (43). This might result in response biases and missing answers, limiting the validity of the scale. The acceptability of the shorter EBPAS-36D appears to be higher as indicated by a substantially lower amount of missing answers in the current study.
In the present sample of German psychotherapists, higher age was associated with less favorable attitudes towards EBP. Consistent with this, licensed psychotherapists reported more negative attitudes compared to psychotherapists in training. While this result is in contrast to two previous studies reporting higher scores on the Requirements and Openness subscales of the EBPAS with increasing age (30,31), it is in accordance with other studies (34)(35)(36)(37). As assumed by van Sonsbeek et al. (33), professionals may rate their own clinical experience higher than EBP with increasing age and experience. However, it should be noted that as yet, age effects may be confounded by cohort effects, since psychotherapy training underwent considerable changes in Germany over the last 30 years. Concerning sex differences, women reported more positive attitudes towards EBP in the present study. Sex differences were apparent for the total scale and the subscales Requirements, Appeal, Fit and Organizational Support. This result adds to other research demonstrating more positive attitudes towards EBP in women (32,30,34,37,33). Still, these sex differences were not found consistently (20,35,36), which is why possible moderators should be investigated in future research.
As demonstrated in previous studies (31,21), provider's attitudes towards EBP were associated with organizational climate for the implementation of EBP.
Since psychotherapists working in private practices were unable to rate the implementation climate of any organization or work group, those participants were asked to rate a parallel version of the ICS, capturing the implementation climate in the German health system. Interestingly, individuals' attitudes towards EBP were not associated with their evaluation of the implementation climate of the health system. One reason for this might be that the German health system does not provide support for EBP to a similar extent as some organizations do, thus not leading to more positive attitudes towards EBP (22). Another reason might be that psychotherapists with a positive attitude towards EBP are able to choose organizations with a better implementation climate for EBPs or affect the implementation climate of the organization they work at, while they are probably unable to choose a health system according to its implementation climate or affect the implementation climate of the health system.
When interpreting the results of the present study, some limitations must be borne in mind. All data are based on self-reports in a cross-sectional online survey.
A large proportion of the sample reported to have ever or currently been working in science and having a cognitive behavioral therapy approach. The ndings refer to a convenience sample that is most likely self-selected for interest in EBP and not representative (65) of the population of mental health providers. This does not affect the evaluation of the psychometric properties of the EPBAS-36D; still future studies with representative samples should seek to con rm the results. Although the dropout rate in the present study can be considered as average for online surveys (66), a selection bias cannot be precluded with only particularly interested participants completing the survey (67). Finally, although a de nition of EBP preceded the questionnaire, some comments indicated that individual participants may have been uncertain about the exact meaning of EBP. Future research should assess the comprehensibility of the de nition to ensure that all participants rate the same construct.

Conclusions
In order to be successful, implementation efforts need to consider providers' attitudes towards EBP. Thus, reliable instruments assessing positive as well as ambivalent attitudes are crucial. Although further validating research is required, the present study con rms good psychometric properties and validity of a German version of the EBPAS-36 in a sample of psychotherapists. Additionally, a second-order model of attitudes towards EBP is proposed. Availability of data and materials The datasets analyzed in the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by funds from BMBF (Bundesministerium für Bildung und Forschung, German Federal Ministry of Education and Research, 01KR1804B). The BMBF had no role in the design and conduct of the study, including the collection, management, analysis, interpretation of the data, or the preparation, review, or approval of the manuscript and decision to submit the manuscript for publication.
Authors' contributions KS, AB and HC designed the study. KS, JT and AB translated the instruments. MR and GAA provided feedback on the translations. KS collected the data. KS and AB analyzed the data and drafted the manuscript. JT provided a rst draft of the introduction. HC, MR and GAA commented on the initial draft and critically revised it. All authors have read and approved the nal version of the manuscript.