Psychometric properties of the rehabilitation treatment beliefs questionnaire for psychosomatic rehabilitation

Abstract Purpose Treatment beliefs play a crucial role for patient satisfaction and the treatment outcome in psychosomatic rehabilitation. The aim of this study was the development and the psychometric evaluation of an indication-specific questionnaire of beliefs about psychosomatic rehabilitation. Materials and methods The study was conducted at a psychosomatic rehabilitation clinic. The primary item list comprised k = 125 items. After a descriptive item analysis, we conducted an exploratory factor analysis. Furthermore, we tested reliability via McDonald’s Omega and construct validity by analyzing correlations of the scales with related constructs. Results Of the N = 264 participants, 50% were female and the mean age was 50.4 (SD = 9.8) years. K = 85 items were suitable for factor analysis, which resulted in k = 30 items constituting six scales, explaining of 57% the overall variance. The corrected item-total correlations were between r = 0.48 and r = 0.83. Internal consistency ranged from ω = 0.81 to ω = 0.86. Conclusion The newly developed questionnaire assesses specific treatment beliefs about inpatient psychosomatic rehabilitation. The psychometric properties of the six scales are acceptable. Further studies should confirm the psychometric results, such as the factorial structure of the questionnaire. IMPLICATIONS FOR REHABILITATION Treatment beliefs are known to play an important role for the adherence, therapy outcome and satisfaction in psychosomatic rehabilitation. We developed and tested a indication-specific questionnaire assessing treatment beliefs in psychosomatic rehabilitation. The questionnaire can be used to explore patient’s rehabilitation-related treatment beliefs, predict treatment outcomes, and to develop interventions attempting to modify these.


Introduction
Mental disorders show a prevalence of 25-30% in Europe [1].Among the most common mental health problems are depressive disorders and anxiety disorders, as well as somatoform and substance dependence disorders [1].Burden associated with mental health disorders is associated with reduced quality of life, unemployment, early retirement, and disability pension [2-5].In the last decade, mental disorders have become the most important reason for early retirement in Germany [6,7].
Psychosomatic inpatient rehabilitation is the second most important inpatient treatment sector for patients with mental disorders in Germany [8].Most commonly, patients with mood disorders, anxiety disorders, adjustment disorders, somatoform disorders, and personality disorders will seek treatment in psychosomatic inpatient rehabilitation over an average course of 38 days [9,10].While acute psychosis or acute suicidality is defined as a contraindication for psychosomatic rehabilitation [11], rehabilitants are treated on a chronic course for their disorders, with an average course of 5-7 years [12].Furthermore, most rehabilitants experience an occupational disability of at least 6 weeks within the last 12 months [13].Therefore, psychosomatic rehabilitation focuses on recovery in terms of functioning and participation, as well as the long-term ability to work.The treatment is based on the principles described in the World Health Organizations' framework entitled the International Classification of Functioning, Disability and Health (ICF), and uses a biopsychosocial, multi-professional, and multidimensional approach [9].It includes medical, psychological, physiotherapeutic, and occupational therapy and aims to obtain or restore functioning and participation [9].
Inpatient psychosomatic rehabilitation in Germany is by the majority of the patients significantly effective, [14], but by up to 30% no response [15].Research on predictors of outcome-related variables is therefore crucial to enhance the effectiveness of rehabilitation.So far, different patient-related, illness-related, and treatment-related factors have been identified as predictors of patient satisfaction, adherence, and treatment outcomes, both in primary healthcare and in the context of rehabilitation.Furthermore, treatment beliefs are of importance, which have been found to be associated with outcomes of pharmacological treatment for different somatic illnesses [16,17], with outcomes of cardiac rehabilitation [18], with adherence to complementary medicine [19], and with patient satisfaction in the context of rehabilitation [20][21][22][23].Treatment beliefs comprise beliefs about the effects of a treatment, as well as side effects, its duration, accessibility, and manageability [24][25][26].Treatment beliefs are linked closely to the construct of treatment-related expectations, which are described as an inherent part of beliefs [26].According to the integrative model of expectations in patients undergoing medical treatment [27], treatment-related expectations of patients differentiate between outcome expectations, process expectations, and structure expectations.Outcome expectations include both benefit expectations and harm expectations.
Many questionnaires assessing treatment beliefs are oriented on the theoretical Necessity-Concerns Framework [16,17], which includes beliefs about concerns and beliefs about necessity, with the Beliefs about Medicines Questionnaire [16] being a wellknown questionnaire utilizing this framework.In addition to concerns and necessity, some instruments, such as the Treatment Beliefs Questionnaire [18], the Treatment Representation Inventory [28], and the Lower Back Pain Treatment Beliefs Questionnaire [29], operationalize treatment beliefs via the perceived suitability of the treatment and practical barriers [18], satisfaction with the decision for the type of treatment [28], accordance between patients' expectations and the actual treatment, and the aspect of how convincing patients find their treatment [29].Such questionnaires were designed for specific indications such as cardiac rehabilitation [18], coronary artery disease [28], and low back pain [29].Two generic questionnaires, which are based on the integrative model of expectations in patients undergoing medical treatment [27], include previous treatment experiences and current treatment effects [30], and additionally differ between probabilistic expectations and value-based expectations, e.g., desires vs. fears [31].Generic treatment belief questionnaires that were developed for the field of rehabilitation in Germany often include outcome expectations and process expectations [32][33][34].
However, despite the great relevance of psychosomatic rehabilitation for the treatment of mental disorders and despite the relevance of treatment beliefs for various treatment outcomes, there is no indication-specific questionnaire about treatment beliefs for use in inpatient psychosomatic rehabilitation yet.Therefore, the aim of our study was the development and the psychometric evaluation of an indication-specific questionnaire of beliefs about psychosomatic rehabilitation.

Transparency and openness
The name of the study is: "Patient satisfaction and rehabilitation outcome in psychosomatic rehabilitation.

Development of item list
We generated an item list based on the items of the Beliefs about Rehabilitation questionnaire [BRQ; 33], based on the literature, and on the results of N ¼ 10 semi-structured qualitative interviews with rehabilitation patients.The interviews were conducted by telephone 2 weeks before the start of rehabilitation to explore treatment beliefs about psychosomatic rehabilitation [36].Prospective rehabilitants were asked about their expectations about the planned psychosomatic rehabilitation in regards to the expected results, clinic structure, therapy processes, and concerns.A qualitative content analysis approach based on Mayring [37] revealed the following nine themes of treatment beliefs regarding psychosomatic rehabilitation: reasons for rehabilitation, conditions within the clinic, rehabilitation planning, organization of the rehabilitation, content of the rehabilitation, results of the rehabilitation, concerns, expectations of one's own behavior, and contact to other patients.Based on these themes, and taking into account the current state of research, we generated an initial item list consisting of k ¼ 125 items, which covered outcome expectations (including psychosocial and occupational aspects), process expectations, structural expectations, and expectations of one's own engagement in therapy.We also included items on concerns and perceived necessity.Furthermore, we added items on current emotions regarding rehabilitation and expected emotions after completion of the rehabilitation.
Items were formulated beginning with "I expect that … " for expectations (including expected emotions) and "I am concerned that … " for concerns.Necessity items were formulated as statements, for example, "Without rehabilitation, my illness will become worse."The response scale ranges from 0 ("not true at all") to 4 ("completely true").

Recruitment of participants
Study participants were recruited consecutively between April 2019 and January 2020 in a psychosomatic rehabilitation clinic.An on-site study coordinator sent information about the study, a consent form, and the questionnaire to eligible patients about 2 to 3 weeks before their rehabilitation.All rehabilitants aged 18 years or older were asked to participate.No exclusion criteria were defined.Of the N ¼ 832 patients contacted during the course of the study, N ¼ 268 (32%) patients agreed to participate.Reasons for non-attendance remained unknown.The participants, who agreed to participate, filled in the consent form and the questionnaire at home.N ¼ 264 patients completed the questionnaire.Upon patient admission to the clinic, the on-site study coordinator received both the consent form and the questionnaire in a separate, closed envelope to ensure pseudonymization.Sample size calculations were based on the main research question within the project (prediction of treatment outcome), and no separate calculations were made for the psychometric testing of the questionnaire.

Additional measures
In addition to the item pool assessing beliefs about psychosomatic rehabilitation with k ¼ 125 items, we assessed sociodemographic characteristics (age, sex, marital status, education, employment status, migration background, and social status) for the description of the sample.We did not assess race or ethnicity in this study.
To assess construct validity, we used the measures regarding illness beliefs, job-related treatment needs, and psychosocial health, and tested hypotheses about the relationship between the scales of our questionnaire and these measures.

The illness perception questionnaire revised [IPQ-R]; German version
The IPQ-R measures beliefs about illness using 64 items in nine dimensions (identity, timeline, timeline cyclical, consequences, personal control, treatment control, coherence, emotional representations, and causes) [38][39][40].The psychometric values of the German version are good [40].In our study, the internal consistency was acceptable to good, with Cronbach's Alpha ranging between a ¼ 0.69 and a ¼ 0.87.

Screening-Instrument Arbeit und Beruf, screening instrument work and occupation [SIBAR]
The SIBAR assesses job-related characteristics in the following three scales (I, risk of early retirement; II, job-related stress; III, jobrelated treatment needs) [41].The questionnaire contains 28 items.The psychometric and prognostic parameters of the SIBAR are good.Internal consistency is known for scale I. Scale II and scale III consist of only one item each.[41].In this analysis, we only use scale III for validation analysis

Hamburger module zur erfassung allgemeiner aspekte psychosozialer gesundheit f€ ur die therapeutische praxis; Hamburg modules for the assessment of psychosocial health in clinical practice [health-49]
With 49 items in 9 scales, the HEALTH-49 measures depressiveness, phobic anxiety, somatoform complaints, psychological well-being, interactional problems, self-efficacy, activity and participation, social support, and social stress [42].The three scales of depressiveness, phobic anxiety, and somatoform complaints can be combined into one overall score "psychological and somatoform complaints."In our study, the internal consistency measured by Cronbach's Alpha was acceptable to good, with a ¼ 0.77 to a ¼ 0.94.

Statistical analyses
In the context of psychometric testing, we first conducted a descriptive analysis, and then we assessed construct validity and the reliability of the questionnaire.

Descriptive analyses
We explored item characteristics by conducting a descriptive analysis including characteristic values of distribution (means, variances, and floor/ceiling effects), normality testing (Q-Q-Plot, skewness, and kurtosis), difficulty (difficulty index according to Dahl [43]), and item homogeneity (intercorrelations and squared multiple correlations).We examined the intercorrelation matrix to identify weak intercorrelations.Items that met the following criteria, were deleted: missing > 5%, items with severe non-normality (skewness > ±2, kurtosis > ±7 [44]).Some items were deleted due to theoretical considerations (see Results).In order to test for the suitability for factor analysis, we calculated the KMO score and Bartlett's test for homogeneity of variances.To inspect multivariate normality we calculated the Mardia coefficient for skew and kurtosis, both were significant (p < .05),so the assumption of multivariate normality cannot be upheld.

Factor analysis
Factor analysis was conducted using the R package psych [45] based on Pearson's correlations.We used the following criteria to determine the most reasonable number of factors: Scree-Test [46], Parallel-Analysis [47], and Velicer's minimum average partial (MAP) test [48].The factor analysis involved two steps.The first step included all the items that were left after descriptive selection.In this step, all the items with a minimum of one of the following criteria were excluded: factor loading < 0.32, communality < 0.30, at least one crossloading and item complexity > 2.
Step two of the factor analysis included all the items left from step one.In both steps, we used principal axis analysis with oblimin rotation because the factors were not considered independent, neither statistically nor theoretically.We considered a factor with at least three items as interpretable [49].

Reliability analysis
For each extracted factor, we calculated item-total correlations as an index for the discriminatory power of the items and determined reliability measures of McDonald's Omega and Cronbach's Alpha for each scale with the R package semTools [50].For the interpretation of both reliability measures we used the common cut offs for Cronbach's Alpha as recommended [51], with values higher than 0.70 being considered generally acceptable [52].

Calculation of scales
For each scale, the mean was calculated first, and then it was divided by the factor 4 and multiplied by 10 to generate scores ranging from 0 to 10.We calculated the correlations between the scales (Pearson's correlations) to determine the interdependency between the factors.

Construct validity
After completing the factor analysis and the calculation of scales, we formulated hypotheses regarding the expected relationships between scales of our questionnaire and the other constructs measured via the IPQ-R, SIBAR, and HEALTH-49.We calculated Pearson's correlations between scales of the new questionnaire and scales of the other measures.
We expected moderate to high correlations between the perceived necessity of rehabilitation and illness identity measured by the IPQ-R [38] and job-related treatment need measured by scale III of the SIBAR [41].Likewise, we expected moderate to strong correlations between rehabilitation-related concerns and illness consequences and emotional representation of the illness as measured by the IPQ-R [38], as well as psychological and somatoform complaints measured by the three scales of the HEALTH-49 [44].We assumed moderate correlations between rehabilitationrelated outcome expectations and perceived treatment control as measured by the IPQ-R and between expectations regarding rehabilitation-related expectations with respect to information and participation and perceived personal control, also as measured by the IPQ-R.

Sample characteristics
Of the N ¼ 264 participants who completed the questionnaire, 48.8% were female and the mean age was 50.4 (SD ¼ 9.8) years.Further sample characteristics are presented in Table 1.

Results of the factor analysis
After descriptive analysis, k ¼ 31 items from the initial k ¼ 125 items, were excluded from further analyses because of unacceptable characteristics (missing > 5%, skewness > ±2, kurtosis > ±7).Moreover, nine items assessing emotions regarding rehabilitation were deleted because of theoretical considerations.These items had been included in the item pool to represent an emotional factor, but did not contribute to discriminating between concerns and emotions.We deleted these items to ensure differentiation between the concepts.The remaining k ¼ 85 items were suitable for factor analysis; KMO ¼ 0.84, Bartlett's test for homogeneity of variances: Chi-square (4.186) ¼ 15.298,12, p¼.001 [54].Another k ¼ 55 items were discarded in the course of factor analysis.
Factor extraction methods differed in the optimal number of extracted factors: parallel analysis extracted six factors, MAP seven factors, VSS two factors and scree plot four factors.We calculated exploratory factor analyses for different numbers of factors and decided to use the six factor solution.The six factor, solution had a clean simple structure with a minimum of three items per factor and each of the factors were interpretable on a content level.The six factor solution included k ¼ 30 items, explaining 57% of the overall variance.The items showed communalities between h 2 ¼ 0.30 and h 2 ¼ 0.90 and corrected item-total-correlations between r ¼ 0.48 and r ¼ 0.83 (Table 2).
The factors were named "outcome expectations" (6 items), "process expectations: leisure time and social contacts" (4 items), "process expectations: participation and treatment structure" (4 items) and "process expectations: information" (5 items), "concerns" (8 items), and "necessity" (3 items).See Table 2 for an overview of the factors.We named the questionnaire Rehabilitation Treatment Beliefs Questionnaire -Psychosomatic (RTBQ-Psych).Descriptive statistics of the scales are displayed in Table 3, while the full questionnaire can be seen in Supplemental Material 1.

Characteristics of the scales and reliability
We calculated intercorrelations and McDonald's Omega for internal consistency.Intercorrelations of the scales ranged from r¼ À 0.05 to r ¼ 0.52 (Table 4).Internal consistency ranged from x h ¼ 0.81 ("necessity") to x h ¼ 0.86 ("outcome expectations" and "process exceptations: leisure time and social contacts").

Construct validity
We calculated correlations between the scales of the RTBQ-Psych and the scales of other questionnaires assessing related constructs.The results showed significant, moderate to high correlations between rehabilitation-related "concerns," HEALTH-49 "psychological and somatoform complaints," IPQ-R "consequences," and IPQ-R "emotional representation" (Table 5).We found weak, but significant correlations between "necessity," IPQ-R "identity," and SIBAR "job-related treatment need" (low values indicate higher need) in the expected direction, as well as between "outcome expectations" and IPQ-R "treatment control."Two hypotheses were not confirmed.The scales of "process expectations: participation and treatment structure" and "process expectations: information" were not associated with IPQ-R "personal control."

Discussion
The aim of this study was to develop and evaluate a questionnaire of beliefs about psychosomatic rehabilitation, the Rehabilitation Treatment Beliefs Questionnaire (RTBQ-Psych).We developed an item pool based on the results of qualitative interviews with patients about their rehabilitation-related treatment beliefs and based on theoretical considerations.Factor analysis revealed six factors with k ¼ 30 items in the final version representing different aspects of rehabilitation-related treatment beliefs.The scales describe "outcome expectations," "process expectations: participation and treatment structure," "process expectations: information," "process expectations: leisure time and social contacts," and "necessity" and "concerns" regarding inpatient psychosomatic rehabilitation.Internal consistency was good for all scales.The face validity of the scales and their corresponding items were good.We found mostly low to moderate intercorrelations between the scales, except for one strong intercorrelation between the scales of "process expectations: participation and treatment structure" and "process expectations: information."One possible explanation is that expectations of information are closely linked to participation.Patients who expect to engage more in the rehabilitation process might also expect sufficient information about rehabilitation.Although intercorrelation was high, there were no relevant cross-loadings, so we decided to keep the two separate scales in this version of the questionnaire.
Regarding construct validity, we found moderate to high correlations, which were mostly in accordance with our hypotheses regarding between the scales of the RTBQ-Psych and the related constructs considered.Only two out of eight expected associations could not be confirmed.We did not find associations between the perceived personal control of the illness, "process expectations: information," and "process expectations: participation and treatment structure."People who estimated their personal control of the illness to be high did not expect more participation in or more information about rehabilitation.Perceived personal control of the illness might not be directly linked with one's expectations regarding rehabilitation, as one's own engagement and motivation to actively participate and require information might rather emerge from individual characteristics and attitudes.Like questionnaires that are theoretically based on the Necessity-Concern Framework, our questionnaire also includes two scales regarding concerns and perceived necessity.While the concern scales of other questionnaires focus on physical sideeffects [16,18] or social aspects, such as negative consequences at home or at work [33], the concern scale of the RTBQ-Psych assesses psychological side effects, such as embarrassing oneself in front of others.It thereby covers specific concerns regarding treatment components in psychosomatic rehabilitation, for example the side-effects of group psychotherapy.
Regarding perceived necessity, other treatment beliefs questionnaires include statements that the treatment is greatly needed and an inevitable step to improve one's health [16,18,28] and are similar to our items.
According to the integrative model of expectations about undergoing medical treatment [27], treatment expectations comprise outcome expectations, process expectations, and structure expectations.In our questionnaire, one outcome expectations scale and three scales regarding process expectations are derived.The items of the "outcome expectations" scale are specific for psychosomatic rehabilitation, such as to learn to deal better with stress through rehabilitation.
In comparison to other treatment beliefs questionnaires that include more generic scales assessing process expectations [31,33,34], the three process expectations scales of the RTBQ-Psych assess expectations concerning specific processes in psychosomatic rehabilitation, such as a balanced treatment of body and soul, receiving understandable information from the therapists about one's symptoms, or exchanging ideas with other patients about one's illness or the psychosomatic treatment.
Taken together, with the RTBQ-Psych, it is possible to assess a broad range of rehabilitation-related treatment beliefs regarding outcomes, processes, perceived necessity and concerns specific to psychosomatic rehabilitation.With respect to clinical implications, the RTBQ-Psych could be useful to incorporate patient's rehabilitation-related treatment beliefs into interventions.Expectations are known to be a key factor in the development and persistence of mental disorders and thus, frameworks for expectation-focused psychotherapeutic interventions already exist, with different approaches that seek to enhance adherence, engagement and treatment outcome by addressing individual beliefs and expectations [55][56][57].For instance, specific interventions, such as providing individualized information on illness and treatment on the basis of beliefs and needs [58,59], showed positive effects on treatment outcomes in rehabilitation.Furthermore, McAndrew et al. [60] argued, that not only addressing maladaptive beliefs is important, but also working on a concordance between patients and providers beliefs and expectations.The development of the RTBQ-Psych is of special relevance, as treatment beliefs differ in respect to the treatment under consideration and no psychosomatic rehabilitation specific instrument to assess treatment beliefs existed before.It will not only give insight to clinicians into the domains of rehabilitation-related treatment beliefs, but clinicians will be able to address concrete maladaptive beliefs about the process and outcomes of the treatment as well as the perceived concerns and necessity.

Limitations of the study
There are some limitations that should be considered when interpreting our results.We did not evaluate the reproducibility by testing the questionnaire in another sample and further psychosomatic clinics, which should be addressed in following studies.Furthermore, there is no gold standard for the construct of treatment beliefs, and we were not able to evaluate criterion validity [61].
After descriptive analysis, an item pool of k ¼ 85 items was tested with N ¼ 264 patients, so the study sample size was relatively small considering recommendations that the sample size should be at least fivefold higher than the number of items.However, the ratio between number of items and sample size does not necessarily effect the final factor solution [62].Another criterion mentions, that in case of communalities around 0.50, the sample size should be between 100 and 200 [53].A sample size between 100 and 200 was reached, but communalities ranged from 0.30 to 0.90.Furthermore, a participation bias might have influenced the results, as only 32% of the contacted rehabilitation patients participated in the study.Finally, the study sample was derived from the population of only one inpatient psychosomatic clinic.However, the sample was representative for inpatient psychosomatic rehabilitation according to another large study in this field [63], with regard to diagnoses, distributions of sex, education levels, and marital status, except for age.Our study sample was somewhat older (M ¼ 50.37,SD ¼ 9.8 years) in comparison to de Vries et al. (2011; M ¼ 44.2, SD ¼ 8.9 years) and contained more male participants compared to a meta-analysis assessing the effectiveness of psychosomatic rehabilitation [14].At least the higher portion in male participants can be explained by the fact, that the German Pension Insurance Rhineland-Palatinate predominantly occupies the cooperating clinic with patients, where traditionally more men are insured compared to Federal German Pension Insurance.In conclusion, future studies should confirm the reliability and validity of the RTBQ-Psych, as well as the factorial structure of the questionnaire (e.g., by confirmatory factor analysis).Focus of these studies should be the avoidance of a possible participation bias, by conducting a multicenter study and try methods, such as incentives to increase participation rate.One could further argue, that the initial item pool with k ¼ 125 items was a barrier for participation.Thus, using the final RTBQ-Psych with k ¼ 30 items might increase participation rate itself.

Conclusion
The RTBQ-Psych is a questionnaire with 30 items, which assesses specific treatment beliefs of patients in inpatient psychosomatic rehabilitation.It was developed by including the results of qualitative interviews with patients.The psychometric results were acceptable, but should be further tested and validated.Yet, the RTBQ-Psych might be used to incorporate rehabilitation related treatment beliefs into the interventions within the psychosomatic rehabilitation.This might not only enhance the adherence and involvement of patients as well as treatment outcome, but would also add to the patient-centered care in psychosomatic rehabilitation.

Table 1 .
Sample characteristics of the N ¼ 264 participants.

Table 2 .
Factor loadings, communality and corrected item-total-correlation of the final version.

Table 3 .
Descriptive statistics of the RTBQ-Psych scales.

Table 4 .
Intercorrelations of the scales and reliability.

Table 5 .
Correlations of the scales with other instruments (construct validity).SIBAR Scale III: higher values indicate lower treatment need.� Statistically significant correlations, p < .05.