Rehabilitation use in multiple sclerosis: Do illness representations matter?

Abstract Objectives Multidisciplinary rehabilitation improves illness outcomes and is recommended in clinical guidelines for multiple sclerosis (MS). However, many people with MS do not make use of rehabilitation. We do not know much about the barriers to the use of rehabilitation in MS, but in other patient groups, illness representations have proven to be predictors of service utilization. Therefore, the aim of our study was to explore whether, in patients with MS, illness representations are associated with self‐reports of rehabilitation use in the past and the intention to use rehabilitation in the future, beyond sociodemographic and illness‐related factors. Materials and Methods Patients were recruited in a cross‐sectional nationwide online survey in Germany. Hierarchical binary logistic regression analysis was used to analyze whether illness representations are associated with the use of rehabilitation in the past and the intention to use rehabilitation in the future, over and above socio‐demographic and illness‐related variables. Results There were 590 patients, who had MS, participating in the study. Illness representations were correlated to both outcome variables beyond sociodemographic and illness‐related factors: The probabilities of having the intention to use rehabilitation and of making using of rehabilitation were higher in patients who believed that their MS was controllable by treatment and perceived that their MS would have severe consequences. Conclusions Our data suggest that addressing patients’ illness representations may facilitate the intention to use and the use of multimodal rehabilitation, contributing to better illness outcomes.


| INTRODUC TI ON
Multiple Sclerosis (MS) is an autoimmune neurological disease which affects approximately 2.3 million people worldwide (Multiple Sclerosis International Federation, 2013). At the time of diagnosis, 85% of patients have a relapsing-remitting form of MS, and within this subgroup of patients, around 80% will develop secondary progressive MS (Multiple Sclerosis International Federation, 2013). MS is characterized by a high inter-and intraindividual variability (Beer, Khan, & Kesselring, 2012), involving various combinations of physical, cognitive, psychosocial, behavioral, and environmental problems (Khan, Turner-Stokes, Ng, Kilpatrick, & Amatya, 2007). These include impairments related to strength, coordination and vision, spasticity, cognitive deficits, bowel and bladder problems, sexual dysfunctions, pain, and fatigue, which result in limiting activity, for example in mobility and self-care, and in restricting participation in society (Beer et al., 2012;Khan & Amatya, 2017;Khan et al., 2007;Multiple Sclerosis International Federation, 2013). In light of this complexity, there is-besides pharmacological treatment-a need for comprehensive multidisciplinary medical rehabilitation (Beer et al., 2012;Khan & Amatya, 2017;Rasova et al., 2010), which is also recommended in clinical guidelines for MS (DGN, 2014;NICE, 2014). Rehabilitation realizes patient centredness, uses a biopsychosocial approach, and is functionally oriented with the aim of maximizing activity and participation (Beer et al., 2012;Khan et al., 2007). To achieve these goals, comprehensive rehabilitation programs include a large body of functional interventions, especially physical treatment methods, occupational, speech, and swallowing therapy, as well as psychological interventions (Beer et al., 2012;Rasova et al., 2010). In Germany, multimodal, multiprofessional MS rehabilitation is mostly provided in an inpatient setting. It lasts from 3 to 6 weeks, and the patient has generally multiple therapy sessions a day on workdays.
Despite a lack of high-quality evidence for many rehabilitative treatments seen individually (Khan & Amatya, 2017), evidence for multidisciplinary rehabilitation in terms of improvements in activity and participation is supported (Khan & Amatya, 2017;Khan et al., 2007).
Nevertheless, as in other European countries (Helland, Holmøy, & Gulbrandsen, 2015), many people in Germany with MS do not make use of rehabilitation (Nebe & Naumann, 2015), although multidisciplinary rehabilitation is widely available for these patients.
Not much is known about the barriers to the use of rehabilitation in MS. However, as Helland et al. (2015) showed in their qualitative study about the barriers and facilitators related to rehabilitation stays in MS, patients' assumptions and expectations about rehabilitation were important factors influencing the use of MS rehabilitation. Covering patients' assumptions about illness (relating to its symptoms, cause, timeline, control/cure, and consequences ), the patients' illness representations, which are theoretically embedded in the well-evaluated Common Sense Model of self-regulation (CSM; (Leventhal, Leventhal, & Cameron, 2001)), have initiated a large body of research. Consistent with the assumptions of the CSM, illness representations have proven to be important predictors of various aspects of self-regulation, such as adherence, adjustment, and work participation in many acute and chronic illnesses (Hagger & Orbell, 2003;Hoving, van der Meer, Volkova, & Frings-Dresen, 2010;. Furthermore, in the context of cardiac rehabilitation, illness representations turned out to be predictors of attendance at rehabilitation (French, Cooper, & Weinman, 2006;Whitmarsh, Koutantji, & Sidell, 2003) and dropout from outpatient rehabilitation (Yohannes, Yalfani, Doherty, & Bundy, 2007).
Therefore, the aim of our study was to explore whether, in patients with MS, illness representations are associated with selfreports of rehabilitation use in the past and the intention to use rehabilitation in the future, beyond sociodemographic and illnessrelated factors.

| Design and participants
We conducted a cross-sectional nationwide online survey in

| Measures
Illness representations were measured using the German version of the Brief Illness Perception Questionnaire (B-IPQ; Broadbent, Petrie, Main, & Weinman, 2006). The B-IPQ is a 9-item scale designed to rapidly assess the cognitive and emotional representations of illness. The scale measures patients' cognitive and emotional representations of their illness, including consequences, timeline, personal control, treatment control, identity (symptom burden), coherence, concern, emotional response, and causes. The B-IPQ items (except the cause item, which uses free text) range from 0 to 10, and a greater score on an item represents a larger value in the measured dimension. According to a systematic review and meta-analysis, the B-IPQ is widely used and has good psychometric properties (Broadbent et al., 2015). In our survey, we excluded the cause item and the timeline item, as other studies with MS samples revealed this item to be extremely negatively skewed (Dennsion et al., 2010).
The behavioral variable "use of rehabilitation" in the past (previous attendance) and the intention to use rehabilitation in the future were assessed via the two items "Have you ever made use of rehabilitation because of your MS?" and "Would you make use of rehabilitation because of MS if necessary?" Both items had to be answered dichotomously ("yes" or "no").
Furthermore, patients completed a questionnaire covering sociodemographic variables (age, sex, education, and family status) and self-reported information about illness-related factors (diagnostic subgroup, time since the last exacerbation, time since the first MS symptoms, and time since MS diagnosis).

| Data analysis
First, we conducted descriptive analyses to explore the relationships between the included independent variables and the outcome variables: We analyzed the interrelationships between the B-IPQ items, the relationships between the B-IPQ items and the sociodemographic variables and illness-related variables, respectively, and the relationships between all the independent variables and outcome variables (use of rehabilitation and the intention to use rehabilitation) using Pearson correlation coefficients. Correlations of .1 are interpreted as small, correlations ≥.3 as medium, and correlations ≥.5 as strong (Cohen, 1988). Secondly, we applied two separate hierarchical binary logistic regression procedures to analyze whether illness representations are associated with the use of rehabilitation in the past or with the intention to use rehabilitation in the future, over and above sociodemographic and illness-related variables. In the first step, sociodemographic variables were dummy-coded and entered into the models. The second step added the dummy-coded illness-related variables, and the third step added illness representations. To avoid multicollinearity, the only variables entered during the first and second steps were those that correlated bivariately at p < .05 with the outcome variables. The following coefficients were interpreted: Nagelkerke R 2 as an index of the quality of the overall model and the percentage of the variance explained by stages 1-3.
Values >.20 can be defined as acceptable, values >.40 are interpreted as good, and values >.50 are interpreted as very good (Backhaus, Erichson, Plinke, & Weiber, 2003). Furthermore, the Wald coefficients and their p-values, odds ratios, and the respective 95% confidence intervals are shown for all included variables of stage 3.
All data analyses were performed using IBM SPSS Statistics 23 (IBM Corp, 2015).

| RE SULTS
A total of N = 590 patients with MS participated in the study. The vast majority of participants (>90%) were recruited via the DMSG (website, newsletter, or Facebook site), while only 2% of participants were recruited via MS clinics. Women comprised 72.4% of the sample. The mean age was 45.6 years (SD = 10.3). Among the patients, 51.7% had relapsing-remitting MS, and the mean time since the MS diagnosis was 11.0 years (SD = 8.5). Two-thirds of the sample had made use of rehabilitation because of MS in the past and would make use of rehabilitation in the future because of MS if indicated (64.6% and 68.1%, respectively). .237 (<.001) .107 (.010) .301

(<.001)
Personal control  The baseline associations between the B-IPQ items are shown in Table 2. The correlations are predominantly significant, but in the small range. There are medium correlations between personal control and treatment control, concern and identity, concern and emotional response, and consequences and emotional response.
Furthermore, the association between identity and consequences is strong.    The same procedure also controlling for use of rehabilitation in the past (in the first step) showed that perceived past behavior explained 56% of the variance of intention to use rehabilitation.
The explained variance was slightly higher when including the sociodemographic and illness-related variables (+ 1.5%) and including illness representations yielded a further increase in the explained variance, by four percentage points. Besides use of rehabilitation, with the highest Wald coefficient (Wald = 115.099; p < .001), treatment control (Wald = 7.914; p = .005) and concern (Wald = 6.808; p = .009) were also significantly associated with the intention to use rehabilitation.

| D ISCUSS I ON
The results of our study show that certain illness representations are related to the intention to use and the use of rehabilitation in MS, above and beyond sociodemographic and illness-related factors: The probabilities of intention to use rehabilitation and of use of rehabilitation were higher in patients who believed that their MS was controllable by treatment and perceived that their MS would have severe consequences. Illness representations explained a higher proportion of variance than sociodemographic and illnessrelated variables, and the illness representation domain "treatment control" showed the strongest relationship within both regression models. The relevance of illness representations was further underlined in the third regression analysis, where perceived past behavior (rehabilitation use) was controlled for the intention-related outcome variable.
With respect to the outcome domain, our results are largely in line with studies in the context of cardiac rehabilitation. The illness representation domains "cure/control," "consequences," "identity," and "coherence" were significantly associated with attendance at cardiac rehabilitation in patients with acute myocardial infarction (French et al., 2006), and they also turned out to be predictors of dropout from a cardiac rehabilitation program (Yohannes et al., 2007).  (Dennison et al., 2009). Taking into account these considerations, we consider the self-reported measures as reasonable proxies for the measurement of the illness-related variables in our study. However, including a physician-based measurement of illnessrelated variables should be considered as a good way to complement our study design.
Furthermore, our models showed that certain illness representations, namely perceived consequences and treatment control, have the strongest relationship with the use and the intention to use rehabilitation in MS. The consequences domain seems to be associated with many outcomes in MS, such as illness intrusiveness, physical functioning, depression, self-esteem, anxiety (Vaughan et al., 2003), health-related quality of life (Spain et al., 2007), pain severity, and pain interference (Harrison, Silber et al., 2015). Furthermore, the remaining illness representation domains have also been shown to be associated with important outcomes. For example, the identity domain has been identified as associated with the quality of life (Spain et al., 2007;Wilski & Tasiemski, 2016a) and functional impairment (Dennsion et al., 2010), illness coherence with psychological stress (Dennsion et al., 2010), the perceived timeline with pain-related outcomes (Harrison, Silber et al., 2015), and treatment control with selfmanagement (Wilski & Tasiemski, 2016b). Nevertheless, these results  TA B L E 5 Prediction of "intention to use rehabilitation"
In  (Bitzer et al., 2009) and, therefore, present a good opportunity for the explicit inclusion of illness representation interventions.
In patients with chronic back pain, the delivery of such an intervention in a rehabilitation context was seen as promising (Glattacker, Heyduck, & Meffert, 2012). With the aim of facilitating the intention to use rehabilitation, the delivery of such interventions in other settings is required. However, as perceived treatment control seems to play a predominant role as a facilitator of utilization, such interventions should focus on fostering patients' treatment expectationsfor example, through the provision of comprehensive information about the processes and aims of multidisciplinary rehabilitation approaches. However, illness and treatment representations are constantly modified in a broad personal, social, and cultural context (Leventhal et al., 2001), resulting in the fact that illness and treatment representations can show considerable intra-and interindividual variability. Therefore, the context of the development of illness and treatment representations such as the individual's illness history or social interaction with the family or professionals should be taken into account when implementing such interventions.
Focusing on the quality of research on illness representations, the limitations pointed out in the review of Dennison et al. (2009) are still valid in more recent papers: Most studies are cross-sectional (Dennsion et al., 2010;Harrison, Silber et al., 2015;Wilski & Tasiemski, 2016a, 2016b, sample sizes are often small (below 150), and the independent and outcome variables are predominantly measured via self-report questionnaires (Dennison et al., 2009 (Dennison et al., 2009) in a theory-driven way.
A major limitation of our study is its cross-sectional design.
Furthermore, all variables-including diagnostic subgroup-were measured using self-report questionnaires. This implies the potential of a recall bias, for example with respect to our retrospective assessed behavioral outcome variable "rehabilitation use".
Operationalizing the intention to use rehabilitation in a binary way might have been too simplistic. Furthermore, it must be taken into account that an intention does not necessarily predict behavior such as participation in rehabilitation. The recruitment strategy implies the potential for selection bias: No information is available about people with MS who declined to take part in this online study. However, the sample was large and highly comparable with respect to (German) epidemiological data found in the literature (Stuke et al., 2009). Another limitation lies in the fact that we refrained from measuring other independent variables that could have been relevant, such as MS severity or mood-although there is preliminary evidence which shows that illness representations are not just a reflection of mood states in MS (Harrison, Silber et al., 2015;Skerrett & Moss-Morris, 2006). Finally, we excluded the illness representation domain timeline due to the negative skewness in other MS samples (Dennsion et al., 2010) and excluded the cause item due to its open format, which would have involved a different issue and separate analyses (see Bassi et al., 2016).

| CON CLUS ION
Illness representations are related to the intention to use and the use of rehabilitation in MS, above and beyond sociodemographic and illness-related factors. Although, as far as we know, this study is the first of its kind in a German MS population to include a large sample and control for demographic and illness-related variables, the results should be replicated in longitudinal studies to overcome the abovementioned limitations. However, our data suggest that addressing patients' illness representations may facilitate the intention to use and the use of multimodal rehabilitation, which is recommended as an important treatment option in MS, one that contributes to better illness outcomes.

ACK N OWLED G M ENTS
The Project "Attitudes towards medical rehabilitation in patients with Multiple Sclerosis" was conducted on behalf of the German Pension Insurance. We thank all those patients who participated, as well as the German Multiple Sclerosis Society and the MS clinics, for their assistance with recruitment. The article-processing charge was funded by the German Research Foundation (DFG) and the University of Freiburg in the funding program Open Access Publishing.

CO N FLI C T O F I NTE R E S T
Angelika Nebe works at the German Pension Insurance (see affiliation). The authors have no further competing interests to report.