Comparison of preferences of healthcare professionals and MS patients for attributes of disease‐modifying drugs: A best‐worst scaling

Abstract Background The choice between disease‐modifying drugs (DMDs) for the treatment of multiple sclerosis (MS) becomes more often a shared decision between the patient and the neurologist and MS nurse. This study aimed to assess which DMD attributes are most important for the healthcare professionals in selecting a DMD for a patient. Subsequently, within this perspective, the neurologists’ and nurses’ perspectives were compared. Lastly, the healthcare professionals’ perspective was compared with the patients’ perspective to detect any differences that may need attention in the communication about DMDs. Design A best‐worst scaling (BWS) was conducted among 27 neurologists and 33 MS nurses treating patients with MS to determine the importance of 27 DMD attributes. These attributes were identified through three focus groups with MS patients in a previous study (N=19). Relative importance scores (RISs) were estimated for each attribute. Multivariable linear regression analyses were used to compare the different perspectives. Results According to the neurologists and nurses, safety of the DMD was the most important DMD attribute in the treatment decision, closely followed by effect on disability progression, quality of life and relapse rate. Patients with MS agreed with the importance of the last three attributes, but valued safety significantly lower (b=−2.59, P<.001). Conclusions This study suggests that, overall, neurologists and nurses regard the same DMD attributes as important as MS patients with the notable exception of safety. This study provides valuable information for the development of interventions to support shared decision making and highlights which attributes of DMDs may need additional attention.


| INTRODUCTION
Multiple sclerosis (MS) is a degenerative disease of the central nervous system causing physical and cognitive disabilities. 1 The relapsingremitting form of MS (RRMS) can be treated with disease-modifying drugs (DMDs) to reduce progression or even induce improvements.
DMD treatment requires long-term administration with a minimum of missed doses. DMDs are classified as first-, second-or third-line treatments. First-line DMDs are used in patients with mild disease activity. In case of high disease activity or intolerance to the side effects, second-or third-line DMDs may be chosen. These DMDs have more favourable effectiveness rates but less favourable safety profiles. 2,3 Overall, thirteen DMDs are currently approved by the European Medicine Agency and the Food and Drug Administration. 4,5 This variety in DMDs often provides MS patients with more than one acceptable option for treatment.
Therefore, a choice between DMDs needs to be made, which is often difficult. Besides differences in effectiveness rates and safety profiles, the DMDs can differ in other characteristics, or attributes, such as mode of administration and side effects. To make an optimal choice, the patient's preferences for such attributes should be incorporated in the decision. In shared decision making, the patient makes the decision about the treatment together with the healthcare professional. 6 The healthcare professional ensures the patient is informed about the best available treatment options and supports the patient in clarifying the advantages and disadvantages of each of the options.
The patient communicates his or her values and preferences for these advantages and disadvantages, and deliberates with the healthcare professional, in this case a neurologist and/or an MS nurse, about the treatment options to reach a decision. Decision aids, which are specifically designed to support shared decision making, have been shown to improve the patients' involvement in the decision and the quality of the decision. 7 Understanding the preferences of patients and healthcare professionals for treatment options could inform the development of a decision aid. Although the treatment decision should be individualized to every patient and should include the patient's preferences and healthcare professional's opinion tailored to the patient's needs, the evaluation of the average perspectives about the importance of the treatment attributes in the decision could be useful for developing a patient decision aid that meets most patients' and healthcare professionals' informational needs. An assessment of the average preferences could be valuable for selecting the information that needs to be included in the decision aid to ensure effectively supporting the shared decision-making process without making the use of the decision aid too cognitively burdensome for the patient. Preferences of patients for attributes of DMDs have been evaluated in several stated preference studies. [8][9][10][11][12][13][14][15][16] Recently, focus groups and a best-worst scaling were conducted to identify and prioritize a range of attributes that could be important in clinical decision making from the patients' perspective. 8 This study found that the average MS patient regarded the effects on disability progression and quality of life as the most important attributes of DMDs in the decision, followed by the effect on the relapse rate, severity of side effects and the effect on the severity of relapses. 8 Several studies showed that preferences of healthcare professionals for treatment options can differ from the preferences of patients. [17][18][19][20][21] These differences in preferences could result in different expectations and information needed from the decision aid. In MS, one study has assessed the neurologists' perspective regarding the importance of a limited number of five attributes in the DMD choice. 22 To our knowledge, the patients' perspective on important DMD attributes for choosing between DMDs has not yet been directly compared with the healthcare professionals' perspective.
This study aimed first to use a rigorous method, a best-worst scaling, to identify which DMD attributes the healthcare professionals (neurologists and MS nurses) find important to take into consideration in the decision regarding DMD treatment. As the relationship with the patient may differ between neurologists and MS nurses because of the more frequent and informal contact between MS nurses and patients, the second aim of this study was to evaluate whether the perspectives on the importance of DMD attributes differ between neurologists and MS nurses. Third, the overall MS healthcare professionals' (neurologists and MS nurses) perspective was compared with the MS patients' perspective on the importance of DMD attributes in the treatment decision.

| METHODS
A best-worst scaling case 1 as described by Flynn and Marley 23 was conducted to obtain the neurologists' and MS nurses' perspective on the importance of attributes of DMDs. In contrast to discrete choice experiments, a best-worst scaling allows for obtaining the relative importance of attributes for a large number of attributes, regardless of the levels of the attributes. 24 The best-worst scaling was embedded

Key points
• Understanding the healthcare professionals' perspective about the importance of attributes of disease modifying drugs for multiple sclerosis, and comparing this perspective with the patients' perspective, would be important to develop a patient decision aid that is accepted by, and usable for, both patients and healthcare professionals in the shared decision-making process • The healthcare professionals and patients overall agree about the importance of the effects of DMDs on disability progression, quality of life and relapse rate, but safety was valued significantly lower by patients.
• The study results provide guidance for the selection of information to be included in a patient decision aid for disease-modifying drugs in multiple sclerosis to effectively support the patient and healthcare professional in making a shared decision.
in an online questionnaire. Therefore, the Checklist for Reporting Results of Internet E-Surveys was followed for reporting the design and the results. 25 We used the same best-worst scaling that was performed to elicit preferences for patients with MS. 8 The best-worst scaling included 27 attributes, which were identified in a two-stage procedure. First, in an exploratory phase, possible relevant attributes were identified through a literature review and interviews with three neurologists and three nurses. Second, three focus groups (N=19) were conducted to elicit the attributes that patients with MS regarded important in the decision. At the end of each focus group, any additional attributes from the exploratory phase that were not mentioned yet were discussed to evaluate whether these additional attributes were relevant as well. A total of 27 important attributes for the decision about DMD treatment were identified. Using Sawtooth SSI Web version 8.2.0, four best-worst scaling versions, each consisting of 17 unique choice tasks with five attributes per choice task, were created to obtain the most efficient, fractional design. In such a design, it is not needed to include all possible combinations of attributes in the choice tasks, which is far too burdensome to administer. Instead, specific sets of five attributes from the full attribute lists are created to ensure that the relative importance of each attribute can be derived.
Each attribute was presented 12 or 13 times in the four best-worst scaling version, was combined at least once with every other attribute and appeared two to four times in each position in the choice task.
Each respondent randomly received one of the best-worst scaling versions. In every choice task, the respondent was asked to state which attributes were the most and the least important in the decision about DMDs, making a hypothetical trade-off between the attributes. 24 The best-worst scaling was pilot tested among researchers (n=3) and patients with MS (n=3) to ensure comprehensibility of the questionnaire. An example of a choice task for neurologists and MS nurses is presented in Figure 1. Further description on the identification, selection and definition of attributes and on the development of the choice tasks and the questionnaire is described in detail elsewhere. 8 In addition to the best-worst scaling choice tasks, questions about demographical and professional characteristics were included.
The respondents had to answer all questions to complete the questionnaire. At the end of the questionnaire, an optional open question provided the respondents the opportunity to report any important attributes in the decision about DMD treatment they missed in the choice tasks of the best-worst scaling.

| Subjects
Healthcare professionals were eligible for participation in the study if they were neurologists (i.e. MS-specialized neurologists or gen- Additionally, all nurse practitioners and physician assistants in MS (N=15) were also contacted through a mailing list of their professional association. In case of no response among the neurologists and nurses, a reminder email was sent after 1 and 2 weeks after the first email.
The neurologists and nurses were informed that by completing the questionnaire they gave consent for their responses to be used in the study. Cookies were used to prevent the same respondent from filling out the questionnaire twice. The study was fully anonymous, and no data, such as IP addresses, were recorded that could lead to the identity of the healthcare professional. F I G U R E 1 Example of a choice task in the best-worst scaling

| Analyses
Only completed questionnaires were included in the analyses.
Descriptive statistics were used to present characteristics of the neurologists and MS nurses. To evaluate the preferences of the healthcare professionals for the DMD attributes, Sawtooth SSI Web version 8.2.0 was used to perform a hierarchical Bayes model employing a multinomial logit procedure. 26 A raw score for the importance of each attribute was calculated in an iterative process of estimating each respondent's utility score conditionally on estimates of other respondents' utility scores. The raw score was rescaled to a relative importance score (RIS) on a ratio scale. 27  To assess whether the RISs assigned by the neurologists and nurses to the 27 attributes differed from the RISs assigned by the patients with MS, hierarchical Bayes analyses were conducted including responses of the neurologists and nurses, and responses of patients obtained in a previous study. 8 A second linear regression model was built to control for age, sex and level of education in the function between the respondent type (i.e. healthcare professional or patient) and the RISs of the 27 attributes. Level of education was dichotomized into respondents who had completed primary and/or secondary school only and respondents who had completed any additional type of tertiary education. The patients' perspective was obtained from a previous study among 185 patients with relapsing-remitting MS or clinically isolated syndrome, and with a range in experience with DMDs (from having never used a DMD to having used more than three different DMDs) and duration of diagnosis. Further detailed description of the patient recruitment and the patients characteristics can be found elsewhere. 8

| RESULTS
Of the 120 neurologists and an estimated 140 nurses contacted, 79 healthcare professionals accessed the questionnaire between 20 November 2015 and 8 February 2016. In total, 62 healthcare professionals completed the questionnaire (estimated overall completion rate of 24%), of which two healthcare professionals did not meet the inclusion criteria because they were working in Belgium (n=1) or did not treat patients with MS on a regular basis (n=1). The sample of 60 healthcare professionals that met the inclusion criteria consisted of 27 (45%) neurologists and 33 (55%) nurses. Table 1 (Table S1 Most-minus-least counts) did not substantially affect the rankings obtained based on the hierarchical Bayes analyses and therefore confirmed the results. Of the 60 neurologists and nurses, two (3%) neurologists reported additional DMD attributes that were not included in the best-worst scaling: teratogenic properties of the DMD and the certainty of achieving the effects of the DMD.

| Comparison of neurologists with MS nurses
Only few differences were found between neurologists and nurses ( Table 2). When comparing neurologists to nurses, relatively large differences in mean RISs (absolute difference in RISs of 1 or more) were found in effect on current MS symptoms, effect on life expectancy and total costs of the DMD, which were higher for the neurologists, and interaction with other medication, which was higher for nurses. However, when controlling for sex, the years of work experience, the number of patients with MS treated each year and the degree of focus on MS treatment, only a significant difference was found between neurologists and nurses in the low ranked attributes total cost of the DMD (b=−1.27, P=.015) and interaction with other medication (b=2.09, P=.007). These differences resulted in quite substantial shifts in ranking: total cost of the DMD was ranked 17th by neurologists (RIS=1.22) and 25th by nurses (RIS=0. 16), and interaction with other medication was ranked 18th by neurologists (RIS=1.10) and 13th by nurses (RIS=2.97). Based on their RISs though, these attributes did not influence the treatment decision severely.

| Comparison of healthcare professionals (neurologists and MS nurses) with patients
When controlling for age, sex and level of education, there were sig-

| DISCUSSION
In this study, a best-worst scaling was used to assess the neurologists' and MS nurses' perspectives on the importance of 27 DMD attributes in the decision about DMD treatments, which enabled comparison of the healthcare professionals' perspective to the patients' perspective.
A best-worst scaling is a rigorous method for assessing preferences often used in healthcare research, 29 but, to our knowledge, has not been used before to prioritize the information about treatment options for the development of a decision aid.  The healthcare professionals and the patients both valued effect on disease progression, quality of life and relapse rate highly, which is not unexpected because these are the primary aims of DMD treatment. The most notable difference between the two perspectives was found in the importance of safety. The healthcare professionals gave significantly more value to this attribute compared with patients. More specifically, safety was ranked as fourth most important by healthcare professionals, while ranked only eighth by patients, who gave more value to other attributes focused on effectiveness and the severity of more common side effects. In the process of shared decision making, this could be an issue. Healthcare professionals may, for example, be less inclined than patients to choose a DMD for which experience and research has shown risks for life-threatening or severely disabling side effects, such as progressive multifocal leukoencephalopathy.
T A B L E 3 Relative importance scores of the DMD attributes in the treatment decision according to respondent type (i.e. neurologists/nurses or MS patients), and results of the multivariable analyses for respondent type while controlling for other variables

Rank
Relative importance score mean (SD) Rank  Other limitations that apply to the best-worst scaling method and the results of the patients' perspective have been reported elsewhere. 8

| CONCLUSION
This study showed that safety, effect on disability progression and effect on quality of life were the most influential attributes of DMDs for healthcare professionals in the treatment decision. The importance of these and other highly ranked attributes did not differ be- These results provide valuable information for the development of interventions to support shared decision making. This study also demonstrates the feasibility of combining focus groups and a best-worst scaling to identify important attributes that could later be included in a patient decision aid. A best-worst scaling could be an interesting method when having to restrict the number of attributes for inclusion in a patient decision aid.

ACKNOWLEDGEMENT
We would like to thank the discussant Janne Schepers for her help-

CONFLICT OF INTEREST
During the time of the study, PJ Jongen has received honoraria from Bayer, Merck and Teva for contributions to symposia as a speaker or for consultancy activities. The other authors have no other conflicting interest to declare.