Article Text
PDF
Abstract
Objective To compare the value that rheumatologists across Europe attach to patients' preferences and economic aspects when choosing treatments for patients with rheumatoid arthritis.
Methods In a discrete choice experiment, European rheumatologists chose between two hypothetical drug treatments for a patient with moderate disease activity. Treatments differed in five attributes: efficacy (improvement and achieved state on disease activity), safety (probability of serious adverse events), patient's preference (level of agreement), medication costs and cost-effectiveness (incremental cost-effectiveness ratio (ICER)). A Bayesian efficient design defined 14 choice sets, and a random parameter logit model was used to estimate relative preferences for rheumatologists across countries. Cluster analyses and latent class models were applied to understand preference patterns across countries and among individual rheumatologists.
Results Responses of 559 rheumatologists from 12 European countries were included in the analysis (49% females, mean age 48 years). In all countries, efficacy dominated treatment decisions followed by economic considerations and patients’ preferences. Across countries, rheumatologists avoided selecting a treatment that patients disliked. Latent class models revealed four respondent profiles: one traded off all attributes except safety, and the remaining three classes disregarded ICER. Among individual rheumatologists, 57% disregarded ICER and these were more likely from Italy, Romania, Portugal or France, whereas 43% disregarded uncommon/rare side effects and were more likely from Belgium, Germany, Hungary, the Netherlands, Norway, Spain, Sweden or UK.
Conclusions Overall, European rheumatologists are willing to trade between treatment efficacy, patients' treatment preferences and economic considerations. However, the degree of trade-off differs between countries and among individuals.
- Rheumatoid Arthritis
- Treatment
- Patient perspective
- Economic Evaluations
- DMARDs (biologic)
Statistics from Altmetric.com
Background
Traditionally, drug treatment decisions for patients with rheumatoid arthritis (RA) were primarily based on benefits and risks associated with a drug. Patient’s preferences and treatment costs usually received little attention, only.1 The introduction of biologics has increased treatment opportunities substantially. The new treatments are between 20 and 60 times more costly, and therefore raise concerns about affordability of RA care.2 ,3 Today, in most European countries, access to biologics is highly regulated.4 In addition to the economic dimension becoming increasingly important, physicians across Europe are encouraged to actively involve their patients in treatment decisions. Patient-centred care is expected to improve adherence to agreed treatment plans and thus positively influence health outcomes.5 However, it can be a challenge to include patients’ preferences in decisions that are influenced by multiple other aspects.
Although costs and cost-effectiveness (CE) have been given consideration in recent European League Against Rheumatism developed treatment recommendations,6 ,7 recommendations rely mainly on evidence for effectiveness and safety. Optimal patient care across Europe will depend on how recommendations are implemented in the context of differences in the economic situations of countries and differences in the attitudes of physicians and patients. A recent article revealed a strong association between access to (expensive) biologics and the country’s wealth.8 Interestingly, however, another review found that drug coverage for innovative and expensive drugs does not necessarily translate into its use; there were differences in use between countries with comparable reimbursement criteria, and also between regions within a country.9 Differences in attitudes among clinicians likely contribute to differences in treatment choices.10
To our knowledge, limited data exist about the differences in values that rheumatologists attach to various treatment characteristics beyond efficacy and safety. Little is known about how rheumatologists assess economic consequences of a treatment choice: Do they take into account relative CE considerations or primarily use absolute costs to make economic trade-offs? Moreover, there is limited research on how rheumatologists consider patient’s preferences.
In this study, we aim to assess if rheumatologists across Europe are willing to trade-off treatment efficacy and safety for economic considerations and patients’ preferences. Further, the study aims to determine different preference profiles of rheumatologists and whether preference profiles are associated with the country rheumatologists are located in.
Methods
A discrete choice experiment (DCE) was designed to investigate relative preferences of rheumatologists when choosing drug treatments in RA. In the DCE, rheumatologists were presented with a series of choices and asked in each to select the preferred drug treatment among two hypothetical treatment options (A or B). The treatment options were described by a set of attributes, further specified by attribute levels.
Selection of attributes and levels and patient profile
Selection of attributes and levels is fundamental in the design of a reliable DCE study.11 ,12 We followed a stepwise approach. First, potentially important attributes, attribute definitions and levels were identified from the literature.11 ,13–23 Second, an expert group (n=6) consisting of rheumatologists and experts in the field of economic evaluations, DCE and decision making agreed on an initial list of attributes/definitions/levels. Third, the proposed candidate attributes/definitions/levels were discussed with eight rheumatologists from three countries to ensure that they reflected clinical realities.
Five attributes were selected for the DCE: efficacy (status and improvement of disease activity score of 28 joints, safety (risk of a serious adverse event (AE)), patient’s preference (level of agreement with treatment choice), annual drug costs and CE (incremental cost-effectiveness ratio (ICER) compared with usual care expressed as incremental cost of one quality-adjusted life year gain)). For each attribute, three levels were agreed (figure 1).
Attributes and levels and example choice set of DCE experiment. (A) Attributes and levels describing drug treatment options in DCE experiment. (B) Example choice set – repeated 17 times with varying attribute levels. *In the choice sets, also changes of the individual DAS28 components (tender joint count (TJC), swollen joint count (SJC), erythrocyte sedimentation rate (ESR) and patient global assessment of disease activity (PGA)) were presented. **ICER is expressed in costs (€) per QALY gained for the selected treatment compared with usual care. DAS28, disease activity core 28; DCE, discrete choice experiment; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year.
The patient profile (possibly influencing treatment choices) was selected to imply a clear need for change in the treatment and a decisional problem in terms of balancing previously agreed attributes. The patient profile allowed for a switch to expensive (biological) medications in the majority of countries to ensure that respondents do not feel intuitively restricted in their decision by local healthcare regulations. The agreed profile described a patient with anticitrulline antibody positive RA and moderate disease activity despite two conventional synthetic disease-modifying antirheumatic drugs. The full patient profile (see online supplementary table S1) has been approved by experts and rheumatologists.
Supplemental material
Experimental design
Based on the attributes and levels, there are 243 (35) possible treatment combinations – too many to ask each rheumatologist to evaluate.24 We reduce the number of choice sets using a Bayesian efficient experimental design (Ngene software25). A Bayesian design aims to maximise precision of estimated parameters for a given number of choice questions26 by incorporating a priori information about sign and value of parameters. The a priori information was gained from a preliminary DCE with 10 rheumatologists. Implausible or unrealistic treatment options were excluded from the design (eg, efficacy is low, costs are high, but the CE is best). In addition, three choice sets were included to test if respondents made sensible choices. (1) A dominance test—a choice set with one treatment option that is clearly better than the other and (2) two repeated choices – choice sets that repeat earlier choices to assess the stability of respondents’ choices. A total of 17 choice sets were included in the questionnaire. An example of a choice set is shown in figure 1.
The questionnaire
The questionnaire consisted of two parts: (1) the DCE task and (2) questions collecting sociodemographic data (eg, age, gender and work environment) to facilitate interpretation of the results. The full questionnaire was piloted among rheumatologists (n=14) from two different countries to ensure that choice sets, attributes and levels were relevant, clear and plausible across countries. Economic attributes were presented in local currencies.27 An online survey was distributed by email through a national principle investigator per country.
Country selection
As attitudes of rheumatologists may be influenced by external factors such as the country’s culture, wealth or healthcare environment, countries from all geographical regions in Europe were invited: Belgium (BE), France (FR), Germany (GE), Hungary (HU), Italy (IT), the Netherlands (NL) Norway (NO), Portugal (PT), Romania (RO), Spain (SP), Sweden (SE) and UK.
Statistical analysis
Rheumatologists’ responses were considered for data analyses when at least 50% of choice sets were completed and they successfully passed the dominance test. A three-step approach was used to analyse data.
First, a mixed logit (ML) model was used to assess the relative importance of attributes, using Nlogit, V.5.28 The following utility model was estimated:
U represents the observable relative preference of rheumatologist (i) for a treatment choice (j) which can be defined as a sum of preference scores for attributes/levels. β0 is the constant, β1–8 are the mean attribute utility weights (physician's preferences) for the respective attribute and ɳi represents the random parameter for rheumatologist i. Dummy coding was used to describe variables β1–6. Reference categories for efficacy, safety and patient’s preference were moderate response, very rare AEs and neutral patient attitude, respectively. The signs of the β coefficients indicate whether the attribute has a negative or positive effect on the rheumatologists' preference. An ML model allows model parameters to vary between respondents, which is reflected in the random parameter. If the SD of this random parameter is significantly different from zero, this is interpreted as evidence of significant preference variation for the attribute within the population.29 Results were considered significant for p<0.05.
The relative importance of attributes was calculated based on the range of the level coefficients per attribute using the method described by Malhotra and Birks.30 More specifically, the relative importance weights were derived by dividing the range of the level coefficients for one attribute by the sum of ranges of coefficients of all attributes.
Second, to assess potential similarities, countries were grouped on the basis of their similarity in mean relative importance weights across the five attributes. Hierarchical Ward’s linkage with a squared Euclidean distance measure was used (using STATA V12 (StataCorp. 2011. Stata Statistical Software: Release 12. College Station TSL)). To determine the optimal cluster number, a second clustering method (k-means) was applied and results were compared with hierarchical Ward’s linkage to verify if cluster structures were reproducible. Introduction of additional clusters was stopped where both clustering methods provided different results. Further, we based the decision on numbers of clusters on the rule of thumb suggesting 
, with k describing the number of suggested clusters and the number of observations.
Third, a latent class model (LCM, using Nlogit) was used to determine preference profiles of all individual rheumatologists. LCM can be used to identify the existence and number of classes in the population based on their treatment preferences. The LCM can also be used to explore if covariates (such as clusters of countries) influence the probability to belong to a particular class. Class membership is latent in that each respondent belongs to each class up to a modelled probability.31 To determine the number of classes, we selected the model with the best fit based on the Akaike information criterion. To further understand the role of country, previously developed country clusters were added as covariate to the LCM in an additional step. Statistically significant parameter estimates (p<0.05) indicate that the covariate (ie, the dummy of country cluster) contributes to the explanation of latent classes. For example, if the parameter estimate for the covariate ‘country cluster’ is positive and significant for a certain class, it indicates that rheumatologists from the countries that belong to the country cluster are more likely to belong to that particular class.
Results
Respondent's sociodemographic characteristics
Overall 559 rheumatologists from 12 European countries were included in the analysis. Mean age was 48.0 years and 49% of rheumatologists were females (table 1). Test–retest reliability was in line with existing literature29 with 82.3% (retest 1) and 80.8% (retest 2) of respondents having chosen the same alternative in the test–retests. Five respondents failed in the dominance test and were therefore excluded from the analysis.
Characteristics of rheumatologists for the total group and each country separately
Attribute preferences when selecting a new treatment, overall and at a country level
The main results of the ML model are summarised in table 2a (coefficients) and b (relative importance) for all countries and each country separately. The detailed results overall and per country from ML models can be extracted from online supplementary table S2.
Results from the discrete choice experiment
Coefficients (table 2a) revealed that in all countries good response and especially remission were significantly more desirable than moderate response (βremission>βgood response). When coefficients were significant, probability of a serious AE decreased preference for that given treatment option (βuncommon<βrare). The patient’s agreement with the treatment was universally associated with an increased preference and a stronger, negative impact on preference was seen when patients dislike treatments (|βdisagreement|>|βagreement|).The contribution of patient’s disagreement with a treatment choice was significant in all countries. The economic aspect also played an important role. Higher economic burden was consistently associated with decreased preference. CE was not significant in three countries (FR, IT, RO), medication costs however were significant in all countries.
When comparing contribution of different attributes across countries (table 2b), results revealed that efficacy dominated the treatment choice in all countries with comparatively low variations between countries (range 39%–52%). High variability was observed for the relative contribution of safety (range 1%–20%), whereas patient’s preference played role in all countries (range 7%–21%). Similarly, absolute costs played a role in all countries (range 10%–24%) whereas contribution of CE varied importantly with several countries largely disregarding CE (range 4%–21%). In the majority of countries, absolute costs were more important than relative CE considerations.
Cluster of countries according to the preference profile
Two clusters were identified (table 2). In cluster 1 (BE, GE, HU, NL, NO, SE, SP, UK), all attributes, except safety, had a relevant contribution to treatments choice. In cluster 2, including (FR, IT, PT RO), there was a higher emphasis on safety (range 18%–20%) and lower emphasis on relative CE considerations (range 3%–8%).
Classes of individual rheumatologists according to the preference profile
Four classes of rheumatologists have been identified according to their preference profile (table 3). In all classes, efficacy (both levels) and patient’s disagreement with a treatment were significant determinants of treatment choice. In class 1, rheumatologists accounted for both economic attributes (absolute costs and relative CE) but largely disregarded safety. Overall, 43% of rheumatologists were estimated to belong to class 1. Rheumatologists belonging to one of the remaining classes 2–4 disregarded ICER.
Results latent class model (LCM) distinguished four classes of respondents
Respondents of class 3 balanced safety and total medication costs, while those in classes 2 and 4 accounted only for either safety (class 2) or total medication costs (class 4).
The probabilities to belong to classes 2, 3 and 4 were estimated with 22.6%, 31.1% and 11.1%, respectively. Rheumatologists from country cluster 2 (FR, IT, PT RO) were significantly more likely to be located in classes 2, 3 and 4, with CE playing a minor role.
Discussion
The study revealed that across countries rheumatologists are willing to trade-off efficacy against patients’ preferences and/or economic aspects in treatment choices. Forty-three per cent of rheumatologists balanced all attributes, including CE when choosing a therapy, while for all other rheumatologists, CE did not play a substantial role. Medication costs however remained relevant for 77% of clinicians. Two groups of countries could be distinguished. One group, which represented the majority of countries, was more likely to balance all treatment attributes including CE, while the other group placed little emphasis on CE.
In all countries, rheumatologists considered economic, indicating that rheumatologists support efficient use of limited healthcare resources. However, latent class analyses showed that about half of individual rheumatologists are estimated to largely disregard the CE. This can be explained by clinicians having limited access to CE data of relevant drugs or methodological complexities and weaknesses in existing data.32 ,33 Given the importance of ICER, we need to ensure that trustful and comparable good quality data are provided to clinicians to support its use in clinical decision making. In addition, the direct healthcare environment often rather supports economic trade-offs based on absolute costs34 even though CE data provide more valuable information on efficient treatment choices for society. The present findings confirmed that in UK CE and more specifically ICER plays a more decisive role in reimbursement decisions compared with many other European countries.35Of note, CE studies are frequently limited to selected countries although transferability of data across countries is not always possible, further limiting its use.32
Patients’ preferences were taken into account by all rheumatologists, despite variations observed among individuals and between countries. Cultural differences in the physician–patient relationship or the physician’s attitude may partly explain the differences. A recent review, for example, found that clinicians are more likely to support shared decision making when their direct work environment supports the concept.1 Interestingly, however, rheumatologists from all countries considered in particular patient’s disagreement, possibly to avoid poor drug adherence. Although perception towards shared decision making can vary, this study for the first time revealed consensus among individuals and across countries on the importance of a patient’s disagreement, and is the first to address patient’s preferences in the context of other treatment attributes.
Some limitations have to be considered for interpretation of results. First, the sample size was lower in some countries and differences in sociodemographics across countries were observed, thus selection bias and limitations in generalisability of results cannot be excluded. Subgroup analyses of the results from NL (results not reported), revealed no significant differences in preferences for different subgroups. Further an efficient experimental design was developed to optimise response efficiency. Second, pharmaceutical industry36 or local healthcare regulations (in particular the reimbursement criteria) may influence the preferences. However, most participating countries do not restrict access to expensive drugs for patients described in the DCE; therefore, this type of effect is likely limited to countries with very strict criteria (UK, HU and RO). Third, although the presented attributes were confirmed as most relevant for treatment decisions in RA and a strong methodology was used to select and defined attributes, it cannot be excluded that further attributes play a role in some countries. Further research could contribute to a better understanding on other relevant factors across European countries. Fourth, presenting two economic attributes—costs and CE—in the DCE has its limitation as also absolute costs together with efficacy and safety provide insights into CE of a drug. However, only including both economic attributes allowed to raise awareness on the fact that clinicians are still predominantly focused on costs.
Finally, this study provides only insight into stated preferences. It remains unknown whether rheumatologists reach decisions in the same way in clinical practice as they state in the theoretical framework of the DCE. Also, in clinical reality decisional problems may exceed the complexity reflected in the study design. Revealed preferences would be additionally informative; however, design and execution of such a study would be very difficult.
Despite limitations, this study reveals that values rheumatologists attach to treatment characteristics vary importantly between countries and among individual rheumatologists. The complexity of today’s treatment decisions justifies more research on behaviour of prescribers, as the clinicians' attitude is one of the important factor contributing to quality and equality in healthcare. We hope the current study will raise awareness on the importance of physicians' behaviours and open discussion on how clinicians are expected to trade-off various aspects of a treatment decision.
Conclusion
Overall, rheumatologists take efforts to balance multiple aspects of a treatment choice including the patient’s potential disagreement and the economic consequences. However, important differences in the assessment of economic aspects were observed between and within countries. In most countries, clinicians still focus on absolute costs while CE would provide more valuable information on the societal consequences of a decision.
Acknowledgments
The authors would like to thank all participating rheumatologists for their contribution to the study and the following clinical rheumatologists for their support in the development of the study design and questionnaire development: Tubergen van AM, Vosse D, Linden van der JMJP, Onna van MGB, Walravens MJF, Saritas AM, Schoonbrood THM, Bours SPG, Denissen ECM, Spaetgens B, Essers IMM, Stolwijk C, Antoaneta I, Dadoune S, Rozenberg S and Molto A. Further the authors would like to thank the following individuals and organisations for their support in the study coordination: France: CRI-IMIDIATE network and the SNMR (Syndicat National des Médecins Rhumatologues); Germany: Barbara Gundelach (German Society of Rheumatology); Vera Höhne-Zimmer, The Netherlands: Dutch Society of Rheumatology; Norway: Norwegian Society of Rheumatology; Romania: Romanian Society of Rheumatology.
References
Footnotes
Handling editor Hans WJ Bijlsma
Twitter Follow Carlo Scirè at @rthritis
Competing interests MH contributed during an unpaid extended maternity leave (2013–2016) agreed with Hexal AG, Germany.
Provenance and peer review Not commissioned; externally peer reviewed.