Rheumatoid arthritis patients and rheumatologists approach the decision to escalate care differently: Results of a maximum difference scaling experiment

Abstract

Objective

Anti-rheumatic drugs are frequently not appropriately modified, according to ACR Guidelines, in patients with active rheumatoid arthritis (RA) as defined by a DAS28 score greater than 3.2. The objective was to determine which factors most strongly influence patients’ and rheumatologists’ decisions to escalate care.

Methods

We administered a Maximum Difference Scaling survey to 106 rheumatologists and 213 RA patients. The survey included 58 factors related to the decision to escalate care in RA. Participants answered 24 choice tasks. In each task, participants were asked to choose the most important factor from a set of five. We used Hierarchical Bayes modeling to generate the mean relative importance score (RIS) for each factor.

Results

For rheumatologists, the five most influential factors were: number of swollen joints (RIS: 5.2; sd: 0.4), DAS28 score (RIS 5.2; sd 0.5), physician global assessment of disease activity (RIS 5.2. sd 0.6), worsening erosions over the last year (RIS: 5.2; sd: 0.6) and RA disease activity now compared to three months ago (RIS 5.1; sd: 0.6). For patients, the 5 most important factors were: current level of physical functioning (RIS: 4.3; sd: 1.1), motivation to get better (RIS: 3.5; sd: 1.4), trust in their rheumatologist (RIS: 3.5; sd: 1.6), satisfaction with current DMARDs (RIS: 3.4; sd: 1.4) and current number of painful joints (RIS 3.4; sd: 1.4).

Conclusions

Factors influencing the decision to escalate care differ between rheumatologists and patients. Better communication between patients and their physicians may improve treatment planning in RA patients with active disease.

INTRODUCTION

Several controlled studies have shown that adhering to standardized protocols aimed at minimizing disease activity improves outcomes in rheumatoid arthritis (RA) (1-7). Given these data, guidelines recommend that physicians monitor disease activity with validated instruments, such as the DAS28, and escalate care to achieve and maintain “tight control”. While thresholds for escalating care differ, a DAS28 score of 3.2 or greater is considered to indicate active disease and protocols using this cut-off result in improved clinical outcomes (7). Despite these data, several studies have shown that care is frequently not escalated in RA patients when clinically indicated by a DAS28 score of 3.2 or greater (5;13).

A number of studies have described how patients and rheumatologists make treatment decisions in RA (5;13-21). These studies have examined a limited set of factors from either physicians’ or patients’ viewpoints, and none directly compared the two perspectives. To gain a better understanding of the full spectrum of factors influencing medical decision making from both the physicians’ and patients’ perspectives, we previously conducted a qualitative study (22). In this study, we found that numerous factors (n=58) influenced decision making, and that rheumatologists and patients consider different factors in their treatment decisions.

Qualitative data, however, do not permit measurement of the impact of each of these factors on decision making. The objective of this study was to quantify the relative impact of each of the factors generated in the qualitative study on both rheumatologists’ and patients’ decision making and to subsequently compare the two. Improved insight into the factors which most strongly influence both patients and rheumatologists’ decision making is required to develop interventions targeted at improving adherence to guidelines and the quality of care delivered in RA.

MATERIAL AND METHODS

Subjects

We surveyed Dutch rheumatologists and patients with RA. Rheumatologists were asked to participate during the Dutch Rheumatology Congress in September 2009. Consecutive rheumatologists were approached by one of two medical researchers (LvH and WK) to complete the survey on a laptop computer. Dutch rheumatologists that did not participate during the congress were contacted by e-mail and provided with a link to the Maximum Difference Scaling (MDS) survey between October and November 2009. Rheumatologists that did not respond received a reminder within two weeks.

To recruit patients, clinic staff from six outpatients rheumatology clinics (two in general hospitals, two in teaching hospitals and two in university hospitals) were asked to consecutively approach RA patients during patient’s consultation until at least 200 patients were willing to participate. RA patients currently taking at least one traditional DMARD, and/or a biologic or corticosteroids were eligible to participate. Patients with hearing, visual and cognitive impairments were excluded. All patients that agreed to be contacted were approached by one of the researchers (LvH or RvB) and a telephone appointment to complete the survey was scheduled. Patients completed the survey using a paper or electronic version. In both cases, a research assistant assisted by telephone. The researchers used a standard definition list, based on an official Dutch patient information site, to clarify specific terms as needed. Patients were contacted between November 2009 and March 2010.

Factors included in the Maximum Difference Scaling survey

The factors included in the MDS survey were selected based on those discussed during preparatory focus groups. The focus group discussions (2 focus groups with rheumatologists and 3 focus groups with RA patients) aimed to explore all the factors that could influence treatment-related decisions in patients with active RA as defined by a DAS score of 3.2 or greater (22). RA patients were recruited for the focus groups by specialized rheumatology nurses working at the Radboud University Nijmegen Medical Centre. Of the 22 patients that initially agreed to participate, 15 were able to attend the focus group meeting on the scheduled date. The mean age of the RA patients was 59 years (sd: 9). Ten patients were female and five were male. Rheumatologists were purposefully recruited to ensure representation of two prescribing patterns noted in a previous study (24): those who escalated care (increased dose, added or changed DMARDs) in at least one third of the visits with patients having a DAS28 score ≥ 3.2 and those that did not.

Ten of the 14 rheumatologists approached were able to participate in the focus group on the scheduled date. The other four rheumatologists could not participate due to time schedule constraints. Three rheumatologists were female and 40% worked in an academic hospital. Their mean age was 50 years (sd: 8) and they had a mean number of years of experience of 11 years (sd: 7). During each focus group a moderator and observer were present. The moderator used a script which was developed to include prompts to elicit barriers described in Cabana’s model (25). The complete list of factors generated during the focus groups is provided in the Appendix.

Design of the Maximum Difference Scaling Survey

We designed an MDS survey to quantify the relative importance of each factor (26;27). MDS was developed based on random utility theory as an alternative to rating and ranking tasks by Jordan Louviere in 1987. MDS surveys ask subjects to choose the best item from a set of items originated from a master list. The combination and ordering of items differ per task. MDS was chosen for this study because it simplifies ranking tasks for the subject, is well suited for phone interviews, is able to effectively discriminate between ratings of different factors involved in complex decisions and is not influenced by scale-related biases.

We used Sawtooth Software’s SSI Web platform (version 6.6) to design and administer the survey and to analyze the data. Participants were asked to answer 24 choice sets each composed of a set of five factors from the master list of 58 items. A representative example of a MDS question including the introduction text is provided in Figure 1.

Figure 1.

a Introduction text questionnaire

b Example of a MDS question

* Combination of 5 items was different for every MDS question

In order to create an efficient design, the software took into account the following principles:

1. Orthogonality: Every item was shown approximately an equal number of times and each item is paired with each other item an equal number of times.

2. Minimal overlap: The number of times each item appears within the same set across the entire design is reduced to a minimum.

3. Positional balance: Each item appears approximately an equal number of times in each position to avoid order bias.

4. Connectivity: Each item is directly or indirectly linked to another item.

5. Stability: 300 different versions of the questionnaire were used to increase the variation in the way items are combined in order to reduce potential context bias.

We also measured the following physician characteristics: age, gender, number of years of experience, working setting (general, teaching, or university), how they spend the majority of their time (patient care, clinical research, administrative duties), and patient characteristics: age, gender, highest education level, disease duration, patient global assessment of disease activity and pain intensity, both using an 11-point numerical rating scale (NRS), and currently prescribed DMARDs.

Analyses

Preference data were analyzed using Hierarchical Bayes (HB) modeling (28;29). First, crude importance scores per item are determined for each individual under the logit rule: the probability of choosing the i^th item as most important from a set containing i through j items is equal to Pi=e^ui/sum(e^ij). The estimates are updated by an iterative process. For each iteration, an estimate is made for each item, conditional on current estimates of the other iterations. The process then converges and generates raw scores which can be interpreted as interval data. The scores are then rescaled to a 0 to 100 on a ratio scale and importance scores sum to 100.

We calculated the mean relative importance scores for each item and subsequently compared the scores between rheumatologists and patients. In addition, 20 independent sample t-tests were performed to compare the top ten of importance scores between rheumatologists and RA patients. We calculated the absolute difference between the importance scores as well as the relative importance differences by dividing the mean RIS scores of rheumatologists by the mean RIS score of patient to obtain a ratio. Because of multiple testing, a conservative p-value of < 0.001 (0.05/20=0.003) was considered as a statistical significant difference.

RESULTS

Participants

One hundred-six of the 258 approached rheumatologists (41%) agreed to participate; 35 completed the survey during the Dutch Rheumatology Congress and 71 completed the survey over the Internet. 279 patients were willing to be contacted of whom 16 were not reached by phone after several attempts and 50 patients refused to participate, leaving 213 (76%) patient participants. Participants’ characteristics are described in Table 1. The study design did not permit data collection describing non-participants.

Table 1.

Characteristics of patients and rheumatologists

Characteristics patients (n=213)	Total Number (n=213)
Mean age (sd)	60.0 (11.6)
Median disease duration (IQ-range)	7.0 (12.5)
Female (%)	69.5
Highest educational degree (%)
Primary school	11.3
High school	68.1
College/University	20.7
Current medication use (%)1
MTX	63.4
Hydroxychloroquine	18.8
Leflunomide	9.4
Adalimumab	10.8
Sulphasalazine	11.7
Etanercept	6.6
Rituximab	3.3
Number of DMARD (%)
1 DMARD	63.4
2 DMARDs	31.1
3 DMARDs	1.4
4 DMARDs	0.5%
None	3.8
Mean disease activity score over the last week on NRS2 (sd)	5.3 (2.6)
Mean pain score over the past week on NRS* (sd)	5.2 (2.8)
Characteristics rheumatologists (n=106)	Total Number (n=106)
Mean age (sd)	47.2 (9.3)
Female (%)	49.1
Type hospital (%)
General	37.7
Teaching2	34.9
University	27.4
Majority of time spending (%)
Patient care	92.5
Clinical research	6.6
Administrative duties	0.9
Number of years of experience	15.2 (9.4)

¹Patients could give multiple answers

²these items were assessed using a numerical rating scale (NRS)

³Teaching hospital: a hospital in which doctors are educated to become rheumatologist

Patients’ and rheumatologists’ relative importance scores

The mean relative importance scores for rheumatologists’ and patients’ are presented in Table 2 and Table 3, respectively. Rheumatologists’ decision-making was most strongly influenced by objective measures of disease activity such as the number of swollen joints (mean RIS: 5.24; sd: 0.39) and the DAS28 (mean RIS 5.19; sd: 0.54). Physicians’ global assessment, patients’ functional status, poor prognostic features and patients’ willingness to escalate care were also considered as important influences (Table 2).

Table 2.

The top ten of reasons that are considered by rheumatologists in case of active disease-activity

	Mean Relative Importance Score (sd)		Ranking
Most important reasons to change for rheumatologists	Rheumatologists	Patients	Rheumatologists	Patients
Swollen joints	5.24 (0.39)	2.57 (1.63)	1	12
DAS28	5.19 (0.54)	2.40 (1.56)	2	17
Rheumatologist’s impression of overall disease activity	5.17 (0.55)	3.03 (1.61)	3	8
Worsening erosions past year	5.15 (0.53)	2.12 (1.62)	4	27
Disease activity now compared to 3 months ago	5.12 (0.60)	2.37 (1.69)	5	19
Risk factors for more severe RA	4.58 (1.04)	2.17 (1.49)	6	22
Physical functioning and mobility	4.38 (0.97)	4.30 (1.07)	7	1
Presence erosions most recent X-rays	4.21 (1.44)	2.17 (1.59)	8	23
Worsening of deformities last year	3.86 (1.19)	1.88 (1.63)	9	30
Patient’s willingness to change DMARDs	3.61 (1.24)	1.13 (1.21)	10	37

sd=standard deviation

Table 3.

The top ten of reasons that are considered by patients with RA in case of active disease-activity

	Mean Relative Importance Score (sd)		Ranking
Most important reasons to change for patients	Rheumatologists	Patients	Patients	Rheumatologist
Physical functioning and mobility	4.38 (0.97)	4.30 (1.07)	1	7
Patient’s motivation to get better	1.93 (1.60)	3.55 (1.41)	2	23
Patient’s trust in their physician	0.37 (0.56)	3.46 (1.63)	3	45
Patient’s satisfaction with current DMARDs	2.28 (1.40)	3.41 (1.37)	4	21
Painful joints	2.98 (1.56)	3.37 (1.38)	5	13
Rheumatologist’s opinion to change DMARDs	3.25 (1.36)	3.13 (1.53)	6	11
Patient’s general health	3.12 (1.38)	3.12 (1.56)	7	12
Rheumatologists’ impression of overall disease activity	5.17 (0.55)	3.03 (1.61)	8	3
Patient’s level of comfort in expressing concerns	0.13 (0.16)	2.69 (1.58)	9	54
Presence of generalized bodily pain	0.31 (0.57)	2.67 (1.51)	10	46

sd= standard deviation

Patients were also strongly influence by physicians’ global assessment and function (Table 3). In addition to patient reported outcomes, the following factors were highly valued by patients: motivation to get better (mean RIS3.55; sd: 1.41), current satisfaction with DMARDs (mean RIS: 3.41; sd: 1.37) and factors related to the patient-physician relationship, such as trust (mean RIS: 3.46; sd: 1.63) and level comfort in expressing concerns (mean RIS: 2.69; sd: 1.58).

Differences between patients and rheumatologists

The ten factors with the largest absolute mean differences between physicians and patients are illustrated in Table 4. It shows for example that “trust in the physician” (mean difference: 3.09; 95% CI: 2.84-3.33)) and “the level of comfort in expressing concerns” (mean difference: 2.56; 95% CI: 2.34-2.77) was far more important for patients compared to rheumatologists. In contrast, prognostic features for RA and objective disease activity features, such a the worsening of erosions over past year (mean difference: 3.03; 95% CI: 2.79-3.27) and swollen joint count (mean difference 2.67;95% CI: 2.43-2.90), were more important to rheumatologists than to patients.

Table 4.

Ten most important differences in importance between rheumatologists and patients

			Ranking
Reasons to change	Differences in mean scores (95%CI)	Ratio Rheumatologists / Patients	Rheumatologists	Patients
Patient’s trust in the physician	−3.09 (−3.33;−2.84)	0.11	45	3
Worsening erosions past year	3.03 (2.79;3.27)	2.43	4	27
DAS28	2.79 (2.56;3.03)	2.16	2	17
Disease activity now compared to 3 months ago	2.75 (2.50;3.01)	2.16	5	19
Swollen joints	2.67 (2.43-2.90)	2.04	1	12
Patient’s level of comfort in expressing concerns	− 2.56 (−2.34;−2.77)	0.05	54	9
Patient’s willingness to change DMARDs	2.48 (2.19;2.76)	3.19	10	37
Risk factors for more severe RA	2.41 (2.13;2.70)	2.11	6	22
Presence of generalized bodily pain	− 2.36 (−2.13;2.60)	0.12	46	10

^*Negative values indicate that this factor was more important according to patients

Only current level of physical function and mobility and physician impression of overall disease active were included among both patients’ and physicians’ top ten factors, and the latter was rated significantly higher for physicians than patients (mean difference: 2.14; 95% CI: 1.91-2.39). Four of the rheumatologists’ top five items were at least twice as important for rheumatologists compared to patients.

The relative importance scores for all items from the rheumatologist’s and patient’s point of view are presented in Figure 2. Further details about the differences in importance of factors are provided in the Appendix. Convenience features, such as the need for regular blood tests were far more important for patients than for rheumatologists [14.3 times more important (1/0.07)] in their decision whether or not to escalate care. The quality of the patient-rheumatologist relationship was one of the most important items for patients and was far less important for rheumatologists [6.4 times more important (1/0.16)]. From the rheumatologists’ perspective, a patient’s history of cancer was eight times more important for them compared to patients. However, it should be noted that history of cancer was not ranked highly by either group. It was the 54^th item in the patient’s ranking list and the 22^nd item in the rheumatologist ranking list.

Figure 2.

Mean relative importance scores for rheumatologists and patients

DISCUSSION

In this study we found that rheumatologists and patients differ significantly in how they approach treatment decisions for patients with active disease as indicated by a DAS28 score of 3.2 or greater. Rheumatologists are most strongly influenced by objective markers of disease activity, physical functioning and mobility, prognostic markers and patient willingness to change DMARDs. In contrast, patients are most strongly influenced by physical functioning and mobility, coping skills, as well as features reflective of the patient-physician relationship including trust in their rheumatologist, comfort in expressing concerns as well as the rheumatologist’s opinion. There was no overlap between rheumatologists’ and patients’ top ten-ranked factors except for physical function and physician global assessment of disease activity.

A number of studies have sought to understand how patients and rheumatologists make treatment decisions in RA (5;13-21). These studies were based on specific hypotheses to evaluate a limited set of factors. To the best of our knowledge, this is the first study to quantify the importance of a comprehensive set of factors that influence treatment decisions in RA, and to directly compare patients’ and rheumatologists’ viewpoints. Despite the differences in the methods used across studies, the results described in this paper are supported by related findings in the literature (5;13-18;21;30). Two previous surveys among rheumatologists found that “a decrease in disease activity compared to previous visit” and “patient refuses changes in treatment” were important reasons for not changing treatment (5;13). Another survey noted the influence of physicians’ recommendations (16). It showed that a majority of patients did want to follow physicians’ suggestions regarding treatment (71.5%). Martin et al. found that patient’s trust in their physician increased patient’s confidence in treatment decisions (16;30). However, unlike previous reports, the present study included the breadth of factors influencing both patients’ and physicians’ decision making and required subjects to make trade-offs between the competing factors resulting in quantification of preferences and a rank ordered list.

The finding that patients and physicians differ substantially in the information they use to decide whether or not to change medications may help explain why treatment is frequently not modified according to clinical guidelines. For example, if a rheumatologist explains how DMARD escalation may help prevent further joint damage but the patient does not endorse this factor as a reason to escalate care, the patient may be reluctant to change treatment. Moreover, it is important to note the significance that patients place on the quality of the patient-physician relationship. This finding suggests that patients’ trust and comfort with their rheumatologist is a necessary, albeit not sufficient, factor for them to consider adopting a change in treatment.

This study has several limitations. Our MDS survey included 58 items (27;31). Since 100 points were divided over 58 items, the range of importance scores was limited, and a specific cut-off separating reasons that are not important cannot be determined. Secondly, as with other studies involving self-reported data, the stated importance of reasons in treatment considerations cannot be assumed to accurately reflect decision making in clinical practice. This study was performed among a sample of Dutch rheumatologists and Dutch RA patients which may limit the generalizability of our results. Given that the DAS was developed in the Netherlands, it is possible for example that rheumatologists and patients participating in this study might be more strongly influenced by the DAS score. In addition, we were unable to evaluate the impact of previous DMARD use on participants’ decision-making. The rheumatologists’ participation rate can be judged as moderate compared to other surveys among rheumatologists (17;21;31). Although we could not recruit a random sample of patients, we attempted to limit selection bias by inviting consecutive patients to participate.

In conclusion, we found that patients and physicians approach treatment-related decisions in RA differently. These differences might explain why guidelines to achieve and maintain tight control of disease activity are frequently not adhered to. Our findings indicate that the quality of the therapeutic relationship is very important to patients indicating that effective communication between physicians and patients is requisite for improving the process of decision making, and ultimately of clinical outcomes, in patients with active RA. Further research is needed to determine if decision aids can help patients and physicians better to communicate about the decision to escalate care in order to achieve and maintain tight control in RA (32).