Introduction

Subjects with a spinal cord injury (SCI) at cervical level have impaired upper extremity (UE) function. Therapy for the UE is most important to achieve maximal independence in activities of daily living. Therapy for the UE of subjects with tetraplegia can be divided into three stages: the acute stage, the sub-acute stage and the reconstructive stage.1 In the latter stage, when no further neurological or functional improvement is expected, reconstructive interventions can be considered in carefully selected subjects with tetraplegia to improve their UE function and skills. The available reconstructive interventions are reconstructive surgery with or without the implantation of functional electrical stimulation (FES) devices.2, 3, 4 Since the introduction of FES implants at the end of the last century further development and application of the devices ceased because the manufacturer stopped production. Although initially good results were achieved,3 the device gained only limited popularity. This underlines that the decision whether or not to have reconstructive interventions is complex. Clinicians are often focused on the potential effect of treatment on daily functioning, while patients also consider other factors like burden of treatment and health service delivery in their decision.5

An understanding of the importance of both health outcomes as well as non-health outcomes for the patient is crucial in establishing shared decision making between patients and clinicians.6 Although many preference elicitation methods exist in decision science, methods based on explicit trade offs between positive and negative aspects of treatment are preferred to opinion-based techniques.5

In the present study, one such method, namely conjoint analysis (CA), was used for a systematic and general assessment of the importance of non-health-related factors and health outcomes on the preference of patients with a cervical SCI for reconstructive UE interventions.

Methods

Conjoint analysis

Traditionally CA has been a collective term, covering both the theory and methods of a variety of different approaches that can be used to design experiments and analyse the individual response data derived with these experiments. During the 1990s, CA was introduced in the elicitation of patient preferences in health care. An extended overview of the method, underlying theory and the application of CA in health-care research have been discussed previously.5, 7, 8, 9

A CA comprises of five stages, including (1) identification of typical elements, also called attributes, of the therapy which is investigated; (2) assignment of various levels to the attributes; (3) design of a series of hypothetical treatment scenarios using combinations of attribute levels; (4) establishing response tasks in which the treatment scenarios are judged by respondents; (5) analysis of the response data. The stages of the present study are described in the following section.

Identifying attributes and levels

A draft version of the attributes and levels in the present study was defined by a local rehabilitation expert team and based on the literature and clinical experience. Attributes must be relevant, comprehensive and tradable. A balance must be found between the number of attributes necessary to describe treatments and the feasibility of the decision task. Although FES was not an actual treatment option at the time of this study, it was included to assess the desirability of implanting medical devices on the preference for treatment. Some of the ‘none options’ formulated in the levels, for example, for risk of complications, may be clinically unrealistic, yet they are methodologically important to access trading between attributes and levels. The draft version of attributes was adapted after being reviewed by five (international) experts. After a pilot test in six SCI subjects, a final version was applied in the present study (Table 1).

Table 1 Attributes and levels
Full size table

The CA response task

A choice-based CA software package by Sawtooth Software (Sawtooth Software Inc., Sequim, WA, USA) was used to design the treatment scenarios. Seventeen sets of two scenarios were randomly composed for each subject using the minimal overlap method. With this type of CA, the occurrence of equal levels for one attribute in both treatment scenarios included in the response set is limited to a minimum. The choice-based CA also allowed calculation of main effects of attributes and levels at group level. Subjects were asked to select the scenario in each set, which they would prefer if they were considering UE reconstructive therapy. For an example of a choice set see Figure 1.

Figure 1
figure 1

Example of scenario set.

Full size image

Internal validity and consistency

Three fixed choice sets were included in the response task. The first composed a positive scenario with all levels set at minimum opposed to a negative scenario with all levels set at maximum. The second and third presented identical scenarios in a reversed order (the mirror set). Subjects were expected to choose the positive scenario as well as the same scenario twice in the mirror sets. Inconsistent subjects were excluded from final analysis.

Research setting and patient selection

Potential subjects were selected from the outpatient records of seven rehabilitation centres in the Netherlands, specialized in the treatment of SCIs. We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during the course of this research. The inclusion criteria were a motor complete C5, C6 or C7 SCI, according to the guidelines of the American Spinal Injury Association (ASIA),10 with at least one arm classified as motor group 1–4 according to the International Classification for Surgery of the Upper Limb in Tetraplegia.11 The motor groups are listed in Table 2. This classification also determines sensory function of the hand, which is important for actual planning of surgery in individual patients but was not considered as an exclusion criterion for expressing preference for treatment. Subjects had to be medically and neurologically stable, at least 1 year after the initial injury, and they should be potential candidates for surgical reconstruction of elbow extension and palmar and/or lateral grasp function based on their motor function. Subjects were excluded if they previously had surgery to improve UE function, or if they profusely had declined reconstructive treatment in the past. After informed consent, one of the authors (JvT) visited the subjects at home. Gender, age, time since injury, and SCI and UE classification according to the guidelines were established. General and standardized information about UE reconstructive surgery was provided with a PowerPoint presentation and oral explanation. The information was not tailored to the actual clinical status of the patients but described surgical possibilities in a general manner. Information, based on actual state of the art,2, 4 was given about following topics:

  • The possibility to improve active elbow extension and the importance or even prerequisite of elbow extension for proper use of improved hand function;

  • the possibility to improve key and palmar grasp;

  • the need for and extent of preoperative screening;

  • the surgical procedures necessary for treatment;

  • The need for postoperative splinting and in-patient and outpatient rehabilitation;

  • The risks of postoperative bleeding, infections and adhesions, which could have a negative effect on functional results, were mentioned in general terms as possible complications.

Table 2 Motor groups according to the International Classification for Surgery of the Upper Limb in Tetraplegia
Full size table

After finalization of the CA response task, questions of patients about individual possibilities for UE surgery were answered and they were asked if they would consider such an intervention in the near future.

Data analysis

Calculation showed that data of 50 subjects were required to obtain sufficient statistical power. Multinominal logit regression analysis was used to calculate regression coefficients for each level of all the attributes.

Attributes where the difference, tested with a t-test (P<0.05), between maximum and minimum regression coefficient was significant contributed to the overall decision for treatment.

The relative importance of the attributes, which represents the influence of an attribute compared to the other attributes, and the relative preference for each level within an attribute compared to the other levels of the same attribute were also calculated.

The hierarchical Bayes analysis module of the Sawtooth Software package was used to calculate the relative importance of the attributes for each individual subject. An arbitrary cutoff of 30% was used to determine whether a subject had a dominant preference for one attribute in the decision for treatment (relative preference of 0.30 or above for any of the including attributes).

Results

A total of 57 subjects were selected, four could not be reached or declined from participation. Characteristics of the 53 participants are presented in Table 3. Four subjects were considered inconsistent in their response and omitted from the CA analysis.

Table 3 Demographic data of the study population
Full size table

The results of the analysis are shown in Table 4. On a group level, all attributes significantly contributed to the decision to undergo reconstructive intervention therapy (P<0.01 for all attributes). Figure 2 shows the relative importance of the attributes and their 95% confidence intervals. The in-patient rehabilitation period, the type of intervention, the number of operations, the time spent in cast and the risk of complications are either more important or of the same order of importance as functional outcome at hand level. Only the outpatient rehabilitation period is deemed less important than functional outcome in the choice for treatment. In the individual analysis, 29 of the 49 (59.2%) subjects included considered all attributes important in the decision for treatment. Implantation of an FES device was absolutely rejected by 8.2% (these subjects never choose a scenario that included the implantation of an FES device). Dominant preferences were as follows: intervention type 12.2%, number of operations 6.2%, in-patient rehabilitation time 8.2% and results hand 6.2%.

Table 4 Results of conjoint analysis
Full size table
Figure 2
figure 2

Relative importance and 95% confidence intervals of attributes.

Full size image

Figure 3 shows the preferences for the levels within each attribute. In general, the distribution over the levels is as expected with the best or most attractive levels having the highest preference. Of the patients interviewed, 29 (52%) indicated to be interested in reconstructive surgery.

Figure 3
figure 3

Proportion or relative attractiveness of the levels (for a description of the levels see Table 1).

Full size image

Discussion

Technical innovations in health care are developing rapidly, and also spinal cord medicine is confronted with many new opportunities.12, 13 In the development, application and evaluation of new techniques, the opinion of the public and/or patients is increasingly considered to be very important.9 This is emphasized by the fact that studies show differences in clinician and consumer preferences with respect to health outcome and non-health outcome factors of therapeutic interventions.14 The assessment of acceptance of new technology by patients should therefore not be based on functional outcome alone.

The present study focused on the decision on different types of reconstructive UE intervention in tetraplegia. All attributes investigated in the present study contributed significantly to the preference for reconstructive UE intervention and should be taken into account when subjects are informed about these therapies. Attributes related to the intensity or ‘burden’ of therapy are equally important, or even more important than functional outcome characteristics in deciding about therapy. However, a large interindividual variation in the relative importance of the attributes was found and over 40% of the subjects focused on one specific treatment characteristic. Clinicians should be aware of the possibility that in clinical situations, although all aspects of treatment are discussed as is standard in good clinical practice, some patients do not consider all relevant information. For instance, although scientifically proven and favoured by professionals as a very rewarding procedure,15 the potential benefits of active elbow extension seem to be underestimated by a considerable number of potential recipients of surgical UE reconstruction in this study.

Tendon transfer alone is preferred to tendon transfer combined with electrical implants. However, only 8.2% of the subjects absolutely rejected the implantation of devices. Although the studies by Gorman et al.16 and Rushton et al.17 showed that only a minority of potential candidates for an FES device actually choose for implantation, this might not be a result of the implantation of foreign bodies, but rather because of the combination of uncertain potential benefits and treatment burden.

Subjects seems to prefer one operation over two operations but no preference is expressed for the location at which the postoperative period, while the operated arm is immobilized in a cast, is spent. This might be explained by the uncertainty that patients face as a result of a temporary increased dependency. The length of outpatient treatment was found to be relatively unimportant, except for the longest period of 12 weeks, which was highly unappreciated.

An appropriate choice of attributes and levels is crucial in a CA study. The attributes and levels included in this study were based on the literature and reviewed by international experts. Additionally, a selection procedure of attributes was necessary to keep the response task manageable for patients. As a consequence, conclusions can only be drawn in relation to the attributes included in this study. For instance, a more active ‘flexor-tendon-injury-type’ rehabilitation protocol, which is recently brought into discussion,18 cannot be evaluated based on the results of this study.

Obviously, the subjects in the study population represent a range of UE functions and functional capabilities. However, the main objectives of surgical intervention for patients with an arm classification of M1 to M4 are active elbow extension and restoration of one or two grasp functions.2, 4 In this respect, the sample is adequate for the purpose of this study, and the subjects should be capable of weighting the importance of the selected attributes in decision making for surgery on a general level.

The present study focused on a group of subjects in a stable condition who had their SCI for a longer period of time. A total of 29 subjects (52%) showed interest in reconstructive interventions for themselves. However, in clinical practice only a minority of suitable subjects actually decide to be operated. Research has shown that, in determining life satisfaction in patients with long lasting SCI, psychological and social factors are more important than the severity of the injury.19 In the initial rehabilitation period, however, self-care goals are frequently considered to be most important by the patients.20 Improvement of UE function can contribute to achieving these goals. It would be interesting to see if preferences for therapy and the importance of treatment characteristics change over time and differ in patients with a recent SCI. This needs to be investigated, as the results could be beneficial in the discussion about the timing of the intervention.

In conclusion, this study demonstrates that it is important to consider both health and non-health factors in the decision for treatment of the UE function in patients with tetraplegia. With regard to reconstructive UE interventions, subjects with tetraplegia attach equal importance to non-health outcome factors, indicating the intensity or burden of the treatment as to improvement on hand or elbow function in determining the preference for treatment.

The provision of state of the art information and an individual approach are essential in considering the planning of surgical therapy and postoperative rehabilitation. Clinicians should be aware of the various elements, which determine the motivation of patients and the fact that a considerable number of patients are focused on limited aspects of treatment.

The present study emphasizes the importance of research into the factors that influence the motivation of patients to choose a treatment. Results of such studies can be used to guide further development and practical application of new technology for further improving of the outlook of patients with an SCI.