Background
Quality of decisions in health care is increasingly viewed as sensitive to individual factors such as preferences and beliefs which on the level of groups and of populations are very difficult to determine [
1,
2]. Since concrete choices reveal little about decision quality, the process of making a decision might be a better quality indicator [
3‐
5]. This process is predominantly performed internally and includes negotiating information about possible benefit and side-effects with preferences, values and risk attitudes [
6]. As therefore, quality refers to both highly individual and intrapersonal criteria, the challenge of assessment is obvious. Accordingly, an ideal decision making process would mean implying cognitive and emotional appraisal of relevant information with anticipated related consequences [
7,
8]. The latter has been defined as “informed decision” which can be considered as meeting two conditions: first, the patient should have processed the relevant risk knowledge and second, the choice should reflect the patient’s values [
6]. Based on this definition, the Multimodal Measure of Informed Choice (MMIC) [
3] has been used as an endpoint for studies on decision support. The MMIC comprises three dichotomous measures, knowledge (good, poor), value (positive, negative) and choice (uptake or non-uptake of the intervention under consideration) leading to eight types of choices Only two of these indicate informed choice: either an informed patient’s values are in favour of the intervention which is applied by the patient, or an informed patient’s values are against the intervention which is rejected. Marteau et al. define value as a basic attitude, which referring to Ajzen is a person’s overall evaluation of the behaviour in question [
9]. However, the process of decisions and the mechanisms or moderators by which attitude impacts on behaviour cannot be derived from MMIC, which merely assesses the result of this process. Other intrapersonal characteristics also influence the decision making process [
10]. Furthermore, different internal considerations – such as beliefs about the consequences of an action or the individual’s own ability to control the situation - can lead to the same choices. Frequencies of informed choice decisions measured using Marteau’s method do not explain how they were achieved.
Elaboration of health related decisions i.e. the motivational process an individual goes through when anticipating an action - the so called action regulation - are seen as internal cognitive processes by a group of theories such as the health belief model, the social–cognitive theory, the theory of reasoned action, the theory of planned behaviour (TPB) and the protection motivation theory [
11]. These theories share the assumption that each behaviour is predominantly a function of attitudes and beliefs, as well as expectations of future events and outcomes. Facing various alternatives, individuals will choose the action most likely leading to positive outcomes. Among these theories, the TPB [
9] is one of the best proven to explain a specific health behaviour by a set of domains in a wide number of research areas [
12‐
14]. Prediction of self-reported behaviour is superior to observed behaviour. However, according to the literature TPB is capable of explaining 20% of the variance in prospective measures of actual behaviour (and 27 to 39% of intention) [
15‐
18].
The TBP postulates three conceptually independent domains determining an intention to perform a specific behaviour: "attitude" refers to the degree to which a person has a favourable or unfavourable appraisal of the behaviour in question; "social norm" refers to the perceived social pressure to perform/not perform the behaviour; "perceived behaviour control" refers to the perceived ease or difficulty of performing the behaviour and is assumed to reflect past experiences as well as anticipated impediments and obstacles. The theory is based on the expectancy-value model, assuming that overall evaluation of decisional options often contains two separable sub-domains: an expected outcome and a given value [
7]. For example, the subjective value of a given outcome such as reduction of the number of relapses in multiple sclerosis as the benefit from disease modifying treatment (DMT) affects the attitude in direct proportion to the strength of the belief regarding occurrence of this outcome. A patient might be convinced of the efficacy of DMT regarding reduction of relapse rates but on the other hand might not prioritize this goal against others leading to low impact of this belief on this patient’s attitude. As a general rule, the more favourable the attitude and subjective norm with respect to a specific behaviour and the greater the perceived behavioural control the stronger is an individual’s intention to perform the behaviour under consideration. The relative weight by which the three domains (attitude, subjective social norm, perceived behavioural control) impact on the intention is expected to vary across different behaviours and contexts.
Since, although not conclusive, the evidence for the TPB’s validity to a broad variety of behavioural decisions is promising, we choose the model to validate theoretical assumptions underpinning our developments of decision support strategies for patients with multiple sclerosis [
19‐
21].
Multiple sclerosis (MS) is a chronic-progressive disease of young adults with a presumed autoimmune aetiology [
22]. People affected by the diseases have to deal with pronounced uncertainty regarding prognosis and also regarding effectiveness of available treatments. Hitherto, no curative treatment exists. However, appearance of new relapses, new lesions on magnetic resonance imaging and progression of disability can be slowed down by DMT but at the risk of side effects that can be long lasting and sometime serious [
19]. Therefore, decisions about disease modifying treatments (DMT) are highly sensitive to patient preferences [
23]. In addition, MS patients claim active roles in these decision making processes which seem even higher than in other diseases [
24,
25].
This paper describes development and validation of the questionnaire “Planned behaviour in MS” (PBMS), a patient-reported instrument assessing the process of decision making about DMT. By improving understanding of processes underlying patients’ choices on DMT as well as improving understanding of mechanisms mediating effects of decision support strategies, we aim at tailoring decision aids and other support strategies more to patients’ needs.
Discussion
This study aimed to develop a questionnaire to assess MS patients’ cognitive and emotional representation of decision making processes regarding disease modifying treatments (DMT) based on the theory of planned behaviour (TPB). In pre-testing the questionnaire had already been found to possess convincing item properties and feasibility and encouraging indications of validity. The larger dataset from the RCT yielded further support for the PBMS measurement concept in various regards.
The PBMS was found to possess high predictive power for the target criterion, intention to use DMT. This suggests that the PBMS item pool fully covered the relevant belief composites and that TPB is applicable to this type of decision. PBMS also showed a good sensitivity to differences in beliefs about DMT and the level of corresponding risk knowledge. PBMS response patterns reflected the hypothesised effects of an educational intervention providing factual information about DMT, with responses indicating that the intervention group subsequently showed a more critical attitude and less willingness to comply with social norms. These results correspond with the programme’s underlying aim of enabling patients to make decisions based on personal values and realistic beliefs about expected outcomes in an area where most information is commercially driven. It was expected that the patients’ ability to reflect on these beliefs would be strengthened by providing evidence-based patient information on DMT.
The PBMS showed substantial construct validity by discriminating subgroups with differing DMT status (5 of six domains). Beyond the absolute PBMS score values, the PBMS was also found to be sensitive to the decision support intervention at the level of model fit. The predictive power of the PBMS was higher in the intervention group than in the control group.
This effect is insightful as the basic idea of supporting patients in processing evidence based patient information is to enable patients to reflect on their own priorities when making DMT decisions [
28]. Our results showed the intervention group to possess increased cognitive awareness of the belief composites underpinning individual decisions. Cognitive awareness, however, results from deeper elaboration of scientific and internal motivational information relevant to the decision [
8]. The latter process complies with evidence based medicine and decision support strategies. These aim to enhance communication and information processing by using decision aids, evidence based patient information tools, educational programmes, and shared decision making. Earlier work by our group showed that, although it may not alter treatment choices, evidence-based patient information can alter decision processes [
21]. The results relating to possible shifts in the proportions by which single domains impact on the intention remain ambiguous in the present study. Proportional changes as found in this study could reflect the dynamic character of proportional impact of the TPB domains. However, this dynamic needs further investigation.
The study also has some limitations. First, the selection of TPB out of a pool of more than 30 psychological theories of behaviour change [
29] can be challenged. Although evidence for its applicability to various behaviours is substantial, TPB and the rigorous structure of its components (e.g. expectations multiplying values) have been criticized for ascribing too much importance to rational processes [
14,
30]. As well as rational reasoning, emotional processes and routine behaviour (lacking any decision) also affect motivation when deciding on a specific behaviour [
14]. Regardless of the current understanding that emotional processing is essential for and effectively inseparable from cognitive appraisal, we also think that emotional appraisal is adequately represented within each of the sub-domains of our tool (Additional file
1). For instance, by formulating items such as “
The risk I would be taking by putting off immunotherapy for too long frightens me.” we intended to address emotional aspects of MS-patients’ belief system. On the other hand, the theory’s emphasis on rationality was helpful in focusing on the process of systematic and conscious reflection on the motivation in medical decision making. Our study results confirm our rationale in this regard.
The questionnaire’s additive and multiplicative structure can be challenged for two further reasons. First, the extent of possible hidden redundancy in the items’ underpinning constructs is not yet clear. Second, since our multiple logistic regression analyses were limited to main effects, we cannot exclude interaction effects, e. g “knowledge of risks” may interact with “compliance with norms”. Both would contradict the rigid algorithm to achieve the component scores and the intention / behaviour estimate. However, automatic detection of interactions would have required far greater sample sizes and clear a priori hypotheses for specific interactions. For this initial study we therefore considered it reasonable to adhere to the strategy outlined by Ajzen [
9,
26]. The conversion of an intention into action is determined not only by the person’s intention but also by personal and environmental barriers and by the person’s volitional control. In this regard, our results have to be seen as preliminary. However, we are awaiting data on DMT uptake from the PEPADIP trial which will enable us to validate the PBMS using the actual target behaviour rather than a substitute. It can be argued that some of our results could emerge by chance due to multiple testing. Indeed, the study was not sufficiently powered to keep all reported differences when using an alpha level adjusted to a number of six tests conducted in parallel. However, our testing was driven by clear hypotheses referring to the complete model. Moreover, as these hypotheses refer to a given fixed hierarchical structure, multiplicity of the used parameters is limited. Similarity of the results of pre-test and controlled trial provides further support for their relevance. Finally, it was not possible to employ a test-retest strategy to estimate PBMS reliability because the control group showed considerable change in some of the domains between pre- and post-intervention. However, three and six months follow up data are now available for the PBMS. Here, no further change was detected, either for the control group or for the total sample, and test-retest reliabilities were satisfactory for all six domains (.74; .71; .79, 79; .60; .71) based on 155 data sets. Further studies will be necessary to replicate these results.
Modeling behavioural decisions in terms of TPB conforms to basic assumptions of patient empowerment, since it takes into account the social situation (subjective social norm) and self-efficacy (control beliefs) as well as attitude. Substantial evidence has shown the TPB to be helpful for developing interventions motivating change of specific health behaviours such as condom use, exercise, diet, and medication adherence [
11,
18]. Our study is in agreement with a number of other studies that have used TPB to elucidate inherent processes in individuals making health related decisions [
12‐
14,
30]. When developing and evaluating the effectiveness of theory based interventions, it is useful to be able to identify specific underlying beliefs and measure their impact on informed decision making. This is particularly important, as better decisions, in terms of shared decision making or the informed choice cannot (by definition) result in immediate health improvement. Our study shows how theory can be used to inform the design of an effective intervention and to guide its evaluation. However, since the TPB itself is generic and its domains are quite elementary, we feel that it has great potential for developing specific applications for other decision support contexts as well. This would answer the call for more theoretical foundation for decision support strategies [
31,
32]. Using the tool will help us to further develop end-points for evaluation of patient empowerment and shared-decision making, for example by making measures such as the Multimodal Measure of Informed Choice more specific [
3].
Competing interests
CH has received grants from Biogen-Idec, Merck-Serono, Novartis Pharma and Teva Pharma as well as speaker’s fees. AS has received a board membership fee from Novartis and speaker’s fees from Sanofi-Aventis. SK is supported by a rehab-fellowship grant from the National MS Society, USA. JK and KF have received travel expenses from Merck Serono. IB has no conflict of interest.
Authors’ contributions
Conceived and designed the experiments: JK, SK, CH, KF,NS. Performed the experiments: SK, KF, IB. Analysed the data: JK. Contributed reagents/materials/analysis tools: SK, IB, KF, NS,AS. Wrote the manuscript: JK, SK, CH, AS. All authors read and approved the final manuscript.