We conducted an evaluation study of an interactive, transactional, and evidence-based library of decision aids in primary care physicians and examined the underlying steps of the SDM process, and the influence of prior experience with decision aids.
The subjective duration of consultations was not significantly associated with the subjective appraisal of how detailed the steps of the SDM process were discussed. In 8.9% of consultations physicians said they were unacceptably extended and in 90.6% of consultations, physicians said that a decision could be made. A shared decision was perceived by physicians in 57.1% of consultations. Physicians perceived that a decision was more likely to be made when therapeutic options were discussed “detailed” and that a decision was less likely when therapeutic options were discussed “very detailed”.
Prior experience with the precursor of arriba-lib was not a critical variable within our sample of primary care physicians. We did not find significant associations between prior experience and detailedness of the discussion of the SDM steps, selection of arriba-lib modules, subjective duration of consultations, physicians´ indications that a decision could be made, or physicians´ indications of who had made the decision.
Our study has several limitations. In our evaluation study we had no control group so that we cannot compare our results to the situation of usual care. It is likely that the participating physicians did not consistently perform consecutive patient recruitment. This might have led to a positive selection of patients who were already favourably disposed to SDM. This positive selection bias concerning SDM might also hold true for the participating physicians because just 32% of eligible physicians took part in the study. The wording of our four point scale (not at all”, “hardly”, “detailed”, “very detailed”) might have been problematic. The most common choice might be located between “hardly” and “detailed”. Results of statistical analyses within an evaluation study should always be treated with caution and should be regarded as preliminary [30
]. As we did 10 different analyses, the adjusted significance level would be α=.05/10
.005 according to the Bonferroni method. The α level of our only significant result (p
.01) slightly exceeded this level after correcting for multiple testing.
There is a limited database in Germany showing how physicians perceive the concept of SDM and whether they have the necessary basic communicative skills for SDM [33
]. A telephone survey of 502 physicians and 1512 German citizens revealed that 67% of physicians preferred a shared decision. There were no differences regarding gender or speciality, but younger physicians were more likely to favour SDM. In their nationally representative sample of U.S. physicians, Murray et al. found that three quarters preferred shared decision making with their patients [34
]. In our sample, physicians say in about 75% of consultations based on shared decision making that they were extended in an acceptable time frame. This finding is also corroborated by Nannenga et al. [35
]. There seems to be a threshold in physicians´ perceptions when a decision can be made. This is supported by our finding that, according to our physicians, a decision is more likely when therapeutic options are discussed “detailed” and it is less likely when therapeutic options are discussed “very detailed”. A very detailed discussion of therapeutic options might lead to an information overload in patients and therefore exceed a threshold of indecision. It may further indicate that physicians and/or patients feel uncertain and need more time to discuss possible options which may not necessarily result in a decision. The analysis of log files, which is presented in detail in another publication [36
], showed that the average consultation time was 8 minutes. In Germany, primary care physicians are mainly paid for patient contacts of 10 minutes. Therefore, the use of decision aids did not extend the average consultation time. We found discrepancies between these subjective appraisals of the detailedness of shared decision making steps and the log data, which represents user interactions with our electronic library of decision aids. It was possible to record the time that was spent with a certain option within the modules (e.g. emoticons) and we were therefore able to calculate the proportion of consultation time spent with specific features. In the cardiovascular prevention module, 35 of 122 consultations (28.7%) spent 100% of consultation time in the history part of the programme, which includes risk presentation. These consultations were shorter than average. In the other modules with weigh scales, 15 of 62 consultations (24.2%) spent 100% of consultation time in the history part; 11 of these consultations used the oral antidiabetics module. Again, these consultations were shorter than average. In contrast, all of the physicians indicated in their subjective appraisals of the detailedness of shared decision making steps that therapeutic options were discussed. In these consultations, physicians obviously discussed therapeutic options with their patients without using the respective modules which points to a reduced fidelity in this point [5
]. They might have preferred different ways of discussing them, or they did not agree with the evidence-based options presented in the modules which are sometimes in opposition to German guidelines [6
Having no prior experience with a decision aid was not an implementation barrier in our study. In their updated systematic review, Légaré et al. found time constraints, patient characteristics, and the clinical situation to be the most often reported barriers for the implementation of shared decision making [37
]. The most often reported facilitators were provider motivation, positive impact on the clinical process, and patient outcomes. Studies do not show that shared decision making necessarily requires more time [35
]. We were also able to show that the length of most of the consultations was acceptable. A suspected negative impact on the doctor-patient relationship, a perceived disregard of professional status, and a possible threat to professional autonomy are implementation barriers discussed by Kaplan [17
]. The results of the qualitative study of Watson et al. reveal that in order to implement decision aids in primary care a challenge might be the reconfiguration of the physician´s role in the physician-patient relationship. The reordering of power within this relationship might require more support than just training in implementation strategies [41
Concerns about the comprehensiveness and up-to-dateness of decision aids might be another potential barrier for implementation [1
]. There was a close cooperation between developers of patient decision aids and medical experts in the process of designing arriba-lib. Ease of use and a balanced presentation of evidence-based information emphasizing the freedom to choose might have resulted in our observation that prior experience with decision aids was not a critical variable in the implementation of an electronic library of decision aids. These conclusions are supported by a qualitative study on a computer delivered, theory based intervention for guideline implementation in general practice, although this is not a main goal of decision aids based on SDM. McDermott et al. found that the emerging reminders regarding guideline adherence were more likely to be accepted when physicians considered them to offer support and choice [42
]. Physicians said that information should be presented in a condensed way and in an easy-to-understand format. The information should be tailored to the individual patient and physicians demanded to be able to choose among the presented information. Information on clinical topics, although evidence-based, must be offered in a way that patients and physicians maintain the impression of having the freedom to choose. Nevertheless, some physicians in our study raised concerns against some modules, e.g. atrial fibrillation, in which they saw a discrepancy between the evidence base and guidelines. This obviously resulted in a reluctance to use them.
Colombet et al. conducted a focus group study among general practitioners on an electronic decision aid that, for example, provided personalised risk estimates on cardiovascular prevention and diagnosis of depression [43
]. Mentioned topics were the handling of the programme, the understanding of contents, and the acceptance of advice provided by the programme. It was shown that the understanding of risk information was highly variable in physicians. The authors advocate for training on the contents of the programme before feasibility testing, which we did in our study. Furthermore, the acceptance of evidence-based information for use in the decision making process should also be considered.
Evidence-based decision aids may offer support for physicians in the management of self-acquired information in patients. Baumgart describes a high ambivalence of physicians regarding the informed patient in her qualitative study [44
]. Physicians report incorrect interpretations of information acquired by patients that often need a time consuming correction. The more complex the disease and the available treatment options, the more physicians appreciate information-seeking initiatives from their patients. Some physicians see a positive challenge in interacting with informed patients who might receive a greater sense of control in dealing with their disease. Modern information-oriented societies require a change from a paternalistic physician to an expert who accompanies patients in their search for and analysis of medical information. Evidence-based decision aids can play an important role in this process.