To the best of our knowledge, this paper is among the first to describe in detail the recruitment process of FMGs, family physicians, and patients to a pilot trial of the DECISION+ program, a CME program in shared decision-making on the optimal use of antibiotics for treating ARIs in primary care practice. Moreover, the results of this trial support the feasibility and acceptability of conducting a large clustered randomised trial involving dyads of family physicians and their patients in SDM regarding the optimal use of antibiotics for ARI. This conclusion is reached even if not all predetermined standards for our criteria were always fully met. Indeed, it has been established that not reaching the preestablished criteria does not necessarily indicate unfeasibility of the trial but rather underlines changes to be made to the protocol .
Identification of FMGs and associated family physicians was a relatively simple process. The necessity for FMGs to be accredited, the relative stability of these FMGs, and the availability of a free web-based registry facilitated this process. Identification of individual family physicians would have been more difficult because available registries are rapidly outdated .
The main difficulty in recruitment appeared during the initial contact by the principal investigator with the medical director of the FMG. Twelve FMGs (57%) declined the invitation even before the face-to-face meeting. Thus, 24% of the eligible FMGs agreed to take part in the study, less than the 50% expected. We were probably too confident when targeting a 50% positive response rate from all identified FMGs. Recent studies regarding appropriate use of antibiotics reported recruitment rates for groups of providers of 56% and 37% [38, 39]. Recruitment rates were poorly reported, when reported at all, in other similar studies [10, 40], despite their importance in providing important information regarding generalisability of results. The main reason for refusal from the FMG was that participation would add unnecessary workload (nine hours of workshops) to already too busy family physicians or that the clinical topic was felt to be irrelevant for their CME activities at that time. In addition, the project was carried out at the time that some FMGs were in the implementation phase, and this reorganisation process might have contributed to the smaller than anticipated number of acceptance.
Nonetheless, it is interesting to note that out of the eight FMGs that agreed to an on-site meeting, four immediately accepted the invitation to participate in the study. This suggests that having the opportunity to clarify face-to-face the nature of the study and the time and efforts required from participants may facilitate the recruitment process when participants are interested in learning about the study. According to Dormandy and colleagues (2008), factors that appeared important in retaining practices in a study are good communication, easy collection methods, and payment of incentives as scheduled . At the time of the first contact, the issue of incentive was not discussed with the FMG. This did not seem to affect their agreement to participate or not after the meeting, as none of the FMGs reported the lack of payment as a reason for refusal. This reinforced earlier findings showing that financial incentive alone is not sufficient for a high recruitment rate [15, 17]. Moreover, the acceptance rate from family physicians from these FMGs was high. Indeed, 75% of family physicians agreed to take part in the study. It should, however, be noticed that acceptance by the FMG was probably dependant on acceptance by a large proportion of its family physicians.
Conducting studies with family physicians and their patients is challenging, mostly as a result of the competing time demands of family physicians for other clinical and administrative tasks [15, 17]. Furthermore, it is one thing to take part in a study and facilitate recruitment of patients for the research team, but to have to complete documentation for each patient after each of their encounters is a major challenge. To cope with these issues, the strategies used by this research were to minimise the burden on the family physicians by having research members take care of all study-related tasks, as done in previous work . This appears to have been very successful since the retention rate of all participants was very high. The number of patients per physician was also kept to a minimum.
As for patients, the challenge was to proceed to recruitment in the FMG waiting room, without disrupting or delaying the patients' consultation flow. Having research assistants recruiting patients minimised a potential selection bias if recruitment had been driven by family physicians and maximised the recruitment rate, as recruitment driven by physicians is usually only moderately successful .
Considering their busy schedules, we anticipated that collecting information from physicians after each clinical encounter would be challenging, and therefore, we were particularly interested in the proportion of missing postencounter questionnaires. Surprisingly, all completed the questionnaires (100% completion rate), and there were very few missing data, suggesting that it is possible to collect similar and simultaneous measures from both family physicians and their patients. Factors that might have contributed to this high rate could include the fact that postencounter questionnaires were very short (approximately two to three minutes to complete) and were included with the patient's medical file so that they could be completed at the same time as the medical note. Furthermore, completion is likely to have been enhanced by the presence of a research assistant who reminded physicians of their task.
Based on this pilot trial, the planning of the recruitment of participants for a larger trial needs to be carefully thought out. Although demanding for principal investigators, having a meeting with members of the FMGs seems to be useful, as it could have reinforced the decision to participate in this type of study. Moreover, it could have led to a more informed consent, conducive to sustained participation. Second, recruiting patients particularly in the context of ARI also needs to be carefully planned. One of the data collection periods occurred during summertime, when patients consulting for ARIs are less frequent, and thus, the time for collection period was longer than anticipated. Third, the pilot study was limited to FMGs, as they had a fee-for-service organisation, but other primary care groups might be considered for a larger trial. Family practice teaching units could be an interesting group to target, as they are usually larger in size than FMGs and have more physicians, including residents. Regardless of study type or method, successful identification, selection, and recruitment of participants, whether they are groups of physicians, individual physicians, or patients, are important issues to be considered by all intending researchers , and pilot trials provide such opportunities.
Regarding participation in the DECISION+ program, a total of 46% of all participating family physicians completed the three 3-hour workshops, for a total of nine hours over four to six months. We had made a priori acceptability criteria that a 70% participation rate would be ideal for the success of a larger-scale study. Considerations need to be taken and lessons learned. Findings from a recent systematic review on the adoption of SDM by providers revealed that participation of providers in the intervention varied between 47% and 100%, with a median of 52%, independent of the length of the intervention . Moreover, a recent study reported that a six-hour workshop for practitioners seems the ideal length of time, independent of numbers of workshops . In retrospect, inserting three workshops of three hours over a four- to six-month period was probably too demanding on the part of family physicians in FMGs who usually have other hospital, educational, and administrative duties. In light of these results, having 69% of the family physicians participating in two or more workshops (six hours or more) in this pilot trial seems acceptable. The burden associated with the numbers and lengths of the workshops were also demanding on the trainers who could not comply with all competing demands, which affected the replicability of the intervention in the control group, as reported elsewhere .
Considering all these elements, the length and duration of the workshop should be considered when planning and conducting a larger trial to increase generalisability of the intervention. Introduction of a web-based or self-learning element, reduction of the length and/or numbers of workshops for both participants and trainers, and improvement of train-the-trainers strategies for collaborators needs to be considered.
Limitations need to be acknowledged. First, the pilot study was limited to FMGs, as they had a fee-for-service organisation, but other primary care groups might be considered for a larger trial. Second, the low response rate of FMGs may limit the generalisability of the findings, although the sociodemographic characteristics of participating family physicians were similar to the characteristics provided by the provincial registry (data not shown). Third, we lacked information regarding the number of patients that presented themselves at the outpatient clinics for an ARI and the total number of patients seen by physicians during the study period. Finally, it is possible that FMGs, family physicians, and patients who participated in the study might have had higher interest or motivation in the DECISION+ program than those who did not participate in the study.
Notwithstanding these limitations, rigorous pilot trials are meritorious projects, and their results should be subjected to the same level of scrutiny and requirements as those provided to the results of full-scale trials . However, opinions vary regarding whether the results of the pilot trial should focus on the implementation processes, efficacy outcomes, or both . Publication of findings of pilot trials could certainly help to inform other researchers about methodological or practical challenges related to such studies and does constitute a meaningful contribution to the literature [27, 43, 44].