We used a mixed methods convergent research design to triangulate results obtained from quantitative and qualitative methods [50]. Quantitative analysis involved a quasi-experimental before-and-after study design to evaluate the effect of an online CPD course that is part of the CPD program of the Fédération des médecins spécialistes du Québec (FMSQ). In this design, measurements before the CPD course served as comparators for the same measurements taken after the CPD course for the same participants [50]. We analyzed responses to open questions with a qualitative thematic deductive approach to understand why participants considered integrating the targeted behavior into their professional practice [50]. For both quantitative and qualitative analyses, we used Godin’s integrated model for predicting health professionals’ behavior as a conceptual framework [51] and we used Michie’s conceptual framework on behavior change techniques (BCTs) to explore our qualitative data in detail [52]. As shown in Figure 1 [52,53], these frameworks share the concept that behavioral determinants lead to behavior adoption [51-53]. Both originate from Ajzen’s Theory of Planned Behavior and have many behavioral determinants in common [51-53]. Michie et al [52] propose that BCTs can influence and target behavioral determinants, which in turn translate into behavior change. Possible behavioral change techniques include planning, persuasive communication and others [52]. Details on these 2 conceptual frameworks are available in Multimedia Appendix 1 [51-54]. We reported results following the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guideline for quantitative results (Multimedia Appendix 2) [55] and used the SRQR (Standards for Reporting Qualitative Research) for qualitative results (Multimedia Appendix 3) [56].

The study took place in Quebec, Canada, from March 1, 2022, to June 20, 2024. Targeted participants were mainly specialists from the FMSQ membership. However, general practitioners who subscribe to the FMSQ’s online learning platform MÉDUSE could also participate in the study. The study population was physicians who registered for the course and completed an online consent form. After registering, participants were informed about the study by a system-generated message displayed at the beginning of the online course. The CPD course was not mandatory but was recognized by the Royal College of Physicians and Surgeons of Canada as a 1-hour group learning credit (Section 1) [57]. This convenience sampling was based on time constraints and access to MÉDUSE. Inclusion criteria were being a licensed physician in Quebec and participating in the online CPD course. There were no exclusion criteria.

We describe the intervention according to the Template for Intervention Description and Replication (TIDieR) guidelines [58]. The intervention was a 1-hour asynchronous online CPD course developed by the FMSQ to help physicians approach colleagues in difficulty (eg, burnout, psychological distress, second victim symptoms). Thus, this course focused on a form of peer support. The term “colleague” included anyone in the health and social services sector, including administrative staff. The scientific committee comprised 3 physicians, 2 of whom had peer support experience through their work with the Québec Physicians’ Health Program (QPHP). The QPHP organization also helped with course design. A scenario-based video, filmed with professional actors, illustrates the QPHP’s 5-Step Caring Approach: (1) becoming aware of context and stressors, (2) recognizing the warning signs, (3) initiating a conversation with the colleague, (4) providing compassionate support, and (5) respecting your own limits and practicing self-care [59]. Approaching a colleague in difficulty corresponds to Tier 1 of the Scott Three-Tier Model of Support, which includes immediate team members of a health care worker experiencing distress [21].

A CPD expert with more than 10 years of experience (MT) used Godin’s integrated model for predicting health professionals’ behavior [51] for instructional design to increase participants’ willingness to approach a colleague in difficulty. For instance, the model was used to increase participants’ beliefs in the consequences of approaching a colleague in difficulty. In the first module of the course, viewers are shown how a compassionate approach, which enables a colleague to share emotions related to stressful situations, can help mitigate the impact of these sources of stress. In the second module, physicians are shown the negative consequences that may occur if a colleague in difficulty is not helped or if help is delayed. Participants completed an intervention plan during the CPD course to use when approaching a colleague at their workplace. Other educational strategies include narrated vignettes, knowledge-check questions, additional readings, and reflective writing, which often refers back to the scenario shown in the video. During the online course, physicians are also provided with a self-evaluation tool to identify when they or a colleague may need help. The tool provides individual strategies to maintain or improve mental health at each step along the continuum from being healthy to having important mental health problems [60].

During the course, participants have the option of downloading 10 summaries (ranging from 1 to 3 pages) of course content or supplementary materials [59-67]. Bloom’s taxonomy defined the course’s learning objectives [68]. The CanMEDS role targeted by this course was the professional role, specifically with the objective of emphasizing the commitment to physicians’ health and well-being, which ultimately promotes optimal patient care [69]. The CanMEDS framework highlights skills that are essential throughout medical practice, providing guidance for medical education and ongoing professional development for practicing physicians, all within a competency-based approach [69]. The course was made available on MÉDUSE, the FMSQ’s online CPD platform [70]. Physicians completing the course were sent an automated reminder after 2 weeks of inactivity, and the unfinished course remained available on the physicians’ dashboard on MÉDUSE as a reminder to complete it.

The primary outcome was the intention of physicians to approach a colleague in difficulty. Godin defined intention, from the work of Ajzen and Fishbein on the Theory of Reasoned Action, as the motivation to adopt a specific behavior [51]. In this study, this targeted behavior was approaching a colleague in difficulty.

Intention to adopt a behavior is a precursor of behavior adoption in the integrated model for predicting health professionals’ behavior [51]. Although intention is not a perfect proxy for future behavior, directly observing peer support behavior among health care workers is neither feasible nor ethical, as these conversations are sensitive and confidential [71]. Using self-reported behavioral intention as a proxy allowed for a practical and more timely evaluation of the CPD course [71]. Self-reported behavior 4 months after course completion was also measured but not used as a primary outcome because there is often a loss to follow-up when measures are not taken soon after course completion [72].

For both quantitative and qualitative analyses, we used Godin’s integrated model for predicting health professionals’ behavior as a conceptual framework [51]. In addition, we used Michie’s conceptual framework on BCTs to analyze qualitative responses in detail as well as course content [52]. We analyzed all quantitative data with SAS software (version 9.4; SAS Institute Inc). We performed verification with nonparametric tests using R software (R Foundation for Statistical Computing). The NVivo 15 software (Lumivero) was used for qualitative analyses.

We obtained ethical exemption from a hospital research ethics committee (Comité d’éthique de la recherche sectorielle en santé des populations et première ligne—Centre intégré universitaire de santé et de services sociaux de la Capitale-Nationale [CIUSSSCN]; project #2025-3127). The research project participants had given their free and informed consent to take part in the study. Before each questionnaire, a consent form detailed the objective and intended use of the collected data for research (Multimedia Appendix 4 contains the French version and Multimedia Appendix 5 contains the English version). The database was deidentified to protect the confidentiality of participants. A password was required to access the data saved in an Excel file, which has been stored on a secure server. Data analysis was carried out on a CIUSSSCN computer that was also password-protected. Participants experiencing emotional difficulties could speak with peer supporters from the QPHP. Physicians volunteered to participate and received no financial compensation.

During the 28-month study period, 914 physicians were invited, of whom 792 (87%) agreed to participate (Figure 3). A total of 30 participants were excluded because they did not answer a single CPD-REACTION questionnaire item. We performed analyses on 61% (466/762) of participants who completed the CPD-REACTION questionnaires before and after the course (T1 and T2). Finally, the frequency of self-reported behavior was calculated on 8% (61/762) of participants who completed both the self-reported behavior change questionnaire 4 months after the course (T3) and the CPD-REACTION questionnaire immediately after the CPD course (T2). We excluded 5 participants who did not complete the postcourse survey (T2) from the analysis of data at T3, as it was only at T2 that we could verify their completion of the course.

Among the participants, the woman-to-man ratio was approximately 1:1, and the majority of physicians (655/762, or 86%) were specialists (Table 1). The average age of participants was 48 (SD 12.4) years, 43.5% (332/762) were women, and 86% (655/762) practiced in a specialty. Physicians who responded both before and after the course (n=466) had similar sociodemographic characteristics. We did not have sociodemographic characteristics for the 97 physicians who responded only to the T2 questionnaire without responding to the T1 questionnaire.

Intention means increased from 3.88 (SD 1.73) before to 4.92 (SD 1.40) after the course, as measured on a scale of 1 to 7. This mean difference increase of 1.04 (P<.001) is statistically significant (Table 2). This 1-point increase was also clinically significant, with a Cohen d_z value of 0.72 representing a moderate effect size. Bivariate analyses of changes in the 4 determinants of intention (social influences, beliefs about capabilities, beliefs about consequences, and moral norm) were all statistically significant at an α level of .05. Determinants of intention (beliefs about capabilities, social influences, and beliefs about consequences) increased by statistically significant mean differences ranging from 0.14 to 0.85. Moral norm decreased by a statistically significant mean difference of 0.30 (P<.001). Lower extreme values after the CPD course did not explain the decrease in moral norm. The effect size on the increase in beliefs in capacities and on the increase in social influences was moderate. The effect size on the decrease in moral norm was small, and the effect size on the increase in beliefs about consequences was not clinically significant (Table 2). Mean intention and determinants of intention were similar whether participants had responded to both questionnaires (T1 and T2) or to only 1 questionnaire (Table S2 in Multimedia Appendix 8 [73,74]).

The increase in the mean difference of intention was approximately 1 point and was statistically significant for all categories of gender, age, and domain of practice (laboratory, surgical, and medical domains) that included at least 10 participants (Multimedia Appendix 9). We did not interpret subcategories containing fewer than 10 participants (eg, general practitioners). Results of the multiple linear regression model confirmed that the intention difference was not modified by gender, age, or domain of practice when all variables were included in the same model (Multimedia Appendix 9).

In the GEE model, participation in the CPD course was associated with a higher intention to approach a colleague in difficulty. On average, intention scores were 1.06 points higher in participants after the CPD course compared to before the CPD course (β=1.06; P<.001), after adjustment for determinants and for sociodemographic characteristics (Table S1 in Multimedia Appendix 8). Beliefs about capabilities (β=0.37; P<.001), social influence (β=0.22; P=.002), and beliefs about consequences (β=0.15; P=.03) before the CPD course showed significant positive associations with mean intention, whereas moral norm was not significant. No sociodemographic characteristics were independently associated with intention to approach a colleague in difficulty (Table S1 in Multimedia Appendix 8).

In the final multivariate regression analysis model, beliefs about capabilities (β=0.52; P<.001), social influences (β=0.27; P<.001), and moral norm (β=0.26; P<.03) predicted physicians’ intention to approach a colleague in difficulty (Table 3). Our model accounted for 22% of the variance of intention after the course. Sensitivity analyses verified the stability of the other variables and showed consistent results if we also included physicians who had not responded to the precourse questionnaire (Multimedia Appendix 10), except for the variable belief in consequences, whose statistical significance changed when we modelized physicians who had not responded to the precourse questionnaire.

Postulates of independence of observations and homoscedasticity were respected. Postulates of normality and linearity of residuals were not respected. We verified the correlations between the independent variables and found they did not explain our model’s low predictive capacity (Multimedia Appendix 10).

Of the 251 open-ended responses related to behavioral intentions (T2), 153 verbatims expressed clear behavioral intentions (eg, “I will help my peer”), 10 showed hesitation (eg, “I will try to”), 89 described reasons for intention to approach a colleague in difficulty (eg, “I realize more fully my ability to act”), and 8 referred to other related behaviors. The above categories are not mutually exclusive. Physicians most frequently intended to (1) be more attentive to their colleagues’ signs of distress, (2) approach a colleague with kindness or help or support a colleague, and (3) be more available or listen more to colleagues (Multimedia Appendix 11). The reasons for intention to change behavior physicians most frequently cited were beliefs about capabilities, beliefs about consequences, and knowledge of the behavior (Table 4). For each of the 10 categories of reasons for wanting to adopt the new behavior, intercoder percentage agreement on thematic analysis varied between 98.1% and 99.5%.

The reasons physicians gave for not wanting to change their practice were mostly related to their past behavior, for example, that they were already helping a peer and thus had already adopted the behavior (Multimedia Appendix 11). Other reasons for not intending to change behavior were their environmental context (eg, “I’m now in solo practice”) and that helping a peer was not relevant to their particular medical practice. Four months after the CPD course, the behavioral determinants most frequently given for not approaching a colleague in difficulty were having had no occasion to adopt the behavior, having no colleagues in difficulty, and not being actively in practice or not being in full-time practice (Multimedia Appendix 11).

Qualitative and quantitative results showed most physicians were motivated to approach a colleague in difficulty following the online CPD course. Both qualitative and quantitative results pointed to beliefs about capabilities as being an important determinant of intention. Beliefs about consequences was an important theme in the qualitative data, but it was not an important determinant of postcourse intention to approach a colleague in difficulty. Sensitivity analysis of the 251 physicians who provided open-ended responses related to behavioral intentions (T2) also showed a divergence regarding beliefs in consequences between qualitative and quantitative results.

Among the 66 physicians who completed the self-administered questionnaire at T3, 61 had completed CPD-REACTION immediately after the course. Of these, 41% (25/61; 95% CI 28.6%-54.3%) reported approaching a colleague in difficulty in the 4 months following the CPD course. Those who adopted the behavior had a mean intention after the CPD course of 5.40 (SD 1.01), and those who did not adopt the behavior had a mean intention of 5.00 (SD 1.40). This 0.4 (SD 1.26; P=.24) mean behavioral intention difference is not statistically significant.

A total of 7 BCTs were found in the online CPD course: goal setting, increasing competence, planning, persuasive communication, behavior-related information, modeling, and behavioral experiments (Table 5). Cohen kappa measuring interrater reliability was 0.53 (95% CI 0.14-0.91) for evaluation of BCTs identified the course, which represents a moderate strength of agreement [92]. Of these 7 BCTs, 4 were related to behavioral intentions, 3 involved competence, 2 involved behavioral regulation, and 2 were related to beliefs about consequences (Multimedia Appendix 11). Beliefs about capabilities, knowledge, social influences, and memory, attention, and decision processes were each targeted by 1 BCT.

We assessed the effect of an online CPD course on physicians’ intention to approach a colleague in difficulty and analyzed why participants considered integrating this peer support approach or not into their practice. Mean behavioral intention increased significantly after the course. Postcourse, significant determinants of intention were social influences, beliefs about capabilities, and moral norm. Most frequently stated reasons for intention to change behavior were beliefs about capabilities, beliefs about consequences, and knowledge of behavior. Triangulation showed an alignment of quantitative and qualitative results on beliefs about capabilities as a key factor but diverged on beliefs about consequences. In the 4 months following the CPD course, 41% (25/61) of physicians reported approaching a colleague in difficulty. There was no significant difference in the intention to approach a colleague in difficulty between physicians who adopted this behavior at the 4-month follow-up and those who did not. Those who did adopt the behavior had a nonstatistically significant higher intention after the CPD course than those who did not. Finally, we identified 7 BCTs in the online CPD course: goal setting, increasing competence, planning, persuasive communication, behavior-related information, modeling, and behavioral experiments. These results lead us to make the following observations.

First, we found that, on average, physicians reported a stronger intention to approach a colleague in difficulty after completing the CPD course compared to before. This 1-point intention increase was statistically significant and also clinically significant, having a moderate effect size. This increase was observed in bivariate analysis and confirmed with multivariate analysis. This suggests that the CPD course motivated physicians to approach colleagues in difficulty. Enhancing peer support among physicians through CPD courses is particularly important as it can play a protective role regarding psychological distress in physicians [17,18].

Also, to the best of our knowledge, this 1-point intention increase represents a greater increase in intention than has been observed in most other studies using the CPD-REACTION to evaluate changes in health professionals’ intentions following an intervention [72,85,93-102]. The before-after difference of intention was more than double the difference observed in a study by Ayivi-Vinz et al [85] who evaluated intention before and after 9 online CPD courses taken by Quebec physicians, all on topics other than this one. The 95% CIs of difference in intention between their study and ours do not overlap [85]. Thus, the observed difference between the 2 studies is statistically significant. An important difference is that Godin’s integrated model for predicting health professionals’ behavior was used by the senior instructional designer (MT) to design the course, while the 9 courses in the study by Ayivi-Vinz et al [85] were developed by physicians who may not have used behavior change models in course design. This may give support to the importance of using theory to produce CPD courses. Another explanation is that preintervention intention among the Quebec physicians in the study by Ayivi-Vinz et al [85] was already high at 5.5 out of 7, while in ours it was 3.9 out of 7, and thus there was more room for improvement. Overall, this result suggests that there are benefits of using behavior change theories, especially for CPD courses targeting behaviors that physicians are initially less inclined to adopt. It also suggests that it is possible to increase behavioral intention in physicians using online training formats.

Social influences, beliefs about capabilities, and beliefs about consequences increased. An unexpected postcourse decrease in moral norm occurred. Moral norm had the highest mean among the 4 determinants of intention both before and after the intervention, indicating that participants already felt a moral obligation to adopt the targeted behavior and thus there was little room for improvement. Also, the apparent decrease had a small effect size, which may not translate into clinical significance. One possible explanation for this unexpected decrease in moral norm is that participants may have given more nuanced answers after the CPD course as they better understood the ethical and deontological implications of approaching a colleague in difficulty. In the precourse questionnaire, 38.3% (177/462) of the physicians answered with the maximum score of 7/7 for moral norm, compared to none after the intervention. Nevertheless, further qualitative data analysis is needed to explore this result. Analyzing which BCTs target moral norm could also be important, as this is not studied by Michie et al [52].

Second, we also examined whether participant characteristics modified the impact of the course on their behavioral intention. The before–after intention difference was the same regardless of age, gender, and domain of practice (laboratory, surgical, and medical domains). Thus, the course successfully reached physicians regardless of their sociodemographic characteristics and does not need to be adapted for a specific population. For example, a previous study found that surgeons were the least likely physicians to use peer support [34], hence the importance of our finding that surgeons also responded favorably to the online CPD course.

Third, we found that determinants of intention to approach a colleague in difficulty after the CPD course included social influences, beliefs about capabilities, and moral norm. Beliefs about capabilities was the strongest predictive variable of the intention to approach a colleague in difficulty. This means improving physicians’ perception of their self-efficacy in adopting a behavior, as well as their perceived control over performing that behavior, is important to motivate them to approach a colleague in difficulty. Thus, CPD course developers should target beliefs about capabilities through BCTs for peer support training. Beliefs about capabilities was also the most significant determinant of intention in Godin’s systematic review on health care workers’ behaviors [53]. BCTs that are known to be effective for this include: (1) increasing skills, (2) coping strategies, (3) repetition or practice of the behavior, (4) feedback, (5) planned self-affirmations, (6) motivational interviewing, and (7) graded tasks (starting with tasks that are easy to accomplish and gradually making them more difficult). In previous studies, beliefs about capabilities, moral norm, and beliefs about consequences were all predictive of intention to adopt behaviors related to patient care (eg, “Prescribe opioids in a safe manner when managing perioperative pain in my patients”) postcourse [72,85]. Our results differ slightly from these, with social influence also being predictive of this intention. Approaching a colleague in difficulty is a social behavior that can be adopted by a whole departmental team [38]. Our results suggest that the social context may be a more important influence on team-based behaviors than on clinical behaviors, which are associated more with individual clinicians.

Our model had a low predictive capacity (R²=0.22), explaining only 22% of the variance in intention. In comparison, previous literature showed that the predictive capacity of models analyzing 1 single behavior varied between 19% and 62%, depending on the type of behavior [53]. Bakwa et al’s [72] study of multiple behaviors related to patient care among Quebec physicians explained 82% of the variance in physicians’ intention to adopt the clinical behavior. However, Bakwa et al [72] included multiple behaviors in the same model, while we studied only 1 behavior, and moreover one that is specific to the well-being of physicians rather than geared to a clinical task. Indeed, existing studies mainly predict behaviors related to patient care [53,72]. While variables of Godin’s integrated model for predicting health professionals’ behavior are sufficient to predict such behaviors [51], approaching a colleague is not a clinical behavior and may be influenced by factors not measured by Godin’s model such as lack of time, lack of privacy, and organizational culture [36,103]. This is significant because the Quintuple Aim—adopted by many health care systems—now emphasizes 5 key outcomes, and not all of them clinical: patient outcomes, patient experience, efficiency, clinician well-being, and equity [13]. Therefore, further development is needed regarding fostering behaviors related to clinician well-being.

Fourth, triangulating quantitative with qualitative results confirmed that beliefs about capabilities was a key determinant of intention to approach a colleague in difficulty. Interestingly, results regarding beliefs about consequences diverged slightly: while it was the second most frequent reason for intention to change behavior stated in physicians’ open-ended responses, it was not a statistically significant determinant of intention in our quantitative results. One explanation is that statistical significance varied according to the other variables included in the model, perhaps because the P value for beliefs about consequences was so close to the critical α value of .05. Another possible explanation is that beliefs about consequences, as described by Godin [51], seems to refer to the consequences for oneself. Items for measuring beliefs about consequences quantitatively were worded as follows “Overall, I think that for me approaching a colleague in difficulty would be....” [73]. In contrast, in our qualitative analysis, beliefs about the consequences of adopting the behavior could also refer to its consequences for the colleague being approached, the team, the department, or for society (Table 5). Indeed, many physicians taking part in this study perceived that the consequences of approaching a colleague in difficulty would be beneficial for the team and the department, a theme often mentioned in the open-ended responses. Thus the perception of the consequences of team-focused care may include organizational benefits, team benefits, patient benefits, and benefits to individual team members [104]. Also, approaching a colleague in difficulty is a team-focused behavior that takes place between 2 health care workers, while behaviors related to patient care occur between a clinician and a patient (eg, prescribing medication). We hypothesize that such team-focused behaviors may depend on a wider variety of determinants, such as organizational culture, than do individual behaviors related to patient care [36,104,105]. Whether peer support strategies can be applied without fear of judgment or retaliation within the organization may depend on its climate of psychosocial safety, or the perception among employees that their psychological health is a priority for the management [106,107]. A study by Tolin et al [38] reported that physicians’ perception of institutional support had a positive impact on implementation of a physician-focused peer support program. Broadening the item statements on beliefs about consequences to include effects on others and adding questions with regards to role and identity may be a good starting point for improving measuring tools regarding team-based behaviors.

Other determinants to measure in future studies should include social and professional role or identity and past behavior or habits related to the behavior, both mentioned by participants as determinants of intention and present in Godin’s integrated model for predicting health professionals’ behavior (Table 4) [51]. Our qualitative analysis identified the following further potential determinants: (1) knowledge; (2) action planning; (3) memory, attention, and decision processes; (4) environmental context and resources; and (5) emotion related to behavior. The first 3 of these determinants are also behavioral domains targeted by the BCTs we identified in the CPD course.

Fifth, 4 of 10 physicians reported having approached a colleague in difficulty in the 4 months following the CPD course, and those who did so had a higher intention to adopt the behavior after the CPD course. No statistically significant difference in intention to approach a colleague in difficulty was found between those who did and those who did not adopt the behavior (5.4 vs 5.0; P=.24). This unexpected finding is possibly due to the limited statistical power of the analysis, estimated at 23%. Also, environmental factors, such as practicing solo, were the barriers most mentioned by physicians. Some of these physicians who might have been willing to approach a colleague in difficulty were simply not in the right environment to be able to do so. Other studies report on peer support courses that are part of comprehensive peer support programs that include an organization-wide referral system [15,22,26,27,30,34,38-40,43]. Thus, a physician could be the only physician in a department and still participate in a peer support program thanks to the referral system. It would be valuable to further evaluate whether formalizing peer support through these programs helps overcome certain environmental barriers or, conversely, introduces new barriers, compared to informal peer support outside of a program, as occurred in our study.

Lastly, we identified 7 BCTs in the online CPD course. This is the first time that BCTs have been identified in an FMSQ course and triangulated with before-and-after measures of CPD-REACTION. The results suggest that the BCTs effectively targeted determinants of intention, such as beliefs about capabilities and social influence, which had changed after the course. However, this study, which analyzed only 1 course, cannot make recommendations on which BCTs to use for which behaviors. Nevertheless, our results offer significant methodological progress in evaluating CPD course content, which could lead to innovations in CPD course development. Further studies on the use of theory-informed BCTs in CPD course design to enhance clinician adoption of targeted behaviors could further our understanding of effective course development strategies.

First, the quasi-experimental before-and-after study design is vulnerable to confounding concomitant events that could influence study outcomes because the design comprises only a single group [50]. However, the short duration of the online CPD course (1 hour) made it less likely for other events to introduce bias. We chose this before-and-after study design as it is better suited to the constraints of practice settings than randomized control trials and better reflects the real-world CPD context [50]. Furthermore, we confirmed bivariate analysis with the GEE model to better measure the effect of the CPD course. Triangulation of quantitative with qualitative methods improved the internal validity of the study [50,108].

Second, our results mainly apply to specialist physicians who chose to follow a nonmandatory CPD course and thus had some interest in the subject. The effect of the CPD course could have differed if it were mandatory for all physicians, as seen in other education programs [109]. Furthermore, general practitioners were underrepresented in this study (n=9). Thus, we do not know how general practitioners would respond to the CPD course but have no reason to think that they would behave differently than specialists. In addition, the well-being of all health care professionals is paramount, and it is not clear if the online CPD course would have any impact on other types of health professionals.

Third, there was a possible social desirability bias whereby physicians overrated their intention to approach a colleague in difficulty [50]. Confidentiality and anonymity were ensured to limit this bias. Of note, we measured intention using the same instrument used by Bakwa et al [72] and by Ayivi-Vinz et al [85] in their studies, so the desirability bias would have applied to their results as well and does not explain the higher before-and-after intention difference we observed in this study.

Fourth, this study does not allow us to confirm the reasons for which results regarding beliefs about consequences diverged, reasons to change behavior that were prominent in qualitative results but unstable or nonsignificant in the multiple linear regression. Although hypotheses are detailed in the Discussion, further studies would be needed to validate them.

Fifth, another limitation is that we used a backward elimination procedure, which can increase the risk of overfitting and biased effect estimates. However, results in the full model and fitted model were similar.

Sixth, we lacked statistical power to measure the difference in intention to approach a colleague in difficulty between those who did and those who did not adopt the behavior.

Lastly, self-reporting of behavior may also overestimate behavior adoption, as participants may not have used the approach taught in the course and may still report having adopted the behavior. Furthermore, only a small proportion of participants completed the self-reported behavior questionnaire 4 months after the course. This may indicate a selection bias in favor of the most motivated physicians, which could overestimate the proportion of those adopting the behavior. The sensitivity analysis in Multimedia Appendix 8 shows that this potential selection bias on number of questionnaires answered does not affect the primary outcome (mean difference in intention to approach a colleague in difficulty) but only the secondary outcome of self-reported behavior 4 months after the training.

To the best of our knowledge, this is the first study that uses validated before-and-after intervention measures to examine the effect of a CPD course on physicians’ intention to approach a colleague in difficulty (experiencing distress, burnout, or second-victim symptoms after a patient adverse event). The CPD course produced an increase in physicians’ intention regardless of age, gender, or domain of practice (laboratory, surgical, and medical domains). Beliefs about capabilities was an important determinant of intention to approach a colleague in difficulty. Although the CPD course increased physicians’ intention to approach a colleague in difficulty, there is still a gap in understanding what other factors predict this intention. Within 4 months of the CPD course, 41% (25/61) of the physicians reported approaching a colleague in difficulty using strategies taught in the online CPD course. Identifying BCTs in the CPD course design was a methodological innovation for FMSQ course evaluation.

Although the CPD course demonstrated a positive effect on physicians’ intention to approach a colleague in difficulty, more research is needed to better understand the mechanisms through which the intervention works. This could be done by translating the course into English, making it available to English-speaking physicians in Quebec and other Canadian provinces or internationally, and measuring the effect of the CPD course on approaching a colleague in difficulty in other contexts. Furthermore, instructional designers should focus on enhancing physicians’ beliefs about their capabilities to facilitate their adoption of this peer support behavior. Instructional designers could consider using BCTs, especially when targeted behaviors are less widely accepted by clinicians or perceived as more difficult to adopt. Evaluating the effect of BCTs on behavioral intention could deepen our understanding of interventions aimed at changing clinicians’ behaviors.

Our results suggest that peer support courses help build a culture of care among physicians. By enhancing our knowledge of interventions aimed at improving the peer support capabilities of physicians, this project indirectly aimed to improve the well-being of health care professionals. Peer-supported professionals could then be better able to contribute to resilient health care systems and to providing quality care to patients.