Investigating the peer Mentor-Mentee relationship: characterizing peer mentorship conversations between people with spinal cord injury

Abstract Purpose This study aimed to: (1) develop a coding manual to characterize topics discussed and conversation techniques used during peer mentorship conversations between people with spinal cord injury (SCI); (2) assess the reliability of the manual; and (3) apply the manual to characterize conversations. Materials/Methods The study was conducted in partnership with three Canadian provincial SCI organizations. Twenty-five phone conversations between SCI peer mentors and mentees were audio-recorded and transcribed verbatim. Ten transcripts were inductively analyzed to develop a coding manual identifying topics and techniques used during the conversations. Inductive technique codes were combined and deductively linked to motivational interviewing and behaviour change techniques. Two coders independently applied the coding manual to all transcripts. Code frequencies were calculated. Results The coding manual included 14 topics and 31 techniques. The most frequently coded topics were personal information, recreational programs, and chronic health services for mentors and mentees. The most frequently coded techniques were giving personal information, social smoothers, and closed question for mentors; and giving personal information, social smoothers, and sharing perspective for mentees. Conclusion This research provides insights into topics and techniques used during real-world peer mentorship conversations. Findings may be valuable for understanding and improving SCI peer mentorship programs. Implications for Rehabilitation SCI peer mentorship conversations address a wide range of rehabilitation topics ranging from acute care to living in the community. Identification of the topics discussed, and techniques used in SCI peer mentorship conversations can help to inform formalized efforts to train and educate acute and community-based rehabilitation professionals. Identifying commonly discussed topics in SCI peer mentorship conversation may help to ensure that peer mentors are equipped with the necessary knowledge and resources, or the development of those resources be prioritized. Developing a method to characterize the topics discussed and techniques used during SCI peer mentorship conversations may aid in designing methods to evaluate how rehabilitation professionals provide support to people with SCI.


Introduction
Spinal cord injury (SCI) has lasting implications for all aspects of a person's life. Evidence has emerged highlighting the importance and value of peer mentorship in the rehabilitation of people living with an SCI to adjust, adapt, and thrive after an injury [1][2][3][4][5][6][7][8][9]. Peer mentorship is defined as "a peer interaction that aims to provide encouragement, counsel, and/or information to individuals who share similar lived experiences" [2]. It is critical to examine how peer mentorship conversations are carried out in real-life settings to understand how peer mentorship interactions support people living with an SCI.
informational support from peer mentors may improve self-efficacy, quality of participation, and well-being in peers living with an SCI [2,8].
Research has examined additional aspects of the peer mentorship relationship that foster supportive relationships [4], characteristics of quality peer mentors [15], psychological and leadership theories and approaches (e.g., transformational leadership, selfdetermination theory, and motivational interviewing) [7, 16,17], and outcomes of peer mentorship [2,3,5,6,18]. While these studies provide an initial understanding of the peer mentorship relationship and associated outcomes, they are limited in that they do not examine the topics discussed and techniques used during peer mentorship conversations. Little is known about the nature of the information being provided or how this information is provided during real-world SCI peer mentorship conversations. Creating a manual to identify conversation characteristics may facilitate efforts to understand, evaluate, and ultimately improve the delivery of SCI peer mentorship programs to support the rehabilitation of people living with an SCI.
One potential approach to examining peer mentorship conversations is to characterize the conversations in terms of the topics (i.e., subjects of conversation) discussed and techniques (i.e., observable conversation strategies) used. To our knowledge, there are currently no established coding manuals that have been specifically developed to characterize topics discussed during SCI peer mentorship conversations. Peer mentorship conversations are often unstructured and may cover a wide variety of topics and goals. Therefore, a coding manual to characterize the possible topics of discussion is necessary before conversations can be examined. Such a coding manual could then be used to examine how the topics that arise during peer mentorship conversations align with available information resources. Comparing conversation topics to existing evidence-based resources may clarify peer mentors' scope of practice while highlighting key areas of interest that may require additional training.
In contrast to characterizing topics, several coding manuals exist for identifying conversation techniques or intervention content delivered in one-to-one interactions, such as behaviour change techniques (BCTs) and interpersonal strategies [19][20][21][22]. BCTs are conceptualized as the active and irreducible components of an intervention that act to change a behaviour [23]. BCTs may be an important component of SCI peer mentorship conversations given the growing body of literature demonstrating that peer-led interventions are effective at changing behaviours in people living with an SCI, such as physical activity and self-care [17,[24][25][26]. However, it is unclear how, and importantly, if peer mentors living with an SCI use BCTs in real-world settings. Identifying the BCTs used by mentors during SCI peer mentorship conversations may indicate the extent to which peer mentorship conversations target behaviour change.
The BCT version 1 (BCTTv1) is a cross-behavioural, hierarchically-structured taxonomy consisting of 93 distinct BCTs [23]. Characterizing interventions in terms of BCTs may help identify the mechanisms that underpin an intervention's success or failure [27]. BCT taxonomies have been used extensively to examine behaviour change interventions in published manuscripts, reports, as well as in recorded conversation transcripts in a wide variety of contexts [22,[28][29][30]. Given the diverse applicability of BCT taxonomies, the BCTTv1 may be a useful coding manual to use in the SCI peer mentorship setting to identify the techniques that peer mentors use to support mentees living with an SCI.
While the BCTTv1 describes techniques that practitioners can apply, it does not describe the relational factors used to apply those techniques or how the client responds. The Motivational Interviewing Skill Code (MISC) may address these limitations as it has been used to identify intervention content and interpersonal strategies used in one-to-one interactions [31]. The MISC [31] was developed to evaluate the use of motivational interviewing, a person-centred counselling approach that aims to enhance clients' intrinsic motivation towards behaviour change [32]. The MISC offers a valuable method to evaluate the practitioner-client relationship by characterizing practitioner and client statements [31]. This dyadic coding approach can be used to indicate the level of alliance and collaboration between practitioners and clients, which is considered a critical element of the peer mentor-mentee relationship [7,15,16,32].
Although the BCTTv1 and MISC appear to be promising approaches to examining peer mentorship conversations, they have not yet been applied to SCI peer mentorship conversations. Thus, it is unknown if peer mentors use these or other strategies in peer mentorship conversations as peer mentors are often not trained or expected to deliver BCTs or MI strategies in community-based settings. Applying coding manuals such as the BCTTv1 and the MISC alongside inductive analyses to identify novel techniques and identifying the topics discussed will provide a coding manual to describe what occurs naturally during peer mentorship conversations. This coding manual could be applied to future peer mentorship conversations to further elucidate peer mentors' scope of practice and identify helpful and harmful interpersonal strategies that are used. Developing a tool to reliably characterize peer mentorship conversations may help to optimize the evaluation and delivery of SCI peer mentorship programs.
The present study aimed to develop a coding manual to describe the conversation strategies and topics of real-world SCI peer mentorship conversations between people living with an SCI. This study then aimed to apply the coding manual, which integrates the BCTTv1 and MISC to peer mentorship conversations to characterize the topics and techniques used in peer mentorship conversations. Specifically, the research team aimed to: 1. develop a coding manual to categorize topics and techniques used in peer mentorship conversations; 2. assess the reliability of the coding manual, and; 3. apply the coding manual to characterize the topics discussed and techniques used in peer mentorship conversations.

Design
Ethics approval was received from the University of British Columbia Okanagan. The study design was observational using audio-recordings collected at one time point. This study was designed and conducted in partnership with three SCI community-based organizations with established peer mentorship programs, including SCI Alberta, SCI British Columbia, and SCI Ontario, using an integrated knowledge translation approach [33]. A detailed description of our partnership approach has been published previously [15]. Additional information about our partnership can be found on Open Science Framework (see osf.io/qszr9) and in the Supplementary File.

Participants
All partner SCI organizations recruited SCI peer mentors and mentees via social media, newsletters, and direct emails. To be eligible to participate, mentors and mentees needed to be at least 18 years of age, identify as someone living with an SCI, and speak English. In total, we aimed to recruit 40 people to participate (i.e., 20 transcripts of peer mentorship conversations). This sample size is similar to previous studies that aimed to establish coding methods for characterizing practitioner (n ¼ 15 transcripts) and client (n ¼ 18 transcripts) statements in behavioural support sessions [20,22].

Procedures
Participants responded to an online eligibility questionnaire. If eligible, participants provided informed consent, information about their peer mentorship experiences, and demographic information via an online questionnaire. SCI peer mentors and mentees were then matched using pre-determined criteria. These criteria were determined through discussion with the three SCI organizations and included gender and/or availability of participants. These criteria were minimal and were not used in all cases (e.g., participant requested a peer mentor with specific experience such as aging) to align with criteria used by the SCI organizations. Each pair was asked to participate in a one-on-one conversation over the phone.
Participants were asked to talk freely and were told there were no expectations or requirements for the length or topic of their discussion. These instructions were intended to foster a conversation that may mimic a real-world peer mentorship conversation. All conversations were audio-recorded and transcribed verbatim by the research team. Immediately following the interaction, participants completed an online questionnaire assessing their perceptions of the conversation. Copies of questionnaires are provided on Open Science Framework (see osf.io/qszr9).

Demographics questionnaire
Participants were asked to respond to demographic questions, including age, sex, province of residence, education, and marital status. Participants were also asked to provide information regarding their SCI, including years since injury, injury level, injury type, completeness of injury and injury cause.

Peer mentorship experiences questionnaire
Participants were asked to respond to questions regarding their experiences with peer mentorship, including the number of years they have received or provided mentorship and the number of individuals to whom they have provided peer mentorship or from whom they have received peer mentorship. Peer mentors were asked to indicate their confidence in their ability to provide peer mentorship on a 7-point Likert scale, ranging from 1 ("not confident at all") to 7 ("completely confident"). Peer mentees were asked to rate their satisfaction with their previous peer mentorship experience on a 7-point Likert scale, ranging from 1 ("not at all satisfied") to 7 ("completely satisfied").

Conversation perceptions questionnaire
Participants were asked to indicate whether the recorded session was their first interaction with the peer mentor or mentee. Participants were also asked to rate their agreement as to how similar their recorded peer mentorship session was to a typical peer mentorship session. Participants rated their agreement on a 7-point Likert scale, ranging from 1 ("strongly disagree") to 7 ("strongly agree").

"Important others" health care climate questionnaire (IOCQ)
This six-item questionnaire was adapted from the IOCQ to evaluate peer mentees' perception of the autonomy supportiveness of their peer mentor [34]. Mentees were asked to rate the peer mentorship conversation in terms of autonomy, feeling understood, fostered confidence, feeling heard, and ability to ask questions. Mentees rated their agreement on a 7-point Likert scale, ranging from 1 ("strongly disagree") to 7 ("strongly agree"). The value for Cronbach's Alpha for the survey was a ¼ .84, indicating "good" internal consistency.

Analysis
Inductive and deductive analysis approaches were conducted in three phases to address the research question (see Figure 1). Analyses of the conversation transcripts were conducted in three phases. Each phase aligned with the three objectives of this study.

Phase 1: coding manual development
The coding manual was developed inductively to identify topics discussed and both inductively and deductively to identify techniques used in peer mentorship conversations. First, the topic coding manual was developed using applied thematic analysis [35]. Two researchers (RM, EG) independently conducted a thematic analysis across ten transcripts to identify the topics that were discussed [35]. The researchers independently coded the ten transcripts and identified common themes. Through discussion, the two researchers merged common themes and developed definitions for the merged themes. Similar to methods in previous research [20], two researchers (RM, HG) conducted an inductive sort task of the resulting themes to identify redundancies and overlap across topics. This task involved the researchers independently sorting topics into groups of common topics. Through discussion, topics with common themes were refined (i.e., combined or more clearly differentiated), redundant topics were removed, and overlapping topics were combined to consolidate the topic coding manual further. Next, both inductive and deductive analyses were conducted to develop the technique coding manual. Two researchers (RM, EG) used the same steps of applied thematic analysis across the same ten transcripts to identify the techniques that were used by peer mentors or mentees, resulting in a peer mentor-and mentee-specific manual [35]. Identified techniques used by SCI peer mentors and mentees were deductively analyzed for overlap with the BCTTv1 and MISC. Two researchers (RM, EG) independently coded the inductive technique codes using the BCTTv1 and MISC to determine overlap. Any BCTTv1 or MISC code that overlapped with an inductive code was integrated with that inductive code. Any BCTTv1 or MISC code that appeared relevant to SCI peer mentorship conversations but did not correspond with an inductive code was added to the technique coding manuals. In alignment with previous research using the BCTTv1 and MISC, inter-rater reliability was calculated using percentage agreement, Kappa statistic and the Prevalence-Adjusted Bias-Adjusted Kappa (PABAK) [36]. PABAK adjusts for the high prevalence of negative agreement (i.e., when both coders agree that codes are not present) and bias presented in the "real-world" [36]. Agreement was registered when both coders assigned the same code(s) to a technique. If researchers assigned different code(s), disagreement was registered. Inter-rater reliability values of 0.60-0.79 indicated substantial reliability and values above 0.80 were considered outstanding for both Kappa and PABAK statistics [36]. Inter-rater reliability was calculated using identical methods for each phase of analysis.
The overlap between the peer mentor and mentee coding manuals was discussed amongst the research team, and the two manuals were subsequently integrated. Two researchers independently coded for overlap between the peer mentor and peer mentee codes and inter-rater reliability between coders was calculated. Inter-rater reliability was calculated using percentage agreement, Kappa, and PABAK [36]. Finally, the resulting manual was presented to the broader research team, including all community partners, to test face validity and finalize all topic and technique codes.

Phase 2: assess the reliability of the coding manual
All transcripts were coded using the coding manual developed in Phase 1. Two researchers (RM, KB) identified topics discussed and techniques used in mentor statements and mentee statements in five waves. Five transcripts were coded for topics first and then techniques in each wave. In the first wave, coders independently coded the five transcripts one-by-one, completing the following steps for topics then techniques for each transcript: 1) inter-rater reliability was calculated, 2) discrepancies were resolved through discussion, and 3) the coding manual was adapted to improve agreement (i.e., refining definitions, writing notes, and adding new codes when relevant). The purpose of coding transcripts one by one in the first wave was to familiarize coders with the coding manual. In subsequent waves, coders independently coded all five transcripts, then followed the steps above. The final coding manual is included in the Supplementary File and on Open Science Framework (see osf.io/qszr9) and described in the results. Interrater reliability for every mentor and mentee statement was calculated using percentage agreement, Kappa, and PABAK [36].

Phase 3: characterize mentor and mentee verbal statements
For each transcript, the frequency that the mentor and mentee discussed each topic and used each technique was counted. Across the 25 transcripts, the percentage of statements characterized by each topic in the manual was calculated for both mentors and mentees. The same calculation was completed for techniques. A Mann Whitney-U test was used to compare peer mentor and mentees' mean frequency of topic and technique codes [37]. Effect sizes were calculated using Pearson's r and were categorized as small (< 0.10), medium (0.10-0.30), or large (>0.50) [38].

Participants
Eighty-two individuals expressed interest in participating in this study. A post hoc decision was made to include 30 mentors and 30 mentees for a total of 30 recordings to maintain the feasibility of the study. Thirty-one recordings were scheduled, and six were cancelled or removed due to participant dropout after scheduling (n ¼ 5) or technical difficulties with recordings (n ¼ 2). Participant dropout was due to illness (n ¼ 2), changes in availability after scheduling the peer mentorship conversation (n ¼ 6), or the participant could no longer be reached (n ¼ 2). In total, 25 peer mentors and 25 peer mentees completed a recording (mean duration ¼ 47.7 ± 16.1 min; range ¼ 16 to 97 min). Twenty-three of the 25 recordings consisted of same-gendered pairs. Nine pairs consisted of two women, and 14 pairs consisted of two men. Participants' demographic information can be found in Table 1. Sixteen of the 25 peer mentors had received some form of training in peer mentorship, with the majority of training being provided by SCI organizations. Table 2 includes descriptions of previous experience indicated by peer mentors and mentees. Sixteen participants indicated it was their first conversation with the peer mentor or mentee, three indicated it was not the first, and six did not respond. Peer mentees were somewhat satisfied with their previous peer mentorship experiences with a mean score of 4.6 ± 1.9 out of 7. Participants agreed that the recorded conversations were typical of a peer mentorship conversation with a mean score of 5.5 ± 1.0 out of 7. Table 3 includes descriptions of mentees' perceptions of the peer mentorship conversations.

Phase 1: coding manual
Following the inductive analysis, the first iteration of the coding manual included nine topic codes, 21 mentor-specific technique codes, and 37 mentee-specific technique codes. When the inductive peer mentor manual was compared to the deductive coding manuals, 14.0% of the peer mentor manual overlapped with BCTTv1 and 68.7% with MISC. Inter-rater-reliability for coding using the BCTTv1 at minimum was substantial (Percent agreement: 59.0%; Kappa: 0.77; PABAK: 0.96). Inter-rater reliability for coding using the MISC was outstanding (Percent agreement: 83.3%; Kappa: 0.90; PABAK: 0.97). A 10.8% overlap between the mentee coding manual and the MISC was found. Inter-rater reliability for coding using the MISC was outstanding (Percent agreement: 89.2%; Kappa: 0.92; PABAK: 0.98). As minimal overlap was found between the mentor coding manual and the BCTTv1, only relevant MISC codes were combined with the mentor and mentee manuals. After adding relevant codes, 22 techniques were included in the mentor manual, and 37 techniques were included in the mentee manual. There was a 27.0% overlap between the mentor and mentee manuals. Thirty-eight techniques were included after combining the two manuals. The full coding manual was consolidated to 31 techniques and 14 topics following the inductive sort task. No techniques or topics were added following consultation with the larger research team. Topic names and definitions were refined through discussion (see the Supplementary File). Changes included renaming functional outcomes to physiological outcomes to account for topics such as temperature regulation, autonomic dysreflexia, or other secondary health conditions that may not relate to functional outcomes. Emphasis on sexuality was added to the definition of relationships.

Phase 3: characterize mentor and mentee verbal statements
In total, there were 7782 topic codes across all transcripts, including 4104 peer mentor topic codes and 3613 peer mentee topic codes. The most commonly coded of the 14 topics included in the coding manual were personal information, recreational programs, and chronic health services, accounting for 41.3%, 10.1%, and 9.0% of the total topic codes, respectively. The same topic codes were most common for peer mentors and mentees. Personal information, recreational programs, and chronic health services accounted for 40.2%, 9.7%, and 8.7% of the topic codes for peer mentors, respectively and 42.0%, 10.6%, and 9.3% for total topic codes for mentees. See Table 4 for total, peer mentor and mentee topic frequencies across all transcripts and Figure 2 for proportions of topic codes across all transcripts. The topics discussed did not statistically differ between peer mentors and mentees. The frequency of each technique code across all transcripts in total, for peer mentors and mentees, is presented in Table 5. Table 5 also includes the frequency of technique codes for both SCI peer mentors and mentees. In total, there were 9628 technique codes across all transcripts, including 5223 peer mentor technique codes and 4394 mentee technique codes. Overall, and for SCI peer mentors, the most commonly coded of the 31 techniques were giving personal information, social smoothers, and closed questions. Overall, giving personal information, social smoothers, and closed questions accounted for 26.4%, 15.3%, and 9.1% of the total technique codes. These techniques accounted for 22.0%, 16.3%, and 10.0% of mentor technique codes, respectively. The most commonly coded techniques for mentees were giving personal information, social smoothers, and sharing perspectives. These techniques accounted for 31.8%, 14.1%, and 8.1% of the techniques used by mentees, respectively. Significant differences were found between peer mentor and mentee's discussion of topics and use of techniques, which are outlined in Table 4 and Table 5. A Mann Whitney-U test indicated that 21 of the 31 techniques statistically differed in frequency of use between peer mentors and mentees. A Mann Whitney-U test indicated no topics statistically differed in frequency of use between peer mentors and mentees. See Table 5 for total, peer mentor, and mentee technique code frequencies across all transcripts and Figure 3 for proportions of technique codes across all transcripts.

Discussion
This research demonstrated a systematic approach to develop, validate, and apply an integrated coding manual to reliably characterize the topics discussed and techniques used during SCI peer mentorship conversations. Peer mentorship conversations appear to address a wide range of topics related to SCI, rehabilitation and participants' lives more generally. Additionally, the techniques employed during SCI peer mentorship conversations in this study emphasize interpersonal connection as a means of communication aligned with the MISC rather than overt behaviour change strategies aligned with the BCTTv1. Our findings highlight that SCI peer mentorship conversations are distinct from behavioural support or counselling interactions and may support people living with an SCI in many areas of their lives.
The scope of conversations spanned 14 different topics, including topics related to emotional outcomes, recreation, accessibility, and health care. These topics aligned with priorities previously identified in the SCI peer mentorship literature [39,40]. For example, all but one topic covered by the evidence-  [41] was included in the coding manual. The breadth of topics discussed in these conversations suggests that peer mentorship holds impressive potential for translating knowledge on a variety of topics related to SCI. This finding indicates that peer mentors require a wide range of knowledge and skills specific to these topics. Peer mentors may also need to know when a mentee should be referred to another resource and what resources are available. For instance, peer mentors should be aware of the limits of their scope and knowledge on topics of physiological, psychological, and emotional outcomes of SCI and ensure mentees are directed to the appropriate professional health resources. When developing trainings, the content of peer mentorship conversations should be considered to enhance the breadth and quality of what peer mentors can communicate about and may help to define their scope of practice.
Training in strategies to navigate potentially sensitive topics may be a valuable addition to training in topical content. For example, financial assistance was discussed frequently and required specific knowledge related to government and other organizational funding mechanisms. However, using mentor strategies such as providing personal information may not always be appropriate to discuss the topic of finances as there are significant disparities in financial assistance for people living with an SCI, which may serve as a point of contention. This finding further highlights the relevance and benefit of motivational interviewing and relational-based conversation strategies that promote collaboration and convey empathy as compared to BCTs. Indeed, the resulting coding manual primarily overlapped with conversation techniques used in motivational interviewing identified by the MISC and minimally overlapped with BCTs identified by the BCTTv1. These findings indicate that peer mentorship interactions may not correspond with typical behavioural support interactions, which commonly involve multiple BCTs delivered within a dyadic conversation between a practitioner and a client [20]. In recent years, there has been attention on using peer mentorship conversations to promote health behaviour changes such as selfcare and physical activity [17,[24][25][26]. Peer mentors in these peerbased health behaviour interventions were rigorously trained in BCTs and interpersonal strategies grounded in motivational interviewing and behaviour changes theories such as social cognitive theory and self determination theory [17,[24][25][26]. When these BCTs and interpersonal strategies were implemented, the peer-based interventions were effective in promoting health behaviour change [24][25][26]. However, our findings suggest the most frequent techniques used during SCI peer mentorship conversations focus on sharing stories and building rapport, indicating these conversations focus on relational factors rather than behaviour change. The intent of peer mentorship within services delivered by community-based organizations may not be behaviour change, and behaviour change strategies may not necessarily be delivered as frequently, if at all, in peer mentorship programs delivered in community settings. Thus, efforts to evaluate peer mentorship delivered in communitybased settings may need to focus on relational outcomes rather than or alongside health or behavioural outcomes.
The techniques included in the coding manual closely align with person-centred concepts that show promise for supporting people living with an SCI, including motivational interviewing, self-determination theory, and transformational leadership [7, 16,17,[42][43][44]. For example, techniques such as providing advice with permission and open-ended questions are suggested to promote peer mentees' sense of autonomy as they provide mentees with choice and consider their individual needs [7,16]. Additionally, techniques such as giving personal information may foster a sense of trust and relatedness between mentees and mentors [7]. These findings align with a previous communitybased Delphi study that established consensus on characteristics of quality peer mentors living with an SCI [15]. The characteristics identified were also closely aligned with these person-centred concepts and underscored the importance of emotional intelligence, autonomy support, and communication skills [15]. Together, these findings emphasize the potential value of providing peer mentors with training focused on building person-centred communication skills and conversational styles.
A consideration for developing peer mentorship training based on these research findings may include guidance in the communication skills used in motivational interviewing. Research suggests that practitioners who use motivational interviewing consistent techniques are more likely to improve the sense of collaboration between practitioners and clients and promote clients' sense of autonomy [32,[45][46][47][48]. In contrast, techniques that do not align with the motivational interviewing approach tend to evoke a sense of defensiveness and may ultimately reduce the sense of collaboration between practitioners and clients [32]. Peer mentors in this study were found to use a variety of motivational interviewing consistent techniques (e.g., reframe). However, motivational interviewing inconsistent techniques (e.g., advice without permission) and techniques that may serve as a conversational roadblock (e.g., closed question) were used more frequently. Thus, it may be useful to train peer mentors to use motivational interviewing consistent techniques to improve collaboration between peer mentors and mentees.
The techniques used by peer mentors and mentees also align with research pointing to the value of personal narratives and sharing of stories for providing social support [49][50][51]. Narratives are suggested to be a useful method for explaining complex  Significance set at p < 0.05. � p < 0.05, �� p < 0.005, ��� p < 0.0001. r < ¼ 0.1 small, 0.1 > r < ¼ 0.3 medium, r > 0.5 ¼ large [38].
topics [49]. Stories of others with similar lived experiences appear to deliver more than just information to peer mentees living with an SCI and may provide mentees with a sense of reassurance and support [49]. Peer mentees may also be affected by how peer mentors respond to their narratives [50]. A peer mentor's responses to a mentee's narratives may serve to either validate or deny their experiences [50]. Given the frequency with which peer mentors and mentees discuss personal information during SCI peer mentorship conversations, it may be important to understand how peer mentors share their stories and respond to mentees' stories.
Our research also suggests that mentees may also play an important supportive role in peer mentorship conversations, as both roles used all but one technique. These results indicate the role of mentor versus mentee may not be static or fixed. Individuals who initially identified themselves as the mentee in the conversation may 'switch' to being the mentor, using techniques related to providing support (e.g., advice with permission) rather than receiving support (e.g., follow/neutral) depending on the topic or expertise of the mentee or mentor. This conversation style differs from what is typically seen in motivational interviewing-based interactions in which the practitioner and client roles are discrete. In a typical motivational interviewing counselling session, practitioners generally use techniques that aim to promote change or provide support, and clients respond with techniques that may indicate their motivation to change or not to change [18,47]. The difference between these conversation styles emphasizes the fluid nature of peer mentorship conversations and the desire for peer mentees and mentors to reciprocate support through their own expertise and lived experience.

Limitations and future directions
Some limitations of this study may impact the generalizability of our findings. First, participants only had one peer mentorship conversation. Although participants in this study had a wide range of experience with peer mentorship and some had previous relationships, our findings may differ from techniques used at different stages in the peer mentor-mentee relationship. Second, participants were only recruited from three SCI peer mentorship programs across Canada. While our design was co-developed to be broadly generalizable to these programs, our findings are not necessarily generalizable to peer mentorship programs offered in other contexts [37]. Finally, all conversations took place over the phone, while peer mentorship conversations in real-world settings often occur in-person or using audio-video technology (e.g., Zoom). Related to the validity and reliability of the resulting coding manual, only three coders who had prior experience with abductive analyses were involved in this study and only the facevalidity of the coding manual was assessed. Coder agreement when comparing the inductive codes to the BCTTv1 was low and should be taken into consideration when interpreting the findings. Low coder-agreement may relate to the training of the coders or lack of clarity of the inductive code definitions. Although both coders were trained in using the BCTTv1 using the BCT online training (see: https://www.bct-taxonomy.com), the recency of training or subsequent practice may have influenced coder skill. Additionally, our analysis only accounted for the frequency of topics and techniques and did not account for the duration. For example, a topic may have only been coded twice, but participants may have spoken about that topic at length. Finally, six participants (20%) dropped out of the study due to illness, scheduling, or technical difficulties. This dropout rate may suggest significant selection bias and should be considered when interpreting the findings.
To address these limitations, future research should examine peer mentorship conversations at varying durations of a peer mentor-mentee relationship, using different modes of delivery, and in different contexts. Our findings should also be expanded by considering the impact of non-verbal communication in peer mentorship conversations that occur in person or using audiovideo technology. Research should further examine the validity and reliability of this coding manual. Next steps should include assessing measures beyond face validity and the extent to which our coding manual can be used reliably by multiple coders with varying levels of coding experience. Future research should also investigate additional analyses of the topics and techniques used during peer mentorship conversations, such as the duration of codes or the patterns in which they are used. Using a dynamic systems approach may be useful to understand how these techniques and topics are applied dyadically [52]. For example, Shaw et al. [53] applied this coding manual to understand the influence of conversation modality on the conversation techniques used during peer mentorship conversations.

Practical applications
The resulting coding manual may have several practical future applications. The topics and techniques identified in this research may provide an important first step in informing what should be included in trainings for SCI peer-mentors. The resulting coding manual highlighted several topic areas where peer mentors should be equipped with the appropriate knowledge and resources and clearly understand their scope of practice. This research also suggests that training in relational conversation styles such as motivational interviewing may be more valuable than training in BCT delivery. For example, the findings were used to developed a brief motivational interviewing seminar that was delivered to peer mentors at SCI Alberta, SCI British Columbia, and SCI Ontario with the support of a University of British Columbia Community-University Engagement Support (CUES) Fund award. The seminar was based on the spirit of motivational interviewing and the foundational techniques to build upon the conversation techniques identified in this research.
However, it is important to note that this coding manual only describes what could occur during peer mentorship conversations, not necessarily best practices. Best practices could be identified by applying the coding manual to mentorship conversations in real-life and research-based settings alongside outcome evaluations using methods such as those adopted by Shaw et al. [53]. Future efforts that adopt this research design would help to identify topics and techniques associated with positive outcomes and uncover how peer mentors help people adjust, adapt, and thrive after an SCI.

Conclusion
This research provides new insights into the topics discussed and techniques used in real-world peer mentorship conversations between people living with an SCI. The findings of this research are valuable for understanding the mechanisms of peer mentorship conversations that may enhance the rehabilitation process and help people living with an SCI adjust, adapt and thrive. Providing a manual to characterize peer mentorship conversations may support the evaluation of peer mentorship programs, which can ultimately improve how SCI peer mentorship programs support people living with an SCI. Insights from this research may also inform future examinations to understand the value of peer mentorship for rehabilitation in other populations.

Data deposition
The dataset generated and analyzed during the current study are available on Open Science Framework (see osf.io/qszr9).

Ethical approval
We certify that all applicable institutional and governmental regulations concerning the ethical participation of human volunteers were followed during the course of this research. All methods were approved by the University of British Columbia Research Ethics Board.