Discussion
Only three counsellors had 3 or more audio-recorded sessions for each of the 11 assessment periods and were thus eligible for inclusion in the study. This factor severely limits the ability to draw conclusions based on the results. However, including counsellors with fewer recorded sessions from each assessment period would have made the MITI rating of counsellors' performance too vulnerable to difficulties experienced in individual sessions, such as particularly "difficult" clients, and compromised the study's usefulness as a model for large scale replication.
The findings of the present study suggest that in spite of its theoretical simplicity not all practitioners necessarily learn MI easily. Any successful implementation in a naturalistic clinical setting requires ongoing supervision of counsellors, including feedback and monitoring clinical practice. These findings are in line with previous research that sought to evaluate MI training and supervision interventions [
9‐
11].
Like several other studies on MI training using objective outcome measures, the present study finds improvements in MI skill, for example, an increased use of reflections, after the initial workshop [
7‐
9,
28,
29]. In a recent review only 1 out of 11 studies reported no significant difference post training relating to MI skill [
7]. Moreover, our findings show that counsellors improved their MI skill compared to baseline level of skill in most of the MITI variables, though some variables took longer to improve. Most interestingly, the MITI variables continuously and steadily improved over the two-and-a-half year study period. From a client point of view, this means that the counsellor sessions were on average conducted more skilfully each assessment period.
Although the mean for the three counsellors included steadily improved throughout the study period, there were great variations in skill among counsellors and in learning time. There were no differences in background of counsellors included in the study that might explain the variance in acquisition of MI skill and performance over time. Literature in the field has found little or no correlation between the level of education or degree of work experience and positive outcomes, although variations in outcome have been found across counsellors [
30‐
32]. However, none of this helps to explain the causes of particular counsellors' skill in certain variables.
According to counsellor comments, reduced in-session MI performance might depend on having "a bad day", tiredness and personal issues that negatively affect the ability to maintain appropriate focus, external disturbances of counsellor sessions in the workplace, and variations in difficulty of client problems. Such factors may account for some of the variations, both across counsellors and in counsellors over time. Moreover, differences in acquisition of MI skills could relate to the degree of motivation or readiness to change in counsellors. The study design might be improved by including an independent measure of counsellor readiness to change, or motivation to learn MI. More research is needed into possible causes for differences in skill and outcome in clients across clinicians.
The variability in acquisition and performance of MI skills, and the fact that performance within counsellors for specific skills varies over time, highlight the challenges facing those who wish to implement MI in real-world settings. The differences in ease of learning and learning time are of sufficient significance to raise the issue of a selection procedure for MI training (as has been suggested in the literature) [
31,
33], as well as rendering deeper consideration of training and supervision methods necessary. It may be that organisations wishing to implement MI would better apply their resources to ongoing supervision and treatment integrity assessment of MI practice, as opposed to only investing in costly MI workshop training - or at least develop a programme for continuous development of MI skill.
With regard to which MI skills were most and least easily acquired by counsellors, the Ratio of Reflections to Questions demonstrates a rapid improvement, suggesting that reflective listening may not be difficult for counsellors to learn. After approximately six months, the mean for the nine sessions coincided with the MITI Manual's recommended threshold for Beginning Proficiency. This is encouraging, as use of Reflections contributes to Empathy, which is thought to set a necessary stage for MI sessions and has been shown to be predictive of outcome [
34,
35]. From a pedagogic viewpoint, this finding favours MI training beginning with reflection or "OARS" (open questions, affirmations, reflections, summaries) type skills, which were the focus of all training schemes reviewed by Madson et al [
7].
In the present study, it took two-and-a-half years for all three counsellors to reach the recommended threshold for Beginning Proficiency in the MI Spirit variables, which have also been found to have bearing on outcome [
3]. MI training might benefit from greater concentration on skills associated with MI Spirit, such as eliciting change talk, by adding training designed to meet these needs after the initial workshop. Another finding, however, was that there were great variations in MI Spirit skill acquisition across counsellors, with C2 reaching Beginning Proficiency after a few months, and C1 only reaching Beginning Proficiency after two-and-a-half years of practice and supervision. This suggests that the ability to learn to use some MI techniques will vary considerably across potential MI practitioners.
The MITI Direction variable was the only of the global variables that remained roughly constant for all counsellors throughout the study. This is consistent with the backgrounds of the counsellors included in the study. All counsellors included were highly experienced in ordinary smoking cessation counselling. Unlike the other of the global MITI variables, a high score for Direction does not necessarily reflect a better use of MI, although the reverse is true, that is, in order for a session to reflect good use of MI, it must have a high Direction rating [
2].
SNTQ counsellors have a manual comprising a number of Closed Questions, concerning, for example, client tobacco consumption and smoking patterns, which might have affected the proportion of Open Questions to Open and Closed Questions index negatively. Moreover, coders did not code counsellor questions on practical issues such as whether clients wanted material sent to them, or their contact details, since some counsellors included this as part of the session whereas others did not.
The data represents counsellor performance as acquired and retained over an extended period of time, in an ordinary clinical setting of the type where MI might be ordinarily implemented, thus providing a view of MI that is substantially different from typical RCTs demonstrating its efficacy. The great variation of MI skill in counsellors who have undertaken the same amount of training and supervision suggests that treatment integrity is crucial, in research as well as in ordinary clinical practice. In research aiming to accurately assess the efficacy of MI, the first step must be to ensure that practitioners responsible for delivering the treatment are indeed using MI (or a treatment which bears sufficient similarity to MI). The present findings suggest that attendance of initial MI training is not a good measure of MI skill. In a recent review of efficacy of MI in smoking cessation counselling, only 1 of 14 selected studies used a validated treatment integrity instrument [
5]. Failure to account for the extent to which the proposed treatment intervention has been successfully implemented, through treatment integrity assessment, risks discrediting the treatment method itself, whereas the absence of positive outcome may instead be attributable to failed implementation. It is submitted that the use of MI in ordinary clinical practice ought to be subject to (at least) periodic assessment of treatment integrity using a validated instrument. Results of such assessment would not only serve as a trigger for "refresher" sessions where MI was not successfully implemented, but also provide a valuable learning tool for continued development of MI skills. Treatment integrity assessment would also help to ensure that clients receiving purported MI treatment actually receive that to which they consent.
One possible limitation is the nature of the SNTQ setting (self-initiated contact) that suggests considerable readiness for change among clients. However, it has been suggested that if a client is ready for change, the use of MI may stifle their desire to change [
2]. It is possible that this is common in SNTQ clients, and identified by (some) SNTQ counsellors. Consequently, for clients who no longer display resistance or ambivalence towards smoking cessation, some counsellors may switch their approach to cognitive behavioural therapy inspired counselling. Such in-session behaviour would result in lower MITI scores, but not necessarily negatively influence outcome. To this extent, the validity of MITI is limited insofar as it is not constructed to capture complex therapist competence manifested in intentional and strategic use of MI [
13], therefore potentially underestimating practitioner skill.
In spite of the limitations resulting from the small number of included counsellors, the present study has several important advantages with regard to design. The authors are not aware of any other studies where MI training has been followed by continual supervision and monitoring of counsellor MI skill in an ordinary clinical setting over as long a time period as the present study. Most other studies into the acquisition and retention of MI skills are limited to workshop format training efforts and pre- or post-training evaluations. Only 3 studies of 27 included in a recent systematic review by Madson et al. covered workshop and supervision [
7], and only 1 described coaching after initial MI training [
9]. This is despite the suggestion that a workshop-only format, although yielding improvements in knowledge and confidence, tends not to achieve competency or longevity in all MI variables [
36,
37]. Given these findings, and the considerable resources spent on attempts at en masse implementation of MI using workshop-only formats, studies covering longer time periods that examine training and supervision interventions in relation to acquisition and retention of MI skills in participants are essential, and the present study may provide a blueprint for future large scale research.
The study is one of the first to provide empirical data relating to Miller and Moyers' "Eight Stages in Learning Motivational Interviewing" [
38]. However, contrary to what is suggested by Miller and Moyers, the results suggest that the OARS skills that form part of the second stage in learning MI are the skills that counsellors exhibit first in their clinical practice. The present study's design would be suitable for large sample replication to test the authors' hypotheses. Such a replication study has the potential to influence the MI training literature in an important manner. Similarly, the present study's design might serve as a basis for (large sample) replication to empirically test the recommended thresholds for Beginning Proficiency suggested in the MITI 3.0 manual, if outcome measures are related to MI skill.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
LF drafted the manuscript, and made substantial contributions to the interpretation and analysis of results. LGF contributed to the conception and design of the study, in addition to the interpretation of the data and the initial drafting of the manuscript. HL contributed to the statistical analysis of data with regards to tobacco abstinence in clients, and made helpful comments on previous drafts of the manuscript. AH contributed with the material from the SNTQ, in addition to helpful comments on previous drafts of the manuscript. All authors approved the version to be published.