Clinician attitudes and behaviours related to brief intervention
To our knowledge, this study is the third residency-based SBI training program to demonstrate positive changes in clinician attitudes and behaviours related to alcohol intervention. Researchers at the University of Massachusetts previously reported increases in readiness to intervene and in actual performance of brief interventions performed by residents and faculty physicians in a similar program which also provided clinician training, routine screening, and clinician prompting with screen-positive patients [24, 37]. Wilk and Jensen  reported increases in brief interventions by residents in interviews with unannounced standardized patients following brief intervention training. In our study, both faculty and residents showed significant increases in their certainty in diagnosing PD after training. Residents showed greater increases in diagnostic certainty and intervention rates than faculty, an important finding in light of evidence from our study and others  that residents are less likely than faculty to perform brief interventions. Because previous studies have demonstrated that not all alcohol-related discussions include advice to reduce drinking, a key element of SBI cited by the US Preventive Services Task Force , our evaluation was focused on the percentage of patients who actually received advice to reduce drinking. During the project's intervention phase, clinicians reported a modest increase in providing advice to reduce drinking (from 6.8% to 8.6%). While this increase did not reach statistical significance (p = .287), a significant increase was seen in the percent of perceived problem drinkers receiving such advice (50% to 75%, p = .047). These findings, although based on small numbers of encounters with problem drinkers, are consistent with previous studies indicating that clinicians who have received SBI training are more confident in their ability to conduct brief interventions and more likely to intervene with problem drinkers [26, 39, 40]. In contrast to some earlier studies which found younger clinicians to be more willing to intervene than older clinicians [39, 41, 42], this study found no impact of age or gender on confidence in diagnoses of problem drinking or advice given to reduce drinking. Reasons for this finding are unclear, but could be related to the relatively young age of the overall group (mean age 37, with only two clinicians over age 50) or to the fact that clinicians of all ages received intensive training, which has been shown to correlate with greater clinician confidence and performance levels [40, 42, 43]. Further research is needed to determine which of the training program's multiple components – experiential training, implementation of routine alcohol screening performed by nurses, prompting clinicians with positive screening results and assessment data, or compliance feedback regarding intervention rates – were most critical in achieving increased intervention rates.
Overall, our study indicated that faculty prescribed reduction in drinking to more patients than residents did, but that resident intervention rates showed greater increases after training than faculty rates. Resident interventions did, however, show a non-significant trend toward decline toward the end of the one-year study. While this could represent a loss of training effect over time, evaluation results also may have been confounded by conducting the final exit interview evaluation in July, when skilled third-year residents had just graduated and newly-promoted residents were struggling to manage increased patient volumes. Interestingly, intervention prompt forms reviewed for the previous report on this project  indicate that there was no decline in resident interventions when prompted with positive screening results during this period: brief interventions were performed in 75% (6/8) of cases. Regardless of the reason for the decline, findings suggest that reinforcement methods such as booster sessions are needed to maintain behaviours taught in the initial training sessions.
One of this study's most encouraging findings is the fact that the number of patients receiving advice to reduce drinking after SBI training actually exceeded the number of patients felt to be problem drinkers. This finding suggests that the SBI program was successful in legitimizing and normalizing conversations about alcohol, such that clinicians felt more at ease in addressing alcohol use in a variety of clinical scenarios, and indicates that at-risk drinkers as well as problem drinkers received brief advice to reduce their drinking.
Recognition of problem drinking
In contrast to our expectation, clinician exit questionnaires did not reflect increased recognition of problem drinking after program implementation. While this could be due to the relatively high levels of clinician recognition at baseline, it could also reflect decreased clinician vigilance once routine screening protocols were in place. This finding, if confirmed in other studies, has implications for future SBI training programs, especially in light of the fact that most "routine" screening systems are not effective in detecting all risky drinkers. Clinician training should include reminders that significant numbers of risky drinkers may remain unscreened or escape detection via questionnaire screening, whose sensitivity rarely exceeds 85%, and clinicians must remain alert to clinical clues related to hazardous or problem drinking.
Limitations of this pilot study
It is important to take into account possible methodological limitations of this pilot study. First, the sample size was not large, particularly during the baseline assessment period. However, the numbers in each group were sufficient to detect between-group differences of at least 10% or greater in recognition of problem drinking and at least 10% or greater in advice to reduce drinking. Secondly, a change in methodology during the study's second implementation phase could have confounded study results. During this period, exit questionnaires were attached to each patient's routing form, and clinicians were requested to complete them on each patient seen. The return rate was low (44%), and selection bias may have occurred, perhaps in favour of patients whom clinicians identified as problem drinkers or in favour of patients who received an intervention. Nonetheless, both clinicians' recognition (6.7% of patients) and intervention rates (8.0%) during this phase are within the range of the other phases of the study and do not suggest selection bias. Thirdly, this study lacked a criterion diagnosis for confirming problem drinking in patients considered by clinicians to be problem drinkers. While some patients may have been incorrectly diagnosed, rates of problem drinking recognition by clinicians throughout this study are similar to this study's previously-published problem drinking estimates obtained by AUDIT-C questionnaire screening, and only slightly lower than the estimated U.S. problem drinking prevalence of 11% in primary care . Future studies comparing clinician's impressions with the results of standardized diagnostic interviews for problem drinking could help to clarify this issue.