Background
Screening and brief intervention (SBI) for risky alcohol use in primary care settings has demonstrated efficacy. In a meta-analysis of 23 controlled studies, patients identified by screening in primary care settings who received BI consumed less alcohol on average (3.6 fewer drinks per week) and were less likely to exceed limits (12% fewer) than control patients at one-year follow-up [
1]. However, despite overall positive findings, there are high levels of variability in BI effects, with significant unexplained heterogeneity in outcomes [
1‐
3]. In fact, the number of studies that do and do not find significant BI effects are nearly evenly split [
4]. This variability may be attributable in part to differences in the intervention components and communication style that underlie tested BI protocols, including fidelity to motivational interviewing (MI).
MI is a patient-centered communication style that seeks to draw out and enhance patients’ own motivation to change health behaviors [
5]. Studies of SBI frequently report that the tested BI protocol was conducted in a manner that was consistent with an MI approach, but additional methodological details are often lacking. For example, Richmond et al. [
6] briefly mention that “Patient counseling included techniques of motivational interviewing,” Senft et al. [
7]. mention that research staff “were trained in principles of motivational interviewing,” Fleming et al. [
8] included a workbook “based on the principles of motivational interviewing,” and Reinhardt et al. [
9] provided counseling sessions “based on motivational Interviewing,” in which “the main content of each intervention was the enhancement of motivation.” None of these examples provide a specific description of how MI was incorporated into the intervention.
There are several potential ways in which MI principles or skills could be integrated into BI. For example, interventionists could be trained to take a collaborative or empathic stance, to avoid potential roadblocks to communication such as labeling and unsolicited advice, to use specific frequencies of communication skills such as complex reflections, to conduct specific exercises, to adhere to different decision rules regarding patient readiness levels, or some combination of the above. Lack of methodological details regarding the content and structure of brief interventions makes it difficult to assess the moderating impact of MI dose on outcome across trials, and primary care-based meta-analyses to date have been unable to report on the impact of this construct [
1‐
4].
Although MI has been found to be effective for addressing alcohol issues in medical care settings [
10] and aligns with several principles of medical ethics [
11], there are many barriers to implementing MI in brief medical interactions. Healthcare providers are faced with many competing priorities and limited time with patients, creating pressure to conduct preventive and behavioral health interventions as efficiently as possible [
12]. Conducting MI, even its briefest applications, is more time consuming than simple advice giving. In addition, MI is a complex set of skills that requires flexible application and contingent responding to moment-to-moment changes in what patients say [
5]. Proficiency in the approach requires extensive training time, ongoing supervision and resources that are often limited among medical trainees and clinicians. A recent review of MI training for healthcare professionals found that average training duration exceeded 1 day and in many cases such training failed to yield clinical improvements in MI proficiency [
13]. Lastly, medical clinicians themselves often perceive barriers to implementing MI, including concerns about lack of time, low levels of perceived responsibility, low confidence, and lack of support [
14].
To date, there is a dearth of empirical work examining mechanisms of behavior change in SBI for risky alcohol use [
15]. Of the available research that has focused on the impact of MI within SBI, results have yielded mixed findings [
16]. For example, Bertholet et al. [
17] re-analyzed combined data from three controlled BI trials for proficiency in important MI behaviors and found no consistent predictors of outcome. In a similar secondary analysis designed to test the mediating link between MI-consistent behaviors such as open-questions, affirmations, and reflections, client change talk, and outcome proposed by the prevailing MI model [
18], this hypothesized causal chain was only supported in specific conditions, such as with therapists who had more than 3 years experience [
19].
There have been some attempts to compare different SBI approaches prospectively. For example, the Effectiveness of Screening and Brief Alcohol Intervention in Primary Care Trial (SIPS trial) compared three brief intervention conditions for hazardous drinkers in primary care settings: an informational leaflet, a brief advice session, and an MI-informed brief lifestyle counseling session [
20]. Results of the SIPS showed no significant differences in drinking reduction between the three conditions at follow-up, although patients in the lifestyle counseling group had significantly increased readiness to change and satisfaction with the intervention. While this trial did compare different intensities of intervention, it was confounded by content and not specifically designed to allow for inferences regarding the role of MI strategies on outcome, as treatment modalities were additive, with all participants first receiving brief advice. Additionally, the content and structure of brief advice was not tightly controlled and study authors report that study interventionists, who were clinic-based providers, often extended into more therapeutic tasks.
More work has been done with regards to alcohol treatment, which may or may not be generalizeable to SBI in primary care. Here, findings are mixed, with some studies supporting the use of specific MI structure and content, and others not. For example, Morgenstern and colleagues attempted to dismantle the active components of MI treatment by prospectively comparing 4-sessions of MI, to non-directional MI, to a no treatment control condition and found similar outcomes across condition [
21]. On the other hand, Field and colleagues compared brief advice, brief MI, and brief MI plus a feedback based telephone booster among trauma center patients who recently consumed alcohol and found that the BMI plus booster condition outperformed other treatments [
22]. Apodaca conducted a recent secondary analysis of a brief intervention for college students with problems drinking and found that therapist use of Mi microskills such as reflections and open-questions produced both change and sustain talk [
23]. This trend was also observed on a recent multivariate meta-analysis of MI process and outcome, which found that both therapist microskills and global ratings predicted change and sustain talk and that sustain talk was predictive of poor outcome, providing equivocal support for the proposed MI theory of change [
24].
While we have evidence to suggest that primary care-based brief advice may be just as effective as MI with smoking cessation [
25], the evidence is not clear for BI for risky alcohol use. One strategy for determining the impact of MI on BI efficacy is by prospectively comparing SBI protocols with varying doses of MI content. While mechanisms of action research have revealed variability in MI within clinical trials, to date, it has not been demonstrated that is possible to create interventions that are reliably differentiable while still maintaining consistency within a general BI structure. In the current study we sought to develop SBI protocols that were differentiable in terms of dose of MI, with the goal of prospectively comparing them in a future controlled trial.
Results
Interrater agreement
Of the 45 sessions, 9 (20%) were co-rated to assess interrater agreement. Rates of interrater agreement were excellent for the fidelity checklist (ICC = 0.93, 95% CI = 0.72, 0.98). All MITI global scores (evocation, collaboration, autonomy support, direction, empathy) were also in the excellent range, with the exception of direction, which could not be calculated because there was no variability in rating. All behavior count threshold (% open questions, % complex reflections, reflection to question ratio, and % MI-adherent) interrater agreement estimates were also in the excellent range.
Participant flow
Of the 502 patients approached by researchers, 413 agreed to be screened for eligibility and 92 (22.3%) screened positive for risky drinking. Among ineligible screens, 291 did not report any past year risky drinking, and 5 exceeded the AUDIT-C upper cutoff. Other common reasons for ineligibility were lack of English fluency (n = 15) and pregnancy (n = 3). Of those who screened positive, 45 (48.9%) enrolled in the study: 15 were assigned to NIAAA, 14 were assigned to the BA, and 16 were assigned to MI. The primary reason for non-enrollment was lack of time.
Participant characteristics
Participants were predominantly female (60%) and 42% were Hispanic. On average, participants reported between 1 and 156 heavy drinking days in the past year (mean 26 (SD = 42)). AUDIT-C scores ranged from 3 to 7 (mean 4.6 (SD = 1.3)).
Timing
Mean time to complete SBI varied across conditions, with BA taking, on average, 7 min (SD = 6, Range 4–14), NIAAA 9 min (SD = 3, Range 6–18), and MI 17 min (SD = 6, Range 5–32).
SBI fidelity
Fidelity to intervention components, as measured by the fidelity checklists was high across all conditions. Overall, the clinician completed between 92 and 96% of protocol elements, with little variability within or across conditions. The mean percent fidelity did not differ across BA, NIAAA, and MI, which had respective scores of 96% (SD = 7%), 94% (SD = 6%), and 92% (SD = 6%) (p = 0.32).
MI differentiability
Significant differences between conditions were found for nearly all MITI constructs (Table
1). All differences were in the expected direction with MI showing the highest doses of MI, then NIAAA, and then BA. The MI condition met or exceeded MI competency standards on all measured domains with the exception of reflection to question ratio. BA and NIAAA did not contain levels of MI that are normally defined as meeting MI competency standards on any of the assessed domains, with the exception of percent complex reflections for the NIAAA condition.
Table 1
MITI scores among conditions
Evocation | 4 | 1.9 (1.0) | (1.2, 2.5) | 3.1 (1.0) | (2.4, 3.7) | 4.8 (0.4) | (4.2, 5.4) | Yes | Yes | Yes |
Collaboration | 4 | 1.7 (0.9) | (1.1, 2.3) | 3.0 (0.9) | (2.5, 3.5) | 5.0 (0.0) | (4.5, 5.5) | Yes | Yes | Yes |
Autonomy support | 4 | 2.9 (0.5) | (2.5, 3.3) | 3.6 (0.7) | (3.2, 4.0) | 5.0 (0.0) | (4.6, 5.4) | Yes | Yes | Yes |
Direction | 4 | 5.0 (0.0) | – | 5.0 (0.0) | – | 5.0 (0.0) | – | – | – | – |
Empathy | 4 | 1.8 (0.9) | (1.1, 2.4) | 2.9 (1.1) | (2.3, 3.5) | 5.0 (0.0) | (4.4, 5.6) | Yes | Yes | Yes |
Percent open questions | 70% | 0.0 (0.1) | (− 0.1, 0.1) | 0.2 (0.1) | (0.1, 0.3) | 0.4 (0.1) | (0.3, 0.4) | Yes | Yes | Yes |
Percent complex reflections | 50% | 0.4 (0.2) | (0.2, 0.5) | 0.5 (0.2) | (0.4, 0.6) | 0.7 (0.1) | (0.6, 0.8) | No | Yes | No |
Reflection to question ratio | 2 | 0.6 (0.2) | (− 0.1, 1.2) | 1.0 (1.2) | (0.4, 1.5) | 1.5 (0.6) | (0.9, 2.0) | No | No | No |
Percent MI adherent | 90% | 0.1 (0.2) | (− 0.1, 0.3) | 0.3 (0.4) | (0.1, 0.4) | 1.0 (0.0) | (0.8, 1.2) | No | Yes | Yes |
Discussion
In this study, we found that it is possible to conduct high fidelity SBI with differing doses of MI. When implemented with risky drinking primary care patients our three SBI protocols yielded high, moderate, and low doses of MI based on MITI score assessments. This study lays for the groundwork for future research to prospectively compare the efficacy of SBI protocols with varying doses of MI. The protocols can be used in their current form to conduct such work and the MI protocol can provide a standardized MI-based SBI protocol for other work. The availability of these protocols may help to shine light on the unclear picture regarding the role of MI in SBI for risky alcohol use in primary care. Similarly, in the area of treatment and in other settings where the impact of MI structure and content on outcome is somewhat equivocal, the development of highly differentiable interventions may allow for improve prospective and retrospective analysis of effects.
The study has several strengths and limitations. The study sample was all English speaking and result may not be generalizeable to other clinic populations. The methods were tightly controlled resulting in strong internal validity and could be easily translated to a future prospective comparison of the interventions using an efficacy trial design. However, the current study does not necessarily reflect rates of fidelity or MI differentiability that might be observed in an effectiveness trial that involves real-world clinicians. Additionally, training and oversight of the study interventionist was much more intensive than might be feasible with non-research staff.
The protocols in the current study were also primarily conducted by one study interventionist. While this factor eliminates the risk of therapist effects confounding observed differences across SBI protocols, it may limit the generalizability of findings. Set the stage for other research that attempts to dismantle without clearly demonstrating idffereintiability.
Overall, this study represents an important contribution to the research base focused on mechanism of action of SBI for risky alcohol use in primary care. Conducting SBI with high doses of MI takes more time and requires more training and oversight, resulting in many feasibility barriers in the current medical training and practice environment. The current study provides researchers with a tool to address this important question. Future research should seek to compare the efficacy of the three SBI protocols, as well as determine the feasibility of training non-research clinicians to conduct the interventions with fidelity to both structure and MI dose, determine the intensity of training and support needed to yield such results and compare the protocols using a effectiveness trial design.
Authors’ contributions
JEH, SAC, KSI, AR, PJL, RS, CMMK, VAC, JEH, SAC, KSI, and PJL conceptualized the work; JEH, SAC, AR collected the data; JEH and CMK analyzed the data; JEH, SAC and AR drafted the manuscript; JEH, SAC, KSI, AR, PJL, RS, CMMK, VAC provided critical revisions to the manuscript; JEH, SAC, KSI, AR, PJL, RS, CMMK, VAC approved the final manuscript and agree to be accountable for its content. All authors read and approved the final manuscript.