|Hettema J, Steele J, Miller W R
This review assessed the effectiveness of motivational interviewing techniques to prepare individuals for changes in behaviour. The authors concluded that there is a strong evidence base to support the use of motivational interviewing in changing addictive and health behaviours. However, effect sizes should be regarded with caution given the large variability between the pooled studies and outcomes.
To summarise the evidence for motivational interviewing (MI) techniques to prepare individuals for changes in behaviour.
PsycINFO was searched using the term 'motivational interviewing'; the dates of the search were not reported. In addition, the authors searched the reference lists of the motivational interview website and previous reviews for further studies.
Study designs of evaluations included in the review
Any study design was eligible for inclusion in the review, but only comparative studies such as randomised controlled trials (RCTs), quasi-RCTs, non-randomised controlled trials and cohort studies were eligible for inclusion in the meta-analysis.
Specific interventions included in the review
Participants included in the review
The authors did not specify any inclusion criteria for the participants. The participants included in the review were drawn from the following behavioural domains: alcohol use, smoking, human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS), drug abuse, treatment compliance, gambling, intimate relationships, water purification/safety, eating disorders, and diet and exercise. The age of the included participants ranged from 16 to 62 years (mean 34.11).
Outcomes assessed in the review
Only studies that assessed at least one post-treatment outcome were eligible for inclusion in the meta-analysis. The number of outcome variables reported in the included studies varied between 1 and 12 per study. Variables focused on measures of addictive behaviours and health behaviours; treatment adherence was also reported. Studies varied in the behavioural domains and scales reported. Further details of the individual study outcomes were provided.
How were decisions on the relevance of primary studies made?
The authors did not state how the papers were selected for the review, or how many reviewers performed the selection.
Assessment of study quality
The methodological quality of the studies was assessed according to 12 criteria taken from a published coding manual: method of treatment assignment; use of treatment quality control; rate of follow-up; duration of follow-up; method of data collection; collateral verification of self-report data; objective verification of follow-up data; inclusion of drop-outs in the analysis; adequate consideration of losses to follow-up; blinded outcome assessment; appropriate statistical methods; and use of multiple study sites.
Each study was awarded a quality score from 0 (low) to 16 (high). The authors did not state how many reviewers performed the validity assessment.
Two independent reviewers coded each study and extracted data according to an adapted coding manual, based on a tool used in previous outcome reviews (see Other Publications of Related Interest no.1). Any discrepancies were resolved by consensus. The amount and type of MI training provided and the specific components of MI included were extracted.
For each study, effect sizes with 95% confidence intervals (CIs) were calculated for each outcome measure where there was sufficient data available (i.e. mean, SD or sample size). Significance tests were also used to estimate effect sizes. Study authors were contacted for missing information and, in the event that effect sizes were still unavailable, data quoted in previous meta-analyses were used (see Other Publications of Related Interest nos.2-4). Zero effect sizes were assigned where the p-value was greater than 0.05 and no effect sizes could be calculated. In addition to individual effect sizes for each outcome, a combined effect size for each individual study was calculated by averaging all variables at each follow-up point using weighted linear combinations. The variance of these combined effect sizes was minimised by assigning weights inversely proportional to the variance of each effect size to each variable included in the analyses. Follow-up periods were classified into five groups and data were combined within these groups: 1 to 3 months; 4 to 6 months; 7 to 12 months; 13 to 24 months; and longer than 2 years. The effect sizes for each time interval were then compared with the baseline level for each study variable.
Methods of synthesis
How were the studies combined?
The studies were combined according to the target behaviour under investigation (i.e. alcohol use, smoking, HIV/AIDS, drug abuse, treatment compliance, gambling, intimate relationships, water purification/safety, eating disorders, and diet and exercise) and comparison group (untreated control, active treatment, MI given in addition to another treatment). Where possible, differentiations were made between pure, stand-alone MI and MI combined with other therapies. Combined effect sizes were calculated and reported with 95% CIs.
How were differences between studies investigated?
Statistical heterogeneity between studies used in the calculation of between-group effect sizes was assessed using the Q statistic. Regression and correlational analyses were performed to assess the effects of MI duration, purity, counsellor training, post-training support, quality score, number of outcome variables, longest follow-up point, type of comparison group or problem area, gender, age, ethnicity and problem severity on the study outcomes. Potential variables studied in alcohol-related studies were blood alcohol level and alcohol-related problem categories. For smoking-related studies, abstinence variables and quit variables were used. HIV variables included knowledge, behavioural intentions and sexual risk behaviours.
Results of the review
Seventy-two studies (n=14,267) were included in the review.
The quality scores ranged from 4 to 16 points (mean 10.8, SD=2.4).
The combined effect sizes (for all outcome variables and follow-up points) for each individual study ranged from -0.19 to 3.25 (mean 0.43, SD=0.62). Thirty-eight of the 72 studies (53%) showed a significant effect favouring MI (p<0.05).
Based on meta-analyses of all comparative studies in each behavioural area, the effect sizes combined for all variables and follow-up time periods were: alcohol, 0.26 (95% CI: 0.18, 0.33; 31 studies); smoking, 0.14 (95% CI: 0.09, 0.20; 6 studies); HIV, 0.53 (95% CI: 0.24, 0.81; 5 studies); drugs, 0.29 (95% CI: 0.15, 0.43; 13 studies); treatment adherence, 0.72 (95% CI: 0.56, 0.89; 5 studies); gambling, 0.29 (95% CI: 0.16, 0.42; 1 study); water purification/safety, 0.30 (95% CI: 0.05, 0.55; 4 studies); and diet and exercise, 0.78 (95% CI: 0.41, 1.16; 4 studies). The only behaviour in which MI did not appear beneficial overall (all effect sizes over all time points) was eating disorders (effect size -0.07, 95% CI: -0.42, 0.26); however this was not a significant finding and was based only on 1 study.
Overall, the strongest evidence base was for addictive behaviours, with the largest effect sizes (>0.7) observed in studies comparing MI with no treatment, waiting-list control or education, or studies adding MI to other standard treatment. However, MI did not appear effective in smoking cessation. In terms of health behaviours, large but inconsistent effects were seen in HIV studies, encouraging effects were observed for diet and exercise programmes, and large effects were seen for encouraging water purification/safety in African villages. Large effect sizes were also reported for studies of treatment adherence; the effects appeared to persist or increase with time when MI was added to an active treatment. Variations in the delivery of MI made substantial differences to the effect sizes. A trend for decreasing effect sizes over time was also apparent for combined effect sizes over all studies: the effect size was 0.77 (95% CI: 0.35, 1.19) at 0 to 1 month post-treatment, 0.39 (95% CI: 0.27, 0.50) at 1 to 3 months, 0.31 (95% CI: 0.23, 0.38) at 3 to 6 months, 0.30 (95% CI: 0.16, 0.43) at 6 to 12 months, and 0.11 (95% CI: 0.06, 0.17) at longer than 12 months.
Combined effect sizes for combined variables across all time points for alcohol studies were 0.22 (95% CI: 0.10, 0.34) for blood alcohol concentration and 0.08 (95% CI: -0.02, 0.19) for alcohol-related problems. For smoking studies, the effect sizes were 0.15 (95% CI: -0.06, 0.23) for abstinence and 0.11 (95% CI: 0.00, 0.21) for quit attempts. For HIV studies, the effect sizes were 1.46 (95% CI: -0.54, 3.45) for knowledge, 0.88 (95% CI: 0.05, 1.72) for behavioural intentions and 0.07 (95% CI: -0.05, 0.19) for sexual risk behaviours.
In the regression analyses, the only significant correlates were found to be use of an MI manual, which produced smaller effect sizes (trend, t=1.53; p=0.28), and ethnicity, where larger effect sizes were associated with minority groups (significant trend, t=-0.39; p<0.05).
Further details of the results were provided.
There is a strong evidence base for MI in preparing individuals for behavioural changes in the areas of addiction and health. In addition to being useful as a stand-alone intervention, it also appears to improve outcomes when added to other treatment approaches.
This review was based on a very broad research question, with broad inclusion criteria. The literature searches were limited to one search term in one database and cross-checks with reference lists from previous reviews. In addition, the search dates were not reported. Given the limited searches and the lack of specific attempts to identify unpublished material, it is possible that relevant material was missed. It was also unclear whether the authors took appropriate steps to reduce errors and bias in the selection of studies and assessment of study quality, although two independent reviewers were responsible for extracting the study data, which suggests that the potential for errors and bias was reduced in this process.
Overall, given the variation between the studies and the use of pooled outcomes and time periods, the overall effect sizes should be treated with caution. The authors stated that they used the Q statistic to assess the level of heterogeneity in their analyses, but results were not reported in this publication. The large amount of data and the use of supplementary material were unavoidable given the size of the project, but it makes it difficult for the reader to check the review's findings. In general, the direction of the effect sizes do support the reviewers' conclusions and steps were taken to try and investigate the effects of potential confounding variables, but caution is advised and further research is required.
Implications of the review for practice and research
Practice: The authors did not state any implications for practice.
Research: The authors stated that further research to clarify the processes behind their observations and to investigate methods to further maximise effects is required, in particular, methods to help practitioners improve their proficiency in delivering the intervention. Further research is also needed to determine the true effects of MI in smoking cessation and the reasons behind the discrepant findings identified in this review. Investigations to identify factors which may influence the effectiveness of MI are also required. Any future studies should, however, ensure that they adequately report how those implementing the intervention were trained. Studies should also ensure that the delivery of MI is validated and that process measures related to outcomes are included. The authors stated that the finding that manual-guided MI was associated with smaller effect sizes should be investigated further.
Hettema J, Steele J, Miller W R. Motivational interviewing. Annual Review of Clinical Psychology 2005; 1: 91-111
Other publications of related interest
1. Miller WR, Wlbourne PL. Mesa grande: a methodological analysis of clinical trials of treatments for alcohol use disorders. Addiction 2003;97:265-77. 2. Bien TH, Miller WR, Tonigan JS. Brief intervention for alcohol problems: a review. Addiction 1993;88:315-36. 3. Burke BL, Arkowitz H, Menchola M. The efficacy of motivational interviewing: a meta-analysis of controlled clinical trials. J Consult Clin Psychol 2003;71:843-61. 4. Dunn C, Deroo L, Rivara FP. The use of brief interventions adapted from motivational interviewing across behavioral domains: a systematic review. Addiction 2001;96:1725-42.
Subject indexing assigned by NLM
Humans; Interview, Psychological; Motivation; Psychotherapy /education /methods; Regression Analysis; Substance-Related Disorders /therapy; Treatment Outcome
Database entry date
This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.