A Meta-Analysis of Research on Motivational Interviewing Treatment Effectiveness (MARMITE) Jennifer Hettema Julie Steele William R. Miller Annual Review of Clinical Psychology Vol 1, 2005 (in press)
funded by a grant from The Robert Wood Johnson Foundation
Adoption Curve for Innovations Source: Everett M. Rogers Diffusion of Innovations
Number of MI Publications Source: MARMITE
MI Outcome Trials Source: MARMITE
Inclusion Criteria for MI Trials For within-group effect sizes: At least one treatment group including MI At least one treatment group including MI At least one post-treatment outcome measure At least one post-treatment outcome measure For between-group effect sizes: At least one control or comparison condition without MI components At least one control or comparison condition without MI components Procedure for creating pre-treatment equivalence of groups Procedure for creating pre-treatment equivalence of groups MARMITE
All studies double-coded for: Methodological quality on 12 dimensions Methodological quality on 12 dimensions Other study characteristics Other study characteristics Attributes of the MI intervention(s) Attributes of the MI intervention(s) Within-group effect sizes Within-group effect sizes Between-group effect sizes Between-group effect sizes MARMITE
Effect sizes were computed: For all reported outcome variables For all reported outcome variables At all reported follow-up points At all reported follow-up points For all between-group contrasts For all between-group contrasts With 95% confidence intervals With 95% confidence intervals Correcting for small sample bias Correcting for small sample bias MARMITE
72 studies included so far: Alcohol (31) One study each: Alcohol (31) One study each: Drug Abuse (14)Gambling Drug Abuse (14)Gambling Smoking (6)Eating Disorders Smoking (6)Eating Disorders HIV Risk (5)Relationships HIV Risk (5)Relationships Treatment Compliance (5) Treatment Compliance (5) Water purification (4) Water purification (4) Diet and exercise (4) Diet and exercise (4) MARMITE
Types of Comparisons MI vs. Specified Treatment (25) MI vs. Specified Treatment (25) MI vs. Treatment as Usual (6) MI vs. Treatment as Usual (6) MI vs No Treatment / Placebo (21) MI vs No Treatment / Placebo (21) MI added to Specified Treatment (7) MI added to Specified Treatment (7) MI added to Treatment as Usual (5) MI added to Treatment as Usual (5) Mixed Designs (6) Mixed Designs (6) Within-Group Only (2) Within-Group Only (2) MARMITE
Methodological Quality Compared to 361 alcohol treatment trials: MQS Mean = vs (ns) MQS Mean = vs (ns) Intervention quality control78% vs 57% Intervention quality control78% vs 57% Multisite trials:28% vs. 5% Multisite trials:28% vs. 5% Follow-up > 12 months18% vs. 51% Follow-up > 12 months18% vs. 51% Follow-up completion > 70%45% vs. 75% Follow-up completion > 70%45% vs. 75% MARMITE
Outcome (Dependent) Measures Mean of 3.3 outcome variables per study Mean of 3.3 outcome variables per study Range: 1 to 12 Range: 1 to 12 To avoid capitalization on change, we computed a combined effect size (d) averaging across all reported outcome variables in each study To avoid capitalization on change, we computed a combined effect size (d) averaging across all reported outcome variables in each study MARMITE
Specified Characteristics of MI Being collaborative Being collaborative Client centered Client centered Nonjudgmental Nonjudgmental Building trust Building trust Reducing resistance Reducing resistance Increasing readiness Increasing readiness Increasing self-efficacy Reflective listening Increasing discrepancy Eliciting change talk Exploring ambivalence Expressing empathy MARMITE
Specified Characteristics of MI Of 12 possible characteristics of MI, The average number mentioned was 3.6 Range: 0-12 MARMITE
Treatment “Dose” of MI Average “dose” of 2 sessions (2.2 hours) The contrasts in dose varied from: Comparison group 25 hours longer than MI to MI 6 hours longer than no-treatment MARMITE
Quality Control of MI Average training time: 10 hours (N=13) Average training time: 10 hours (N=13) Manual-guided 74% Manual-guided 74% Post-training supervision 29% Post-training supervision 29% Fidelity checks 36% Fidelity checks 36% MARMITE
Where was MI tested? Outpatient clinics (15) Outpatient clinics (15) Inpatient facilities (11) Inpatient facilities (11) Educational settings (6) Educational settings (6) Community organizations (5) Community organizations (5) G.P. offices (5) G.P. offices (5) Prenatal clinics (3) Prenatal clinics (3) Emergency rooms (2) Emergency rooms (2) Halfway house (2) EAP Telephone (3) In home (1) Jail (1) Mixed (7) Unspecified (8) MARMITE
Who delivered MI? Paraprofessionals / students (8) Paraprofessionals / students (8) Master’s level (6) Master’s level (6) Psychologists (6) Psychologists (6) Nurses (3) Nurses (3) Physicians (2) Physicians (2) Dietician (1) Dietician (1) Mixed (22) Mixed (22) MARMITE
Sample Characteristics (N = 14,267) N = 21 to 952Mean = 198 N = 21 to 952Mean = 198 Males = 54.8% Range = 0 to 100% Males = 54.8% Range = 0 to 100% Mean Age = 34Range = 16 to 62 Mean Age = 34Range = 16 to 62 Ethnic minorities: 43%(N = 37) Ethnic minorities: 43%(N = 37) MARMITE
Some Generalizations Wide variability in effect size across studies, within problem areas (e.g., for alcohol problems, d varies from 0 to 3.0) Wide variability in effect size across studies, within problem areas (e.g., for alcohol problems, d varies from 0 to 3.0) Effects of MI appear early Effects of MI appear early Effects of MI diminish over time, except in additive studies Effects of MI diminish over time, except in additive studies d =.77 at post-treatment d =.31 at 4-6 months d =.30 at 6-12 months MARMITE
Effect Size of MI Over Time MARMITE Controlled Additive Comparative
Effect size was not predicted by: Number of MI attributes mentioned Number of MI attributes mentioned Methodological quality of study, except Methodological quality of study, except Use of a manual to guide MI did predict effect size: Studies not using a manuald =.65 Studies using a manuald =.37 Demographic characteristics, except: Demographic characteristics, except: Anglo/Caucasian samplesd =.39 Minority samplesd =.79 MARMITE
Effect size varied with outcome measures Alcohol: Quantity of drinkingd =.30 Quantity of drinkingd =.30 Frequency of drinkingd =.31 Frequency of drinkingd =.31 BAC estimatesd =.22 BAC estimatesd =.22 Negative consequencesd =.08 Negative consequencesd =.08 HIV Risk: Knowledged = 1.46 Knowledged = 1.46 Behavioral Intentionsd =.88 Behavioral Intentionsd =.88 Sexual risk-takingd =.07 Sexual risk-takingd =.07 MARMITE
Mean Combined Effect Size by Problem Area ( N=72 Clinical Trials) MARMI TE
Conclusions 1. Robust and enduring effects when MI is added at the beginning of treatment MI increases treatment retention MI increases treatment adherence MI increases staff-perceived motivation MARMITE
Conclusions 2. The effects of motivational interviewing emerge relatively quickly (This is also true of other treatments) MARMITE Project MATCH Outcomes
Conclusions 2a. The effects of motivational interviewing emerge relatively quickly This may not be true for certain problem areas or dependent measures where “sleeper” effects occur (e.g., effects of diet and exercise) MARMITE
Conclusions 3. The between-group effects of motivational interviewing tend to diminish over 12 months This is also true of other treatments Between-group differences diminish in part because control/comparison groups “catch up” over time This may not be true of MI’s additive effects with other treatment MARMITE
Conclusions 4. The effects of MI are highly variable across sites and providers This is also true of other treatments, but may be more true with MI Provider baseline characteristics do not predict effectiveness with MI Treatment process variables do Manuals may not be a good idea MARMITE