bs_bs_banner

RESEARCH REPORT doi:10.1111/j.1360-0443.2012.03845.x

Identification of behaviour change techniques to

reduce excessive alcohol consumption add_3845 1..10

Susan Michie1, Craig Whittington1, Zainab Hamoudi2, Feri Zarnani2,

Gillian Tober3 & Robert West4
Centre for Outcomes Research and Effectiveness, Department of Clinical, Educational and Health Psychology, University College London, London, UK,1 Division
of Psychology and Language Sciences, University College London, London, UK,2 Leeds Addiction Unit, Leeds, UK3 and Cancer Research UK Health Behaviour
Research Centre, Department of Epidemiology and Public Health, University College London, London, UK4

ABSTRACT

Background Interventions to reduce excessive alcohol consumption have a small but important effect, but a better
understanding is needed of their ‘active ingredients’. Aims This study aimed to (i) develop a reliable taxonomy of
behaviour change techniques (BCTs) used in interventions to reduce excessive alcohol consumption (not to treat
alcohol dependence) and (ii) to assess whether use of specific BCTs in brief interventions might be associated with
improved effectiveness. Methods A selection of guidance documents and treatment manuals, identified via expert
consultation, were analysed into BCTs by two coders. The resulting taxonomy of BCTs was applied to the Cochrane
Review of brief alcohol interventions, and the associations between the BCTs and effectiveness were investigated using
meta-regression. Findings Forty-two BCTs were identified, 34 from guidance documents and an additional eight from
treatment manuals, with average inter-rater agreement of 80%. Analyses revealed that brief interventions that
included the BCT ‘prompt self-recording’ (P = 0.002) were associated with larger effect sizes. Conclusions It is pos-
sible to identify specific behaviour change techniques reliably in manuals and guidelines for interventions to reduce
excessive alcohol consumption. In brief interventions, promoting self-monitoring is associated with improved out-
comes. More research is needed to identify other behaviour change techniques or groupings of behaviour change
techniques that can produce optimal results in brief interventions and to extend the method to more intensive
interventions and treatment of alcohol dependence.

Keywords Alcohol, behaviour change techniques, brief intervention, taxonomy.

Correspondence to: Susan Michie, Centre for Outcomes Research and Effectiveness, Department of Clinical, Educational and Health Psychology, University
College London, 1–19 Torrington Place, London WC1E 7HB, UK. E-mail: s.michie@ucl.ac.uk
Submitted 27 July 2011; initial review completed 19 October 2011; final version accepted 10 February 2012

INTRODUCTION abstinence [7]. The study of the active ingredients of

Excessive alcohol consumption is a significant public
health problem [1–4]. Interventions to reduce excessive
alcohol consumption, also known as behavioural coun-
selling and behavioural support, have been found to have
a small but important effect in adults in primary care
settings [5]. These interventions are distinct from treat-
ment for alcohol dependence and often involve a single
session of less than 30 minutes involving a patient and a
health professional. Where single sessions are involved,
the term ‘brief intervention’ is usually adopted. The most
common goal is to reduce alcohol consumption to safe
levels [6], but some interventions seek to achieve alcohol
behavioural treatments is important to alcohol interven-
tions, having both theoretical and clinical implications
[8]. It allows investigation of the mechanisms of action of
specific intervention components and is a starting-point
for assessing in what circumstances different types of
intervention are likely to be effective [8].
We know little about what constitute the ‘active ingre-
dients’ of these interventions. This is partly because a
systematic method does not currently exist for reporting
intervention content in terms of specific behaviour
change techniques (BCTs). Despite guidelines for report-
ing evaluation studies that require authors to pro-
vide ‘precise details’ of their interventions [9,10], the

© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction
2 Susan Michie et al.
terminology used to describe the content of alcohol inter-
ventions is inconsistent across studies, with the same
labels used to describe different BCTs or different labels
used for the same BCT [11].
A ‘common language’, that is, a defined, standardized
terminology for describing component BCTs, is desirable
[12]. A BCT is defined as a ‘concrete description’ of the
methods used by practitioners to change behaviour,
defined precisely to allow for future replication [13]. It
is a specific, irreducible, component of an intervention
designed to change behaviour and a putative active ingre-
dient in an intervention. BCTs can be used alone or in
combination with other BCTs and their effectiveness can
be assessed.
There are several methods of specifying the content of
interventions for reducing excessive alcohol consump-
tion, but none do so at the level of specific BCTs [14,15].
The UK Alcohol Treatment Trial Process Rating Scale
(UKATT PRS) [16,17] has been used reliably to rate the
delivery of video-recorded motivational enhancement
therapy (MET) and social behaviour and network therapy
(SBNT) interventions in terms of manual adherence and
practitioner competence, with 11 and nine items, respec-
tively. Some of the items relate to general therapeutic
style (e.g. ‘task oriented’, ‘collaboration’), and those that
relate more directly to behaviour change are relatively
generic (e.g. ‘ambivalence’, ‘create conflict’). The Yale
Adherence and Competence Scale (YACS) [18], developed
from the MATCH Tape Rating Scale (MTRS) [19,20],
is reliable at the level of six subscales of adherence
and competence in ‘assessment’, ‘general support’,
‘goals’, ‘clinical management’, ‘Twelve-Step facilitation’
and ‘cognitive behavioural therapy’. The Motivational
Interviewing Treatment Integrity Scale (MITI) is reliable
for eight scales that cover information, adherence, type
of question and type of reflection [15]. The Adherence-
Competence Scale for IDC for Cocaine Dependence [21] is
reliable for five very general subscales: ‘monitoring drug-
use behaviours’, ‘encouraging abstinence’, ‘encouraging
12-Step participation’, ‘relapse prevention’ and ‘educat-
ing the client’.
Reliable taxonomies of specific BCTs have been devel-
oped in the fields of smoking [22,23], physical activity
and healthy eating [13,24]. These have enabled advances
in understanding associations between intervention
content and outcome [12,25]. To our knowledge, no such
taxonomy exists for interventions for excessive alcohol
consumption.
The reliable taxonomy for smoking cessation [22] was
developed by identifying BCTs from UK clinical guidance
documents of recommended practice for delivering
smoking cessation interventions. Of 43 BCTs identified,
12 were similar, and 31 additional, to the BCTs developed
for physical activity and healthy eating interventions
© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
[24]. The taxonomy of smoking cessation interventions
grouped BCTs according to four functions: addressing
motivation, maximizing self-regulatory capacity and
skills, adjuvant activities and general aspects of the inter-
action (e.g. communication techniques). Development of
a similar taxonomy for alcohol interventions could enable
similar research to be conducted in this domain and allow
comparison of BCTs across behavioural domains which
may help to identify potentially effective BCTs that are not
used in alcohol interventions. This could improve the
quality of primary and secondary research and the devel-
opment of good practice in designing and implementing
treatment to patients and interventions to the general
population.
A robust evidence synthesis has been undertaken for
brief alcohol interventions [11], and this provides an
opportunity to apply any newly developed BCT taxonomy
to investigate whether specific BCTs can be identified as
being associated with better outcomes. Such an analysis
would not necessarily generalize to more intensive inter-
ventions or to treatment of alcohol dependence. However,
it would be a useful starting-point for this line of research.
This study aimed to:
1 develop a reliable method for identifying BCTs
used in interventions to reduce excessive alcohol
consumption;
2 identify BCTs used to reduce excessive alcohol con-
sumption from guidance documents and treatment
manuals; and
3 investigate associations between inclusion of specific
BCTs in brief alcohol interventions and effect sizes of
those interventions.
METHOD
Taxonomy development and identification of BCTs used
in guidance documents and treatment manuals
Identifying the full set of intervention manuals and
guideline documents available world-wide would not be
feasible. As a starting-point for this kind of research it
was decided to approach 11 international alcohol and
drugs experts, recommended by the Editor of the inter-
national journal Addiction, and invite them to suggest
guidance documents and treatment manuals or service
protocols for ‘behavioural support/counselling/advice to
reduce excessive alcohol consumption (not treatment for
alcoholism)’. The experts were drawn from two countries,
nine from the United Kingdom and two from the United
States. To complement this top–down approach we used
an opportunistic sample of treatment manuals used
by local health-care organizations: 13 London National
Health Service (NHS) primary health-care organizations
with e-mail addresses on their websites. This produced 11
Addiction

documents: three clinical [26–28] and five self-help
[29–33] guidance documents, two treatment manuals
[34, Tober et al.; unpublished observations], identified by
the experts, and one treatment manual provided by an
NHS organization. The process was designed to capture as
many as possible of the BCTs that could be included in
interventions to reduce excessive alcohol consumption
rather than characterize the world-wide domain of
alcohol interventions. Any BCTs missed by this process
could be added as they become known or are introduced
into the field.
The documents were analysed independently by two
psychology researchers, Z.H. and F.Z., to identify BCTs.
This process was guided by the coding manuals for the
smoking and healthy eating/physical activity taxono-
mies. It involved reading the documents line by line,
marking the exact words signifying a BCT and assigning
them either a label from a previously developed taxonomy
or, where the BCT was deemed to be novel, creating a new
label. When both researchers agreed that a BCT was
present, the relevant excerpt was underlined and the label
of the BCT was placed next to it. Discrepancies were dis-
cussed and, if a resolution was not achieved, a behaviour
change expert, S.M., was consulted. If an extract met the
definition of a BCT but was not included in the previous
taxonomies, it was given a label and definition, guided by
the APA Dictionary of Psychology [35].
In order to maximize standardization across behav-
ioural domains, the wording of the BCTs and definitions
from the previously developed taxonomies were kept con-
stant, except for minor modifications relating to alcohol
interventions. For example, the BCT ‘Provide information
on consequences of smoking and smoking cessation’
was modified to ‘Provide information on consequences of
excessive alcohol consumption and reducing excessive
alcohol consumption’. If a BCT was present in both the
physical activity/healthy eating and smoking cessation
taxonomies, the technique wording from the smoking
taxonomy was used, as it is the most recent taxonomy.
To assess inter-rater reliability, Z.H. and F.Z. coded ran-
domly selected documents independently: four guidance
documents and four RCTs from the Cochrane Review,
using BCT coding instructions (available from the first
author). Percentage agreement of identifying presence
or absence of BCTs for the same document excerpts was
assessed. BCTs were identified in the nine guidance docu-
ments, the two published national treatment manuals
regarding MET and SBNT [34, Tober et al.; unpublished
observations] and one NHS treatment manual.
Effectiveness of BCTs in brief interventions
Data (quantity of drinking measured as grams per week)
from 18 randomized controlled trials (RCTs) included in
© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
Behaviour change and excessive alcohol use 3
the Cochrane Review of brief alcohol interventions [11]
were re-analysed using STATA version 9.2 [36]. To
conduct the meta-regression we used the revised metareg
command. In our analyses, the regression coefficients
(b) are the estimated increase in the effect size (mean
difference in quantity of drinking) per unit increase in the
covariate (dummy-coded as 1 = used the BCT or 0 = not
used BCT). As with the original Cochrane Review, nega-
tive effect sizes indicate that the intervention had a better
outcome than the control group, and negative regres-
sion coefficients indicate that studies that used the BCT
produced a larger pooled effect size than studies that did
not use the BCT.
To examine statistical heterogeneity, the Q statistic
and I2 [37] were used as well as a visual inspection of
the forest plots. The I2 statistic was interpreted in the
following way, based on [38] 0–40%: might not be impor-
tant; 30–60%: may represent moderate heterogeneity;
50–90%: may represent substantial heterogeneity;
75–100%: considerable heterogeneity.
To examine the association between the BCTs and
intervention effectiveness, we constructed three multi-
variate meta-regression models, based on clusters found
to be reliable in a taxonomy of BCTs used in behavioural
support for smoking cessation [22]. The first model
included BCTs that maximized motivation to abstain or
reduce excessive alcohol consumption: ‘provide informa-
tion on consequences of drinking and drinking cessation’
(BCT1), ‘boost motivation and self efficacy’ (BCT3),
‘provide normative information about others’ behaviour
and experiences’ (BCT4), ‘prompt commitment from the
client there and then’ (BCT8), ‘provide rewards contingent
on effort or progress’ (BCT7) and ‘motivational inter-
viewing’ (BCT9). The second model included BCTs that
maximized self-regulatory capacity and skills: ‘facilitate
action planning/know how to help identify relapse triggers’
(BCT15), ‘facilitate goal setting’ (BCT14), ‘prompt review
of goals’ (BCT18) and ‘prompt self-recording’ (BCT20).
The third model included BCTs that did not clearly form
part of the motivation and self-regulation clusters: ‘assess
current and past drinking behavior’ (BCT31) and ‘offer/
direct towards appropriate written materials’ (BCT30).
To be included in the analysis, each BCT was required to
be evaluated by at least three separate studies, and not be
correlated highly with other techniques in the cluster.
Where the meta-regression suggested the presence of a
potentially important BCT, we used subgroup analyses to
investigate the data further, and to estimate an effect size
for each subgroup (i.e. studies using the technique versus
studies not using the technique).
To examine how much of the heterogeneity was
accounted for by the covariates(s) included in each
model, we examined the adjusted R2 produced by the
metareg command in STATA. The adjusted R2 gives the
Addiction
4 Susan Michie et al.
proportion of between-study variance explained by the
covariates included in the model. Negative values of R2
were truncated to zero. Sensitivity analyses were used to
explore the effect of removing studies with results classi-
fied as outliers, determined by the Sample-Adjusted Meta-
Analytic Deviancy (SAMD) statistic [39]. We assessed the
possibility of publication bias using the STATA metabias
command. Where there was evidence of significant
asymmetry in the funnel plot (as judged by the Begg &
Mazumdar adjusted rank correlation test) [40], we used
the STATA metatrim command to perform the Duval &
Tweedie non-parametric ‘trim and fill’ method [41]. This
method was used to examine the impact of the missing
studies by adjusting the meta-analysis to take into
account the theoretically missing studies.
RESULTS
Taxonomy development and identification of BCTs
Thirty-four BCTs were identified in the guidance
documents. Eight additional BCTs identified were in the
treatment manuals. Inter-rater reliability of applying the
taxonomy was good: a mean of 80% (range 74–83%).
The final taxonomy of 42 BCTs is shown in Table 1.
Thirteen BCTs addressed motivation, 12 addressed self-
regulation, two performed adjuvant functions and 15
addressed other aspects of the interaction, such as
general communication.
Association between individual BCTs and
intervention effectiveness
Nineteen of the 42 BCTs were found in at least one inter-
vention covered by the Cochrane Review of RCTs on brief
alcohol interventions (Table 1). Data from the 18 RCTs
(n = 7183) showed substantial between-study heter-
ogeneity of effect sizes (I2 = 61%; Q = 43.33, P < 0.001),
suggesting that it was appropriate to use meta-regression
to explore the reason for this variation. No trials were
detected as outliers based on the SAMD statistic, and
there was no evidence of significant asymmetry in the
funnel plot (P = 0.45), suggesting that publication bias
was unlikely (Fig. 1).
Overall, 83% of trials used at least one BCT from the
motivational cluster; 67% used BCT1 (provide infor-
mation on consequences of drinking and drinking
cessation), 50% used BCT3 (boost motivation and self-
efficacy), 28% used BCT4 (provide normative information
about others’ behaviour and experiences), 28% used
BCT9 (‘motivational interviewing’), 17% used BCT8
(‘prompt commitment from the client there and then’)
and no studies used BCT7 (‘provide rewards contingent
on effort or progress’). For the meta-regression of the
motivation cluster, five techniques (BCT1, BCT3, BCT4,
© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
BCT8, BCT9) were entered into a multivariate model,
which accounted for 34.93% of the between-study het-
erogeneity (Table 2). The joint test for all covariates was
not statistically significant (P = 0.14). I2 was 45.20%,
revealing that there was moderate heterogeneity remain-
ing among studies. Only one technique, BCT8 (‘prompt
commitment from the client there and then’), was statis-
tically significant (without any adjustment for multiple
testing) (P = 0.025).
Overall, 78% of studies used at least one BCT from the
self-regulation cluster; 50% used BCT15 (facilitate action
planning/help identify relapse triggers), 39% used BCT20
(prompt self-recording), 33% used BCT14 (facilitate goal
setting) and 17% used BCT18 (prompt review of goals).
All four techniques (BCT14, BCT15, BCT18, BCT20)
were entered into a multivariate model, which accounted
for 97.81% of the between-study heterogeneity (Table 3).
The joint test for all covariates was statistically significant
(P = 0.02). I2 was 29.37%, suggesting that no important
heterogeneity remained among trials. One technique,
BCT20 (prompt self-recording), was statistically signifi-
cant (P = 0.002).
With regard to the BCTs that were not included in
either the motivational or self-regulation clusters, all
the interventions used BCT31 (assess current and past
drinking behaviour) and 44% used BCT30 (offer/direct
towards appropriate written materials). BCT30 was
entered into a univariate model, which accounted for
0% of the between-study heterogeneity (Table 4). I2
was 63.19%, indicating that substantial heterogeneity
remained among trials.
Before conducting a subgroup analysis to examine
BCTs shown to be associated with outcome, we created a
new meta-regression model including only BCTs that
were shown to be statistically significant in the models
reported above (i.e. BCT8 and BCT20). The results of this
model showed that, taken together, the two techniques
explained 73.77% of the between-study heterogeneity,
with low remaining heterogeneity (I2 = 36.94%).
Based on the individual regression coefficients, BCT20
[b = -36.43, 95% confidence interval (CI): -80.32 to
7.46] appeared to be more important than BCT8
(b = -8.37, 95% CI: -58.02 to 41.28). A subgroup
analysis using a random-effects model showed that
the pooled effect size for the seven trials using BCT20
was -59.70 (95% CI: -85.24 to -34.15) compared with
-18.03 (95% CI: -31.57 to -4.48) for 11 trials not using
the technique.
DISCUSSION
A taxonomy of BCTs used for alcohol interventions was
developed, and it was found that it could be used reliably
to code guidelines, treatment manuals and published
Addiction
 Behaviour change and excessive alcohol use 5

Table 1 Behaviour change techniques (BCTs) present in guidance documents and treatment manuals and in the Cochrane Review of
brief alcohol interventions, grouped according to function.

Guidance Treatment Cochrane

BCT documents manuals Review

Address motivation
1. Provide information on consequences of excessive alcohol consumption and reducing ✓ ✓ ✓
excessive alcohol consumption
2. Identify reasons for wanting and not wanting to reduce excessive alcohol consumption ✓ ✓ ✓
3. Boost motivation and self-efficacy ✓ ✓ ✓
4. Provide normative information about others’ behaviour and experiences ✓ ✓ ✓
5. Provide feedback on performance ✓ ✓ –
6. Provide information on withdrawal symptoms ✓ ✓ –
7. Provide rewards contingent on effort or progress ✓ ✓ –
8. Prompt commitment from the client there and then – ✓ ✓
9. Conduct motivational interviewing – ✓ ✓
10. Provide rewards contingent on successfully reducing excessive alcohol consumption ✓ ✓ –
11. Prompt use of imagery ✓ ✓ –
12. Model/demonstrate the behaviour – ✓ –
13. Explain the importance of abrupt cessation ✓ – –
Address self-regulation
14. Facilitate goal setting ✓ ✓ ✓
15. Facilitate action planning/help identify relapse triggers ✓ ✓ ✓
16. Advise on avoidance of social cues for drinking ✓ ✓ ✓
17. Behaviour substitution ✓ ✓ ✓
18. Prompt review of goals ✓ ✓ ✓
19. Facilitate relapse prevention and coping ✓ ✓ ✓
20. Prompt self-recording ✓ – ✓
21. Facilitate barrier identification and problem-solving ✓ ✓ –
22. Advise on environmental restructuring ✓ – –
23. Set graded tasks ✓ – –
24. Advise on conserving mental resources ✓ – –
25. Change routine ✓ – –
Promote adjuvant activities
26. Advise on/facilitate use of social support ✓ ✓ –
27. Give options for additional and later support ✓ ✓ –
Address general aspects of the interaction
28. Emphasize choice ✓ ✓ ✓
29. Assess current readiness and ability to reduce excessive alcohol consumption ✓ ✓ ✓
30. Offer/direct towards appropriate written materials ✓ ✓ ✓
31. Assess current and past drinking behaviour ✓ – ✓
32. Assess past history of attempts to reduce excessive alcohol consumption ✓ ✓ –
33. Assess withdrawal symptoms ✓ – ✓
34. Explain expectations regarding treatment programme ✓ ✓ –
35. Tailor interactions appropriately – ✓ ✓
36. Build general rapport ✓ ✓ –
37. Use reflective listening ✓ ✓ –
38. Provide reassurance ✓ ✓ –
39. Summarize information/confirm client decisions – ✓ –
40. Elicit and answer questions – ✓ –
41. Elicit client views – ✓ –
42. General communication skills training – ✓ –

reports of interventions to reduce excessive alcohol con-
sumption. Forty-two distinct specific BCTs were identified
reliably, many of which were the same as those used
in behavioural support/counselling for other areas of
behaviour change. Of all the BCTs included in RCTs
investigated, ‘prompting self-recording’ was found to be
associated with a greater effect size. ‘Prompting commit-
ment’ from the client was also found to be associated with
greater effect size, but not after adjusting for prompting
self-recording.

© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction
 6

Table 2 Multivariate meta-regression analysis for the motivational cluster.a

Use of BCT Multivariate model

Covariate Yes k (n) No k (n) b (95% CI) P-value Model F (d.f.) P-value I2 residual Adj R2
Susan Michie et al.

Provide information on consequences of drinking and drinking cessation (BCT1) 12 (5048) 6 (2135) 14.13 (-29.98, 58.25) 0.498 2.06 (5, 12) 0.14 45.20% 34.93%
Boost motivation and self efficacy (BCT3) 9 (3795) 9 (3388) 32.84 (-7.92, 71.60) 0.107
Provide normative information about others’ behaviour and experiences (BCT4) 5 (2746) 13 (4437) -38.34 (-85.58, 8/89) 0.102
Prompt commitment from the client there and then (BCT8) 3 (1478) 15 (5705) -55.75 (-103.19, -8.32) 0.025
Conduct motivational interviewing (BCT9) 5 (1893) 13 (5290) -7.99 (-57.99, 42.00) 0.734

BCT: behaviour change technique; CI: confidence interval; k: number of studies; n: number of participants. aBCT7 (provide rewards contingent on effort or progress) was not included in the model because no studies used this
technique.

© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
Table 3 Multivariate meta-regression analysis for the self-regulation cluster.

Use of BCT Multivariate model

Covariate Yes k (n) No k (n) b (95% CI) P-value Model F (d.f.) P-value I2 residual Adj R2

Facilitate action planning/help identify relapse 9 (3398) 9 (3785) 14.78 (-15.02, 44.58) 0.303 4.20 (4, 13) 0.02 29.37% 97.81%
triggers (BCT15)
Facilitate goal-setting (BCT14) 6 (1992) 12 (5191) -21.75 (-61.34, 17.83) 0.256
Prompt review of goals (BCT18) 3 (761) 15 (6422) -28.50 (-72.51, 15.51) 0.185
Prompt self-recording (BCT20) 7 (2807) 11 (4376) -50.49 (-78.00, -22.99) 0.002

BCT: behaviour change technique; CI: confidence interval; k: number of evaluations; n = number of participants.

Addiction
 Behaviour change and excessive alcohol use 7

Aalto 2000

Cordoba 1998

Crawford 2004

Curry 2003

Diez 2002

Fleming 1997

Fleming 1999

Fleming 2004

Gentillelo 1999

Heather 1987

Kunz 2004

Lock 2006

Maisto 2001

Ockene 1999

Richmond 1995

Scott 1991

Senft 1997

Wallace 1988

Combined

-30 0 30 60
Mean Difference

Figure 1 Forest plot of 18 randomized controlled trials (RCTs) comparing a brief alcohol intervention with a control

Table 4 Multivariate meta-regression analysis for the non-theory-based cluster.a

Use of BCT Multivariate model

Covariate Yes k (n) No k (n) b (95% CI) P-value Model F (d.f.) P-value I2 residual Adj R2

Offer/direct towards appropriate 8 (4169) 10 (3014) -7.53 (-48.91, 33.84) 0.703 0.36 (2, 15) 0.36 63.19% 0%
written materials (BCT30)

BCT: behaviour change technique; CI: confidence interval; k: number of evaluations; n: number of participants. aBCT31 (assess current and past drinking
behaviour) was not included in the model as all studies used this technique.

The fact that specific BCTs can be coded reliably offers
a realistic prospect that this method of characterizing
intervention content can be used to provide standards in
clinical trials and intervention descriptions and allow
the investigation of which components of interventions
work. It also provides a method of considering the extent
to which intervention content in guidance documents
reflects research evidence and the extent to which recom-
mended or evidence-based content is reflected in treat-
ment protocols and delivered in practice. Further, in
systematic reviewing, specifying content by individual
BCTs provides a method to guide evidence syntheses. In
the field of smoking, this approach has laid the basis for
developing a reliable taxonomy of competences required
by behaviour change practitioners to deliver the BCTs
[22,42]. Thus, this methodological tool has the potential
to further scientific advance and to foster the develop-
ment and implementation of more effective interven-
tions, and to avoid the implementation of ineffective, or
even harmful, ones.
The finding that prompting self-monitoring was iden-
tified as associated with greater effect sizes from brief
interventions suggests that this BCT should be considered
for brief interventions in this area. However, this explor-
atory finding is in need of a priori testing, ideally in an
RCT. Until such studies are conducted this may be the
strongest form of evidence that is likely to be available for
the foreseeable future, and it is noteworthy that this one
BCT accounted for almost three-quarters of the hetero-
geneity in effect sizes.
The study had a number of limitations. First, it is pos-
sible that our method of identifying treatment manuals

© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction
8 Susan Michie et al.
and guidelines failed to identify a number of BCTs in use.
BCTs may need to be added as they become known or
invented.
Secondly, we are unable to draw conclusions about
the ideal level of generality of specification of interven-
tion content. It may be that some of the BCTs in this list
could be separated usefully into smaller categories and
that others might be combined to form larger ones. It may
also be that one or more of the existing coding systems
would yield similar or greater associations with outcome.
However, even if they were to do so there remains the
question of how far they could be used to specify the
content of interventions because, as noted in the intro-
duction, they are set at a higher level of generality; they
would need to be translated into more specific content to
be able to be communicated and used in training. Never-
theless, it would be useful to compare the current BCT
coding system with existing coding systems in predicting
treatment effectiveness.
Thirdly, treatment fidelity is a significant issue, and it
is not clear how far these documents characterize treat-
ment as delivered. Nor is it clear how well the coding
system could be applied to rating transcripts or video or
audio recordings of treatment sessions. It will be impor-
tant, therefore, to develop a parallel coding system for
treatment as actually delivered. This will provide a basis
for examining fidelity and associations between treat-
ment as delivered and outcomes. This work is being
conducted currently in relation to BCTs used in behav-
ioural support for smoking cessation, within the research
programme of the UK’s National Centre for Smoking
Cessation and Training (http://www.ncsct.co.uk).
A fourth limitation is that investigating associations
between BCTs and outcomes relies upon sufficient varia-
tion in treatment/intervention content across manuals.
There are also doubts about the completeness of the
BCT descriptions in the treatment manuals. This can be
addressed partially by better descriptions of interventions
in published reports [12,43]. To the extent that there is
limited variation and incomplete descriptions of inter-
ventions, our method will fail to detect important asso-
ciations that may be present. Thus, we cannot say that
BCTs which were not associated with outcomes in our
study are not important. Although this method will not
be able to detect all the important interactions between
BCTs, our method has been used to investigate clusters of
BCTs that are predicted theoretically to act synergistically
[12].
Finally, with regard to the studies used in the meta-
regression, ability to detect associations between BCTs
and effect sizes may have been limited by other sources
of variation in the studies, such as setting, partici-
pant characteristics and measurement of outcomes.
The studies were considered sufficiently similar to be
© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
included in the Cochrane Review, but reference to pp.
32–56 in that Review shows that there were some
differences that may have obscured differences attribut-
able to BCTs [11].
Future research is needed to extend these findings.
First, it would be useful to expand the search for treat-
ment manuals and guidelines to see whether other BCTs
are being recommended, and to determine how far these
relate to different patient groups and treatment goals.
Secondly, it will be important in future for journals to
require research reports to include full treatment and
intervention manuals as supplementary material to expe-
dite the process of extracting the BCTs. Indeed, it would
be helpful if researchers themselves could characterize
their interventions in terms of BCTs. Thirdly, given that
effectiveness of interventions may depend on patterns
of BCTs, as the database of studies grows and theory
advances it should become possible to use theoretical
principles to predict specific patterns of BCTs that are
likely to prove most effective and to test those hypotheses
directly. Finally, it will be important to extend this
research into more intensive interventions to treat
alcohol dependence.
In conclusion, to our knowledge this study is the
first to use a systematic method to develop a taxonomy
of BCTs to reduce excessive alcohol consumption
(not including treatment of alcohol dependence). This
taxonomy is a step in the direction of having a set of
standardized and operationalized BCTs to facilitate inter-
vention design, reporting, replication and implemen-
tation. It found that in brief interventions, prompting
self-recording of alcohol consumption was associated
reliably with greater intervention effectiveness. Research
is needed to extend this approach and establish a more
complete characterization of individual BCTs and groups
of BCTs that can help to reduce excessive alcohol con-
sumption and treat alcohol dependence.
Declarations of interest
R.W. has received research funding and undertaken
consultancy for companies that manufacture smoking
cessation medications.
References
1. Brick J. Medical consequences of alcohol abuse. In: Brick J.,
editor. Handbook of Medical Consequences of Alcohol and Drug
Abuse. Binghamton, NY: Haworth Medical Press; 2004,
p. 7–31.
2. Goetzel R. Z., Hawkins K., Ozminkowski R. J., Wang S. The
health and productivity cost burden of the ‘top 10’ physical
and mental health conditions affecting six large U.S.
employers in 1999. J Occup Environ Med 2003; 45: 5–14.
3. Grant B. F., Dawson D. A., Stinson F. S., Chou S. P., Dufour
M. C., Pickering R. P. The 12-month prevalence and trends
Addiction

in DSM-IV alcohol abuse and dependence: United States,
1991–1992 and 2001–2002. Drug Alcohol Depend 2004;
74: 223–34.
4. National Health Service (NHS). Statistics on Alcohol:
England, 2009. London: Information Centre Lifestyle
Statistics; 2009.
5. Bertholet N., Daeppen J. B., Wietlisbach V., Fleming M.,
Burnand B. Reduction of alcohol consumption by brief
alcohol intervention in primary care: systematic review and
meta-analysis. Arch Intern Med 2005; 165: 986–95.
6. Raistrick D., Heather N., Godfrey C. A Review of the Effective-
ness of Treatment for Alcohol Problems. London: National
Treatment Agency; 2006.
7. Moyer A., Finney J. W. Brief interventions for alcohol prob-
lems: factors that facilitate implementation. Alcohol Res
Health 2004; 28: 44–50.
8. Moos R. Theory-based active ingredients of effective treat-
ments for substance use disorders. Drug Alcohol Depend
2007; 88: 109–21.
9. Des Jarlais D. C., Lyles C., Crepaz N., TREND group. Improv-
ing the reporting quality of nonrandomized evaluations of
behavioral and public health interventions: the TREND
statement. Am J Public Health 2004; 94: 361–6.
10. Moher D., Schulz K. F., Altman D. G., CONSORT group.
The CONSORT statement: revised recommendations for
improving the quality of reports of parallel group random-
ized trials. BMC Med Res Methodol 2001; 1: 2.
11. Kaner E., Dickinson H. O., Beyer F. R., Campbell F.,
Schlesinger C., Heather N. et al. Effectiveness of brief alcohol
interventions in primary care populations. Cochrane Data-
base Syst Rev 2007; (2): CD004148.
12. Michie S., Abraham C., Whittington C., McAteer J., Gupta S.
Effective techniques in healthy eating and physical activity
interventions: a meta-regression. Health Psychol 2009; 28:
690–701.
13. Michie S., Abraham C. Interventions to change health
behaviours: evidence-based or evidence-inspired? Psychol
Health 2004; 19: 29–49.
14. Baer J. S., Ball S. A., Campbell B. K., Miele G. M., Schoener
E. P., Tracy K. Training and fidelity monitoring of behav-
ioral interventions in multi-site addictions research. Drug
Alcohol Depend 2007; 87: 107–18.
15. Madson M. B., Campbell T. C. Measures of fidelity in moti-
vational enhancement: a systematic review. J Subst Abuse
Treat 2006; 31: 67–73.
16. Middleton W., Tober G., Frier E., Finnegan O. UKATT
Process Rating Manual. Draft 8. Adapted from Behr, H. M.,
Bisighini, R. M., Carroll, K. M., Maclean, R. & Nuro, K. F.
(1994). Rater’s Manual for Match Tape Rating Scale. New
Haven, CT: Yale University; 2001.
17. Tober G., Clyne W., Finnegan O., Farrin A., Russell I.
Validation of a scale for rating the delivery of psycho-social
treatments for alcohol dependence and misuse: the
UKATT Process Rating Scale (PRS). Alcohol Alcohol 2008;
43: 675–82.
18. Carroll K. M., Nich C., Sifry R. L., Nuro K. F., Frankforter
T. L., Ball S. A. et al. A general system for evaluating
therapist adherence and competence in psychotherapy
research in the addictions. Drug Alcohol Depend 2000; 57:
225–38.
19. Carroll K. M., Connors G. J., Cooney N. L., DiClemente
C. C., Donovan D. M., Kadden R. R. et al. Internal validity of
Project MATCH treatments: discriminability and integrity.
J Consult Clin Psychol 1998; 66: 290–303.
© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
Behaviour change and excessive alcohol use 9
20. Carroll K. M., Kadden R. M., Donovan D. M., Zweben A.,
Rounsaville B. J. Implementing treatment and protecting
the validity of the independent variable in treatment match-
ing studies. J Stud Alcohol Suppl 1994; 12: 149–55.
21. Barber J. P., Mercer D., Krakauer I., Calvo N. Develop-
ment of an adherence/competence rating scale for indi-
vidual drug counseling. Drug Alcohol Depend 1996; 43:
125–32.
22. Michie S., Churchill S., West R. Identifying evidence-based
competences required to deliver behavioural support for
smoking cessation. Ann Behav Med 2011; 41: 59–70.
23. Michie S., Hyder N., Walia A., West R. Development of a
taxonomy of behaviour change techniques used in indi-
vidual behavioural support for smoking cessation. Addict
Behav 2011; 36: 315–9.
24. Abraham C., Michie S. A taxonomy of behavior change
techniques used in interventions. Health Psychol 2008; 27:
379–87.
25. West R., Walia A., Hyder N., Shahab L., Michie S. Behavior
change techniques used by the English Stop Smoking Ser-
vices and their associations with short-term quit outcomes.
Nicotine Tob Res 2011; 12: 742–7.
26. Anderson P., Gual A., Colom J. Alcohol and Primary Health
Care: Clinical Guidelines on Identification and Brief Interven-
tions. Barcelona: Department of Health of the Government
of Catalonia; 2005.
27. Babor T. F., Higgins-Biddle J. C. Brief Intervention for Hazard-
ous and Harmful Drinking: A Manual for Use in Primary care.
Geneva: World Health Organization; 2001.
28. US Department of Health and Human Services. Helping
Patients Who Drink Too Much: A Clinician’s Guide. Bethesda,
MD: National Institute of Alcohol Abuse and Alcoholism;
2005.
29. Kaner E., Heather N. How much is too much? Patient
booklet. Newcastle: Institute of Health and Society,
Newcastle University; 2007.
30. Kaner E., Heather N. How much is too much? Level 2
Extended Brief Interventions. Institute of Health and
Society: Newcastle University; 2006.
31. Kaner E., Heather N. How much is too much? Level 1 Simple
Structured Advice. Institute of Health and Society: New-
castle University; 2006.
32. National Health Service. Your drinking and you. The facts
on alcohol and how to cut down. London: Department of
Health; 2008.
33. National Institute on Alcohol Abuse and Alcoholism. How
to Cut Down on Your Drinking. Bethesda, MD: Department of
Health and Human Services; 2001.
34. Copello A., Orford J., Hodgson R., Tober G. Social behaviour
and network therapy manual. Adapted from: Mattson, M. E.
A clinical research guide for therapists treating individuals
with alcohol abuse and dependence. Bethesda, MD: Depart-
ment of Health and Human Services; 2002.
35. VandenBos G., editor. APA Dictionary of Psychology.
Washington, DC: American Psychological Association;
2007.
36. StataCorp. STATA statistical software: release 9.2
[computer software]. College Station, TX: StataCorp;
2007.
37. Higgins J. P., Thompson S. G. Quantifying heterogeneity in a
meta-analysis. Stat Med 2002; 21: 1539–58.
38. Higgins J. P. T., Green S. Cochrane Handbook for Systematic
Reviews of Interventions. Version 5.0.2 [updated September
2009]. Oxford: The Cochrane Collaboration; 2008/2009.
Addiction
10 Susan Michie et al.
39. Huffcut A. I., Arthur W. Development of a new outlier
statistic for meta-analytic data. J Appl Psychol 1995; 80:
327–34.
40. Begg C. B., Mazumdar M. Operating characteristics of a
rank correlation test for publication bias. Biometrics 1994;
50: 1088–101.
41. Duval S., Tweedie R. A nonparametric ‘trim and fill’ method
of accounting for publication bias in meta-analysis. J Am
Stat Assoc 2000; 95: 89–98.
© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction
42. Amrhein P. C., Miller W. R., Yahne C. E., Palmer M.,
Fulcher L. Client commitment language during motiva-
tional interviewing predicts drug use outcomes. J Consult
Clin Psychol 2003; 71: 862–78.
43. West R. Providing full manuals and intervention descrip-
tions: addiction policy. Addiction 2008; 103: 1411.
Addiction