Content uploaded by Nikos Ntoumanis
Author content
All content in this area was uploaded by Nikos Ntoumanis on Feb 04, 2020
Content may be subject to copyright.
Content uploaded by Nikos Ntoumanis
Author content
All content in this area was uploaded by Nikos Ntoumanis on Jan 17, 2020
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rhpr20
Health Psychology Review
ISSN: 1743-7199 (Print) 1743-7202 (Online) Journal homepage: https://www.tandfonline.com/loi/rhpr20
A meta-analysis of self-determination theory-
informed intervention studies in the health
domain: effects on motivation, health behavior,
physical, and psychological health
Nikos Ntoumanis, Johan Y.Y. Ng, Andrew Prestwich, Eleanor Quested, Jennie
E. Hancox, Cecilie Thøgersen-Ntoumani, Edward L. Deci, Richard M. Ryan,
Chris Lonsdale & Geoffrey C. Williams
To cite this article: Nikos Ntoumanis, Johan Y.Y. Ng, Andrew Prestwich, Eleanor Quested, Jennie
E. Hancox, Cecilie Thøgersen-Ntoumani, Edward L. Deci, Richard M. Ryan, Chris Lonsdale &
Geoffrey C. Williams (2020): A meta-analysis of self-determination theory-informed intervention
studies in the health domain: effects on motivation, health behavior, physical, and psychological
health, Health Psychology Review, DOI: 10.1080/17437199.2020.1718529
To link to this article: https://doi.org/10.1080/17437199.2020.1718529
© 2020 The Author(s). Published by Informa
UK Limited, trading as Taylor & Francis
Group
Accepted author version posted online: 27
Jan 2020.
Published online: 03 Feb 2020.
Submit your article to this journal Article views: 72
View related articles View Crossmark data
A meta-analysis of self-determination theory-informed
intervention studies in the health domain: effects on motivation,
health behavior, physical, and psychological health
Nikos Ntoumanis
a,b
, Johan Y.Y. Ng
c
, Andrew Prestwich
d
, Eleanor Quested
a,b
, Jennie
E. Hancox
e
, Cecilie Thøgersen-Ntoumani
a,b
, Edward L. Deci
f,g
, Richard M. Ryan
h
,
Chris Lonsdale
h
and Geoffrey C. Williams
i
a
School of Psychology, Curtin University, Perth, Australia;
b
Physical Activity and Well-Being Group, Curtin University,
Perth, Australia;
c
Department of Sports Science and Physical Education, Chinese University of Hong Kong, Hong
Kong;
d
School of Psychology, University of Leeds, Leeds, UK;
e
Division of Primary Care, School of Medicine, University
of Nottingham, Nottingham, UK;
f
Department of Psychology, University of Rochester, Rochester, USA;
g
School of
Management, University of South-east Norway, Oslo, Norway;
h
Institute for Positive Psychology and Education,
Australian Catholic University, Sydney, Australia;
i
Department of Medicine, Psychology, and Psychiatry, Center for
Community Health and Prevention, University of Rochester Medical Center, Rochester, USA
ABSTRACT
There are no literature reviews that have examined the impact of health-
domain interventions, informed by self-determination theory (SDT), on
SDT constructs and health indices. Our aim was to meta-analyse such
interventions in the health promotion and disease management
literatures. Studies were eligible if they used an experimental design,
tested an intervention that was based on SDT, measured at least one
SDT-based motivational construct, and at least one indicator of health
behaviour, physical health, or psychological health. Seventy-three
studies met these criteria and provided sufficient data for the purposes
of the review. A random-effects meta-analytic model showed that SDT-
based interventions produced small-to-medium changes in most SDT
constructs at the end of the intervention period, and in health
behaviours at the end of the intervention period and at the follow-up.
Small positive changes in physical and psychological health outcomes
were also observed at the end of the interventions. Increases in need
support and autonomous motivation (but not controlled motivation or
amotivation) were associated with positive changes in health behaviour.
In conclusion, SDT-informed interventions positively affect indices of
health; these effects are modest, heterogeneous, and partly due to
increases in self-determined motivation and support from social agents.
ARTICLE HISTORY
Received 5 November 2019
Accepted 16 January 2020
KEYWORDS
Need support; psychological
needs; autonomous
motivation; wellness
Applications of Self-Determination Theory (SDT; Deci & Ryan, 1985; Ryan & Deci, 2017) in the health
domain have increased substantially in the last 15 years. Although the majority of early SDT-based
studies employed observational designs, in recent years there has been a considerable increase in
the volume of intervention studies that aim to foster health-conducive behaviours (e.g., increased
physical activity, healthy eating, abstaining from use of tobacco) or support health treatments
(e.g., medication adherence, diabetes self-management). Such intervention studies are needed,
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://
creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the
original work is properly cited, and is not altered, transformed, or built upon in any way.
CONTACT Nikos Ntoumanis nikos.ntoumanis@curtin.edu.au
Supplemental data for this article can be accessed https://doi.org/10.1080/17437199.2020.1718529
HEALTH PSYCHOLOGY REVIEW
https://doi.org/10.1080/17437199.2020.1718529
given the difficulty people have in initiating and maintaining healthy behaviours over time, and the
alarming global statistics on the causes of ill-health. For example, in 2018, the World Health Organ-
ization estimated that non-communicable diseases accounted for 71% of worldwide deaths in 2016.
The vast majority of deaths attributed to non-communicable diseases were caused by cardiovascular
disease (44%), cancer (22%), chronic respiratory disease (9%), and diabetes (4%). Changes in lifestyle
can prevent or delay the onset of these diseases, improve their management, and increase psycho-
logical wellbeing. Hence, health researchers have used a number of different approaches, including
SDT-informed interventions, to support positive changes in health behaviours and, indirectly,
improve physical and psychological health.
A brief overview of SDT
Both biomedical ethics (Beauchamp & Childress, 2009) and medical professionalism (Project of the
ABIM Foundation, ACP-ASIM Foundation, and European Federation of Internal Medicine, 2002)
have elevated personal autonomy to the highest-level outcome of health care, equivalent to enhan-
cing patient well-being and social justice. Such developments make SDT-based interventions that
intend to enhance personal self-determination highly relevant to health care.
According to Ryan and Deci (2017), human behaviours are influenced to a great extent by personal
and contextual motivational factors. With regard to personal factors, experimental applications of
SDT in the health domain have focused on two: types of motivation and psychological needs. Motiv-
ation has been conceptualised and measured within SDT as a multifaceted construct with several
regulatory styles lying on a continuum of relative autonomy or self-determination (e.g., see Fig. 8.1
in Ryan & Deci, 2017). On the self-determined end of this continuum is intrinsic motivation, reflecting
behavioural engagement as a result of enjoyment and personal interest in the behaviour. In contrast,
extrinsic motivation is comprised of several regulatory styles that are varied in their relative auton-
omy. Specifically, integrated and identified regulations, although extrinsic motivations, are highly
self-determined regulatory styles. Integrated regulation represents reasons for behavioural enact-
ment that align with one’s identity and core values; identified regulation refers to motivation stem-
ming from personal values and endorsement of a behaviour or its outcomes. For example, a person
might comply with a difficult regimen of diet and physical activity because he or she understands and
endorses its value for long-term health. The behaviours entailed would not be intrinsically motivated,
but would be autonomous and experienced as volitional. However, extrinsic motivation can have
controlled forms of regulation. The first one, introjected regulation concerns being motivated by con-
tingent self-esteem and desire for self-or other-approval. The second controlled type of extrinsic
motivation is external regulation, which is the least self-determined as it represents behaviours motiv-
ated by external pressures or contingent rewards.
These diverse regulatory styles are applicable in the health domain as individuals can be motiv-
ated to engage in a health behaviour (e.g., be more physically active) for a diverse array of
reasons, including enjoyment of exercise, experiencing its health benefits, avoiding letting oneself
or others down by not exercising, or being pressured by a spouse or a health professional to be
active. Lastly, in addition to intrinsic and extrinsic motivation, Ryan and Deci (2017) identified amo-
tivation, a state in which individuals lack any type of intention or motivation to engage in a given
behaviour. Typically, researchers in the SDT literature in the health domain have either measured
each of the aforementioned regulations separately (e.g., Wilson, Rodgers, Blanchard, & Gessell,
2006), or have combined them into composites for autonomous motivation (intrinsic, integrated,
and identified regulations) and controlled motivation (introjected and external regulations; e.g.,
Rouse, Duda, Ntoumanis, Jolly, & Williams, 2016), or used indices representing relative autonomous
motivation (autonomous minus controlled motivation) (e.g., Duda et al., 2014).
The second personal motivational dimension studied in the SDT-applications literature for health
is that of basic psychological needs. Three key needs have been identified by Ryan and Deci (2017):
autonomy (feel a sense of choice about one’s behaviour); competence (being able to bring about
2N. NTOUMANIS ET AL.
positive changes in desired outcomes); and relatedness (feeling accepted by one’s social milieu). By
and large, the majority of SDT-based work in the health domain has studied how the satisfaction of
these three psychological needs predicts autonomous motivation, adaptive behaviours, and health
(e.g., Kinnafick, Thogersen-Ntoumani, & Duda, 2016), although there is growing research on how
the frustration of these needs can result in controlled motivation, amotivation, and ill-being (e.g.,
Ng, Ntoumanis, Thogersen-Ntoumani, Stott, & Hindle, 2013). Some of the work in the health
domain (e.g., Williams, Freedman, & Deci, 1998) has used the term ‘perceived competence’instead
of ‘competence need satisfaction’; however, from a measurement perspective, the two constructs
have been operationalised in very similar ways.
Ryan and Deci (2017) have also highlighted the role of social environments in supporting or
thwarting one or more of the three psychological needs, and in turn influencing the degree to
which motivation is autonomous, and concomitant health behaviours and health-related outcomes
are positive. A broad distinction has been made between behaviours of significant others (e.g., health
practitioners, romantic partners, parents) that are supportive of the three aforementioned needs, and
behaviours that thwart such needs. For example, a health practitioner can support weight loss
attempts by offering meaningful choices, providing positive and informative feedback, and
empathising with and acknowledging the patient’s perspective. In contrast, a parent can try to encou-
rage his/her overweight child to lose weight by using pressure, conditionally accepting the child, or
offering tangible rewards if the child agrees to sign up for a weight loss programme. Interventions
stemming from SDT have focused on enhancing perceptions of need support, often by training sig-
nificant others to utilise behaviours that facilitate experiences of psychological need satisfaction and
foster self-determined motivation for behavioural engagement (e.g., Ntoumanis, Thogersen-Ntou-
mani, Quested, & Hancox, 2017).
Reviews of SDT applications in the health domain
Ng et al. (2012) published the first meta-analysis of applications of SDT in the health domain. They
identified 184 independent datasets, primarily non-experimental studies. The included studies exam-
ined relations between SDT constructs and health behaviours (e.g., physical activity, smoking absti-
nence), and indices of health (e.g., dental hygiene, depression, quality of life). The identified effect
sizes ranged from small to medium. Ng et al. also tested a path model utilising meta-analysed cor-
relations, based on a conceptual model by Ryan, Patrick, Deci, and Williams (2008). Results showed
that perceptions of autonomy support predicted reports of autonomy (β= .41), competence (β
= .33), and relatedness (β= .47) need satisfaction. In turn, the three psychological needs predicted
autonomous motivation, although the standardised beta coefficient was substantially larger for com-
petence (β= .35) than those for autonomy (β= .13), and relatedness (β= .15). Competence also had
direct effects on psychological health (β= .39) and physical health (β= .20). The effects of auton-
omous motivation on psychological health (β= .06) and physical health (β= .11) were small.
Overall, competence emerged as the major predictor of motivation and health outcomes in the
path analysis. A potential limitation of the Ng et al. (2012) meta-analysis is that it combined
indices of physical health with health behaviours. Another limitation was that it included many
non-experimental studies. Experimental design was a moderator with respect to the effect sizes
between autonomy and physical activity, autonomy and intrinsic motivation, and autonomy and
external regulation; experimental studies had larger effect sizes than non-experimental studies.
A more recent review of the SDT literature by Gillison, Rouse, Standage, Sebire, and Ryan (2019)
meta-analysed 74 intervention studies to promote motivation and need satisfaction for health behav-
iour change. The results of effect size calculations showed that such interventions resulted in changes
in autonomy support (g= 0.84), autonomy satisfaction (g= 0.81), competence satisfaction (g= 0.63),
relatedness satisfaction (g= 0.28), and autonomous motivation (g= 0.41). Gillison et al. also coded the
included studies in terms of use of 18 SDT-based techniques (e.g., choice, provision of meaningful
rationales) to promote need satisfaction. Meta-regressions did not identify particular strategies
HEALTH PSYCHOLOGY REVIEW 3
that induced meaningful changes in need satisfaction; hence, the authors concluded that a combi-
nation of such strategies might be necessary to promote need satisfaction. The meta-analysis by Gil-
lison et al. did not calculate effect sizes pertaining to some important SDT-based constructs
(controlled motivation and amotivation) or associated health-behaviour, physical-health, or psycho-
logical-health outcomes. In fact, several of the included studies had only motivation-related variables
as outcomes. Further, the meta-analysis by Gillison et al. did not identify the extent to which changes
in SDT-based constructs were associated with changes in behavioural, physical, or psychological
health outcomes. Lastly, the meta-analysis by Gillison et al. included studies from sport in which
the emphasis was on performance and not on health (e.g., Fransen, Boen, Vansteenkiste, Mertens,
& Vande Broek, 2017) and did not establish the robustness of possible moderators by taking into
account potential confounding between moderators.
Aims of the present study
Advancing the SDT literature in the health domain, we present a meta-analysis of experimental
studies in that context. We extended both the Ng et al. (2012) and Gillison et al. (2019) meta-analyses
by addressing the limitations identified above. We included experimental studies that tested changes
in at least one SDT variable and at least one health-behaviour, physical-health outcome, or psycho-
logical-health outcome. Our primary aim was to calculate effect sizes pertaining to changes in these
variables at the end of the intervention and at the latest follow-up reported in the studies. Based on
the two aforementioned reviews, we hypothesised that SDT interventions would produce medium to
large effect sizes in changes in perceptions of need support and competence; small to medium effect
sizes in changes in autonomy satisfaction, relatedness satisfaction, and autonomous motivation; and
small effect sizes in changes in controlled motivation, health behaviours, and health outcomes. We
also tested, via meta-regression, whether changes in SDT constructs would be associated with
changes in health behaviours, physical- and psychological-health outcomes. We hypothesised
small effect sizes for such associations. In an exploratory fashion, we also aimed to test, via meta-
regressions, predictors of heterogeneity in such effect sizes, such as specific features of the SDT-
based interventions (e.g., the use of specific autonomy, competence, and relatedness supportive
strategies). We also coded 43 behaviour change techniques (BCTs), using the taxonomy proposed
by Michie et al. (2013), to examine whether the moderating role of SDT-based strategies was con-
founded with the co-delivery of specific BCTs. Further sensitivity analyses accounted for the potential
impact of outliers.
Method
Eligibility criteria
We aimed to include all experimental studies testing an intervention based on tenets of SDT to
improve behaviours or outcomes related to the physical and psychological health of participants.
Specifically, studies were eligible if they (1) used an experimental design, such as randomised con-
trolled trials or quasi-experimental studies; (2) tested an intervention that, according to the
authors, was (partially) designed according to SDT principles of motivation and behaviour change;
(3) measured at least one SDT-based motivational construct, and at least one of the following: a
health behaviour (e.g., physical activity), an indicator of physical health (e.g., glycosylated hemo-
globin, HbA1c) or psychological health (e.g., perceived quality of life), at a time point which occurred
after the completion of the intervention. We excluded studies that used SDT-based measures but
employed Motivational Interviewing (Miller & Rollnick, 1991) as their guiding framework, with no
reference to SDT. For a study to be included, the authors had to explicitly mention SDT as the
guiding conceptual framework. SDT and Motivational Interviewing share many similarities, but
there are still issues of contention and debate (e.g., see Deci & Ryan, 2012). No exclusion criteria
4N. NTOUMANIS ET AL.
on publication date or language were employed. However, studies that only employed a qualitative
approach, and therefore did not include any quantitative data, were excluded. Systematic reviews
and other meta-analyses were also excluded. Published journal articles, conference proceedings,
theses/dissertations, and unpublished studies were eligible. For papers that did not include the infor-
mation needed for our analyses (e.g., protocol papers), we contacted the authors directly to request
further details.
Information sources
Database searches were conducted on Medline, PsycINFO, PsycARTICLES, and PubMed. The final
search was completed in November 2018. We also posted a message on the SDT email listserv to
request unpublished studies and scanned reference lists of included studies.
Search
We applied two sets of filters in the database search. Both filters were applied to search for terms in
the titles and abstracts of papers within the databases. The first filter was used to identify studies with
an experimental design (experiment* OR trial* OR manipulat* OR intervention). The second filter was
applied to identify studies that included SDT-based interventions (self-determination theory OR
intrinsic motivation OR basic needs OR basic psychological needs OR autonomy support OR auton-
omy supportive OR need support OR need supportive OR need of autonomy OR need for autonomy
OR autonomy need OR self-determined motivation OR autonomous motivation OR autonomous self-
regulation OR autonomous regulation OR need of competence OR need for competence OR compe-
tence need OR need of relatedness OR need for relatedness OR relatedness need).
Study selection
Information and the full text (if available) of all studies identified in the database search were
imported into a bibliography management software. After removing duplicated studies, a trained
research assistant screened the studies manually and removed studies that did not meet our
inclusion criteria. Our data set is available at https://osf.io/u8csb
Data collection process and data items
The included studies were coded using a data extraction sheet, initially piloted by three authors of the
paper using ten randomly selected studies identified via the database search. The extraction sheet
was modified after the pilot to clarify ambiguity in the coding protocol. The revised version was
then used to code all included studies. Drawing from a pool of three reviewers, all studies were
coded independently by at least two of those reviewers. Discrepancies were resolved following dis-
cussion among the coders. The data extracted included year of publication, study design, number of
treatment conditions and how the intervention across conditions differed, intervention duration,
venue (e.g., school, clinic) and mode (e.g., face-to-face, phone conversations) of intervention delivery,
contact frequencies and durations, background and training of intervention providers, constructs
measured in the study, and participant demographics (i.e., mean age, percentages of males and
females). We also initially coded for frequency and duration of intervention contact, but encountered
some difficulties in doing so in a systematic manner (e.g., in some studies participants could access
online information according to their own schedule). Hence, we decided not to include intervention
duration and frequency in our analyses.
The theoretical underpinning and the BCTs used in the intervention and comparison conditions
were also coded. Based on our knowledge of the SDT literature and earlier stages of a consensus
effort to build a classification of techniques used in SDT-informed interventions in the health
HEALTH PSYCHOLOGY REVIEW 5
domain (Teixeira et al., 2019), we designed a brief grouping for 17 common need supportive beha-
viours or techniques that were applied in the meta-analysed studies (Gillison et al., 2019, also devel-
oped a grouping of 18 SDT techniques, which only partially overlaps with ours, as both lines of work
developed independently). We categorised the 17 techniques as competence-, autonomy-, or relat-
edness-supportive (with seven, six, and four strategies, respectively; see Supplementary File Table S1).
Behaviour change techniques used in these interventions were also coded using Michie et al.’s(2013)
taxonomy. The need supportive and BCT components of included studies were independently coded
by three researchers (two of them also piloted the coding form); each study was coded by at least two
of those researchers. Coding by individual researchers demonstrated ‘substantial agreement’across
coders for need-supportive behaviours (Kappa = .723 p< .001; Landis & Koch, 1977) and ‘moderate
agreement’for BCTs (Kappa = .508, p< .001); the kappa for BCTs is typical for the BCT literature
(e.g., Michie et al., 2015). Discrepancies were discussed and reconciled.
Risk of bias assessment
The risk of bias of primary studies was assessed using an adapted version of the Cochrane Risk of Bias
Tool (Higgins, Altman, & Sterne, 2011). Specifically, the degree of risk of bias was assessed based on
(1) generation of randomisation sequence; (2) concealment of group allocation; (3) blinding of (i) par-
ticipants, (ii) individuals responsible for data collection, (iii) researcher(s) who analysed the data, and
(iv) intervention providers; (4) handling of incomplete or missing data; (5) selective reporting of
results; and (6) any other potential threats to the accuracy of the results.
Summary measures
Analyses were conducted with Stata (version15; StataCorp., 2017) using a random-effects model.
Hedges’gwas used to reflect effect sizes of comparisons between the experimental and comparison
conditions. Absolute values of gbetween 0.2–0.5 are considered small, 0.5–0.8 are medium, and over
0.8 are large (Cohen, 1988).
Synthesis of results
When a study included multiple intervention conditions, Hedges’gwas calculated by comparing the
group receiving the most versus the group receiving the least SDT-based need supportive com-
ponents (based on the coded information). We conducted two separate sets of analyses for outcomes
measured (1) immediately after the completion of the intervention, and (2) at follow-up time points
after the completion of the intervention. When a study measured outcomes at multiple post-inter-
vention follow-up time points, only data from the final time point was used. When pre-intervention
data were available, effect sizes were adjusted for baseline values. If the primary studies contained
multiple effect sizes under any category, they were combined using methods recommended by Bor-
enstein, Hedges, Higgins, and Rothstein (2011). This step requires the use of correlation coefficients
between the constructs; if these coefficients were not available from the original studies, an estimate
of r= .50 was used. Further, sample size adjustments, using intraclass correlation coefficients, were
applied when clustered designs were used (Borenstein et al., 2011). If an intraclass correlation in a
study was unavailable, a value of 0.05 was used for the adjustment (Michie, Abraham, Whittington,
McAteer, & Gupta, 2009).
To test whether SDT-constructs, health behaviours, physical health, and psychological health can
be changed, separate analyses were conducted for (1) perceived need support (overall or combined
across specific need-support dimensions, depending on what was reported in the primary studies),
(2) psychological need satisfaction (i.e., competence, autonomy, relatedness; overall or combined
across the three needs), (3) autonomous motivation (average of intrinsic motivation, integrated regu-
lation and identified regulation, or composite autonomous motivation scores), (4) controlled
6N. NTOUMANIS ET AL.
motivation (average of introjected regulation and external regulation, or composite controlled motiv-
ation), (5) amotivation, (6) health behaviour outcomes (e.g., physical activity, tobacco abstinence), (7)
physical health outcomes (e.g., HbA1c, blood pressure), (8) psychological health outcomes (e.g.,
quality of life, depression). In all analyses, positive gvalues represent more positive changes in the
experimental group over the comparison group.
To test whether changes in SDT-related constructs engender changes in other SDT-related con-
structs, health behaviour, physical health and psychological health, a set of meta-regressions were
conducted. To this end, effect sizes of the interventions on the SDT-related constructs were used
as predictors of effect sizes of the interventions for behavioural or health outcomes.
Identifying and exploring heterogeneity
Heterogeneity of synthesised effect sizes was explored using the Qand I
2
statistics. Specifically, a sig-
nificant Qand an I
2
value close to 100% would suggest heterogeneity. In such cases, the effects of
potential moderators were tested using meta-regressions.
We conducted meta-regressions with each need supportive technique and the BCTs utilised in the
included studies (Michie et al., 2013) as predictor variables. A set of meta-regressions examined
whether the relative presence of a specific need-supportive technique or BCT was associated with
larger or smaller effect sizes. Three variables, one each for competence, autonomy, or relatedness,
were created and coded as follows: if competence, autonomy, or relatedness-need support tech-
niques were applied only in the intervention condition (+1), in both or neither groups (0) or only
in the comparison condition (−1). These three variables were summed to create a further variable
reflecting the total range of need supports applied in the intervention vs. comparison conditions
(coded as +3 to −3). Another set of three variables were created to indicate relative autonomy-, com-
petence-, or relatedness-need support between the two comparison groups, by summing the
number of competence-, autonomy-, or relatedness-supportive techniques (from the list of 17),
respectively, present in the intervention condition and subtracting the equivalent number in the
comparison conditions. Finally, the difference in overall need support in the comparison condition
was subtracted from the overall need support in the intervention condition. The meta-regressions
for competence, autonomy, and relatedness support as predictors were conducted separately
because there was insufficient statistical power to include multiple predictors (i.e., less than 30
effect sizes included, therefore, the ratio of effect size to number of covariates would be smaller
than 10–1; see Borenstein et al., 2011).
The impact of a range of other moderator variables were also considered using meta-regressions,
including the study design (randomised controlled designs versus quasi-experimental designs), pub-
lication type (journal article versus theses/unpublished dataset/conference abstract), intervention
provider (investigators: yes vs. no/unclear; trained trainers: yes vs. no/unclear), mode of delivery
(e.g., face-to-face component: yes vs. no/unclear), treatment duration (in days), participant character-
istics (mean age; percentage of male participants), risk of bias (e.g., allocation sequence concealed:
yes vs. no/unclear).
Small-study bias
Small-study bias is suggested when observed effect sizes increase with smaller sample sizes (and thus
larger standard errors). A potential cause underlying this bias is publication bias (where the likelihood
of publication is affected by the results of studies). Small-study bias was examined using Egger’s test.
Sensitivity analyses
Sensitivity analyses were applied to examine the robustness of the synthesised results. To test the
potential impact of outliers, analyses were repeated by removing outliers. After calculating
HEALTH PSYCHOLOGY REVIEW 7
Sample-Adjusted Meta-Analytic Deviancy (Huffcutt & Arthur, 1995) scores for each study, potentially
outlying studies were detected on resulting scree-plots (see Supplementary File Figures S1–S22). This
approach identifies the influence of each study on the overall effect size by calculating the effect size
without the study present and takes into account the sample size of the study. We also examined
whether any of the BCTs were associated with the effect sizes from individual studies. If this was
found, chi-square analyses (Fisher’s Exact Test, when appropriate) tested whether the significant
BCTs were associated with the significant moderators. Where associations were detected, multi-
variate meta-regressions in which the previously identified moderators were entered alongside
each related BCT were conducted to examine whether the moderators remained significant. The
results from the main analyses were considered to be robust if the sensitivity analyses did not
yield results that led to different conclusions.
Results
Study selection
Using our database search protocol, 2,622 citations were identified. An additional journal article was
included from our request sent through the SDT email listserv. Ten other studies were included via
personal contacts with authors in the field. After the removal of 994 duplicated items, our initial pool
consisted of 1,639 publications. A trained research assistant filtered the list to 77 entries by reading
the full text of publications and discarding irrelevant ones. Two studies were excluded from the final
publication pool, as the statistical information required for our analyses was not available in the pub-
lished document, and we were unable to collect the required data from the authors. Another two
studies were excluded because the results were based on duplicated datasets in other included
papers. Therefore, the final publication pool included 73 studies (there were no studies with multiple
datasets); see Figure 1 for the PRISMA flowchart.
Figure 1. PRISMA flowchart of study selection.
8N. NTOUMANIS ET AL.
Study characteristics
Of the 73 included studies, 68 were published journal articles, three were PhD theses, one was a con-
ference abstract, and one was an unpublished study. In terms of study design, 58 studies used a ran-
domised controlled design, with 20 of these using clusters as the unit of randomisation. The
remaining 15 studies used a quasi-experimental design. A total of 30,088 participants were included
in these studies (average sample size = 412), with approximately 36.6% of participants being male.
Mean age of participants was 35.4 years (ranging from 10.1 to 82.5 years). The experimental
groups included on average 7.4 (SD = 4.6) additional SDT-based strategies relative to the control
groups. There was a large range in the duration over which the intervention was delivered (mean
= 133.4 days; SD = 180.3 days). The final follow-up period ranged from one week to 30 months
post-intervention. The characteristics of each study are summarised in Supplementary File Table
S2. The majority of studies reported adequate randomisation procedures (74.0%), allocation conceal-
ment (60.3%), adequate handling of incomplete data (76.7%), and were free from selective outcome
reporting (79.5%). However, only a few studies blinded key personnel to study condition (Participants:
26.0% of the included studies; Data Collector: 11.0%; Data Analyzer: 8.2%; Intervention Provider:
4.1%). An overview of the risk of bias for each study is presented in Supplementary File Table S3.
The breakdown of specific health behaviours, physical health, and psychological health outcomes
coded in the meta-analysis is reported in Supplementary File Table S4.
Can interventions enhance SDT constructs?
The results suggest that the following constructs were positively changed, based on assessments
taken at the end of intervention (see Table 1): need support g= 0.64; competence g= 0.31; autonomy
g= 0.37, combined need satisfaction g= 0.37; and autonomous motivation g= 0.30. Overall, there
was no effect of the interventions on relatedness (g= 0.20), controlled motivation (g= 0.07), or amo-
tivation (g=−0.07). At follow-up, the effect sizes for need support (g= 1.13), competence (g= 0.55),
and combined need satisfaction (g= 0.49) were larger than the corresponding effect sizes at the end
of the intervention, but had a very wide confidence interval and consequently were not significant.
However, following the removal of outliers on the competence (g= 0.33) and combined need
Table 1. Summary of effect sizes and heterogeneity tests for changes in SDT variables, health behaviours, and health outcomes.
kg 95% CI pQpI²
01a. Need support –End of intervention 21 0.643 0.354, 0.932 <.01 193.84 <.01 89.7
01b. Need support –Follow-up 6 1.129 −0.351, 2.609 .13 467.68 <.01 98.9
02a. Competence –End of intervention 22 0.306 0.120, 0.493 <.01 134.60 <.01 84.4
02b. Competence –Follow-up 11 0.547 −0.045, 1.139 .07 417.85 <.01 97.6
03a. Autonomy –End of intervention 17 0.370 0.146, 0.595 <.01 90.66 <.01 82.4
03b. Autonomy –Follow-up 6 0.250 −0.013, 0.512 .06 18.38 <.01 72.8
04a. Relatedness –End of intervention 14 0.202 −0.041, 0.445 .10 71.51 <.01 81.8
04b. Relatedness –Follow-up 6 0.027 −0.199, 0.254 .81 13.81 .02 63.8
05a. Combined need satisfaction –End of intervention 23 0.369 0.187, 0.550 <.01 199.25 <.01 89.0
05b. Combined need satisfaction –Follow-up 11 0.486 −0.048, 1.019 .07 473.93 <.01 97.9
06a. Autonomous motivation –End of intervention 37 0.296 0.169, 0.424 <.01 146.39 <.01 75.4
06b. Autonomous motivation –Follow-up 14 0.181 −0.001, 0.362 .05 41.84 <.01 68.9
07a. Controlled motivation –End of intervention 18 0.071 −0.042, 0.184 .22 30.01 .03 43.4
07b. Controlled motivation –Follow-up 6 0.017 −0.239, 0.273 .90 16.14 <.01 69.0
08a. Amotivation –End of intervention 14 −0.070 −0.281, 0.140 .51 34.56 <.01 62.4
08b. Amotivation –Follow-up 5 −0.255 −0.535, 0.025 .07 8.56 .07 53.3
09a. Health Behaviour –End of intervention 49 0.450 0.329, 0.571 <.01 334.39 <.01 85.6
09b. Health Behaviour –Follow-up 28 0.278 0.172, 0.384 <.01 78.08 <.01 65.4
10a. Physical health –End of intervention 16 0.042 −0.151, 0.234 .67 52.30 <.01 71.3
10b. Physical health –Follow-up 14 0.280 0.033, 0.528 .03 174.12 <.01 92.5
11a. Psychological health –End of intervention 22 0.294 0.135, 0.452 <.01 78.00 <.01 73.1
11b. Psychological health –Follow-up 10 0.137 −0.087, 0.361 .23 36.71 <.01 75.5
HEALTH PSYCHOLOGY REVIEW 9
satisfaction (g= 0.28) outcomes at follow-up, these effects emerged as significant (due to reduced
variation), and this was also the case for autonomous motivation (g= 0.22; see Table 2). All other
effect sizes pertaining to changes in SDT constructs at follow-up were non-significant.
Few intervention characteristics were significant moderators (see Table 3). Of the need suppor-
tive techniques, studies that utilised the competence supportive technique ‘to be positive that
the individual can succeed’generated larger increases in controlled motivation and larger
reductions in amotivation, compared to studies that did not. Moreover, these studies achieved
marginally larger increases in need support, autonomy satisfaction, and autonomous motivation,
all of which became significant following the removal of outliers (need support: B=1.09,SE =0.38,
t=2.88, p= .01; autonomy satisfaction: B=0.73, SE =0.27, t= 2.73, p= .02; autonomous motiv-
ation: B= 0.49, SE = 0.15, t= 2.78, p= .009). ‘Identifying barriers to change’was associated with
increases in autonomous motivation and ‘conveying a person is valued’was associated with
increases in autonomy satisfaction, reductions in amotivation, and marginal increases in related-
ness satisfaction. Interventions delivered in community settings were more likely to enhance
relatedness and reduce amotivation than interventions delivered elsewhere. There were no
other intervention characteristics that significantly increased or decreased the magnitude of
the effect sizes for autonomy support, competence satisfaction, autonomy satisfaction, combined
need satisfaction, autonomous motivation, or controlled motivation at conventional levels of sig-
nificance. The above moderator effects were largely robust to the influence of outliers, with the
exception of two additional effects emerging once outliers were removed: the technique to
‘provide a meaningful rationale’was positively associated with larger effect sizes for autonomy,
B=.
60, SE = .25, t=2.40, p= .03, and combined need satisfaction, B=.49, SE = .19, t=2.54, p
= .02. Finally, two study quality characteristics significantly moderated effects: adequate allo-
cation concealment reduced effect sizes representing the effect of the intervention on auton-
omous motivation, while blinding the intervention provider increased the effect of the
intervention on relatedness. Various BCTs were associated with increased effect sizes for
various SDT constructs (see Table 4). The potential confounding roles of these BCTs are con-
sidered in the Sensitivity Analyses section below.
Table 2. Summary of effect sizes and heterogeneity tests for changes in SDT variables, health behaviours, and health outcomes
following outlier removal.
kg 95% CI pQpI²
01a. Need support –End of intervention 19 0.739 0.445, 1.033 <.01 149.42 <.01 88.0
01b. Need support –Follow-up$ 6 1.129 −0.351, 2.609 .13 467.68 <.01 98.9
02a. Competence –End of intervention 20 0.267 0.100, 0.435 <.01 90.30 <.01 79.0
02b. Competence –Follow-up 10 0.329 0.046, 0.611 .02* 58.08 <.01 84.5
03a. Autonomy –End of intervention 16 0.404 0.174, 0.633 <.01 87.30 <.01 82.8
03b. Autonomy –Follow-up$ 6 0.250 −0.013, 0.512 .06 18.38 <.01 72.8
04a. Relatedness –End of intervention 13 0.242 −0.008, 0.493 .06 68.69 <.01 82.5
04b. Relatedness –Follow-up$ 6 0.027 −0.199, 0.254 .81 13.81 .02 63.8
05a. Combined need satisfaction –End of intervention 21 0.343 0.172, 0.514 <.01 152.49 <.01 86.9
05b. Combined need satisfaction –Follow-up 10 0.276 0.037, 0.514 .02* 60.19 <.01 85.0
06a. Autonomous motivation –End of intervention 35 0.334 0.211, 0.457 <.01 116.10 <.01 70.7
06b. Autonomous motivation –Follow-up 13 0.223 0.071, 0.375 <.01* 22.2 .04 45.9
07a. Controlled motivation –End of intervention$ 18 0.071 −0.042, 0.184 .22 30.01 .03 43.4
07b. Controlled motivation –Follow-up$ 6 0.017 −0.239, 0.273 .90 16.14 <.01 69.0
08a. Amotivation –End of intervention 13 −0.074 −0.257, 0.174 .71 32.27 <.01 62.8
08b. Amotivation –Follow-up$ 5 −0.255 −0.535, 0.025 .07 8.56 .07 53.3
09a. Health Behaviour –End of intervention 46 0.402 0.288, 0.515 <.01 221.72 <.01 79.7
09b. Health Behaviour –Follow-up 27 0.267 0.163, 0.371 <.01 72.90 <.01 64.3
10a. Physical health –End of intervention 15 0.130 0.003, 0.257 .04* 21.22 .10 34.0
10b. Physical health –Follow-up 13 0.245 −0.012, 0.502 .06 114.64 <.01 89.5
11a. Psychological health –End of intervention$ 22 0.294 0.135, 0.452 <.01 78.00 <.01 73.1
11b. Psychological health –Follow-up$ 10 0.137 −0.087, 0.361 .23 36.71 <.01 75.5
Note: $ denotes the absence of outliers, hence the values reported in this row as the same as those in Table 1.
10 N. NTOUMANIS ET AL.
Table 3. Intervention characteristics meta-regressed on SDT-based outcomes at the end of the intervention.
Study Characteristic
Need Support Competence Autonomy Relatedness Combined Need Satisfaction Autonomous Motivation Controlled Motivation Amotivation
(k= 21) (k= 22) (k=17) (k= 14) (k= 23) (k= 37) (k= 18) (k= 14)
Treatment duration −0.002 −0.001 −0.000 −0.002 −0.000 0.000 0.000 −0.000
Need support techniques
Intervention vs. comparison
Competence support techniques
Optimal challenge −0.42 0.33 0.25 0.48 0.38 0.09 0.16 −0.04
Be positive 0.74
†
−0.09 0.52
†
0.35 0.07 0.33
†
0.30* −0.68*
Info/positive feedback 0.13 −0.01 0.30 0.37 0.06 0.15 0.19 −0.26
Identify barriers −.19 0.11 0.43 0.61
†
0.16 0.37** 0.03 −0.15
Skills/problem solving 0.03 −0.07 0.20 0.37 −0.02 0.14 −0.08 −0.13
Develop plan −0.28 −0.37 −0.19 −0.35 −0.37 0.09 −0.14 0.09
Reframe failures –−0.08 –– −0.15 −0.12 ––
Other 0.01 −0.15 −0.02 0.38 −0.19 −0.10 −0.07 0.18
Autonomy support techniques
Provide rationale 0.01 0.11 0.53
†
0.35 0.26 0.17 0.13 −0.30
Acknowledge feelings 0.38 −0.09 0.07 0.02 −0.12 −0.00 −0.00 0.12
Offer choices 0.10 0.04 0.23 0.45 0.08 0.03 0.12 −0.26
Explore values −0.19 0.03 0.44 0.52 0.12 0.08 −0.09 −0.09
Support self-change −0.24 −0.25 0.00 −0.17 −0.20 0.07 0.03 0.47
Non-controlling language 0.15 −0.12 0.10 0.11 −0.22 0.12 0.18 0.17
Other −0.11 0.04 0.09 0.22 0.00 0.03 −0.01 0.05
Relatedness support techniques
Develop empathy 0.13 0.14 0.19 0.15 0.05 0.25
†
0.25 −0.09
Warmth/inclusion −0.31 0.05 0.12 0.33 0.03 0.05 0.18 −0.17
Convey value 0.07 0.23 0.54* 0.56
†
0.27 0.21 0.15 −0.52*
Convey respect −0.24 −0.07 0.17 0.34 −0.08 0.08 −0.29
†
−0.00
Other −0.43 −0.18 −0.19 0.04 0.17 0.01 −0.16 0.21
Competence (min. 1 strategy) −0.46 −0.28 −0.32 0.27 −0.25 −0.07 −0.04 0.25
Autonomy (min. 1 strategy) −0.21 −0.13 0.23 0.48 0.00 0.06 0.02 −0.04
Relatedness (min. 1 strategy) −0.04 0.09 −0.02 0.33 0.02 0.20 0.07 0.07
No. of needs targeted −0.19 −0.08 −0.07 0.23 −0.06 0.04 0.01 0.06
Diff. in competence strategies −0.01 −0.01 0.12 0.18 0.02 0.08 0.04 −0.12
Diff. in autonomy strategies 0.01 −0.03 0.12 0.09 −0.01 0.04 0.04 −0.02
Diff. in relatedness strategies −0.04 0.04 0.20 0.14 0.03 0.06 0.04 −0.10
Diff. in total SDT strategies −0.00 −0.00 0.06 0.07 0.01 0.03 0.02 −0.04
Venue (1 = yes; 0 = no)
Clinic 0.38 −0.33 0.06 –−0.31 0.02 −0.18 −0.39
Community 0.65 0.39 0.68
†
1.01** 0.40 −0.02 0.18 −0.56*
Fitness/Sports 0.56 0.28 −0.10 −0.00 0.14 −0.14 0.03 0.29
(Continued)
HEALTH PSYCHOLOGY REVIEW 11
Table 3. Continued.
Study Characteristic
Need Support Competence Autonomy Relatedness Combined Need Satisfaction Autonomous Motivation Controlled Motivation Amotivation
(k= 21) (k= 22) (k=17) (k= 14) (k= 23) (k= 37) (k= 18) (k= 14)
School −0.53 −0.16 −0.30 −0.45 −0.17 −0.19 −0.13 0.06
University −0.77 −0.26 −0.35 −0.67 −0.18 0.22 −0.08 0.47
Mode (1 = yes; 0 = no)
Face-to-face −0.51 −0.25 0.04 −0.68 −0.21 0.18 0.09 0.03
Phone –−0.33 −0.22 −0.12 −0.35 0.18 −0.20 0.56
One-to-one 0.64 −0.04 −0.32 −0.12 −0.16 0.10 −0.25 0.47
One-to-many −0.50 −0.02 0.02 −0.39 −0.04 −0.12 0.13 0.04
Many-to-many –0.12 –– 0.52 0.33 0.23 –
Provider (yes = 1; no/unclear = 0)
Investigators −0.11 0.27 0.13 0.52 0.27 −0.02 −0.20 0.27
Trained trainers −0.20 −0.41
†
−0.34 −0.49 −0.44
†
−0.22
†
−0.12 0.07
Design
(RCT = 1; Quasi = 0) −0.20 −0.45
†
−0.38 −0.32 −0.33 −0.25
†
−0.30
†
0.24
Analysis (yes = 1; no = 0)
Accounted for baseline 0.09 −0.15 −0.18 −0.28 −0.14 −0.07 −0.03 0.06
Low Risk of Bias (yes = 1; no/unclear = 0)
Sequence generation −0.27 −0.02 −0.22 0.10 0.05 0.03 0.02 0.29
Allocation concealment −0.39 −0.11 −0.35 −0.27 −0.10 −0.32* −0.11 0.14
Participants blinded −0.04 0.16 −0.20 −0.04 0.12 −0.25
†
−0.01 0.23
Data collector blinded −0.63 −0.48 −0.03 −0.30 −0.32 −0.32 −0.05 0.17
Data analyst blinded −0.72 −0.34 −0.40 −0.25 −0.40 −0.36 −0.05 0.17
Provider blinded 0.65 0.84 0.96 1.42* 0.98
†
−0.23 −0.12 −0.18
Missing/incomplete data −0.18 −0.30 −0.20 0.11 −0.23 −0.08 0.02 0.29
Selective reporting −0.18 −0.30 −0.19 0.11 −0.23 −0.11 0.11 0.12
Other
Participant age (years) 0.01 −0.004 0.01 0.01 −0.002 0.002 0.01 −0.01
Participant sex (% male) −0.01 −0.001 −0.01 −0.01 −0.000 −0.001 0.001 0.001
Journal publication (yes = 1; no = 0) −0.02 0.42 –– 0.49 −0.001 ––
Note: †p< .10; *p< .05; **p< .01; ***p< .001. ‘Other’for need supportive techniques refers to technique reported as being autonomy, competence or relatedness supportive but without sufficient
information as to what exactly it entailed.
12 N. NTOUMANIS ET AL.
Table 4. Behaviour change techniques meta-regressed on SDT-based outcomes at the end of the intervention.
Behaviour Change Technique
Need Support Competence Autonomy Relatedness Combined Need Satisfaction Autonomous Motivation Controlled Motivation Amotivation
(k= 21) (k= 22) (k= 17) (k=14) (k= 23) (k= 37) (k= 18) (k= 14)
1.1 Goal setting (behaviour) −0.20 −0.10 0.02 0.06 0.01 0.02 −0.06 −0.22
1.2 Problem solving −0.07 0.02 0.43 0.61
†
0.07 0.25
†
0.07 −0.24
1.3 Goal setting (outcome) 0.43 0.11 0.32 1.15* 0.22 0.59** 0.23 −0.09
1.4 Action planning −0.32 −0.06 0.12 −0.20 −0.06 0.25 0.06 −0.39
1.5 Review behaviour goals −0.34 −0.18 −0.18 0.02 −0.25 0.12 −0.20 0.56
1.6 Goal-beh. Discrepancy 0.43 −0.08 0.32 –0.08 0.31 0.16 –
1.8 Behavioural contract −0.40 ––– – 0.12 −0.13 −0.17
1.9 Commitment –0.84 1.11* 1.15* 0.93
†
0.64* 0.66** −0.67
2.2 Feedback on behaviour 0.31 0.19 0.30 0.46 0.25 0.08 −0.05 −0.00
2.3 Self-monitoring beh. −0.41 −0.17 −0.10 −0.40 −0.01 0.07 0.22 −0.50
2.4 Self-monitor outcomes –−0.24 −0.32 −0.34 −0.37 −0.30 −0.23 –
2.6 Biofeedback –0.10 −0.03 −0.12 −0.10 −0.27 ––
2.7 Feedback on outcomes −0.16 −0.12 −0.10 −0.25 −0.10 0.41 0.25 0.26
3.1 Soc. Supp. (unspecified) 0.56 0.18 0.39 0.61 0.36 0.33
†
0.03 −0.18
3.2 Soc. Supp. (practical) −0.66 0.41 0.70 0.39 0.37 0.24 0.37
†
−0.28
3.3 Soc. Supp. (emotional) −0.12 0.40 0.51 0.75* 0.33 0.07 0.10 −0.02
4.1 Beh. instruction −0.55 −0.39 −0.43 −0.42 −0.32 −0.17 −0.04 0.11
4.2 Info. on antecedents −0.51 −0.67 −0.63 –−0.67 0.18 −0.18 0.45
5.1 Info. health cons. 0.53 0.28 0.43 0.49 0.26 0.06 −0.16 −0.10
5.3 Info. soc. cons. −0.34 −0.54 0.32 −0.30 −0.25 −0.54 ––
5.4 Monitor emo. cons. 0.43 –0.32 –0.32 0.68* 0.16 –
5.6 Info. emo. cons. 1.00 0.15 0.64 −0.17 0.12 0.21 −0.23 −0.55
6.1 Demo. of behaviour −0.72
†
−0.45 −0.39 −0.56 −0.25 −0.09 0.03 −0.03
6.2 Social comparison 0.43 –0.32 –0.32 0.68* 0.16 –
7.1 Prompts/cues –0.05 0.46 0.25 0.09 0.06 −0.36 0.57
8.1 Beh. practice/rehearsal −0.37 −0.25 −0.25 −0.42 −0.09 −0.01 −0.03 0.14
8.3 Habit formation −0.35 −0.03 –– −0.10 0.06 ––
8.7 Graded tasks −0.69 0.12 –– 0.52 −0.09 0.08 0.27
9.1 Credible source 0.12 0.04 0.31 0.45 0.05 0.13 0.07 −0.07
9.2 Pros and cons −0.11 −0.27 −0.13 –−0.17 0.29
†
−0.03 0.10
10.3 Non-specific reward 0.43 0.84 0.74
†
1.15* 0.65 0.79** 0.36* −0.67
10.4 Social reward 1.14 0.84 1.11* 1.15* 0.93
†
0.74** 0.66** −0.67
10.6 Non-specific incentive 0.43 –0.32 –0.32 0.68* 0.16 –
10.8 Incentive (outcome) 0.43 –0.32 –0.32 0.68* 0.16 –
10.9 Self-reward 0.43 –0.32 –0.32 0.68* 0.16 –
10.10 Reward (outcome) 0.43 –0.32 –0.32 0.68* 0.16 –
13.1 Identify as role model −0.67 −0.34 −0.39 −0.25 −0.40 −0.09 0.02 0.10
13.2 Framing/reframing −0.12 −0.21 0.23 −0.67 0.02 0.46 0.16 –
(Continued)
HEALTH PSYCHOLOGY REVIEW 13
Table 4. Continued.
Behaviour Change Technique
Need Support Competence Autonomy Relatedness Combined Need Satisfaction Autonomous Motivation Controlled Motivation Amotivation
(k= 21) (k= 22) (k= 17) (k=14) (k= 23) (k= 37) (k= 18) (k= 14)
13.5 Identify with beh. –––– – 0.42 ––
15.1 Verbal persuasion capab.0.43 –0.32 –0.32 0.68* 0.16 –
15.2 Mental rehearsal –––– – 0.26 −0.27 0.32
15.4 Self-talk –0.84 1.11* 1.15* 0.93
†
0.72* 0.66** −0.67
16.3 Vicarious consequences −0.69 ––– – −0.44 −0.26 0.27
Note: †p< .10; *p<.05;**p< .01; ***p< .001. Beh. = behaviour; Soc. = Social; Supp. = Support; Info. = Information; Emo. = emotion; Cons. = consequences; Demo. = Demonstration; Phys. = physical;
Environ. = environment; Capab. = capability.
14 N. NTOUMANIS ET AL.
Can SDT-based interventions change health behavior?
SDT-based interventions promoted health behaviours relative to comparison groups, with a medium
effect size at the end of intervention period (g= 0.45), and a small effect size at follow-up (g= 0.28).
Both of these effects were heterogeneous. These effects were marginally larger at the end of inter-
vention and significantly larger at follow-up in the studies which utilised the technique of ‘being posi-
tive that a person can succeed’. These effects were robust to the exclusion of outliers. No other
intervention characteristic increased or decreased the effect sizes on health behaviour prior to the
removal of outliers (see Table 5). After removing outliers, interventions delivered in fitness/sports
centres yielded larger effects sizes of health behaviour at the end of intervention period, compared
to interventions delivered elsewhere, B= .63, SE = .30, t= 2.08, p= .04. Quasi-experimental trials gen-
erated larger effect sizes on health behaviours than RCTs (see Table 5). Some BCTs were associated
with increased effect sizes for health behaviour at follow-up (see Table 6). The potential confounding
effects of these BCTs are considered in the Sensitivity Analyses section.
Can SDT-based interventions change physical and psychological health outcomes?
Although there were no immediate effects of the SDT-based interventions on physical health out-
comes at the end of intervention period (g= 0.04), there was a small benefit at follow-up (g=
0.28). These effects were somewhat robust, once outliers were removed (end of intervention: g=
0.13; follow-up: g= 0.25; see Table 2). At the end of the intervention period, SDT-based interventions
promoted psychological health relative to the comparison groups (g= 0.29), but there was no benefit
at follow-up (g= 0.14); see Table 1.
In terms of moderators for effect sizes associated with physical health (see Table 5), ‘acknowled-
ging feelings’yielded larger intervention effect sizes on physical health at follow-up, as did ‘being
positive that a person can succeed’(albeit the latter only when outliers were removed, B= .61, SE
= .25, t= 2.46, p= .03). Surprisingly studies with longer treatment duration yielded smaller effect
sizes on physical health, B=−.003, SE = .001, t=−2.94, p= .02, though only following the removal
of outliers. Effect sizes, which were calculated taking into account baseline scores, yielded larger
effect sizes on physical health at the end of the intervention, although the moderation effect was
marginal, following the removal of outliers: B= .37, SE = .18, t= 2.02, p= .06.
For moderators of effect sizes associated with psychological health, studies that utilised two tech-
niques considered to provide autonomy support (‘using non-controlling language’and ‘providing a
meaningful rationale’) produced larger effect sizes at the end of intervention and at follow-up than
studies that did not. However, studies that used various competence-support type strategies (‘ident-
ify barriers to change’;‘skills/problem solving’;‘develop plans appropriate to ability’) yielded smaller
effect sizes on psychological health than studies that did not. Studies that incorporated a one-to-
many approach for intervention delivery yielded larger benefits in psychological health outcomes
at follow-up than studies that used other delivery modes. Also, adequately concealing allocation
sequence yielded smaller effects on psychological health, compared to studies with no or unclear
allocation sequence concealment (see Table 5). A number of BCTs were associated with effect
sizes relating to physical and psychological health (see Table 6), and their potential confounding
role are considered in the Sensitivity Analyses sections.
Are changes in effect sizes of SDT-based constructs associated with changes in effect sizes
of health behavior?
Changes in the effect sizes of any of the SDT-constructs were not associated with changes in effect
sizes of health behaviours at the end of the intervention when taking into account all available
studies. However, when identifying potential outliers, one study (Ha, Lonsdale, Ng, & Lubans, 2017)
was found to be a negative outlier (yielding smaller effect sizes) on changes in SDT constructs at
HEALTH PSYCHOLOGY REVIEW 15
Table 5. Study characteristics meta- regressed on health behaviour, physical health and psychological health at the end of the
intervention and follow-up.
Study Characteristic
Health Behaviour Physical Health Psychological health
End Follow-up End Follow-up End Follow-up
(k= 49) (k= 28) (k= 16) (k= 14) (k= 22) (k= 10)
Treatment duration 0.000 −0.001 −0.002 −0.001 −0.001 −0.002
Need support techniques
Intervention vs. comparison
Competence support techniques
Optimal challenge 0.08 −0.12 0.25 −0.03 0.23 −0.06
Be positive 0.34
†
0.35** 0.23 0.52
†
−0.01 0.24
Info./positive feedback 0.16 0.12 −0.12 −0.17 −0.09 −0.18
Identify barriers −0.02 −0.13 −0.10 −0.33 0.11 −0.53*
Skills/problem solving −0.22 −0.03 −0.05 −0.15 −0.04 −0.81**
Develop plan −0.20 −0.01 −0.01 −0.44 −0.37* −0.87**
Reframe failures −0.25 −0.07 0.03 −0.14 −0.44 −0.07
Other −0.03 −0.08 0.19 0.25 −0.06 0.09
Autonomy support techniques
Provide rationale 0.17 −0.12 0.04 0.18 0.37* 0.49
†
Acknowledge feelings 0.08 0.15 0.20 0.60* 0.14 0.24
Offer choices −0.05 −0.20 −0.10 −0.14 0.19 0.39
Explore values −0.32
†
−0.25
†
−0.23 −0.28 −0.26 −0.49
Support self-change −0.11 −0.08 0.42
†
0.09 −0.02 −0.11
Non-controlling language 0.21 0.09 0.02 0.40 0.44** 0.74**
Other 0.13 −0.08 −0.15 0.06 0.25 −0.32
Relatedness support techniques
Develop empathy −0.06 −0.04 0.22 0.12 0.32 −0.20
Warmth/inclusion 0.03 −0.10 −0.09 −0.07 −0.00 −0.12
Convey value −0.09 0.01 −0.06 0.09 0.15 −0.20
Convey respect −0.19 −0.10 −0.27 −0.21 −0.15 −0.20
Other −0.12 −0.22 −0.20 −0.54 −0.03 0.05
Competence (min. 1 tech.) −0.07 −0.03 −0.14 0.15 −0.29 −0.57
Autonomy (min. 1 tech.) 0.18 −0.20 −0.06 −0.11 −0.15 0.45
Relatedness (min. 1 tech.) 0.03 −0.12 −0.02 0.04 0.23 −0.12
No. of needs targeted (−3–+3) 0.02 −0.09 −0.04 0.01 −0.03 −0.01
Diff.in Competence −0.00 0.00 0.01 −0.04 −0.01 −0.09
Diff. in Autonomy −0.00 −0.03 0.04 0.06 0.10 0.12
Diffin Relatedness −0.03 −0.03 −0.02 −0.01 0.04 −0.06
Diff. in total needs targeted −0.00 −0.01 0.01 −0.00 0.02 −0.01
Venue (1 =yes; 0 =no)
Clinic 0.13 0.12 −0.17 0.22 −0.17 −0.09
Community −0.11 −0.03 −0.28 −0.09 0.41
†
–
Fitness/Sports 0.61
†
−0.30 –−0.13 0.04 0.04
School −0.11 −0.19 −0.08 −0.30 −0.18 −0.34
University −0.13 −0.21 0.34 –0.12 –
Mode (1 =yes; 0 =no)
Face-to-face 0.33 −0.07 0.03 0.24 0.16 0.07
Phone 0.18 −0.16 0.51 −0.16 −0.03 −0.34
One-to-one 0.14 0.14 0.29 0.54
†
−0.30 −0.13
One-to-many 0.06 −0.03 0.00 0.35 −0.02 0.53*
Many-to-many 0.45 −0.18 0.34 0.10 0.11 –
Provider (yes =1; no/unclear =0)
Investigators −0.25 −0.10 0.13 −0.20 −0.31 −0.19
Trained trainers 0.21 0.01 −0.31 −0.28 −0.08 −0.20
Design
(RCT = 1; Quasi = 0) −0.57** −0.26 −0.12 −0.09 −0.13 −0.40
Analysis (yes =1; no =0)
Accounted for baseline −0.02 0.16 0.70** 0.29 −0.16 0.03
Low Risk of Bias (yes =1; no/unclear =0)
Sequence generation 0.01 0.06 −0.23 0.09 −0.24 −0.54
†
Allocation concealment −0.21 −0.09 −0.24 −0.27 −0.37
†
−0.74**
Participants blinded 0.02 0.15 0.03 0.24 −0.11 −0.34
Data collector blinded −0.29 0.09 0.17 −0.13 −0.30 −0.19
Data analyst blinded −0.25 −0.10 −0.00 −0.13 −0.41 −0.19
(Continued)
16 N. NTOUMANIS ET AL.
the end of the intervention but a positive outlier (yielding larger effect sizes) on behaviour at the end
of the intervention. After removing this multivariate outlier study, increases in autonomous motiv-
ation and need support at the end of the intervention were each associated with positive changes
in health behaviours and psychological health at the end of the intervention (see Table 7). Moreover,
changes in these two SDT constructs were also found to predict health behaviours at follow-up (see
Table 8).
Are changes in effect sizes of SDT-based constructs associated with changes in effect sizes
of physical and psychological health outcomes?
Increases in effect sizes of autonomous motivation, combined need satisfaction, relatedness, auton-
omy, and competence need satisfactions, and need support effect sizes at the end of the intervention
were positively associated with increases in psychological health effect sizes at the end of the inter-
vention (see Table 7). Changes in SDT-based construct effect sizes were unrelated with changes in
physical health effect sizes at the end of the intervention. Few studies assessed SDT-based constructs
at the end of the intervention in conjunction with physical health and psychological health at follow-
up; in fact, fewer than three studies assessed the same SDT constructs at the end of the intervention
alongside a physical or psychological health outcome at follow-up, meaning meta-regression models
were not calculable. There was a trend for changes in effect sizes of health behaviours at the end of
the intervention to predict changes in effect sizes of physical health (β= .41; p< .10), but it failed to
reach significance, likely due to the small number of available studies.
Small-study bias
Based on measures taken at the end of the intervention, Egger’s test suggested that small-study bias
may be present for health behaviour outcomes (p= .01) and psychological health (p= .006), but not
for any of the SDT constructs or physical health outcomes.
Sensitivity analyses
While the findings were generally robust to the impact of outliers (as outlined above), additional sen-
sitivity analyses provided some evidence of confounding. Specifically, the moderating role of ‘being
positive that a person can succeed’on autonomous motivation appeared to be confounded with
BCT10.3 (non-specific rewards); when entered into a multivariate meta-regression, only BCT10.3 pre-
dicted the outcome, B= 0.78, SE = 0.28, t= 2.82, p= .008, but ‘being positive that a person can
succeed’did not, B= 0.02, SE = 0.19, t= 0.10, p= .92. The moderating role of ‘identify barriers to
change’on autonomous motivation, B= 0.20, SE = 0.15, t= 1.38, p= .18, was similarly confounded
Table 5. Continued.
Study Characteristic
Health Behaviour Physical Health Psychological health
End Follow-up End Follow-up End Follow-up
(k= 49) (k= 28) (k= 16) (k= 14) (k= 22) (k= 10)
Provider blinded −0.49 0.43
†
–0.99
†
−0.28 0.75
†
Missing/incomplete data 0.04 0.02 −0.07 0.22 −0.15 −0.20
Selective reporting −0.03 0.20 −0.37 0.29 −0.16 −0.73
Other
Participant age (years) 0.01 0.00 0.01 0.01 0.00 0.00
Participant sex (% male) −0.00 0.00 0.00 −0.00 −0.00 −0.00
Follow-up duration –−0.00 –0.00 –−0.00
Journal publication (yes = 1; no = 0) 0.26 −0.43 −0.06 –– –
Note: †p< .10; *p< .05; **p< .01; ***p< .001. ‘Other’for need supportive techniques refers to technique reported as being auton-
omy, competence or relatedness supportive but without sufficient information as to what exactly it entailed.
HEALTH PSYCHOLOGY REVIEW 17
Table 6. Behaviour change techniques meta- regressed on health behaviour, physical health and psychological health at the end of
the intervention and follow-up.
Study Characteristic
Health Behaviour Physical Health Psychological health
End Follow-up End Follow-up End Follow-up
(k= 49) (k= 28) (k= 16) (k= 14) (k= 22) (k= 10)
1.1 Goal setting (behaviour) −0.08 −0.07 0.44 −0.10 −0.25 −0.42
1.2 Problem solving −0.09 −0.05 −0.15 −0.33 0.03 −0.53*
1.3 Goal setting (outcome) 0.31 –−0.22 –0.42 –
1.4 Action planning −0.16 −0.06 0.07 −0.37 −0.24 −0.53*
1.5 Review behaviour goals 0.01 0.09 0.28 −0.18 −0.24 −0.39
1.6 Goal-beh. Discrepancy 0.04 0.03 0.43 −0.08 ––
1.7 Review outcome goals 0.52 0.40 0.47 0.12 −0.05 0.09
1.8 Behavioural contract −0.23 0.08 −0.18 −0.54 −0.29 −0.20
1.9 Commitment 0.28 –−0.22 –0.42 –
2.1 Beh. monitor by others –−0.13 ––– –
without feedback
2.2 Feedback on behaviour 0.05 −0.09 −0.04 −0.53
†
−0.01 −0.28
2.3 Self-monitoring beh. −0.00 −0.16 0.24 −0.25 −0.20 −0.36
2.4 Self-monitor outcomes −0.19 −0.14 −0.12 –−0.07 −0.34
2.5 Monitor outcomes −0.03 –––– –
without feedback
2.6 Biofeedback −0.24 −0.14 −0.27 –0.78* –
2.7 Feedback on outcomes 0.04 −0.05 0.12 −0.11 −0.45 −0.19
3.1 Soc. Support (unspecified) −0.15 −0.11 0.30 −0.22 0.17 −0.13
3.2 Soc. Support (practical) 0.09 −0.09 −0.05 0.03 0.28 0.25
3.3 Soc. Support (emotional) −0.08 −0.10 −0.09 –0.49
†
–
4.1 Beh. instruction −0.12 −0.08 0.39* 0.44
†
−0.02 0.22
4.2 Info. on antecedents 0.07 0.22 –0.36 –0.39
4.4 Behavioural experiments −0.73 –– –– –
5.1 Info. health consequences −0.17 −0.12 −0.49* −0.11 0.19 −0.17
5.2 Salience of consequences −0.73 –– –– –
5.3 Info. soc. consequences −0.36 −0.15 −0.26 −0.27 1.18* –
5.4 Monitor emo. Consequences 0.01 –––– –
5.6 Info. emo. Consequences −0.06 −0.21 ––0.46 0.73
6.1 Demonstration of behaviour −0.09 −0.30* 0.10 −0.51 −0.07 −0.34
6.2 Social comparison −0.06 −0.18 –−0.29 ––
6.3 Info. others’approval −0.48 −0.18 –−0.29 ––
7.1 Prompts/cues −0.13 0.04 0.05 −0.13 −0.08 −0.06
7.3 Reduce prompts/cues ––−0.40 –1.18* –
8.1 Beh. practice/rehearsal −0.29
†
−0.16 0.12 0.56 −0.01 0.17
8.2 Beh. substitution −0.47 0.37 −0.04 –−0.34 –
8.3 Habit formation −0.52 –0.10 –– –
8.6 Generalisation target beh. −0.73 –––– –
8.7 Graded tasks −0.30 −0.17 0.13 −0.13 0.01 −0.19
9.1 Credible source 0.14 −0.29* −0.09 −0.10 0.78** 0.04
9.2 Pros and cons 0.08 −0.12 −0.50 −0.47 −0.36 −0.49
9.3 Imagining future outcomes −0.73 –––– –
10.3 Non-specific reward 0.78* –−0.22 –0.67 –
10.4 Social reward −0.03 0.04 0.27 −0.01 0.51* −0.20
10.6 Non-specific incentive 0.39 –––– –
10.8 Incentive (outcome) 0.39 –––– –
10.9 Self-reward 0.39 –––– –
10.10 Reward (outcome) 0.39 –––– –
11.1 Pharmacological support –0.37 ––– –
11.2 Reduce negative emo. –−0.13 ––– 0.60
12.1 Restructure phys. environ. −0.73 –––– –
12.2 Restructure soc. environ. −0.73 –––– –
12.5 Add objects to environ. −0.43 −0.25 −0.13 −0.27 ––
13.1 Identify self as role model −0.27 −0.14 −0.13 −0.27 0.08 –
13.2 Framing/reframing −0.28 −0.22 –−0.22 –−0.40
13.4 Valued self-identity −0.35 –––– –
13.5 Identify with changed beh. 0.68 0.48 ––– –
15.1 Verbal persuasion capab. −0.19 –– –– –
15.2 Mental rehearsal −0.27 –––– –
(Continued)
18 N. NTOUMANIS ET AL.
with BCT1.3 (Goal-setting (outcome)), B= 0.45, SE = 0.19, t= 2.31, p= .03, and the moderating role of
‘providing a meaningful rationale’on psychological health at the end of the intervention, B= 0.17,
SE = 0.18, t= 0.96, p= .35, was confounded with BCT9.1 (Credible Source), B= 0.65, SE = 0.28, t= 2.33,
p= .03.
The mode of one interventionist delivering to many recipients tended to be used alongside BCTs
1.2 (Problem Solving) and 1.4 (Action Planning) when considering psychological health at follow-up
as the outcome (both χ
2
(1) = 10.00, Fisher’sp= .008), and ‘being positive that a person can succeed’
tended to co-occur with BCT10.3 (Non-specific Reward), χ
2
(1) = 7.88, Fisher’sp= .04, when consider-
ing controlled motivation. Furthermore, while ‘being positive that a person can succeed’and
‘acknowledging others’feelings/perspectives’were not related to any significant BCT predicting
physical health outcomes at follow-up, they were related to one another, χ
2
(1) = 5.83, Fisher’sp
= .03. Given these outcomes (psychological health at follow-up, physical health at follow-up, con-
trolled motivation) were assessed in only 10, 14 and 18 studies respectively, multivariate meta-
regressions were not conducted in these instances as there would be fewer than 10 studies per
predictor.
The moderating effects of (a) community venues on relatedness satisfaction, (b) ‘being positive
that a person can succeed’on autonomy satisfaction or health behaviour, (c) ‘providing a mean-
ingful rationale’or ‘conveying the person is valued’on autonomy satisfaction, and (d) ‘reducing
controlling behaviour/language’on psychological health, were not affected because these mod-
erators were not associated with significant behaviour change technique moderators. Similarly,
in the absence of significant behaviour change technique moderators on autonomy support
and amotivation at the end of the intervention, the moderating role (a) of ‘being positive that a
person can succeed’on autonomy support or amotivation, (b) of ‘conveying the person is
valued’on amotivation, and (c) of community settings on amotivation, were also not affected
by confounding with BCTs.
Discussion
We present the results of a meta-analysis of 73 SDT-informed interventions in the health domain. This
is the first meta-analysis that examines changes in indices of motivation, health behaviours, physical
health, and psychological health as a result of such interventions, as well as how such changes covary
over time. We found that SDT-based interventions can be delivered to positively change most of the
examined SDT-based constructs. However, more work is needed to identify how such interventions
can help reduce controlled motivation and amotivation, and support relatedness satisfaction The
interventions also positively impacted health behaviours, physical health, and psychological
health. Nevertheless, most of the effect sizes were small or medium, varied in strength over time,
and/or were heterogeneous. Changes in need support and autonomous motivation in particular
were positively correlated with changes in health behaviours and psychological health. In sum,
there was evidence demonstrating modest efficacy of SDT-based interventions to change health
behaviours (primarily physical activity, and to a lesser extent dietary behaviours and smoking absti-
nence), and to improve indices of health.
Table 6. Continued.
Study Characteristic
Health Behaviour Physical Health Psychological health
End Follow-up End Follow-up End Follow-up
(k= 49) (k= 28) (k= 16) (k= 14) (k= 22) (k= 10)
15.3 Focus on past success −0.35 –––– –
15.4 Self-talk 0.45 0.48 −0.22 –0.67 –
16.2 Imaginary reward −0.35 –––– –
16.3 Vicarious consequences 0.09 –––– –
Note: †p< .10; *p< .05; **p< .01; ***p< .001. Beh. = behaviour; Soc. = Social; Info. = Information; Emo. = emotion; Phys. = phys-
ical; Environ. = environment; Capab. = capability.
HEALTH PSYCHOLOGY REVIEW 19
Table 7. Predicting effect sizes of outcomes from effect sizes of predictors (At the End of Interventions).
Outcomes Predictors 1 2 3 4 5 6 7 8 9 10 11
1 Psychological health (k= 22) –0.06 0.20 −0.13 0.48 0.58** 0.41* 0.79** 0.57** 0.39* 0.31*
2 Physical health (k= 16) –0.04 -/ 1.71 −0.64 −0.34 -/ -/ −0.51 -/
3 Health behaviour (k= 49) –−0.67 1.04 0.40 0.04 0.08 0.16 −0.04 0.19
4 Amotivation (k= 14) –−0.60 −0.43 −0.34 −0.16 −0.39
†
−0.32 −0.11
5 Controlled motivation (k= 18) –0.21
†
0.08 −0.02 0.04 0.08 −0.08
6 Autonomous motivation (k= 37) –0.50** 0.51* 0.53** 0.45** 0.27*
7 Combined need satisfaction (k= 23) −0.89*** 0.86*** 0.99*** 0.38**
8 Relatedness (k= 14) –0.75** 1.04*** 0.28
9 Autonomy (k= 17) –0.97*** 0.46**
10 Competence (k= 22) –0.33*
11 Need support (k=21) –
12 Health behaviour (k= 48)$ .47*** ––−0.35 0.73 0.66* 0.35 0.41 0.44 0.33 0.39*
Note: Beta coefficients are reported in the table. knotes the number of studies measuring each outcome. †p< .10; *p< .05; **p< .01; ***p< .001.
-/ number of observations were less than 3; underlined/italic font results- number of observations were between 3 and 5; underlined/normal font results- number of observations between 6 and 9; non-
underlined/normal font results- at least 10 observations. $ denotes results following removal of one multivariate outlier (Ha et al., 2017); this study did not assess physical health.
20 N. NTOUMANIS ET AL.
Effects of SDT-interventions on SDT indices of motivation
We first examined whether SDT-informed interventions can affect motivation-related indices pro-
posed by the developers of this theory (Ryan & Deci, 2017). Many of these interventions aim to
increase the degree to which important others (e.g., healthcare professionals, fitness instructors,
spouses) are supportive of individuals’three key basic psychological needs (autonomy, competence,
and relatedness). We found that manipulations of the need supportive features of the social environ-
ment were successful, with a medium to large effect size at the end of interventions and a large effect
size at follow-up (although the latter was not significant due to a wide confidence interval generated
from a fairly small number of studies). We also found that SDT interventions were successful in
increasing perceptions of overall need satisfaction, and individually for competence and autonomy.
These effects were small to medium in size, significant at the end of the intervention, and at follow-up
after the removal of outliers (although p=0.06 for autonomy). Gillison et al. (2019) reported a much
larger effect for autonomy (g= .84) than we did (see Tables 1 and 2), perhaps reflecting differences in
the inclusion criteria between the two meta-analyses. Our findings are important, as Ng et al. (2012)
showed that autonomy and competence need satisfaction mediated the effects of need support on
self-determined motivation and health outcomes. Such findings are also relevant for biomedical
ethics and medical professionalism. For instance, the European Society of Cardiology and European
Atherosclerosis Society guidelines (Catapano et al., 2016) for the management of dyslipidaemias, and
the American College of Cardiology and American Heart Association clinical practice guidelines
(Grundy et al., in press) on the management of blood cholesterol emphasise the importance of sup-
porting patient autonomy for making their own decisions about their treatment and health, and fos-
tering patient self-efficacy for change.
We found that intervention effects on relatedness satisfaction were small and non-significant;
similar small effects were reported by Gillison et al. (2019). This could be because SDT interventions
in the health domain use techniques that support primarily autonomy and competence and to a
lesser extent relatedness, as shown by the number of listed techniques in Tables 3 and 5. The tech-
niques for relatedness support centre on showing empathy and respect, which might be useful to
support initiation of change but perhaps not maintain it long-term, particularly if the target behaviour
is complex or does not need to take place alongside other people (e.g., being regularly physically
active, eat healthy). In other social contexts (e.g., workplace; Slemp, Kern, Patrick, & Ryan, 2018)
meta-analytic evidence (albeit correlational) shows somewhat stronger effect sizes for relatedness,
comparable to those for the other two needs. It would be interesting for future SDT interventions
in the health domain (and beyond) to manipulate both the relative balance and the intensity with
which they target autonomy, competence, and relatedness supportive techniques.
We also found that SDT interventions had small to medium effects on autonomous motivation,
which were significant at the end of the intervention and at follow-up (in the latter case, after the
Table 8. Prospective analyses predicting effect sizes of outcomes at follow-up from effect sizes of predictors at the end of
intervention.
Predictors Outcomes Health Behaviour Physical Health Psychological health
1. Amotivation -/ -/ -/
2. Controlled motivation −0.22 -/ -/
3. Autonomous motivation 0.67* -/ -/
4. Combined need satisfaction −0.04 -/ -/
5. Relatedness 0.07 -/ -/
6. Autonomy 0.06 -/ -/
7. Competence −0.09 -/ -/
8. Need support 0.35* -/ -/
9. Health behaviour –.26 .41†
10. Physical health –– .24
Note: Beta coefficients are reported in the table. knotes the number of studies measuring each outcome. †p< .10; *p< .05.
-/ number of observations were less than 3. Underlined/italic font results- number of observations between 3 and 5; underlined/
normal font results- number of observations between 6 and 9; non-underlined/normal font results- at least 10 observations.
HEALTH PSYCHOLOGY REVIEW 21
removal of outliers). Broadly similar findings for autonomous motivation were reported in the pre-
vious two meta-analyses by Gillison et al. (2019) and Ng et al. (2012). Changes in autonomous motiv-
ation are purported by Ryan and Deci (2017) to translate to positive and long-term changes in
behaviour, cognition, and affect; below, we discuss evidence from our meta-analysis for such
effects. We also found that the effects of SDT interventions on controlled motivation and amotivation
were small and non-significant. Gillison et al. did not analyse the effects of such interventions on
these two variables, because they were not considered ‘a positive target for intervention’(p. 116).
However, autonomous motivation, controlled motivation, and amotivation are fairly independent
constructs (the Ng et al. meta-analysis reported correlations in the range of −.26 to +.44). Hence,
future intervention studies in the health domain may need to focus not only on increasing feelings
of enjoyment and personal utility for behaviour change, but also on addressing internal and external
pressures, and on feelings of helplessness for change. Controlled motivation could be an important
predictor of maladaptive health behaviours, not measured in the included studies. For example, a cor-
relational study of eating behaviours by Pelletier, Dion, Slovinec-D’Angelo, and Reid (2004) showed
that autonomous motivation for eating was associated with healthy eating whereas controlled motiv-
ation was related to bulimic symptomatology.
Effects of SDT-interventions on health indices
Extending the Gillison et al. (2019) meta-analysis, we examined the effects of SDT interventions on
health behaviours, psychological health, and physical health. We found that the included interven-
tions had a positive impact on health behaviours, with the effect sizes being small to medium at
the end of intervention and small at follow-up. Further, we also found that changes in autonomous
motivation and perceptions of need support were associated with positive changes in health beha-
viours both at the end of the intervention and at follow-up. This is an important finding which sup-
ports the SDT view about the beneficial outcomes of need support and autonomous motivation for
sustained behaviour change. For example, with regard to autonomous motivation, it has been argued
that it is the internalisation of the value for health and the resultant self-regulation of behaviour that
helps individuals lower their health-risk behaviours and adopt health protective behaviours without
continued intervention support (Ng et al., 2012; Ryan & Deci, 2017). Surprisingly, changes in need sat-
isfaction were not associated with changes in health behaviours at either time point. There were com-
paratively fewer studies that provided the necessary information to conduct such analysis with need
satisfaction variables, which might explain the lack of associations. It is also possible that the effects of
psychological need satisfaction on health behaviours are mediated by autonomous motivation (Val-
lerand, 1997).
SDT interventions had a small to medium effect on psychological health at the end of the inter-
vention, but not at follow-up. The reverse pattern of findings was observed for physical health, with
small and non-significant effects at the end of the intervention, but small to medium significant
effects at follow-up. These findings are consistent with and extend the correlational meta-analyses
by Ng et al. (2012). With regard to physical health, the effect at follow-up is an important one, par-
ticularly when viewed in conjunction with the findings that SDT interventions result in health behav-
iour maintenance. To see benefits in many physical health outcomes, health behaviours need to be
maintained over a period of time (and as reported above, behaviour maintenance is facilitated by
need support and autonomous motivation). For instance, if one stops smoking for 1 year, heart
attacks and strokes are reduced by 50% compared to continued smoking, lung cancer rates fall in
half after 10 years, immune function improves to 50% of normal in 1 year and is returned to baseline
after 5 years (Department of Health and Human Services, 1990,2014). Lowering cholesterol with an
intensive statin medication reduces heart attacks strokes and death by 50% after 1.5 years of treat-
ment (Ridker et al., 2008).
With regard to the effects of the SDT interventions on psychological health, we found that the
positive effects at the end of the intervention were correlated with increases in need support,
22 N. NTOUMANIS ET AL.
autonomous motivation, combined need satisfaction, satisfaction of each individual need, and health
behaviours. However, at follow-up, the effect of the SDT interventions was small and non-significant.
Given that predictors of well-being and quality of life are multi-faceted and not always related to
health behaviours and professional encounters (e.g., quality of personal relationships, financial cir-
cumstances; Chanfreau et al., 2008), this finding is probably not surprising.
Exploratory moderator analyses
Our exploratory moderator analyses were ‘hypothesis generating’rather than ‘hypothesis-testing’,
and as such we did not account for multiple testing. We found that some SDT techniques moderated
intervention effects. For instance, the use of ‘being positive a person can succeed’was associated
with larger effect sizes in terms of intervention effects on various SDT constructs, health behaviour
at follow up, and physical health at follow up. ‘Provision of a meaningful rationale’was associated
with larger effect sizes in autonomy satisfaction, combined need satisfaction (after removal of out-
liers), and psychological health (at the end of intervention and at follow-up). A surprising finding
was that studies that used three competence-support techniques (‘identify barriers to change’;
‘skills/problem solving’;‘develop plans’) yielded smaller effect sizes on psychological health than
studies that did not. It is possible that if these techniques are offered before an individual is auton-
omously motivated, they may lead to perceptions of external control. It is also possible that some of
these techniques were used inappropriately (e.g., problem solving or plan development might not
have been communicated in an autonomy-supportive manner; see also Hagger et al., 2016, and Pre-
stwich, Sheeran, Webb, & Gollwitzer, 2015, for other considerations and suggestions regarding plan-
ning). Quality of intervention delivery, and subsequent fidelity checking is essential in this field to
really understand if/why these techniques work or not (Prestwich, Kenworthy, & Conner, 2017;
Quested, Ntoumanis, Thøgersen-Ntoumani, Hagger, & Hancox, 2017). Interestingly, the total
number of techniques in the intervention vs. control conditions or the difference in the techniques
between the two conditions did not moderate any of the observed effects.
We also explored which BCTs were used in the included studies and how these BCTs were related
to SDT constructs, health behaviours, and health outcomes, at the end of the intervention and at
follow up. Of all SDT constructs, autonomous motivation was related to many more BCTs than any
other SDT construct. There was no particular pattern linking specific BCTs with physical or psychologi-
cal health. We also examined whether any moderating effects of need supportive techniques on SDT
constructs, health behaviours, and health outcomes, would remain significant when accounting for
the BCTs used in the studies. We found three instances where the moderating effects became no
longer significant when accounting for BCTs. The degree to which some SDT techniques overlap
with some BCTs is to be empirically determined. For example, in our analyses, the moderating
effect of ‘providing a meaningful rationale’on psychological health was no longer significant
when accounting for the BCT ‘Credible Source’. Nevertheless, it is important to clarify that, concep-
tually, the need supportive techniques aim to foster need satisfaction and autonomous motivation,
whereas the BCTs primarily aim at changing behaviour. This is important because, for example, the
support of autonomy includes that an individual may autonomously decide not to change their
behaviour (i.e., volitional non-adherence). Volitional or autonomous non-adherence (choosing to con-
tinue to smoke, or to not exercise or not take a recommended medicine) is in line with medical ethics
and reflects the real world of public health (Beauchamp & Childress, 2009).
The analyses pertaining to BCTs should be treated with caution because the BCTs were coded by
us (with moderate inter-rater reliability) based on the description of study methodologies and were
not explicitly identified by study authors. Further, when looking at the effects of specific BCTs on
specific outcomes, some BCTs were always present when other BCTs were present and were
absent when other BCTs were absent. Consequently, it is impossible to determine whether the posi-
tive (or negative) effects of these BCTs are attributable to an individual BCT or combination of BCTs. It
should also be noted that more than half (i.e., 50) of the 93 BCTs were not utilised differently in the