Journal of Health Psychology
2015, Vol. 20(5) 525 –534
© The Author(s) 2015
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1359105315574951
hpq.sagepub.com
Many causes of morbidity and premature mor-
tality are influenced by multiple health-risk
behaviors including unhealthy nutrition and
physical inactivity (Lanas et al., 2007). In fact,
research has shown that certain risk behaviors
occur in combination and tend to cluster within
individuals (e.g. Poortinga, 2007). Therefore, it
has been argued that health behavior interven-
tions should address more than one single
behavior to produce greater impacts on health
(e.g. Prochaska et al., 2008). While multiple
behavior change interventions are increasingly
being implemented, the theoretical basis for
doing so has received less attention (e.g. Noar
et al., 2008).
Usually, theories have been applied in research
on single health behaviors, giving insights into
that specific behavior but providing little under-
standing on relationships between multiple behav-
iors and processes of change (Noar et al., 2008).
To develop effective methods for addressing mul-
tiple behaviors, the multiple behavioral approach
(MBA) (Noar et al., 2008) suggests to concentrate
Cross-behavior associations and
multiple health behavior change: A
longitudinal study on physical activity
and fruit and vegetable intake
Lena Fleig
1
, Carina Küper
2
, Sonia Lippke
3
,
Ralf Schwarzer
4,5
and Amelie U Wiedemann
1
Abstract
This study aimed to examine the interrelation of physical activity and fruit and vegetable intake. The influence
of stage congruence between physical activity and fruit and vegetable intake on multiple behavior change was
also investigated. Health behaviors, social-cognitions, and stages of change were assessed in 2693 adults at
two points in time. Physical activity and fruit and vegetable intake were assessed 4 weeks after the baseline.
Social-cognitions, stages as well as stage transitions across behavior domains were positively interrelated.
Stage congruence was not related to changes in physical activity and fruit and vegetable intake. Physical
activity and nutrition appear to facilitate rather than hinder each other. Having intentions to change both
behaviors simultaneously does not seem to overburden individuals.
Keywords
fruit and vegetable intake, multiple health behavior change, physical activity, stage congruence, transfer
1
Freie Universität Berlin, Germany
2
Humboldt-Universität zu Berlin, Germany
3
Jacobs University Bremen, Germany
4
Australian Catholic University, Australia
5
University of Social Sciences and Humanities, Poland
Corresponding author:
Lena Fleig, Social and Economic Psychology Unit (PF 9),
Health Psychology Division (PF 10), Freie Universität
Berlin, Habelschwerdter Allee 45, 14195 Berlin, Germany.
Email: lena.fleig@fu-berlin.de
574951
HPQ0010.1177/1359105315574951Journal of Health PsychologyFleig et al.
research-article2015
Article
at Freie Universitaet Berlin on May 27, 2015hpq.sagepub.comDownloaded from
526 Journal of Health Psychology 20(5)
on the linkage between at least two health behav-
iors and the association between specific psycho-
logical constructs (e.g. self-efficacy). Further
emphasis should be placed on the question as to
how individuals go about changing more than one
behavior (e.g. sequentially or simultaneously). In
line with the MBA, this study examined how spe-
cific health behavior constructs taken from the
health action process approach (HAPA;Schwarzer,
2008) relate to each other (i.e. cross-behavior
associations) and how individuals change physi-
cal activity (PA) and fruit and vegetable intake
(FVI) (i.e. processes of multiple behavior change).
When looking at behavioral measures, vari-
ous studies provide evidence that healthy nutri-
tion and PA are positively related to each other
(De Vries et al., 2008; Kremers et al., 2004;
Södergren et al., 2012). In other words, both
health behaviors appear to rather facilitate than
hinder each other. Theoretically, both behaviors
may represent two similar contexts between
which psychological resources and strategies
can be transferred (i.e. resource and strategy
transfer; Fleig et al., 2011; Lippke et al., 2012;
Nigg et al., 2009). A “facilitating pattern”
between PA and nutrition should thus not only
transpire on a behavioral level but also with
regard to psychological constructs. Indeed, pre-
vious research suggests that individuals who
believe that they are able to engage in PA
despite barriers also hold more positive beliefs
about their ability to change their nutrition
(Annesi and Marti, 2011; Kremers et al., 2004).
Similarly, individuals who are motivated to
change their PA levels have been found to also
hold the intention to change their nutrition
(Kremers et al., 2004). According to the HAPA
(Schwarzer, 2008), individuals are more likely
to change a behavior if they complement their
intentions and efficacy beliefs with concrete
plans on when and where to change their behav-
ior (i.e. planning). This study therefore aimed
to investigate cross-behavior associations
between motivational (i.e. intention, self-effi-
cacy) and volitional constructs (i.e. planning).
Behavior change toward a particular goal can
be either understood as a continuous process
(such as translating intentions into behavior) or
as a series of distinct stages. These categories
reflect cognitive or behavioral characteristics.
According to the stage assumptions of the HAPA
(Schwarzer, 2008), individuals in the pre-inten-
tion stage (non-intender) are not motivated to
change their behavior. Individuals in the inten-
tion stage (intender) have already set the goal to
engage in a new behavior, individuals in the
action stage (actor) are already active. Studies
following the MBA can investigate linkages
between stages for different health behaviors to
establish cross-behavior associations. Previous
stage-based studies which were mainly con-
ducted within the framework of the transtheoreti-
cal model (TTM) (Prochaska and DiClemente,
1991) provide evidence for a positive association
between stages of change for PA and FVI (e.g.
Clark et al., 2005; Lippke et al., 2012).
Individuals who reside in an advanced stage for
PA are also more likely to be in an advanced
stage for healthy nutrition. The aim of this study
was to replicate these findings within the frame-
work of the HAPA.
In addition, the MBA raises the question as
to whether changes in different behavior
domains are related and, if so, whether individ-
uals modify their health behaviors at the same
time (i.e. simultaneously) or step by step (i.e.
sequentially). One way to answer this question
is to look at associations of stage transitions
across behavior domains. If change processes
of different behavior domains are related, indi-
viduals who show progress with regard to PA
should also be more likely to progress with
regard to healthy nutrition.
Finally, multiple behavior change can be
investigated by comparing the rates of behavior
change in PA and FVI as a function of whether
individuals intend to change both behaviors at
the same time (i.e. simultaneous changer) or
whether they intend to change their lifestyle by
starting with a single behavior (e.g. either PA or
FVI; one-at-a-time changer). This can be easily
conducted by comparing individuals who either
reside in the intention stage for both behaviors
(i.e. double-intender, simultaneous changer) and
those who are in the intention stage for one
behavior but in the pre-intention stage for
at Freie Universitaet Berlin on May 27, 2015hpq.sagepub.comDownloaded from
Fleig et al. 527
another behavior (i.e. single-intender, one-
at-a-time changer). Individuals who intend to
change both behaviors and thus reside in the
same stage for both behaviors can be referred to
as “stage congruent,” whereas those who reside
in different stages can be described as “stage
incongruent.” If rates of success in multiple
behavior change are in favor of either of the sub-
groups, this could be interpreted as one pathway
of change—either simultaneous or one-at-a-
time—being superior to the other.
One-at-a-time changers are assumed to be
more successful in mastering multiple behaviors
as they can focus their resources (Annesi, 2012),
perceive less goal-conflict, and can make use of
cross-behavioral transfer (e.g. Fleig et al., 2014;
Nigg et al., 2009) as compared to simultaneous
changers. On the other hand, simultaneous
changers may maintain an advantage as they
experience synergistic effects when changing
related behaviors (Annesi, 2012; Atkins and
Clancy, 2004) and are more confident about
their ability to regulate multiple goals (Jung and
Brawley, 2013). So far, only very few experi-
mental trials have directly investigated the
impact of targeting two health behaviors simul-
taneously versus one at a time, yielding incon-
sistent results. Whereas three studies (see review
by Prochaska and Prochaska, 2011) indicated no
differences between conditions, two other stud-
ies provided evidence in favor of the sequential
intervention (Hyman et al., 2007; Schulz et al.,
2012). In contrast to that, King et al. (2013)
recently showed that individuals receiving the
simultaneous intervention were more likely to
meet recommendations, but not in the sequential
interventions. With such few and inconsistent
findings, this study explored the role of stage
congruence—as indicative of a sequential ver-
sus a simultaneous pathway—on multiple health
behavior change.
Aims of this study
The primary objective was to examine the
interrelation between PA and FVI. This was
done by looking into three different indicators
of cross-behavior relationships. The following
hypotheses were tested: FVI and PA as well as
its social-cognitive determinants (i.e. intention,
self-efficacy, and planning), and stages are pos-
itively correlated.
A second aim of this study was to investigate
how change in PA was related to change in FVI.
Following a stage theory approach, the follow-
ing hypothesis was tested: there exists a posi-
tive interrelation between the stage progression
concerning PA and FVI.
Finally, the effect of stage congruence on
changes in PA and FVI was explored. More spe-
cifically, the study explored whether or not par-
ticipants in the same stage (stage congruence)
for two behaviors are more successful at chang-
ing than those who have stage incongruence
between two behaviors.
Methods
Design
The research aims were addressed in an online
study with two measurement points in time.
Measures were taken via web-based question-
naires at baseline (Time 1, T1) and reassessed at
a 4-week follow-up period (Time 2, T2). The
study was approved by the Departmental Ethics
Committee and conducted in line with German
Psychological Society ethical guidelines.
Participants and recruitment
Participants were recruited from the German
general population via press releases, announce-
ments on university websites, and mailing lists.
At baseline and at 4 week follow-up period, par-
ticipants filled in an online questionnaire with-
out receiving any incentives. Persons with
medical conditions that conflicted with engag-
ing in regular PA and eating five portions of
fruit and vegetable were excluded. The initial
sample comprised N = 2693 participants. Data
at follow-up were available from n = 1002 par-
ticipants (37.2% of the initial sample). Mean
age of the longitudinal sample was 37.2 years
(standard deviation (SD) = 11.4 years; range
18–78 years), and the sample consisted of more
at Freie Universitaet Berlin on May 27, 2015hpq.sagepub.comDownloaded from
528 Journal of Health Psychology 20(5)
women (77.3%) than men. Of all participants,
38.1 percent were married or in a permanent
relationship, 39.2 percent had at least one child.
Two-thirds of the longitudinal sample (83.6%)
reported having completed high school or a
higher degree, and 53.2 percent were employed.
At baseline, FVI averaged 3.1 (SD = 1.6) daily
servings, and moderate and vigorous PA aver-
aged 131.7 minutes per week (SD = 118.3 min-
utes). The average reported body mass index
was 25.0 (SD = 5.1).
Measures
PA was measured with a modified version of the
Godin Leisure-Time Exercise Questionnaire
(GLTEQ;Godin and Shephard, 1985). This self-
report measure has been validated with physio-
logical and anthropometrical measures (i.e.
VO
2
max and body fat; Jacobs et al., 1993).
Participants were asked to report the average
number of sessions per week and the average
duration of a session regarding vigorous (heart
beats rapidly, sweating) and moderate (not
exhausting, light perspiration) PA in the past
month. Only activities outside of work duties
(at work or at home) and voluntary activities
were addressed. Total PA was calculated by
multiplying the number of sessions per week by
the number of minutes per session. Change
scores for PA were operationalized by subtract-
ing T1 from T2 minutes. In line with the “small
steps approach” (Hill et al., 2003) and similar to
Vandelanotte et al. (2008), successful behavior
change was defined as an increase of at least
10 minutes of moderate and/or strenuous PA per
week on average.
T1 and T2 FVI were measured by the open-
ended question “How many servings of fruit and
vegetable did you eat on a typical day of the last
month?” This item was adapted from an English
measure that has been validated against dietary
biomarkers (Steptoe et al., 2003). One portion
was defined as the amount of food that fits into
the palm of the hand. Participants were asked
not to take into account products made of pota-
toes. Change scores for FVI were calculated by
subtracting number of ingested portions reported
at T1 from T2 intake. For FVI, successful behav-
ior change was defined as an increase of at least
one serving of fruit and/or vegetables on aver-
age per day. To combine behavior change in PA
and FVI, a nominal variable with four categories
was computed: successful change in both behav-
iors, successful change in only one behavior
(either PA or FVI), or no or negative change in
either behavior (Prochaska et al., 2008).
Social-cognitive variables were assessed
using 4-point Likert scales, ranging from not at
all true (1) to completely true (4). PA intention
was assessed with two items matching the two
behavior intensities: “I intend to engage in vig-
orous PA (increased heart rate, sweating) during
the next month” and “I intend to engage in
moderate PA (hardly exhausting, light sweat-
ing) during the next month” (Nigg, 2005;
r = .18). Intentions to consume fruit or vegeta-
bles were measured by two items “I intend to
eat at least five servings of fruit or vegetables
per day during the next month” and “I intend to
eat fruit and vegetables with every meal”
(Lippke et al., 2009; r = .53).
Self-efficacy was measured with seven items
for each behavior domain. The item “I am confi-
dent that …” was supplemented with the follow-
ing exemplary statements for PA: “… I can
manage to be more physically active” (Schwarzer
et al., 2007) or “… I can engage in regular physi-
cal activity even if I have worries or problems”
(Dijkstra et al., 1998; Cronbach’s α = .86) and for
nutrition: “… I can manage to eat five servings
of fruit and vegetables per day” or “… I can
manage to eat five portions of fruit and vegeta-
bles per day even if I have worries or problems.”
(Dijkstra et al., 1998; Cronbach’s α = .92).
Planning was measured with two items per
behavior, such as “I have already precisely
planned when, where, and how to eat five servings
of fruit or vegetables throughout the day” and “I
have already precisely planned where, when, and
for how long to be physically active.” Similar
measures of planning have shown factorial and
predictive validity (Sniehotta et al., 2005).
Stages of change for both behaviors were
measured with validated algorithms for HAPA
stages (Lippke et al., 2009). For both behaviors,
at Freie Universitaet Berlin on May 27, 2015hpq.sagepub.comDownloaded from
Fleig et al. 529
participants were instructed, “Please think
about the last month" and then instructed, “Did
you engage in PA at least 3 days per week for
30 minutes or more?” Regarding nutrition, the
question was “Did you eat at least five portions
of fruit and vegetables per day?” Participants
responded on a rating scale with the verbal
anchors “No, and I don’t intend to do so” and
“No, but I am considering it” referring to the
pre-intention stage, “No, but I seriously intend
to do so” representing the intention stage, and
finally, “Yes, but it is difficult for me” and “Yes,
and it is easy for me” representing the action
stage. Stage transitions were calculated by sub-
tracting T1 stage from T2 stage.
Statistical methods
All analyses were run with SPSS 20. Dropout
analyses compared retained participants and
those lost after T1 using analyses of variance
(ANOVAs) for continuous measures and χ
2
-
tests for categorical measures. Associations
between behavioral and social-cognitive meas-
ures were examined using Pearson’s correla-
tions. To examine the association between
categorical data, Pearson’s chi-square test was
conducted. Multinomial logistic regression
(MLR) was used to evaluate the effect of stage
congruence on behavior change in a subgroup
of individuals who were intending to change
both behaviors (i.e. double-intender, stage con-
gruent group), who were intending to change
PA only (i.e. single-intender, stage incongruent
group A), and those who were intending to
change only their nutrition (i.e. single-intender,
stage incongruent group B). Stage congruence
formed the predictor variable and was dummy
coded, with the stage congruent group as base-
line category. In line with Prochaska et al.
(2008), multiple behavior change was opera-
tionalized by an index with four categories (see
measures). Age, gender, body mass index, and
educational status were entered as covariates.
Odds ratios (ORs) with a 95 percent confidence
interval are reported as estimates of effect size.
Missing values in the data of the longitudinal
sample were imputed using the expectation–
maximization (EM) algorithm.
Results
Descriptive results
Dropout analyses indicated that individuals
who continued in the study were more likely to
be male, unemployed, with a low educational
degree, and in an earlier stage of PA than those
who dropped out (ps < .05). Besides that, no
other differences were found. T1 stage distribu-
tion and stage congruence across the two health
behaviors are displayed in Table 1.
With regard to stage congruence, the major-
ity of participants were assessed as being stage
incongruent (n = 1610, 59.8%). Of the nine pos-
sible stage combinations, most individuals were
allocated to the pre-intention stage regarding
FVI and the action stage regarding PA (i.e. stage
Table 1. Stage distribution and stage congruence in the baseline sample (n = 2693).
Fruit and vegetable intake (FVI)
Non-intender,
n (%)
Intender,
n (%)
Actor, n
(%)
Total, n (%)
Physical
activity
(PA)
Non-intender 375 (13.9) 100 (3.7) 116 (4.3) 591 (21.9)
Intender 365 (13.6) 319 (11.8) 204 (7.6) 888 (33)
Actor 500 (18.6) 325 (12.1) 389 (14.4) 1214 (45.1)
Total 1240 (46.1) 744 (27.6) 709 (26.3) 2693 (100)
Numbers in parentheses represent percent of n = 2693. Shaded sub-groups are included in further analyses on the ef-
fect of stage congruence on multiple behavior change.
at Freie Universitaet Berlin on May 27, 2015hpq.sagepub.comDownloaded from
