Skills Training via Smartphone App for University Students
with Excessive Alcohol Consumption: a Randomized
Controlled Trial
Mikael Gajecki
1,2
& Claes Andersson
3
& Ingvar Rosendahl
1
& Kristina Sinadinovic
1
&
Morgan Fredriksson
4
& Anne H Berman
1
#
The Author(s) 2017. This article is published with open access at Springerlink.com
Abstract
Purpose University students in a study on estimated blood
alcohol concentration (eBAC) feedback apps were offered
participation in a second study, if reporting continued exces-
sive consumption at 6-week follow-up. This study evaluated
the effects on excessive alcohol consumption of offering ac-
cess to an additional skills training app.
Method A total of 186 students with excessive alcohol con-
sumption were randomized to an intervention group or a wait
list group. Both groups completed online follow-ups regard-
ing alcohol consumption after 6 and 12 weeks. Wait list par-
ticipants were given access to the intervention at 6-week fol-
low-up. Assessment-only controls (n=144) with excessive
alcohol consumption from the ongoing study were used for
comparison.
Results The proportion of participants with excessive alcohol
consumption declined in both intervention and wait list groups
compared to controls at first (p < 0.001) and second follow-
ups (p = 0.054). Secondary analyses showed reductions for
the intervention group in quantity of drinking at first follow-
up (−4.76, 95% CI [−6.67, −2.85], Z = −2.09, p =0.037)and
in frequency of drinking at both follow-ups (−0.83, 95% CI
[−1.14, −0.52], Z = −2.04, p =0.041;−0.89, 95% CI [−1.16,
−0.62], Z = −2.12, p = 0.034). The odds ratio for not having
excessive alcohol consumption among men in the intervention
group compared to male controls was 2.68, 95% CI [1.37,
5.25] (Z =2.88,p = 0.004); the figure for women was 1.71,
95% CI [1.11, 2.64] (Z =2.41,p =0.016).
Conclusion Skills training apps have potential for reducing
excessive alcohol use among university students. Future re-
search is still needed to disentangle effects of app use from
emailed feedback on excessive alcohol consumption and
study participation.
Trial Registration NCT02064998
Keywords Randomized controlled trial
.
Problem drinking
.
Alcohol abuse
.
College
.
University
.
Smartphone
.
Mobile
phone
.
eHealth
.
mH
ealth
.
Brief intervention
.
Relapse
prevention
Introduction
With the advent of the smartphone in the 1990s, manufac-
turers began incorporating increasingly powerful computing
and communication capabilities in a handheld format, making
it possible to run native applications, apps, and to view and
interact with advanced content over the Internet. The global
use of smartphones is steadily increasing and it is estimated
that in 2015, there were 1.9 billion users worldwide [1]. The
ubiquity and the capabilities of smartphones have spurred de-
velopers to provide content in a multitude of different areas,
including health care and specifically for mental health care
and substance abuse, where many apps are available but few
have been evaluated scientifically [2].
Electronic supplementary material The online version of this article
(doi:10.1007/s12529-016-9629-9) contains supplementary material,
which is available to authorized users.
* Mikael Gajecki
mikael.gajecki@ki.se
1
Department of Clinical Neuroscience, Centre for Psychiatry
Research, Karolinska Institutet, Norra Stationsgatan 69, plan 7, 113
64 Stockholm, Sweden
2
Stockholm Centre for Dependency Disorders, Stockholm, Sweden
3
Department of Criminology, Malmö University, Malmö, Sweden
4
Liquid Media AB, Stockholm, Sweden
Int.J. Behav. Med.
DOI 10.1007/s12529-016-9629-9
Many apps related to alcohol consumption are com-
mercially available, with the majority being for entertain-
ment purposes, not uncommonly encouraging drinking.
Of those that ad dre ss re duc t ion of dri n ki ng, on l y a few
have been scientifically assessed [3–5]. A recent review
identified only six scientifically researched apps for re-
ducing alcohol use [6]. College and university students,
who engage i n more hazardous drink ing and high con-
sumption events, than do their non-student peers in t he
population [7, 8], are a frequent and easily accessible
target group for intervention efforts. Addressing the neg-
ative consequences of hazardous drinking patterns is of
particular importance to minimize both short-term nega-
tive consequences such as hangovers, aggressiveness,
blackouts and worse, academic performance [9, 10],
alcohol-related injury, and death [11]ortherisksofde-
veloping substance use disorder in later adult develop-
ment [9]. Mobile interventions could be a highly efficient
medium for reaching this group. In a very recent review
on mobile interventions targeting risky drinking among
university students, we found only two studies that exam-
ined the use of smartphone apps [12]. In the first study,
Witkiewitz et al. found that a smartphone app, containing
some c omponents of the Brief Alcohol Scree ning an d
Intervention for College Students (BASICS) program that
has s hown positive effects on university students’ drink-
ing, had no effect on reducing heavy episodic drinking or
simultaneous cigarette smoking and alcohol use at 1-
month-follow-up [13]. The second study compared
assessment-only controls with access to two smartphone
apps, PartyPlanner and Promillekoll, which both offered
real-time estimated blood alcohol concentration (eBAC)
feedback but where PartyPlanner also offered a planning
and follow-up feature. This study, conducted by our re-
search group, found no improvement in any of the inter-
vention groups compared to controls and a negative find-
ing of an increase in frequency of drinking occasions (but
not quantity) in the Promillekoll group. Also, we observed
that about one t hird of the students drank excessively
throughout the trial, beyond public health recommenda-
tions to drink no more than 9 standard drinks per week
for women and 14 standard dr inks for men. Secondary
analyses revealed a gender difference: the increase in
drinking frequency was found only among men [14].
The general sparsity of research findings and the negative
nature of the results in our first study (A) [14]withits6-week
follow-up led us to design two new studies (B and C) [15]. In
study B, the content of both the apps tested in study A was
somewhat improved and follow-up was extended to 12 and
18 weeks to detect any possible delayed effects. In study C,
reported in this article, a sub-group of students with excessive
alcohol consumption at the 6-week follow-up in study B was
offered access to a new smartphone app with skills training
components from the relapse prevention (RP) program [16],
commonly used in face-to-face treatment for problematic al-
cohol use. For an overview of studies A, B, and C, see Fig. 1.
In the present study, our aim was to investigate whether
students who did not respond to apps offering eBAC feedback
might be more responsive to an app with skills training com-
ponents. A fundamental part of RP consists of participants
learning to identify situations where the risk of drinking is
higher, and how to cope with such situations by either altering
their responses to certain situations or avoiding exposure to
the triggers altogether. Common coping skills taught in RP are
urge surfing, a mindfulness exercise for dealing with urges
and cravings, relaxation for coping with stress, and assertive-
ness skills for dealing with social pressure [17] The program
in its entirety has proven effective in face-to-face treatment for
alcohol problems [18]. One of the research challenges for app
studies is selecting suitable components adaptable for delivery
in the app format. The app used in this study packaged select-
ed RP components in an app format and was evaluated among
university students with excessive alcohol consumption.
In summary, earlier studies on smartphone apps for univer-
si
ty students have shown null results [12] although a dose-
response effect was identified where students who accessed
more modules in the BASICS skills-based app were less likely
to drink at all during the 14-day assessment period [13]. Given
that our previous study [14] showed that about one third of the
sample had excessive alcohol consumption but did not change
their behavior after using a real-time eBAC-feedback-based
app, we were interested in investigating the effects of giving
students who reported excessive alcohol consumption, after
having used a real-time eBAC app for 6 weeks (in study B),
access to an in-depth skills training smartphone application,
Telecoach™ as an add-on (in study C). Our primary hypoth-
esis for study C, reported herein, was that the proportion of
excessive alcohol consumption would be lower in the app
group, compared to an assessment-only group. For compari-
son with study A [14], we analyzed alcohol-related outcomes
investigated in that study. In addition, the gender differences
found in that study led us to also conduct gender-based anal-
yses in the present data. The latter were exploratory secondary
analyses, not based on explicit hypotheses.
Methods
Design
A skills training smartphone app, TeleCoach™, was made
accessible to a sample of university students with excessive
alcohol consumption who already had access to one of two
apps offering real-time eBAC feedback (study B) [15]. In
study C, a randomized parallel three-group repeated measures
design, alcohol-related outcomes for a TeleCoach™
Int.J. Behav. Med.
intervention group and a wait list group were compared to an
assessment-only control group. At the 6-week follow-up in
study C, the wait list group was given access to
TeleCoach™, and at both 6 and 12 weeks, alcohol-related
outcomes were compared between the TeleCoach™ interven-
tion group (with access to an eBAC app for 18 weeks and
TeleCoach™ for 12 weeks), the wait list group (with access
to an eBAC app for 18 weeks and TeleCoach™ for 6 weeks)
and an assessment-only control group. The primary outcome
measure was the proportion of students with excessive alcohol
consumption, and secondary measures were quantity, frequen-
cy, binge drinking occasions, average eBAC per week, and
peak eBAC per month, all measured at the 6- and 12-week
follow-ups. The trial was registered at clinicaltrials.gov:
NCT02064998.
Participants
Two hundred fifty-seven university students from three major
universities in the capital area of Sweden were invited via
email to the present study if they reported excessive alcohol
consumption at the 6-week follow-up in study B. The 186
participants who gave their informed consent via an online
form to participating in study C were informed that they
would receive a link to a new web-based smartphone applica-
tion, either immediately after randomization (intervention
group) or 6 weeks later (wait list group), and that they would
be contacted for follow-up assessments 6 and 12 weeks later.
A s ub -samp le from the assessment-only control group i n
study B, matched on excessive weekly alcohol consumption
at the 6-week follow-up, was used as a control group in the
present study. All participants in studies B and/or C were
informed that participation with completion of all follow-ups
meant that they were entered in a lottery offering the opportu-
nity to win one of three iPads.
Randomization
Participants in the present study were randomized to either the
intervention or wait list condition, with a ratio of 1:1 using the
randomization function in the IBM SPSS Statistics for MacOS
X, Version 22 (IBM Corp, Armonk, NY, USA).
Intervention
TeleCoach™ app, a web-based app requiring an Internet con-
nection was developed by the authors and consists of a main
menu with two parts: (a) registration of alcohol consumption
in standard glasses for each day of the past week, resulting in
brief feedback and information on guidelines for hazardous
drinking and (b) a relapse prevention skills training menu
offering two options: Bsay no to alcohol^ or Bfeel better with-
out alcohol.^ The Bsay no^ option leads to additional options
for risk situation analysis or refusal exercises. The risk situa-
tion analysis consists of answering the questions from the
Alcohol Abstinence Self-efficacy Scale (AASE) [19], with
feedback summarizing reported risk situations. Refusal exer-
cises are presented in text form. Selecting the Bfeel better
without alcohol^ option leads to a choice between listening
to recorded relaxation exercises, positive thought exercises, or
urge surfing training. Participants were instructed to use
TeleCoach™ at will. They were also informed that they could
continue using the app previously assigned to them in the
preceding, ongoing study B.
Participants with excessive alcohol
consumption at follow-up invited to Study C
Study A
Published in 2014 (Gajecki et al .,
2014 )
Participants randomized into one of three
groups: Two groups that accessed one of
two eBAC feedback apps, and one
assessment-only control group.
Study B
Under analysis 2016
Study C
Reported in this arcle
Week 0
Week 6
Participants randomized into one of three
groups: Two groups that accessed one of
two eBAC feedback apps, and one
assessment-only control group.
Follow-up and end of study
1
st
follow-up
1
st
follow-up
Waitlist group accessed TeleCoach™
Consenting participants randomized into one
of two groups: One that accessed a skills
training app (TeleCoach™), and one wait-list
group.
Week 12
Week 18
2
nd
follow-up
3
rd
follow-up
and end of
study
2
nd
follow-up
and end of
study
Fig. 1 Figure comparing the time frames and flow of studies A (Gajecki et al., 2014), B (Berman et al., 2016), and C (Berman et al. 2016, and current
article)
Int.J. Behav. Med.
Adverse Event and Technical Limitations
Due to technical problems, the intervention group was given
access to TeleCoach™ 3 weeks after randomization.
Intervention group participants were informed about the delay
about 1 week after randomization; follow-ups were scheduled
for 6 and 12 weeks after access was provided, meaning that
that follow-ups occurred with a 3-week time lag for the inter-
vention group. Another technical limitation is the fact that
objective data on actual app use was not available to the re-
search group.
Seasonality
The study took place between December 2014 and
March 2015. Swedish university education is not based on
the concepts of midterms or finals, so there were no uniform
examination periods during this time. The weeks leading up to
Christmas and New Year’s Eve are associated with parties and
alcohol consumption in Sweden, and during the active study
period, both major public holidays occurred. Also, follow-up
data were collected for the participants in the intervention
group during Easter week.
Assessments
Baselineassessmentandthetwofollow-ups(at6and
12 weeks) were conducted via an online questionnaire that
included the Daily Drinking Questionnaire (DDQ) and a ques-
tion on motivation to reduce alcohol consumption. Links to
the assessment questionnaires were sent by email, with two
reminders sent 2 days a part. The A lcohol Use Disorders
Identification Test (AUDIT; [20]) was part of the baseline
assessment in study B, meaning that AUDIT scores at 6 weeks
prior to registration in study C were available for all study C
participants. The second (and final) follow-up included the
AUDIT, as well as questions on apps used, whether the par-
ticipant had accessed any other means of help for alcohol
consumption and questions on perceived ease of use and suit-
ability of the app for problematic alcohol consumption, while
baseline measurement and the 6-week follow-up only includ-
ed the DDQ and the question on motivation to reduce alcohol
consumption.
Measures
The Daily Drinking Questionnaire (DDQ) [21] was used to
measure quantity and frequency of alcohol consumption. This
instrument was translated into Swedish by Malmö University
in cooperation with the University of Washington. Participants
were asked to consider a typical week during the last month
and state how many standard glasses of alcohol they drank
and over how many hours during each day of this typical
week. They were also asked to report their peak alcohol con-
sumption event during the last month in terms of how many
standard glasses they drank, during a self-reported number of
hours. This measure in a slightly different form has demon-
strated good test-retest reliability in paper format [22] and
good internal consistency (Cronbach’s α =.83)[13].
Estimated blood alcohol concentration (eBAC) was calcu-
lated based on the values from the DDQ in conjunction with
the weight and gender of the participant. The formula used
was the widely known Widmark formula as modified and
used by the US National Highway Traffic Safety
Administration [23]: eBAC (in parts per mille, as is standard
in Sweden) = ([number of standard glasses] × 12 g) / ([body
weight in kg] × C) − ([no. of hours] × 0.15), where C is a
gender specific constant (0.68 for men, and 0.55 for women).
In order to convert the eBAC from parts per mille to percent-
age values for this article, values were divided by 10.
Motivation to reduce alcohol consumption was measured
using a simple question BHow interested are you in reducing
your alcohol consumption?^ on a scale from 1 to 10.
Definitions
Excessive Alcohol Consumption and Binge Drinking.
Excessive alcohol consumption was defined as drinking more
than 14 standard glasses per week for men and more than 9 for
women and binge drinking was defined as 5 or more standard
glasses per occasion for men, and 4 or more standard glasses
per occasion for women [24]. These definitions constitute the
current recommendations of the National Public Health
Agency in Sweden.
Standard Glass
A standard glass was defined as containing 12 g of pure
alcohol [24].
Outcomes
Outcomes in this study were calculated based on the partici-
pants’ DD
Q registrations, with the addition of the participant’s
gender and reported weight for calculating eBAC.
1. Primary outcome: the proportion of participants with ex-
cessive alcohol consumption in each group
2. Secondary outcomes
(a) Quantity—the number of standard glasses consumed
during a 7-day period (based on the DDQ question about
drinking habits in a typical week during the last month)
(b) Frequency—the number of days in a 7-day period during
which the participant consumed alcohol
Int.J. Behav. Med.
(c) Binge occasions—the number of days in a 7-day period
where the participant engaged in binge drinking
(d) Average eBAC—the average eBAC over a 7-day period
(e) Peak eBAC—the eBAC calculated from the peak con-
sumption event during the last 30 days
Statistical Analyses
Descriptive statistics were used to present baseline character-
istics. Analysis of variance (ANOVA) was used to determine
any baseline differences between groups in age, AUDIT [20]
(from baseline assessment in study B), mean eBAC, peak
eBAC, quantity, frequency, and number of binge drinking
occasions. Pearson’s chi-squared tests were used to determine
differences between the groups in gender proportions and the
proportion of participants drinking excessively. Generalized
estimating equations (GEE) [25] with an exchangeable work-
ing correlation structure were used for analyses of longitudinal
data: quantity, frequency, number of binge drinking occasions,
mean eBAC, and peak eBAC. The semirobust Huber-White
sandwich estimator was used to estimate standard errors. The
sandwich estimator makes fewer assumptions than the con-
ventional estimator of variance [26] and therefore increases
the theoretical robustness of the results of GEE analyses, in
relation to a possible incorrect choice of working correlation
matrix [27]. All available longitudinal data were entered into
the GEE analyses. No imputation was carried out as simula-
tion studies comparing the regression coefficients and stan-
dard errors of mixe d models with and without a previous
multiple i mputation have shown very inconsistent results
[27]. In the analysis of the dichotomous outcome, no other
factors or covariates were controlled for because of problems
in ensuring the model would converge. We entered gender,
age, and pre-randomization scores as covariates in the second-
ary analyses to control for possible confounding. Our assump-
tion was that age, gender, and stability in alcohol consumption
before randomization are factors correlated to the app use as
well as being predictors for the outcomes. All GEE analyses
were performed using Stata 14 (StataCorp. 2015. College
Station, TX: StataCorp LP).
Exclusion and Substitution
One participant entered clearly faulty entries in the DDQ rat-
ings at baseline. The DDQ values for this participant were
substituted with the mean sample value, as this was deemed
not to interfere with the statistical calculations in any mean-
ingful way. Other outcome variables relying on DDQ data
were calculated from this substituted value (see Fig. 2 for a
participant flowchart).
Results
Participant Characteristics
Participants were 330 university students with excessive alco-
hol consumption selected from a 6-week follow-up in a par-
allel ongoing trial. Participant characteristics at registration for
the current trial, including age and gender distributions and
mean scores on primary and secondary outcomes are present-
ed in Table 1. It should be noted that AUDIT scores for the
participants, available from pre-randomization, baseline as-
sessment in study B, 6 weeks prior to the invitation to partic-
ipate in the current trial, were indicative of hazardous drinking
[20, 28]. Table 1 shows no overall baseline differences among
participants in the three study groups. Regarding gender, over
two thirds of the participants in this trial were women. Also,
the average consumption in standard glasses was substantially
higher at study rec ruitment among men (M =21.92,
SD = 8.40) compared to women (M = 14.49, SD = 5.61);
(t(328) = −9.46, p <0.001).
Retention
Eighty-seven percent of study participants responded to at
least one or both of the two follow-ups: 72.7% responded to
both follow-ups, while 7.6% responded only to the first
follow-up and 6.7% responded only to the second follow-up.
An ANOVA analysis comparing the three retention categories
and the non-responders showed no differences in baseline
characteristics.
Outcome Analyses
Regarding the primary outcome, the proportion of participants
with excessive alcohol consumption was significantly higher
in the control group (72.7%) compared to both the interven-
tion group (45.3%) and the wait list group (50.0%) at first
follow-up (χ
2
(2) = 17.78, p < 0.001) but not at second
follow-up (χ
2
(2) = 5.85, p = 0.05). At the second follow-
up, the intervention (52.1%) and wait list groups (56.7%)
showed small nominal rises in the proportion of participants
with excessive alcohol consumption, while a nominal decline
was shown in the control group (68.5%). Across both follow-
ups, the odds for not having excessive weekly alcohol con-
sumption in the intervention group were almost twice as high
as for controls, see Table 2.
Participants in the current trial had already been random-
ized to use one of two apps with a focus on eBAC feedback in
study B. As a possible confounding factor in this study could
derive from eBAC app effects, we include a supplementary
table showing an overview of the proportions of participants
with excessive alcohol consumption in relation to their ran-
domization i n study B (Supplementary Table S 1). When
Int.J. Behav. Med.