Content uploaded by Jan-Niklas Kramer
Author content
All content in this area was uploaded by Jan-Niklas Kramer on Jan 17, 2019
Content may be subject to copyright.
RESEARCH ARTICLE
A Cluster-Randomized Trial on Small Incentives
to Promote Physical Activity
D1X XJan-Niklas Kramer, D2X XMSc,
1
D3X XPeter Tinschert, D4X XMSc,
1
D5X XUrte Scholz, D6X XPhD,
2
D7X XElgar Fleisch, D8X XPhD,
1,3
D9X XTobias Kowatsch, D10X XPhD
1
Introduction: There has been limited research investigating whether small financial incentives can
promote participation, behavior change, and engagement in physical activity promotion programs.
This study evaluates the effects of two types of small financial incentives within a physical activity
promotion program of a Swiss health insurance company.
Study design: Three-arm cluster-randomized trial comparing small personal financial incentives
and charity financial incentives (10 Swiss Francs, equal to US$10.40) for each month with an aver-
age step count of >10,000 steps per day to control. Insureds’federal state of residence was the unit
of randomization. Data were collected in 2015 and analyses were completed in 2018.
Setting/participants: German-speaking insureds of a large health insurer in Switzerland were
invited. Invited insureds were aged ≥18 years, enrolled in complementary insurance plans and reg-
istered on the insurer’s online platform.
Main outcome measures: Primary outcome was the participation rate. Secondary outcomes
were steps per day, the proportion of participant days in which >10,000 steps were achieved and
non-usage attrition over the first 3 months of the program.
Results: Participation rate was 5.94% in the personal financial incentive group (OR=1.96, 95%
CI=1.55, 2.49) and 4.98% in the charity financial incentive group (OR=1.59, 95% CI=1.25, 2.01)
compared with 3.23% in the control group. At the start of the program, the charity financial group
had a 12% higher chance of walking 10,000 steps per day than the control group (OR=1.68, 95%
CI=1.23, 2.30), but this effect dissipated after 3 months. Steps per day and non-usage attrition did
not differ significantly between the groups.
Conclusions: Small personal and charity financial incentives can increase participation in physical
activity promotion programs. Incentives may need to be modified in order to prevent attrition and
promote behavior change over a longer period of time.
Trial registration: This study is registered at www.isrctn.com ISRCTN24436134.
Am J Prev Med 2019;56(2):e45
−
e54. © 2018 American Journal of Preventive Medicine. Published by Elsevier Inc.
This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
INTRODUCTION
Strong empirical evidence indicates that physical
inactivity increases the risk of mortality
1
as well
as of noncommunicable diseases, such as coro-
nary heart disease, diabetes, or cancer.
2,3
More than 20%
of the world’s population is inactive (i.e., accumulating
less than 150 minutes of moderate-intensity or 75
minutes of vigorous-intensity aerobic physical activity
From the
1
Center for Digital Health Interventions, Institute of Technology
Management, University of St. Gallen, St. Gallen, Switzerland;
2
Depart-
ment of Psychology, Applied Social and Health Psychology, University of
Zurich, Zurich, Switzerland; and
3
Center for Digital Health Interventions,
Department of Management, Technology, and Economics, ETH Zurich,
Zurich, Switzerland
Address correspondence to: Tobias Kowatsch, PhD, Center for Dig-
ital Health Interventions, Institute of Technology Management, Uni-
versity of St. Gallen, Dufourstrasse 40a, 9000 St. Gallen, Switzerland.
E-mail: tobias.kowatsch@unisg.ch.
0749-3797/$36.00
https://doi.org/10.1016/j.amepre.2018.09.018
© 2018 American Journal of Preventive Medicine. Published by Elsevier Inc.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Am J Prev Med 2019;56(2):e45−e54 e45
per week),
4
illustrating the need for effective promotion
of physical activity at a large scale. Technology may pro-
vide a viable means to address this problem: Driven by
the promise and increasing popularity of wearable devi-
ces,
5
many key stakeholders in health care, including
employers and health insurers, have designed large-scale
physical activity promotion programs that utilize com-
mercial-grade activity trackers.
6
For example, partici-
pants of the U.S. health insurer Humana’s Go365
program can earn points for every 1,000 steps they log
using a smartphone or activity tracker.
7
To provide an effective solution to the high prevalence
of physical inactivity, those programs have to attract the
individuals in need, effectively change behavior, and
engage participants in the long term. Therefore, health
insurers and employers often promote their physical
activity programs using financial incentives like cash,
bonus points, or vouchers.
6,8
This strategy is supported
by empirical evidence: Previous research demonstrated
that financial incentives can increase participation in
health promotion programs
9,10
and change physical
activity−related behaviors, such as exercise session
attendance,
11−13
exercise behavior,
12
and objectively
measured physical activity.
13
However, these studies
mostly use large incentive amounts. A review of RCTs,
for example, calculated a mean maximum incentive
value of $20.75 per week.
11
Because changes in physical
activity are typically not sustained after incentives are
withdrawn,
14−18
larger incentives are unsuitable for
application in large-scale physical activity programs.
A potential solution is the use of financial incentives
that are small enough to be sustained indefinitely, but
little is known about whether small incentives can pro-
duce the desired effects. In previous studies, small
incentives have primarily been shown to change simple
behaviors, such as completing child vaccinations,
19
attending preventive health examinations,
20
or com-
pleting tuberculosis skin test readings
21
primarily in
deprived populations. In studies investigating the effect
of incentives on physical activity, the incentive value
seems to correlate with the effect size,
12
but very few
studies have specifically investigated the effect of small
financial incentives. Thus, it is unclear whether small
financial incentives can promote physical activity and
how they affect behavioral change.
Incentives in the form of donations to charity (charity
financial incentives) are a promising alternative incen-
tive design. Donating to charity activates an additional
neural reward system compared with mere financial
incentives
22
and has been related to happiness in obser-
vational and experimental studies.
23
This “warm glow of
giving”
24
has been used in marketing strategies to
motivate purchase behavior,
25
and recent
17,18
and ongo-
ing
26
studies investigate its potential to promote physical
activity. These previous studies have found mixed
results: In a large RCT in Singapore, weekly personal
financial incentives but not charity financial incentives
($30 for walking 70,000 steps per week) increased daily
steps of full-time workers at 6 months.
18
In a smaller
RCT with older adults in Philadelphia, both weekly per-
sonal financial and charity financial incentives ($20 for
meeting a personalized step goal on at least 5 of 7 days)
were successful in promoting step goal achievement over
the 16-week study period but not during the 4-week fol-
low period.
17
The present study adds to this evidence base by com-
paring the effects of personal financial and charity finan-
cial incentives within a physical activity promotion
program offered by a large Swiss health insurance com-
pany. In contrast to previous research, this study deliber-
ately investigates small incentive values that have the
potential to be sustained indefinitely. In addition to
changes in physical activity, this study also analyzes
effects on participation and retention, because they rep-
resent central outcomes for the success of health promo-
tion programs.
METHODS
The design of this study is described in greater detail elsewhere.
27
Briefly, this study was conducted in cooperation with a large Swiss
health insurance company that implemented a 6-month pilot ini-
tiative to promote physical activity among its insureds from July
to December 2015. In June 2015, eligible insureds were randomly
assigned to a type of incentive (personal financial, charity finan-
cial, or no incentive) and invited via e-mail to participate in the
pilot program. Invitation e-mails contained information about the
promotion program, the incentive condition of the e-mail recipi-
ent, and a link to the insurance’s online platform where insureds
could log in and register for the program (Appendix, available
online). Additionally, the invitation e-mail informed insureds
about the opportunity to buy a compatible activity tracker at a
reduced price if they did not own one already. Insureds who were
not interested in the program were asked to indicate the underly-
ing reasons of their decision by completing a brief survey via a
separate link at the bottom of the invitation e-mail.
On the insurance provider’s online platform, insureds
received detailed information about the pilot program, data
protection policies, and eligibility criteria and could provide
consent to participate. To provide consent, participants had to
confirm that they did not have a medical condition that pro-
hibits increased levels of physical activity by ticking a check-
box. Insureds were advised to consult a physician if they were
in doubt. Insureds did not receive any information about the
existence of different incentive groups. To complete registra-
tion, insureds linked their wearable manufacturer’scustomer
account to the insurance provider’sonlineplatformsothat
their steps would be synchronized daily via an application
e46 Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54
www.ajpmonline.org
programming interface. To facilitate automatic synchroniza-
tion of step counts, compatible activity trackers were limited
to devices developed by the major wearable manufacturers on
the Swiss market Garmin, Jawbone, and Fitbit. Alternatively,
insureds could use the Fitbit smartphone application to track
daily steps. Once registered, participants were able to add
family members who also met the eligibility criteria to the
pilot program.
At the beginning of the program, participants were asked to
complete a web-based survey to collect data on demographic vari-
ables and covariates of physical activity. Participants received 10
Swiss Francs (CHF) for the completion of the survey. The insur-
ance company provided data on age, sex, nationality, and federal
state (canton) of all invited insureds as well as step data of partici-
pants. The IRB of the University of St. Gallen approved the study.
Analyses of primary and secondary outcomes were completed in
2018.
Study Sample
Because of legal regulations in Switzerland, the physical activity
promotion program is not part of the statutory health insurance
but offered to insureds with complementary insurance plans only.
Note, however, that in Switzerland 75% of people are enrolled in
complementary insurance plans.
28
To facilitate enrolment, only
insureds who met the predefined eligibility criteria were invited to
participate: insureds had to be aged >18 years, German speaking,
enrolled in a complementary insurance plan, and registered on
the insurer’s online platform. There was no racial or gender bias
in the selection of invited insureds. Naturally, invited insureds
resided primarily in the German-speaking parts of Switzerland.
No eligibility criteria were defined on the canton level.
Measures
In the personal financial incentive condition, participants received
CHF 10 for each month they walked >10,000 steps per day on
average and CHF 5 for each month they walked <10,000 but
>7,500 steps per day, which matches the approximate minimum
recommendation for daily physical activity.
29
The insurance com-
pany considered a value of CHF 10 as appropriate to be sustained
indefinitely. Charity financial incentives coincided with personal
financial incentives, with the exception that participants could
donate a chosen proportion of their earned reward to one of three
charity organizations (a foundation supporting the rights and
needs of Swiss children and adolescents, a foundation for health
promotion of Swiss adolescents, and an organization committed
to preserving the Swiss hiking track network). Participants in the
control group were informed that participation can enhance health
and well-being but did not receive any incentives during the first 3
months of the program. However, from the fourth month onwards
until the end of the program, the control group was entitled to per-
sonal financial incentives of CHF 20 for walking 10,000 steps on
average and CHF 10 for walking 7,500 steps on average per month.
Thus, participants of all groups could earn or donate a maximum
of CHF 60 during the 6-month pilot program. Participants could
log in to their online account on the insurance company’s website
to view a summary of their daily step counts and the amount of
their earned or donated rewards if applicable. At the beginning of
each month, all participants received an e-mail containing infor-
mation on whether they earned a reward in the past month as well
as a tip on how to increase physical activity.
An important study design consideration was minimizing the
risk of spillover effects between study groups, especially between
the incentivized groups and the control group, to prevent frustra-
tion and dropout among insureds. Insureds were therefore ran-
domized using blocked cluster-randomization based on their
canton of residence (n=20) with a block size of five and an alloca-
tion ratio of 2:2:1 with fewer insureds allocated to the control con-
dition. Each block consisted of two pairs of neighboring cantons
and one single canton. An additional consideration in the study’s
randomization scheme was to account for differences in activity
preferences between urban and rural areas in Switzerland.
30
The
blocks were therefore matched for population density. Next, can-
ton pairs within each block were randomized to the incentive con-
ditions using the toss of a coin, and the remaining canton was
allocated to the control group.
Because the incentive structure changed for the control group
in the fourth month, all outcomes and analyses refer to the first 3
months of the pilot program. Primary outcome was the participa-
tion rate in the three different groups. Insureds were considered
as participants if they provided consent to participate and shared
their step counts at least once during the first 3 months of the
study. Participants’non-usage attrition,
31
daily step counts, and
the proportion of participant days with >10,000 steps during the
first 3 months of the program were analyzed as secondary out-
comes. A participant was coded as non-usage attrition observed,
when she or he stopped synchronizing step counts with the insur-
ance company.
Statistical Analysis
The approach by Gao and colleagues
32
for non-aggregate clus-
ter RCTs with binary outcomes was used to determine the
minimum number of insureds to invite to the pilot program.
Accordingly, a sample size of N=15,822 invited insureds is
necessary to detect a 5% difference in participation rates
between control and incentive groups assuming an a-level of
0.05, a power of 0.80, an intra-cluster correlation of r=0.01
33
and a mean cluster size of 879 (SD=1,326; based on data from
the health insurance company).
Linear mixed models and generalized linear mixed models
34
were fitted to the data to analyze group differences of primary
and secondary outcomes. The model of participation rate
included a fixed effect for incentive condition and a random
intercept for canton. Models of step counts and participant
days with >10,000 steps included fixed effects for time, self-
reported physical activity measures at baseline, covariates of
physical activity, incentive condition, the incentive condition
by time interaction, and a random intercept for participants.
The time variable was mean-centered before entering the
model. The non-usage attrition model included fixed effects
for incentive condition, age, sex, and nationality and a ran-
dom intercept for canton. In addition, all models of secondary
outcomes were adjusted for group differences at baseline. To
be able to adjust models for covariates and group differences
at baseline, only participants who completed the baseline sur-
vey were included in the analyses of secondary outcomes. Sev-
eral sensitivity analyses were performed to assess the
robustness of the results. Differences in participation rates
betweengroupswerefurtheradjustedusingfixed effects for
age, sex, and nationality of participants and cantonal
Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54 e47
February 2019
population density. For the secondary outcomes of steps and
participant days with >10,000 steps, a nested random effect
for canton was added to the model to account for potential
clustering on the canton level. Additionally, sensitivity analy-
ses with regard to missing data were performed for all out-
comes (Appendix Tables 1−4, available online).
The reported results are pooled over separate analyses of 20
imputed data sets created using 50 imputation iterations for
each data set. Data were imputed in wide format using all varia-
bles in the statistical models plus additional auxiliary variables
to set up the imputation models (Appendix Table 6, available
online). Subsequently, convergence of the imputation algorithm
was examined by visual inspection of plots of mean and SD of
imputations across all iterations. Based on data from a Swiss
representative sample,
35
daily step counts >25,000 or <1,000
steps were set to missing before starting the imputation. Data
were analyzed in R, version 3.4.2, using the lme4 package
36
for
fitting (generalized) linear mixed models and the mice pack-
age
37
for multiple imputation.
Because of an error in the randomization process, six of all
26 Swiss cantons were non-randomly allocated to the three
study groups. These cantons contained 831 (3.09%) of invited
insureds and were excluded from all analyses. The study pro-
tocol further specified self-reported health status assessed after
3 months for the control group and after 6 months for both
incentive groups as a further secondary outcome. This com-
parison is not reported because the time difference in meas-
urements does not allow valid group comparisons. Lastly,
participant’s service billings with the insurance company could
not be included in the analyses, because these data were not
available at the time of writing.
RESULTS
In total, N=26,091 insureds (mean age 45.48 years,
SD=14.97, 38.52% female) met the predefined
eligibility criteria and were randomized. Of those,
1,338 (5.13%) participated in the pilot program. Addi-
tionally, 209 family members of invited insureds par-
ticipated, bringing the total number of participants to
n=1,547 (Figure 1). The proportion of women was
higher among participants than among all invited
insureds (47.83% vs 38.52%), indicating that women
were more likely to participate. Participants were on
average aged 42.65 (SD=13.03) years and mostly Swiss
(90.20%). Based on baseline survey data (n=1,223),
43.58% had a university degree and only 3.84%
reported poor or fair health conditions. There were sta-
tistically significant and meaningful group differences
with regard to participants’residential environment,
self-reported health status, and sitting minutes per
week (Table 1). In the charity financial incentive
condition, the mean proportion of donated rewards
was 20.29% (SD=26.81%). However, 310 of 623 partici-
pants (49.76%) chose not to donate any proportion of
their reward to charity.
Among invited insureds, 5.94% participated in the
personal financial incentive group and 4.98% in the
charity financial incentive group compared with 3.23%
in the control group. Differences between incentive
groups and control group were statistically significant
(personal financial incentives: OR=1.96, 95% CI=1.55,
2.49, p<0.001; charity financial incentives: OR=1.59,
95% CI=1.25, 2.01, p<0.001). Contrast analysis revealed
that participation rates also differed significantly
between insureds in the personal financial incentive and
the charity financial incentive group (OR=1.24, 95%
CI=1.06, 1.44, p=0.006).
Figure 1. Flow diagram.
e48 Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54
www.ajpmonline.org
Table 1. Baseline Characteristics of Invited Insurees and Participants
Characteristics p-value
a
SMD
b
PFI CFI CG
PFI
versus CG
CFI
versus CG
PFI versus
CFI
Invited insurees (N=26,091), n11,094 11,133 3,864
Canton level
Number of cantons 8 8 4
Cluster size 1,353.38 (2,080.96) 1,362.75 (1,439.82) 934.00 (1,263.82) 0.907 0.25 0.32 0.01
Population density
(residents/km
2
)
c
619.32 (244.93) 505.92 (1,721.03) 270.15 (826.41) 0.732 0.65 0.19 0.11
Individual level
Age, years 44.81 (14.73) 45.87 (15.12) 46.30 (15.15) <0.001 0.10 0.03 0.07
Female 4,441 (40.03) 4,261 (38.27) 1,348 (34.89) <0.001 0.11 0.07 0.04
Nationality <0.001 0.16 0.07 0.13
Swiss 9,491 (85.55) 9,973 (89.58) 3,486 (90.22)
German 582 (5.25) 384 (3.45) 111 (2.87)
Other 654 (5.90) 476 (4.28) 191 (4.94)
NA 367 (3.31) 300 (2.69) 76 (1.97)
Participants (N=1,547),
d
n784 623 140
Age, years 42.71 (12.99) 42.14 (12.95) 44.63 (12.53) 0.126 0.15 0.18 0.04
Female 382 (48.72) 299 (47.99) 59 (42.14) 0.424 0.13 0.12 0.01
Educational attainment 0.473 0.25 0.19 0.09
Secondary school 12 (1.99) 8 (1.59) 2 (1.72)
Vocational school 181 (30.02) 163 (32.34) 35 (30.17)
High school 92 (15.26) 85 (16.87) 27 (23.28)
University 278 (46.10) 213 (42.26) 42 (36.21)
NA 40 (6.63) 35 (6.94) 10 (8.62)
Residential environment <0.001 0.29 0.15 0.32
Town 85 (14.10) 42 (8.33) 11 (9.48)
Outskirts of town 173 (28.69) 102 (20.24) 24 (20.69)
Village 253 (41.96) 260 (51.59) 64 (55.17)
Countryside 73 (12.11) 88 (17.46) 14 (12.07)
NA 19 (3.15) 12 (2.38) 3 (2.59)
Monthly net income 0.343 0.26 0.24 0.17
≤CHF 2,500 27 (4.48) 30 (5.95) 3 (2.59)
CHF 2,501−5,000 83 (13.76) 80 (15.87) 18 (15.52)
CHF 5,001−7,500 190 (31.51) 147 (29.17) 33 (28.45)
CHF 7,501−10,000 101 (16.75) 78 (15.48) 26 (22.41)
>CHF 10,000 73 (12.11) 46 (9.13) 8 (6.90)
No answer 110 (18.24) 111 (22.02) 25 (21.55)
NA 19 (3.15) 12 (2.38) 3 (2.59)
Health status 0.032 0.23 0.21 0.18
Poor 2 (0.33) 1 (0.20) 0 (0.00)
Fair 15 (2.49) 20 (3.97) 9 (7.76)
Good 230 (38.14) 225 (44.64) 43 (37.07)
Very good 266 (44.11) 203 (40.28) 48 (41.38)
Excellent 68 (11.28) 40 (7.94) 12 (10.34)
NA 22 (3.65) 15 (2.98) 4 (3.45)
Activity tracker brand 0.520 0.13 0.17 0.04
Fitbit 511 (84.74) 432 (85.71) 94 (81.03)
Garmin 67 (11.11) 55 (10.91) 14 (12.07)
Jawbone 25 (4.15) 17 (3.37) 8 (6.90)
(continued on next page)
Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54 e49
February 2019
On average, participants walked 10,709 (SD=4,555)
steps per day and reached the 10,000 steps goal on
54.24% of all days during the first 3 months of the pilot
program. To investigate group differences, Table 2 and
Figure 2 compare steps per day and the probability of
walking >10,000 steps per day between the groups at the
beginning, in the middle, and at the end of the first 3
months of the pilot program. Although participants in
the charity financial incentive group consistently accu-
mulated a higher number of steps than participants in
the personal financial incentive and in the control group,
these step count differences were not statistically signifi-
cant at any time point. The difference between the incen-
tive groups and the control group diminished over time
(Figure 2), but this trend was also not statistically
significant (change of the charity financial incentive
effect over time: −3 steps/day, 95% CI= −6, 1, p=0.16;
change of the personal financial incentive effect over
time: −3 steps/day, 95% CI= −6, 1, p=0.15).
At the beginning of the program, participants
receiving charity financial incentives had a 12% higher
chance (OR=1.68, 95% CI=1.23, 2.30, p=0.004) of
walking >10,000 steps per day compared with the
control group. This difference diminished significantly
over time (change of the charity financial incentive
effect over time: −0.003, 95% CI= −0.01, −0.001,
p=0.003) and was no longer significant 3 months after
the start of the program. After adjusting p-values for
multiple comparisons across time, the probability of
walking >10,000 steps per day did not differ
Table 1. Baseline Characteristics of Invited Insurees and Participants (continued)
Characteristics p-value
a
SMD
b
PFI CFI CG
PFI
versus CG
CFI
versus CG
PFI versus
CFI
Bought an activity
tracker to participate
320 (53.07) 303 (60.12) 62 (53.45) 0.048 0.01 0 .15 0.14
Sitting (minutes/week)
e
2,435.80 (1,378.83) 2,263.40 (1,303.22) 2,165.52 (1,247.31) 0.039 0.21 0.08 0.13
Moderate activities and
walking (MET-minutes/
week), median (IQR)
e
2,628.00
(3,306)
2,745.75
(3,327)
2,079.00
(3,720)
0.243 0.09 0.04 0.05
Note: Boldface indicates statistical significance (p<0.05). Table displays M (SD) for continuous and n(percentage) for categorical variables unless
stated otherwise.
a
Based on one-way ANOVA for normal, Kruskal-Wallis test for non-normal, and x
2
test of independence for categorical variables.
b
Values greater 0.20 defined as small effect size.
38
Non-normal variables were log-transformed before calculating SMD.
c
Reported values are weighted by cluster size.
d
Data on age and sex of par ticipants were provided by the insurance company (n=1,547), data on other variables are based on baseline survey data
(n=1,223).
e
Assessed using the International Physical Activity Questionnaire.
39
CFI, charity financial incentives; CHF, Swiss Francs; CG, control group; IQR, interquartile range; NA, not available; PFI, personal financial incentives;
SMD, absolute standardized difference.
Figure 2. Steps per day and probability of walking more than 10,000 daily steps over time by incentive condition.
e50 Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54
www.ajpmonline.org
significantly between the personal financial incentive
group and the control group at any time point. Like-
wise, the probability of walking >10,000 steps per day
did not differ significantly between the personal and
the charity financial incentive group.
Lastly, non-usage attrition during the first 3 months
was highest in the charity financial incentive group
(10.91%), followed by the personal financial incentive
group (9.62%) and the control group (9.48%). In the
adjusted model, differences between the incentive
groups and the control group were not statistically sig-
nificant (OR=1.05, 95% CI=0.41, 2.69, p=0.92, for the
personal financial incentive group; OR=1.21, 95%
CI=0.48, 3.08, p=0.69, for the charity financial incen-
tive group).
DISCUSSION
This three-arm cluster RCT investigated whether small
personal and charity financial incentives can promote
participation in a physical activity promotion program
utilizing activity trackers. Secondary analyses examined
group differences with regard to physical activity and
attrition of participants. This study is among the first to
investigate the potential of small personal financial and
charity financial incentives for the promotion of physical
activity. A strength of this study is that its pragmatic set-
ting makes it possible to demonstrate real-world effec-
tiveness and to inform the design of large-scale physical
activity promotion programs.
Small monthly personal financial and charity financial
incentives significantly increased participation, with per-
sonal financial incentives being more effective than char-
ity financial incentives. However, the effects of these
incentives on behavior may be limited. The data of this
study suggest that charity financial incentives increase
the likelihood of walking 10,000 steps per day only in
the short term, with effects dissipating after 3 months.
Personal financial incentives did not lead to significant
changes in physical activity. In addition, there was no
evidence that personal financial and charity financial
incentives prevent attrition of participants. Conse-
quently, this study supports utilizing small personal
financial or charity financial incentives to promote par-
ticipation but not behavior change in large-scale physical
activity promotion programs. These findings add to the
understanding of the effects of financial and charity
incentives for the promotion of physical activity.
Regarding physical activity, these results deviate from
previous studies that typically find a stable effect of
incentives as long as they are in place.
17,18
This may indi-
cate that an incentive value of CHF 10 per month could
have been too low to sustain motivation after an initial
Table 2. Group Comparisons Regarding Steps per Day and Probability of Walking More Than 10,000 Steps
Outcome Day 1 p-value
a
Day 46 p-value
a
Day 92 p-value
a
Difference in steps (95% CI)
PFI versus CG 206.49 (−322.13, 735.12) 1.00 91.99 (−415.57, 599.55) 1.00 −22.51 (−557.45, 512.43) 0.934
CFI versus CG 507.24 (−29.48, 1043.96) 0.192 383.81 (−128.35, 895.98) 0.284 260.39 (−280.95, 801.72) 0.346
PFI versus CFI −300.75 (−620.39, 18.90) 0.130 −291.82 (−597.34, 13.70) 0.184 −282.90 (−603.73, 37.94) 0.084
Walking >10,000 steps per
day, OR (95% CI)
PFI versus CG 1.41 (1.03, 1.92) 0.093 1.20 (0.89, 1.61) 0.477 1.02 (0.74, 1.39) 0.919
CFI versus CG 1.68 (1.23, 2.30) 0.004 1.45 (1.07, 1.96) 0.031 1.25 (0.91, 1.71) 0.168
PFI versus CFI 0.84 (0.69, 1.01) 0.063 0.83 (0.69, 0.99) 0.069 0.81 (0.68, 0.98) 0.092
Note: Boldface indicates statistical significance (*p<0.05).
a
p-values are adjusted for multiple testing across time points using the Holm-Bonferroni method. 95% CIs are not adjusted, because Holm-Bonferroni-adjusted CIs are non-informative.
40
CFI, charity financial incentives; CG, control group; PFI, personal financial incentives.
February 2019
Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54 e51
phase of excitement at the beginning of the program had
passed. To the best of the authors’knowledge, the lowest
incentive value that has demonstrated effects on physical
activity in randomized trials is around $1 per day,
15,40,41
which is three times the incentive value of the present
study. In addition, previous studies often used daily
15,41
or weekly
17,18
incentive schedules instead of the monthly
incentives in the present study. Research in the field
of behavioral economics reveals that people tend to
place more weight on immediate and discount future
rewards,
42
a phenomenon known as present bias. The
monthly reward schedule in the present study may have
caused participants to further attenuate the subjective
value of the already small incentives. Both factors, the
small incentive value and the monthly incentive sched-
ule, may explain the study’s limited results on the ability
of incentives to change behavior. Future research could
investigate whether small incentives with immediate or
dynamic incentive schedules lead to sustained behavior
changes or what minimum incentive amount is neces-
sary to produce sustained effects.
Designing physical activity promotion programs that
rely on activity trackers may entail the risk of a favorable
selection effect because individuals who already own a
tracker face less barriers for participation and may be
more likely to meet the recommended activity levels.
Indeed, in the present study, the requirement to pur-
chase an activity tracker was named the main barrier for
participation (selected by 41.05 % of insureds who gave
reasons for refusing to participate; Appendix Table 7,
available online) and 44% of participants already owned
an activity tracker before being invited to the program.
However, although owners of activity trackers were sig-
nificantly younger, better educated, and reported better
health than non-owners (Appendix Table 6, available
online), they accumulated on average 1,004 fewer steps
(95% CI=708, 1,301) than those who purchased an activ-
ity tracker for participation (Appendix Figure 1, avail-
able online). Thus, although the investment in an
activity tracker might have increased commitment to the
program, it remains a big barrier to participation. Rely-
ing on smartphones for tracking physical activity or fur-
ther reimbursing the financial investment (e.g.,
conditional on meeting activity goals after 1 year) could
reduce this barrier and help to better reach people in
need of the program.
Limitations
The present study has a few limitations to consider. First,
because data were collected from a single insurer, results
may not generalize to other programs with other insur-
ers or populations. Second, results regarding secondary
outcomes have to be interpreted with caution because
voluntary registration of participants may have resulted
in selection bias. Indeed, program participants were bet-
ter educated and earned higher wages compared with
the general Swiss population.
43,44
Incentive effects may
differ in populations with lower education or income.
45
Further, nonsignificant results may not be taken as evi-
dence for the absence of meaningful effects because the
study was not powered for secondary outcomes. In addi-
tion, because participants could respond to the baseline
survey after the program had started, it is possible that
some self-reported baseline information was affected by
the incentive strategies. This could potentially lead to
overly conservative estimates of group differences in
adjusted models. Third, participants in the charity incen-
tive group had the opportunity to select the proportion
of their reward that they wish to donate to charity. Thus,
results of this study are not directly comparable to those
of previous studies, which did not offer this option.
Lastly, reported absolute step counts and proportions of
days with more than 10,000 steps are likely to be conser-
vative because consumer-grade activity trackers tend to
underestimate true step counts.
46
CONCLUSIONS
In this large cluster-RCT from Switzerland, small monthly
personal financial and charity financial incentives
increased participation in a physical activity promotion
program utilizing activity trackers. Organizations offering
such programs can thus encourage participation even
when using relatively small incentives. However, the
short-term effects of these incentives on physical activity
and attrition limits their utility in the context of health pro-
motion programs. Incentives may need to be modified in
order to prevent attrition and promote behavior change
over a longer period of time.
ACKNOWLEDGMENTS
We thank the involved employees of CSS insurance, Switzer-
land, for their contribution to the implementation of the study.
We thank Grace Xiao for proofreading the manuscript.
Research reported in this publication was par tly funded by
CSS insurance, Switzerland. We developed the incentive
schemes and the randomization strategy in collaboration with
CSS insurance. CSS insurance supported recruitment and data
collection but had no role in other aspects of the study design,
data analysis and interpretation, or in reviewing and approving
the manuscript for publication.
The IRB of the University of St. Gallen, Switzerland, approved
the study (reference number: HSG-EC-2015-04-22-A).
Tobias Kowatsch and Elgar Fleisch developed the conception
and design of the study. Jan-Niklas Kramer developed the
methodology, performed the data analyses, and wrote the man-
uscript. Peter Tinschert, Tobias Kowatsch, and Urte Scholz
www.ajpmonline.org
e52 Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54
provided critical reviews on earlier drafts of the manuscript and
aided in data analysis and interpretation. All authors approved
the final version of the manuscript before submission.
Elgar Fleisch co-chairs the Center For Digital Health Interven-
tions (CDHI), a joint initiative of the Department of Manage-
ment, Technology, and Economics at ETH Zurich and the
Institute of Technology Management at the University of St.
Gallen, which is funded in part by the Swiss health insurer CSS.
Tobias Kowatsch is the scientific director of the CDHI and Jan-
Niklas Kramer and Peter Tinschert are doctoral researchers at
the CDHI. Elgar Fleisch and Tobias Kowatsch are also co-found-
ers of Pathmate Technologies, a university spin-off company
that creates and delivers digital clinical pathways. No other
financial disclosures were reported by the authors of this paper.
SUPPLEMENTAL MATERIAL
Supplemental materials associated with this article can be
found in the online version at https://doi.org/10.1016/j.
amepre.2018.09.018.
REFERENCES
1. Lear SA, Hu W, Rangarajan S, et al. The effect of physical activity on
mortality and cardiovascular disease in 130 000 people from 17 high-
income, middle-income, and low-income countries: the PURE study.
Lancet. 2017;390(10113):2643–2654. https://doi.org/10.1016/S0140-
6736(17)31634-3.
2. Lee I-M, Shiroma EJ, Lobelo F, et al. Effect of physical inactivity on
major non-communicable diseases worldwide: an analysis of burden
of disease and life expectancy. Lancet. 2012;380(9838):219–229.
https://doi.org/10.1016/S0140-6736(12)61031-9.
3. Kyu HH, Bachman VF, Alexander LT, et al. Physical activity and risk
of breast cancer, colon cancer, diabetes, ischemic heart disease, and
ischemic stroke events: systematic review and dose-response meta-
analysis for the global burden of disease study 2013. BMJ. 2016;354:
i3857. https://doi.org/10.1136/bmj.i3857.
4. Sallis JF, Bull F, Guthold R, et al. Progress in physical activity over the
Olympic quadrennium. Lancet. 2016;388(10051):1325–1336. https://
doi.org/10.1016/S0140-6736(16)30581-5.
5. Steinhubl SR, Muse ED, Topol EJ. The emerging field of mobile health.
Sci Transl Med. 2015;7(283):283rv3. https://doi.org/10.1126/sci-
translmed.aaa3487.
6. HERO. Wearables in wellness—employer use of wearable tracking
devices in wellness programs. http://hero-health.org/wp-content/
uploads/2015/06/HERO-Wearables-in-Wellness-Report-FINAL1.pdf.
Published June 2015. Accessed March 22, 2017.
7. Patel MS, Foschini L, Kurtzman GW, et al. Using wearable devices and
smartphones to track physical activity: initial activation, sustained use,
and step counts across sociodemographic characteristics in a national
sample. Ann Intern Med. 2017;167(10):755–757. https://doi.org/
10.7326/M17-1495.
8. Nyce S. Boosting wellness participation without breaking the bank.
www.towerswatson.com/en/Insights/Newsletters/Americas/insider/
2010/boosting-wellness-participation-without-breaking-the-bank.
Published July 2010. Accessed March 2, 2018.
9. Batorsky B, Taylor E, Huang C, Liu H, Mattke S. Understanding the
relationship between incentive design and participation in U.S. work-
place wellness programs. Am J Health Promot. 2016;30(3):198–203.
https://doi.org/10.4278/ajhp.150210-QUAN-718.
10. Volpp KG, Troxel AB, Pauly MV, et al. A randomized, controlled trial
of financial incentives for smoking cessation. N Engl J Med. 2009;360
(7):699–709. https://doi.org/10.1056/NEJMsa0806819.
11. Strohacker K, Galarraga O, Williams DM. The impact of incentives on
exercise behavior: a systematic review of randomized controlled trials.
Ann Behav Med. 2014;48(1):92–99. https://doi.org/10.1007/s12160-
013-9577-4.
12. Mitchell MS, Goodman JM, Alter DA, et al. Financial incentives
for exercise adherence in adults: systematic review and meta-anal-
ysis. Am J Prev Med. 2013;45(5):658–667. https://doi.org/10.1016/j.
amepre.2013.06.017.
13. Barte JC, Wendel-Vos GW. A systematic review of financial incentives
for physical activity: the effects on physical activity and related out-
comes. Behav Med. 2017;43(2):79–90. https://doi.org/10.1080/
08964289.2015.1074880.
14. Patel MS, Asch DA, Rosin R, et al. Individual versus team-based finan-
cial incentives to increase physical activity: a randomized, controlled
trial. J Gen Intern Med. 2016;31(7):746–754. https://doi.org/10.1007/
s11606-016-3627-0.
15. Patel MS, Asch DA, Roy Rosin M, et al. Framing financial incentives
to increase physical activity among overweight and obese adults: a ran-
domized, controlled trial. Ann Intern Med. 2016;164(6):385–394.
https://doi.org/10.7326/M15-1635.
16. Patel MS, Volpp KG, Rosin R, et al. A randomized, controlled trial of
lottery-based financial incentives to increase physical activity among
overweight and obese adults. Am J Health Promot. 2018;32(7):
1568–1575. https://doi.org/10.1177/0890117118758932.
17. Harkins KA, Kullgren JT, Bellamy SL, Karlawish J, Glanz K. A trial of
financial and social incentives to increase older adults’walking. Am J
Prev Med. 2017;52(5):e123–e130. https://doi.org/10.1016/j.amepre.
2016.11.011.
18. Finkelstein EA, Haaland BA, Bilger M, et al. Effectiveness of activity
trackers with and without incentives to increase physical activity
(TRIPPA): a randomised controlled trial. Lancet Diabetes Endocrinol.
2016;4(12):983–995. https://doi.org/10.1016/S2213-8587(16)30284-4.
19. Banerjee AV, Duflo E, Glennerster R, Kothari D. Improving immuni-
sation coverage in rural India: clustered randomised controlled evalua-
tion of immunisation campaigns with and without incentives. BMJ.
2010;340:c2220. https://doi.org/10.1136/bmj.c2220.
20. Lagarde M, Haines A, Palmer N. Conditional cash transfers for
improving uptake of health interventions in low- and middle-income
countries: a systematic review. JAMA. 2007;298(16):1900–1910.
https://doi.org/10.1001/jama.298.16.1900.
21. Malotte CK, Rhodes F, Mais KE. Tuberculosis screening and com-
pliance with return for skin test reading among active drug users.
Am J Public Health. 1998;88(5):792–796. https://doi.org/10.2105/
AJPH.88.5.792.
22. Moll J, Krueger F, Zahn R, Pardini M, de Oliveira-Souza R, Grafman J.
Human fronto-mesolimbic networks guide decisions about charitable
donation. Proc Natl Acad Sci U S A. 2006;103(42):15623–15628.
https://doi.org/10.1073/pnas.0604475103.
23. Dunn EW, Aknin LB, Norton MI. Spending money on others pro-
motes happiness. Science. 2008;319(5870):1687–1688. https://doi.org/
10.1126/science.1150952.
24. Andreoni J. Impure altruism and donations to public goods: a theory
of warm-glow giving. Econ J (London). 1990;100(401):464–477.
https://doi.org/10.2307/2234133.
25. Varadarajan PR, Menon A. Cause-related marketing: a coalignment of
marketing strategy and corporate philanthropy. J Mark. 1988;52
(3):58–74. https://doi.org/10.2307/1251450.
26. Williams DM, Lee HH, Connell L, et al. Small sustainable mone-
tary incentives versus charitable donations to promote exercise:
rationale, design, and baseline data from a randomized pilot
study. Contemp Clin Trials. 2018;66:80–85. https://doi.org/
10.1016/j.cct.2018.01.005.
27. Kowatsch T, Kramer J-N, Kehr F, Wahle F, Elser N, Fleisch E. Effects
of charitable versus monetary incentives on the acceptance of and
February 2019
Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54 e53
adherence to a pedometer-based health intervention: study protocol
and baseline characteristics of a cluster-randomized controlled trial.
JMIR Res Protoc. 2016;5(3):e181. https://doi.org/10.2196/resprot.6089.
28. Eisler R, L€
uber A, Wie wichtig ist den schweizern eine spitalzusatz-
versicherung? 2016;1−30. www.comparis.ch/»/media/files/medien-
corner/studies/2006/krankenkassen/spitalzusatzversicherungen_stu-
die.pdf. Accessed September 27, 2018.
29. Tudor-Locke C, Craig CL, Brown WJ, et al. How many steps/day are
enough? For adults. Int J Behav Nutr Phys Act. 2011;8:1–17. https://
doi.org/10.1186/1479-5868-8-79.
30. Lamprecht M, Fischer A, Stamm H. Sport schweiz 2014: Sportaktivit€
at
und sportinteresse der schweizer bev€
olkerung. Observatorium Sport
und Bewegung Schweiz c/o Lamprecht & Stamm Sozialforschung und
Beratung AG; 2014.
31. Eysenbach G. The law of attrition. J Med Internet Res. 2005;7(1):e11.
https://doi.org/10.2196/jmir.7.1.e11.
32. Gao F, Earnest A, Matchar DB, Campbell MJ, Machin D. Sample size
calculations for the design of cluster randomized trials: a summary of
methodology. Contemp Clin Trials. 2015;42:41–50. https://doi.org/
10.1016/j.cct.2015.02.011.
33. Donner A, Klar N. Design and Analysis of Cluster Randomization Tri-
als in Health Research. London, UK: Arnold, 2000.
34. Raudenbush SW, Bryk AS. Hierarchical Linear Models: Applications
and Data Analysis Methods. Thousand Oaks, CA: Sage Publications,
2002.
35. Sequeira MM, Rickenbach M, Wietlisbach V, Tullen B, Schutz Y. Physi-
cal activity assessment using a pedometer and its comparison with a
questionnaire in a large population survey. Am J Epidemiol. 1995;142
(9):989–999. https://doi.org/10.1093/oxfordjournals.aje.a117748.
36. Bates D, M€
achler M, Bolker B, Walker S. Fitting linear mixed-effects
models using lme4. arXiv preprint arXiv:14065823. 2014.
37. van Buuren S, Groothuis-Oudshoorn K. Mice: multivariate imputation
by chained equations in R. J Stat Softw. 2011;45(3). https://doi.org/
10.18637/jss.v045.i03.
38. Yang D, Dalton JE. A unified approach to measuring the effect size
between two groups using SASÒ.SAS Global Forum. 2012;2012:1–6.
39. Craig CL, Marshall AL, Sjo
̈stro
̈mM,etal.Internationalphysical
activity questionnaire: 12-country reliability and validity. Med
Sci Sports Exerc. 2003;35:1381–1395. https://doi.org/10.1249/01.
MSS.0000078924.61453.FB.
40. Strohacker K, Gal
arraga O, Emerson J, Fricchione SR, Lohse M, Wil-
liams DM. Impact of small monetary incentives on exercise in univer-
sity students. Am J Health Behav. 2015;39(6):779–786. https://doi.org/
10.5993/AJHB.39.6.5.
41. Shin DW, Yun JM, Shin JH, et al. Enhancing physical activity and
reducing obesity through smartcare and financial incentives: a pilot
randomized trial. Obesity (Silver Spring). 2017;25(2):302–310. https://
doi.org/10.1002/oby.21731.
42. Volpp KG, Asch DA, Galvin R, Loewenstein G. Redesigning employee
health incentives—lessons from behavioral economics. N Engl J Med.
2011;365(5):388–390. https://doi.org/10.1056/NEJMp1105966.
43. Bundesamt f€
ur Statistik. St€
andige wohnbev€
olkerung ab 15 jahren nach
h€
ochster abgeschlossener ausbildung, 2016. www.bfs.admin.ch/bfs/
de/home/statistiken/kataloge-datenbanken/tabellen.assetde-
tail.4242913.html. Published January 2018. Accessed March 13, 2018.
44. Bundesamt f€
ur Statistik. H€
aufigkeitsverteilung (monatlicher nettolohn),
voll- und teilzeitbesch€
aftigte nach lohnh€
ohenklassen - privater und
€
offentlicher sektor zusammen - schweiz. www.bfs.admin.ch/bfs/de/
home/statistiken/kataloge-datenbanken/tabellen.assetdetail.327851.html.
Published April 2016. Accessed March 13, 2018.
45. Mantzari E, Vogt F, Shemilt I, Wei Y, Higgins JP, Marteau TM. Per-
sonal financial incentives for changing habitual health-related behav-
iors: a systematic review and meta-analysis. Prev Med. 2015;75:75–85.
https://doi.org/10.1016/j.ypmed.2015.03.001.
46. Toth LP, Park S, Springer CM, Feyerabend MD, Steeves JA, Bassett
DR. Video-recorded validation of wearable step counters under free-
living conditions. Med Sci Sports Exerc. 2018;50(6):1315–1322. https://
doi.org/10.1249/MSS.0000000000001569.
www.ajpmonline.org
e54 Kramer et al / Am J Prev Med 2019;56(2):e45
−
e54