RESEARCH ARTICLE Open Access
Simple and a ionale-p o iding SMS
eminde s o p omo e accele ome e use: a
wi hin- ial andomised ial compa ing
pe suasi e messages
Ma i T. J. Heino
1,2*
, Keegan Kni le
1
, A i Haukkala
1
, Tommi Vasanka i
3
and Nelli Hankonen
1,2
Abs ac
Backg ound: Li e a u e on pe suasion sugges s compliance inc eases when eques s a e accompanied wi h a
eason (i.e. he “because-heu is ic”). The eliabili y o ou comes in physical ac i i y esea ch is dependen on
su icien accele ome e wea - ime. This s udy es ed whe he SMS eminde s—especially hose ha p o ided a
a ionale—a e associa ed wi h inc eased accele ome e wea - ime.
Me hods: We conduc ed a wi hin- ial pa ially andomised con olled ial du ing baseline da a collec ion in a
school-based physical ac i i y in e en ion ial. O 375 pa icipan s (mean age = 18.1), 280 (75%) op ed o ecei e
daily SMS eminde s o wea hei accele ome e s. These 280 pa icipan s we e hen andomised o ecei e ei he
succinc eminde s o eminde s including a a ionale. Da a was analyzed ac oss g oups using bo h equen is and
Bayesian me hods.
Resul s: No di e ences in o al accele ome e wea minu es we e de ec ed be ween he succinc eminde g oup
(Mdn = 4909, IQR = 3429–5857) and he a ionale g oup (Mdn = 4808, IQR = 3571–5743); W = 8860, p= 0.65, CI95 = −280.
90–447.20. Simila ly, we ound no di e ences in wea ime be ween pa icipan s ecei ing SMS eminde s (Mdn = 4859,
IQR = 3527–5808) and hose no ecei ing hem (Mdn = 5067, IQR = 3201–5885); W = 10,642.5, p= 0.77, CI95 = −424.20–
305.30. Bayesian ANOVA a o ed a model o equal wea ime means, o e one o unequal means, by a Bayes Fac o o 12.
05. Accumula ed days o alid accele ome e wea da a did no di e ei he . Equi alence es ing indica ed ejec ion o
e ec s mo e ex eme han a Cohen’s d (s anda dised mean di e ence) o ±~0.3.
Conclusions: This s udy cas s doub on he e ec i eness o using he because-heu is ic ia SMS messaging, o p omo e
accele ome e wea ime among you h. The because-heu is ic migh be limi ed o ace- o- ace communica ion and
si ua ions whe e no in en ion o o commi men o he beha io has ye been made. O he explana ions o null e ec s
include non- eading o messages, and eminde messages unde mining he sel - eminding s a egies which would
occu na u ally in he absence o eminde s.
T ial egis a ion: DRKS DRKS00007721. Regis e ed 14.04.2015. Re ospec i ely egis e ed.
Keywo ds: Accele ome y, In e en ion, Tex messaging, SMS, Pe suasion, Adhe ence, Beha iou change, Adolescen s,
School-based esea ch, Pa ially andomised ial
* Co espondence: [email p o ec ed]
1
Facul y o Social Sciences, Uni e si y o Helsinki, Helsinki, Finland
2
Facul y o Social Sciences, Uni e si y o Tampe e, Tampe e, Finland
Full lis o au ho in o ma ion is a ailable a he end o he a icle
© The Au ho (s). 2018 Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0
In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and
ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o
he C ea i e Commons license, and indica e i changes we e made. The C ea i e Commons Public Domain Dedica ion wai e
(h p://c ea i ecommons.o g/publicdomain/ze o/1.0/) applies o he da a made a ailable in his a icle, unless o he wise s a ed.
Heino e al. BMC Public Heal h (2018) 18:1352
h ps://doi.o g/10.1186/s12889-018-6121-2
Backg ound
Compliance wi h accele ome e wea ins uc ions
Reliable and alid assessmen is necessa y when e alua ing
whe he public heal h policies o in e en ions change
physical ac i i y (PA) le els in he a ge g oup. Li le con-
sensus exis s abou wha o measu e, when, wi h wha and
o how long in PA esea ch [1,2]. While an inabili y o
indi iduals o accu a ely emembe hei pas PA and so-
cial desi abili y a e clea p oblems wi h sel - epo ed PA
measu es [3], objec i e measu emen s o PA (e.g. pedome-
e s and accele ome e s) ha e issues oo. Zhuang e al. [4]
ound ha missing accele ome y da a was mo e common
in 15- o 17-yea -olds han among younge pa icipan s,
especially du ing weekends (Sundays in pa icula ), wi h
missing da a occu ing inc easingly om he i s eco d-
ing day o he las . This exempli ies a key issue in meas-
u emen : he p opo ion o an indi idual’s day o week
cap u ed by he measu e. An ex eme example would be
an indi idual, who only wea s he measu emen de ice
when unde aking PA. Thus, some guidelines sugges ha
a pe son should wea an accele ome e o a minimum o
10 h daily o a leas 4 days in a 7-day measu emen
pe iod in o de o ob ain an accu a e eading o PA [1,2].
Pa icipan s’compliance wi h ins uc ions on wea ing he
accele ome e is clea ly e y impo an in ob aining accu -
a e PA measu emen s [5].
Resea ch on enhancing accele ome e ins uc ion
compliance a es is a e [2,6], pa icula ly among olde
adolescen s. One s a egy has been mone a y incen i es
con ingen on p ope wea - ime [7]. Sallis e al. [8] used
an al e na i e s a egy, asking pa icipan s o e-wea he
accele ome e i hey had no wo n i o a leas 5 alid
days (> 10 alid hou s o da a) o a minimum o 66 alid
hou s ac oss 7 days.
Ba ak e al. [9] sugges ha new oppo uni ies o p o-
mo e compliance—such as ex messaging (SMS; Sho
Messaging Se ice)—may be mo e eliable and e ec i e
han adi ional me hods, such as w i en o e bal wea
ins uc ions by he in es iga o . Zhuang e al. [4], oo, ec-
ommend SMS eminde s. To age e al. [10]usedSMS
eminde s o inc ease compliance bu did no epo e -
ec s o accep abili y. In a sel -selec ed I ish sample o ad-
olescen s [11], daily SMS eminde s we e associa ed wi h
pu ing on he accele ome e in he mo ning, bu no in
inc eased o e all compliance (de ined as alid days o da a
o minu es o non-wea ). The s udy did no epo le els
o wea o e ec s o he eminde s. The disc epancy be-
ween emembe ing o pu on he de ice and ac ually
wea ing i o a su icien amoun o ime indica es ha
hese may be sepa a e beha io s.
Compliance and he ‘because-heu is ic’
Since he classic “Xe ox machine s udy”by Lange ,
Blank and Chanowi z [12], p o iding easons o
compliance has been discussed in he social in luence li -
e a u e. The s udy indica ed ha placebic o
pseudo- easons [13](“Excuse me, I ha e 5 pages. May I
use he xe ox machine, because I ha e o make copies?”;
93% compliance) could esul in simila compliance a es
as ac ual easons (“[…]because I’m in a ush?”; 94%
compliance) compa ed o he eques only condi ion
(“Excuse me, I ha e 5 pages. May I use he xe ox ma-
chine”; 60% compliance). P a kanis (2007), iden i ied “pla-
cebic easons”in his index o social in luence ac ics, bu
called o u he esea ch in o he subjec . Less ca e ul
a e Cialdini, Golds ein and Ma in [14], who ou he
“unique mo i a ional in luence o he wo d because”, bas-
ing hei claims on he impo ance o easoning in social
in luence. To his day, he xe ox machine s udy emains
ci ed in he p ess as an example o he powe o he wo d
‘because’[15–18].
A well-known p inciple o human beha io says ha
when we ask someone o do us a a o we will be
mo e success ul i we p o ide a eason. People simply
like o ha e easons o wha hey do. [19]
Following he e minology used by Key, Edlund,
Saga ing and Bize [20], he phenomenon o inc eased
compliance by p o iding easons is e e ed o as “ he
because-heu is ic.”Le us acco dingly de ine he naï e
because-heu is ic as “ easons inc ease compliance.”
In he Lange , Blank and Chanowi z s udy 1, his e ec
o easons inc easing compliance was only ound when
he con ede a e asked o ‘asmall a o ’( i e ins ead o
en pages, ansla ing o e ec sizes o d = 0.87 and d =
0.13, espec i ely) [12]. S ill, he esul s in gene al, as
well as hei implica ions ha e been ques ioned [21,22].
A s udy by Folkes sugges s, ha ins ead o he size o
he eques , he e ec is mode a ed by con ollabili y
[21]. Pooling Folkes’ eason condi ions esul s o an e -
ec size o d = −0.026, speaking agains he quo e abo e,
and poin ing ou ha he “powe o easons”e ec is
malleable, in he leas .
To ou knowledge, only one published di ec eplica-
ion o he Lange , Blank and Chanowi z s udy 1 exis s
[20]. The main e ec o he s udy eplica ed (d = 0.67 o
placebic o e no eason and d = 0.69 o eal o e no
eason condi ions), al hough o e 20% (34 ou o 163) o
he pa icipan s needed o be excluded o a ious ea-
sons. Lack o published eplica ion s udies, o cou se, is
no new in he ield o psychology [23].
In a concep ual eplica ion o he phenomenon, in
small eques condi ions, easons (ei he placebic o eal)
inc eased compliance by an equi alen o d = 0.43 (calcu-
la ed om Table 1 o [24]) when including hei add-
i ional pe suasion g oup and d = 0.22 when excluding i .
Ano he concep ual eplica ion [25] ound d = 0.15 o
Heino e al. BMC Public Heal h (2018) 18:1352 Page 2 o 16
eques s pe cei ed as small, and d = 0.21 o eques s
pe cei ed as la ge (as calcula ed om Figu e 3 o [25]).
These s udies seem o empe ea lie claims o he
powe o easons in inc easing compliance. In con as
o he naï e because-heu is ic, le us de ine he weak
because-heu is ic as “ easons inc ease compliance, bu
only i he pe cei ed a ou is small”.
This s udy will in es iga e he e ec s o he because
heu is ic on compliance wi h he physical ac i i y meas-
u emen p ocedu es in he con ex o baseline measu e-
men s o a la ge school-based in e en ion.
The Le ’s Mo e I clus e andomized ial
Inadequa e PA p edic s inc eased mo bidi y and mo ali y
in people o low socioeconomic s a us (SES) [26], wi h
SES di e ences in PA eme ging al eady in adolescence
[27]. Finnish oca ional school s uden s a e less physically
ac i e han hose in high school [28]. The Le ’sMo eI
in e en ion aimed o inc ease PA and dec ease seden a y
beha io s in olde adolescen s in oca ional schools.
The cu en s udy was conduc ed as a sub-s udy o he
clus e andomised e ec i eness e alua ion ial o he
Le ’s Mo e I in e en ion [29]. In a p eceding easibili y
s udy [30], pa icipan s’accele ome e wea imes we e
subop imal; 47% (18/38) o baseline pa icipan s eached
he cu o o 10 h pe day o a leas 4 days, 63% (17/27)
o he i s and 75% (9/12) o he second ollow-up. A
equen ly ci ed explana ion o no wea ing he accele -
ome e was o ge ing o pu on he de ice.
Aims and hypo heses
In his wi hin- ial s udy, we in es iga e SMS- eminde
s a egies o imp o e he du a ion o accele ome e wea
ime. The li e a u e ci ed p e iously lead us o hypo he-
sise ha eminde s would inc ease accele ome e wea
ime and ha ci ing easons would ampli y he e ec . In
addi ion o daily wea hou s, we a e in e es ed in he
numbe o days ou pa icipan s p o ide alid ac i i y
da a (i.e. days o ≥10 h o ac i i y da a). The a ge be-
ha io is hus wo old: 1) pu ing on he accele ome e
in he mo ning o as many days as possible, 2) wea ing
he accele ome e o as long as possible in he waking
hou s each day. In his s udy, wo main esea ch ques-
ions a e posi ed:
1. A e SMS- eminde s associa ed wi h g ea e accele -
ome e wea imes?
The cu en s udy in es iga ed his by compa ing
he compliance a es ac oss a) pa icipan s who op ed
o ecei e SMS eminde s o wea hei accele ome e ,
and b) pa icipan s who op ed no o ecei e he e-
minde s (non- andomised con ol g oup). I o ge ing
is an impo an eason o non-compliance, in he
absence o in e ening ac o s, eminde s should in-
c ease compliance.
S a is ical hypo hesis H
1
: Those who ecei e SMS e-
minde s will ha e highe accele ome e wea imes han
hose who do no .
2. Does o e ing easons o comply a ec
accele ome e wea ime?
I easons inc ease compliance, SMS eminde s con-
aining easons o wea an accele ome e should lead o
g ea e compliance.
S a is ical hypo hesis H
2
: Those who ecei e easons in
he SMS eminde s ha e mo e minu es o accele ome e
wea and mo e days o alid da a (≥10 h o ac i i y) han
hose who do no ecei e eminde s con aining a eason.
An addi ional esea ch ques ion, on whe he p o iding
easons o comply wi h accele ome e wea inc eases
ial e en ion, is omi ed he e. These null esul s a e e-
po ed in [31].
Me hods
The design o his s udy was a wi hin- ial, ou come
-assesso blinded, pa ially andomised con olled ial
(RCT). In addi ion o he andomised expe imen be ween
wo message ypes, quasi-expe imen al da a we e acqui ed
om a sel -selec ed op -ou a m (see Fig. 1). This s udy
was conduc ed du ing he baseline assessmen o he i s
wo ec ui men wa es (ou o six) o he Le ’sMo eI
clus e - andomised con olled ial [29]. This a icle is
based on unpublished wo k a ailable a h ps://os .io/
89mhu/. Addi ional in o ma ion on me hods and esul s,
in addi ion o all analysis code, can be ound in he supple-
men a y websi e a h ps://gi .io/ Nl8X (pe malink p o-
ided in [32]).
Pa icipan s and sampling p ocedu es
To be included in he s udy, he pa icipan s had o ul-
ill inclusion c i e ia o he Le ’s Mo e I s udy [29] and
had o ha e consen ed o he accele ome y measu e-
men s: all we e a leas 16 yea s old and we e oca ional
school s uden s. The eminde a ms consis ed o he
pa icipan s who op ed in o ecei e eminde s o accel-
e ome e wea .
Du ing baseline ec ui men o he i s wo ec ui men
wa es o he Le ’s Mo e I ial, s uden s in wo oca ional
schools we e app oached du ing class and in o med abou
hei school’s s udy pa icipa ion in he s udy. A e he in-
i a ion o pa icipa e in he main ial and collec ion o
signed in o med consen o ms, hose who consen ed
we e gi en an online ques ionnai e o comple e. De ails o
ial p ocedu es a e epo ed in he p o ocol [29].
A e 1–3 days, esea ch assis an s ga e he pa icipan s
a wais -wo n accele ome e (Hookie AM 20, T axmee
Heino e al. BMC Public Heal h (2018) 18:1352 Page 3 o 16
L d., Espoo, Finland) and ins uc ed hem on how o wea
i o a du a ion o se en consecu i e days (including he
day o ecei ing he de ice). The used Hookie accele om-
e e is a i-axial accele ome e ha collec s da a a 100 Hz
sampling a e wi hou p ep ocessing. The measu emen
ange o he accele ome e is ±16 g and he esolu ion is
4 mg (millig a i y). The Hookie accele ome e employs he
same i-axial accele a ion senso componen (ADXL345;
Analog De ices, No wood MA) ha is used in widely used
esea ch-g ade accele ome e s [33]. The alida ion o he
Hookie accele ome e has been epo ed in bo h child en
[34] and adul s [35] in s udies compa ing analysis o aw
accele a ion da a om di e en accele ome e s.
When pa icipan s ecei ed he accele ome e s, hey
we e asked whe he hey would like o ecei e SMS
messages o help hem emembe o pu i on e e y
mo ning. Those who consen ed o he messages we e
subsequen ly andomised o one o wo message condi-
ions, and hose who op ed no o ecei e he eminde s
we e ea ed as a sel -selec ed con ol a m.
A e 7 days, pa icipan s e u ned hei de ices o e-
sea ch assis an s and we e asked o ill ou a sho ques-
ionnai e assessing p ocess measu es (see Addi ional ile 1:
Appendix S1 and Addi ional ile 2: Appendix S2.
Random assignmen
Pa icipan s we e assigned o he eason and succinc
a ms a e hey we e ec ui ed. The i s au ho ex ac ed
he phone numbe s om he lis and used R code o c e-
a e an amoun o andom numbe s equal o he numbe
o new pa icipan s. The ec o o andom numbe s was
hen assigned o he pa icipan s. Pa icipan s wi h a
numbe equal o o smalle han he median o he ec-
o we e alloca ed o he eason-condi ion. O he s we e
alloca ed o he succinc condi ion. Resea ch assis an s
wo king in he ield o assess we e blind o g oup alloca-
ion. Rec ui men and andomisa ion ook place on he
same day, and es ic ions such as blocking o s a i ica-
ion we e no used.
Rec ui men ook place in wo wa es, alongside he e-
c ui men o he main ial. In o de o inc ease he a es
o pa icipan s op ing in o he eminde s, he ec ui -
men p omp was sligh ly modi ied o he second wa e.
The esea ch assis an s p esen ed he SMS eminde s as
he de aul op ion, and asked whe he his is accep able
o he pa icipan s.
Random assignmen was no isible o he pa icipan s
and he esea ch assis an s did no men ion ha di e -
en kinds o messages we e going o be sen . The s a is i-
cian who analysed he aw accele ome e da a was blind
o g oup assignmen .
In e en ions
An impo an issue ega ding he cu en s udy was o
a oid ampe ing wi h he e ec s o he main ial. In
o he wo ds, i should no a ec main ial ou come
measu es in any o he ways excep o inc eased da a
quali y. Ca e was aken o o mula e he SMS messages
o no p essu e pa icipan s o p o oke changes in main
ial ou come measu es such as PA.
Fig. 1 CONSORT low diag am
Heino e al. BMC Public Heal h (2018) 18:1352 Page 4 o 16
We al e ed a p e ious p ocedu e [11] by a ying he
message con en sligh ly each day o educe habi ua ion
and hus expec ed o inc ease he chances o he mes-
sage being ead, o bo h a ms.
The wo a ms ecei ed di e en message con en .
a) Succinc eminde condi ion: 1. a g ee ing –2. a
eminde –3. a hank you
b) Reminde and eason: 1. a g ee ing –2. a eason
beginning wi h “Because…” , ollowed up wi h a
eminde –3. a hank you
Messages a e p esen ed in de ail in Table 1below.
We sen he messages using an SMS Ga eway de ice
MT-SF100-G-EU (Mul iModem iSMS Se e 1-po ) by
Mul i-Tech Sys ems (h p://www.mul i ech.com/b ands/
mul imodem-isms). We used a manu ac u e -designed
guided use in e ace o he i s ec ui men wa e and
a cus om in e ace designed by a local se ice p o ide
o he second wa e.
Regis a ion and de ia ions om egis e ed plan
The s udy plan was e iewed by he E hics Commi ee o
Gynaecology and Obs e ics, Pedia ics and Psychia y o
he Hospi al Dis ic o Helsinki and Uusimaa (decision
numbe 367/13/03/03/2014).
O icial public egis a ion in he Ge man Clinical
T ials Regis e (DRKS-ID: DRKS00007721) was com-
ple ed 3 mon hs a e ec ui men o he i s wa e had
been ini ia ed, bu be o e da a was a ailable.
P e- egis a ion (be o e s a ing da a collec ion) ailed
due o lack o a ailable esou ces a he ime.
The o iginal plan was o es ablish he addi i e e ec o
messages con aining a eason and hose no con aining
one o e a no-message condi ion du ing he baseline
measu emen o he i s ba ch. Wi h he sample size we
expec ed (n= 140), we would ha e had o e 95% powe o
de ec an e ec o d = 0.6 (sligh ly smalle han he one
disco e ed in he Lange , Blank and Chanowi z eplica ion
[20]). We had hen planned o pi he mo e success ul
message ype agains a hi d message in he second wa e.
Ins ead o going o wa d wi h he plan o using a hi d
message, we made he decision o ga he ano he wa e o
pa icipan s wi h he same message ypes a e he da a
om he i s wa e was analysed. This was due o he ac
ha , con a y o ou expec a ions, no di e ence be ween
he wo messages was de ec ed. This is impo an o no e,
as i means we can no longe ely on a long- e m e o
a e o 5% [36]and—as p- alues depend on he sampling
dis ibu ion—de aul p- alues om common s a is ical
p og ams no longe apply [37].
To add ess he issue o inadequa e epo ing in he
sciences [38], he cu en epo complies wi h he
Consolida ed S anda ds o Repo ing T ials (CONSORT)
s a emen [39]. Con ibu o oles a e cla i ied in
Addi ional ile 3: Appendix S3, acco ding o a axonomy
o his pu pose [40].
Ou comes
P ima y ou come measu es
P ima y ou come measu es we e 1) accele ome e wea
ime minu es and 2) days wi h ≥10 h o alid accele om-
e e da a. As his ial was conduc ed wi hin a la ge
ial, se e al o he measu es we e collec ed and a e lis ed
in he Le ’s Mo e I p o ocol [29]. The main ial used a
Table 1 SMS con en , ansla ed o English
Mo ning Reminde wi h a ionale ( he “because heu is ic”) Succinc eminde
1s Mo ning! Because you pa icipa ion is p ecious, please emembe
o pu on he mo ion measu emen de ice and wea i un il you
go o sleep (excep in he showe e c.) - hanks!
Mo ning! This is a eminde o pu on he mo ion measu emen
de ice and wea i un il you go o sleep (excep in he showe e c.)
- hanks!
2nd Hi! Because you’ e aboa d in p oducing e y impo an knowledge,
please emembe o pu on he mo ion measu emen de ice now
and wea i as ins uc ed un il you go o sleep. Thanks a lo !
Hi! Please emembe o pu on he mo ion measu emen de ice
now and wea i as ins uc ed un il you go o sleep. Thanks a lo !
3 d Hello! Because he s udy wouldn’ succeed wi hou you help,
please emembe o pu on he mo ion measu emen de ice again
and wea i un il you go o sleep (excep in he showe e c.) - hanks!
Hello! Please emembe o pu on he mo ion measu emen de ice
again and wea i un il you go o sleep (excep in he showe e c.)
- hanks!
4 h Mo ning! Because he da a you ga he is highly alued, please
emembe o pu on he mo ion measu emen de ice and
wea i un il you go o sleep. Thanks (we’ e al eady pas midpoin )!
Mo ning! Please emembe o pu on he mo ion measu emen
de ice and wea i un il you go o sleep. Thanks (we’ e al eady pas
midpoin )!
5 h Howdy! Because you pa icipa ion p oduces e y impo an
knowledge, please emembe o pu on he mo ion measu emen
de ice and wea i un il you go o sleep (excep in he showe e c.) -
hanks!
Howdy! Please emembe o pu on he mo ion measu emen
de ice and wea i un il you go o sleep (excep in he showe e c.)
- hanks!
6 h Hi! Because e en his las day is impo an , please emembe o
pu on he mo ion measu emen de ice and wea i un il you go
o sleep. Re u n he mo ion measu emen de ice o school
omo ow - hanks!
Hi! Please emembe , e en on his las day, o pu on he mo ion
measu emen de ice and wea i un il you go o sleep. Re u n he
mo ion measu emen de ice o school omo ow - hanks!
Heino e al. BMC Public Heal h (2018) 18:1352 Page 5 o 16

3-axis accele ome e wi h a 2GB in e nal memo y (Hoo-
kie Me e 2.0, Hookie Technologies L d., Espoo,
Finland). The ac i i y da a was egis e ed using aw da a
and a 100 Hz sampling a e.
Implemen a ion assessmen measu es
A one-page ques ionnai e (Addi ional ile 1:AppendixS1;
ansla ion in Addi ional ile 2: Appendix S2) was used o
gain addi ional insigh in o he ecep ion o he messages.
Sel - epo ed message eceip . As we could no
ga he objec i e log da a on he numbe o messages
opened, we asked pa icipan s o assess on how many
mo nings hey had opened and ead he SMS. Response
op ions we e: No on a single mo ning, On 1 mo ning,
On 2–3 mo nings, On 4–5 mo nings and E e y
mo ning.
Manipula ion and con amina ion check. As pa ici-
pan s we e andomised indi idually, as opposed o clus-
e s a school class le el, discussing he SMS messages
wi h hei classma es could ha e led o s uden s inding
ou ha no e e yone ecei ed he same messages, and
pe haps also e eal he s udy hypo heses. We a emp ed
o gauge he ex en o his by asking hem how o en
hey had discussed he messages wi h pee s. Response
op ions we e: No once, Once, 2–3 imes, 4–5 imes and
Mo e o en.
Accep abili y o SMS message con en was assessed
by asking he pa icipan s, how much hey ag ee wi h
he s a emen “I was sa is ied wi h he con en o he
messages”. Response op ions again had a 5-poin scale:
Comple ely disag ee, Somewha disag ee, Do no ag ee
no disag ee, Somewha ag ee and Comple ely ag ee.
S a is ical analyses
All non-Bayesian analyses we e conduc ed using RS udio
unning R [41,42]. Plo s we e d awn using R packages
‘ggplo 2’[43]and‘ya ’[44]. Dis ibu ions be ween he
eason and succinc g oups in he implemen a ion assess-
men ques ions we e compa ed using he chi-squa e es .
Accelome e wea imes we e analysed using boo -
s apping me hods. A 95% boo s ap con idence in e al
o a mean can be acqui ed by esampling obse ed da a
o simula e a sampling dis ibu ion, ob aining he alues
o he 0.025 h and 0.975 h pe cen iles o esampled
means [45]. A ke nel densi y plo , boo s ap con idence
in e al and a boo s ap es o equi alence we e con-
duc ed using R package ‘sm’[46] o di e ences o dis i-
bu ions o he wo eminde a ms. Wilcoxon ank sum
es wi h con inui y co ec ion was used o compa e me-
dians be ween g oups.
ANOVA o equi alence o means be ween he wo e-
minde g oups and he no- eminde g oup, as well as i s
illus a ion, was pe o med using R package ‘use ien-
dlyscience’[47]. Addi ionally, a MANOVA wi h wea
ime minu es and wea days wi h alid da a as
dependen a iables, and SMS g oup as an independen
a iable, was used o es obus ness o esul s.
A 95% Bayesian Highes Densi y In e al (HDI) [44]o
he means o alid wea days was plo ed using R package
‘ya ’. HDI e e s o he mos likely popula ion pa ame e
alues (he e: means) gi en he da a; in o ma ion which is
no deli e ed by equen is con idence in e als [48,49].
Bayes ac o s
Due o ou sampling me hods (e.g. decision o collec mo e
da a was based on obse ed da a), adi ional equen is
s a is ics aced limi a ions. Thus, we also calcula ed Bayes
Fac o s [50–52] o ou main ou come measu es. A Bayes
ac o BF
01
is essen ially he a io o wo likelihoods, an-
swe ing ques ions such as “Gi en he da a, how many imes
mo e likely is he null hypo hesis, compa ed o a speci ic al-
e na i e hypo hesis”. We used he R package BayesFac o
[53]; Fo compa ing means, his package assigns he al e -
na i ehypo hesisaCauchyp io .Weusedap io scaleo
0.3, in acco dance wi h common e ec s in heal h psycho-
logical esea ch [54]. This e lec s a p io belie ha 50% o
he e ec s lie be ween d = −0.3 and 0.3. Fo con ingency a-
bles, p io s a e desc ibed in Jamil e al. [55]. The minimum
alue is 1, and an inc ease e lec s he belie , ha he dis i-
bu ion o obse a ions in he gi en ca ego ies unde H1 is
ela i ely mo e simila o H0. Addi ional in o ma ion on in-
e ence using Bayes Fac o s, and p io obus ness checks
a e oundin hesupplemen a y websi e (h ps://gi .io/
Nl8X).
Equi alence es ing
In he equen is s a is ical pa adigm, suppo o he null
hypo hesis is indica ed by he p ac ice o equi alence es -
ing [56]. Fo a di e ence be ween means, one essen ially
i s es ablishes a egion o equi alence o ze o, hen con-
duc s and combines wo - es s. The i s one es s
whe he he e ec is highe han he lowe bound (in ou
case, −0.3), and he o he es s whe he he e ec is
smalle han he highe bound (in ou case, 0.3). The es s
we e conduc ed using R package “TOSTER”[57].
We did no conduc mul i-le el analyses o accoun
o he in a-class co ela ion o 0.09 o o al accele om-
e e wea ime. He e ogenei y analysis is p esen ed in
he supplemen a y websi e (h ps://gi .io/ Nl8X) ile
unde “He e ogenei y among clus e s”.
Using s anda d de ia ions es ima ed om easibili y
s udy [30] da a, we de e mined a p ac ically signi ican
e ec size o wea ime hou s o be d = 0.42 –enough
o b ing a pe son om 9.5 h o daily da a o each he
cu o o 10 h. Fo ou pu poses, we decided o conside
e ec sizes be ween −0.3 and 0.3 as equi alen o ze o.
Addi ional de ails a e p esen ed in he supplemen a y
websi e unde “S a is ical powe ”.
Heino e al. BMC Public Heal h (2018) 18:1352 Page 6 o 16
Analysis ega ding s a is ical powe is p esen ed in
Fig. 2, holding alpha cons an a 0.05 and sample size a
achie ed le els. As seen om he igu e, we had 90%
powe o disco e an e ec o size d = 0.39, 80% o de-
ec d = 0.3, 60% o de ec d = 0.27 and 40% o disco e
an e ec o d = 0.21. Thus, ype 2 e o p obabili ies
we e small o e ec s nea ou de ined minimal e ec
size o in e es , bu high o small e ec s.
We also e alua ed Type S and ype M e o p obabili ies
[58], and he -s a is ic [59]. The analysis is p esen ed in
he supplemen a y websi e (h ps://gi .io/ Nl8X). In b ie ;
ou design was ela i ely well-equipped o handle
medium-sized e ec s, bu is subjec o conside able bias
unde small e ec s.
Resul s
Desc ip i e da a
A pa icipan low diag am p esen ed in Fig. 1indica es
how he messages we e sen o almos all pa icipan s as
in ended.
O he 375 pa icipan s consen ing o accele ome e
measu emen s as pa o he main ial, 95 op ed ou
o ecei ing eminde s and an addi ional 7 did no e-
cei e messages due o echnical di icul ies. In he
end, he SMS messages wi h easons we e sen o 138
and he succinc messages o 135 pa icipan s. Con-
sen a e o eminde s was 54% (101 ou o 186) in
he i s wa e and 95% o he second wa e (179 ou
o 189).
Table 2shows he sample cha ac e is ics o he base-
line da a.
Implemen a ion and p ocess measu es
Manipula ion and con amina ion check, as well as sa is-
ac ion wi h he messages and discussing hei con en a e
p esen ed in he supplemen . In b ie , we did no de ec
di e ences ac oss any g oups, wi h Bayes Fac o s indica -
ing s ong suppo o he null hypo heses. As shown in
Fig. 3.Se en y ou poin nine pe cen o esponden s e-
po ed ha ing opened and ead he SMS a leas ou
mo nings. Discussing he con en o he messages wi h
pee s was no common; 91.1% answe ed ha ing done so
ne e o jus once Fig. 4.
Open commen s did no e eal un o eseen nega i e
e ec s. In addi ion, 13% (9 ou o 70) o pa icipan s
who answe ed he ques ion explici ly added, ha e-
membe ing o wea he de ice was due o ecei ing
he messages.
Wea imes
Wea ime minu es
Accele ome e wea imes did no indica e meaning ul
di e ences be ween g oups (see Fig 5)
Boo s ap es s o equal densi ies indica ed no di e -
ences in o al wea ime minu es be ween he wo
message ypes (p= 0.28), no be ween hose who
ecei ed and did no ecei e messages (p=0.35). Wil-
coxon ank sum es showed no di e ences in dis i-
bu ions be ween message g oups (W = 8860, p= 0.647,
Fig. 2 S a is ical powe , - es o an unknown eal e ec
Heino e al. BMC Public Heal h (2018) 18:1352 Page 7 o 16
CI95 = −280.90–447.20) o whe he one op ed in he
messages o no (W = 10,642.5, p= 0.771, CI95 = −
424.20–305.30). Di e ences we e nei he de ec ed be-
ween he wo schools (W = 17,398.5, p= 0.051, CI95
=−1.60–619.60) o ec ui men wa es (W = 17,310.5,
p=0.067,CI95=−19.0–586.3).
The iolin plo s in Fig. 6illus a e how wea imes
in all h ee g oups a e dis ibu ed. Bayesian ANOVA
gi es us BF
01
= 12.05, indica ing s ong e idence o
equi alen means, agains a model whe e all means a e
unequal. P io obus ness g aph (see supplemen )
s a ing om = 0 depic ed a con ex unc ion, whe e
BF
01
ises o 10 a = 0.27 and eaches 422.34 a =
2.00. Fu he mo e, BF
01
ela i e o an o de ed model
o Reason > Succinc > Op ou was 23.07 (see sec ion
“In e p e ing Bayes Fac o s”in he supplemen a y
websi e (h ps://gi .io/ Nl8X).
Equi alence es s indica ed, ha he mean wea ime
di e ences be ween message ypes (69.92 min, 90% CI
[−262.37; 402.21]) and he eminde /op ou g oups
(1.98 min, 90% CI [−347.12; 351.08]) we e s a is ically
signi ican ly la ge han d = −0.3 and smalle han d =
0.3. In o he wo ds, he e ec size o he di e ence in
means was deemed less han |0.3|.
Valid measu emen days
Figu e 7shows densi ies and sp ead o alid measu e-
men days by g oup. As can be isually inspec ed om
he HDIs, popula ion means a e equi alen .
Di e ences be ween he dis ibu ions o measu e-
men days wi h > 10 h o da a we e no de ec ed be-
ween he eason and succinc g oups, χ
2
(7) = 7.893,
p= 0.342. A Bayesian con ingency ables es p o ided
BF
01
= 6.96 (Poisson sampling, p io concen a ion =
1.0; p io obus ness es depic s a conca e unc ion
whe e, as concen a ion app oaches 2, BF
01
ap-
p oaches 22.97).
Di e ences we e no de ec ed in alid wea day dis-
ibu ions be ween pa icipan s o whom eminde s
we e sen , and o whom hey we e no : χ
2
(7) = 8.344,
p= 0.303. A BF
01
= 34.79 (Poisson sampling, p io
concen a ion = 1.0; obus ness unc ion is conca e as
be o e. As concen a ion app oaches 2, BF
01
ap-
p oaches 93.50).
Again, equi alence es s o mean di e ences be ween
message ypes (−0.07 days, 90% CI [−0.47; 0.33]) was
s a is ically signi ican ly la ge han d = −0.3 and smalle
han d = 0.3. The mean di e ence be ween eminde and
op ou g oups (−0.18 days, 90% CI [−0.60; 0.24]) was
s a is ically signi ican ly smalle han d = 0.3, bu we
could no ejec he hypo hesis ha he e ec was highe
han d = −0.3.
A MANOVA wi h bo h o al wea ime minu es and
alid wea days as dependen a iables nei he de ec ed
Fig. 3 Opening and eading he SMS. I em s em: “I opened he SMS and ead i on he mo ning i was sen .”
Table 2 Sample cha ac e is ics
SMS g oup
Reason Succinc Op ou Send ailed To al
To al n 138 135 95 7 375
Wea ime
da a a ailable
133 129 83 7 352
Female 28% 30% 27% 43% 30%
M age (SD) 17.9 (1.8) 18.2 (2.6) 18.9 (4.3) 18 (1.4) 18.3 (2.9)
No e: One pe son om bo h SMS g oups missed he i s message due o
phone numbe impu a ion ailu e. This was conside ed o be o no p ac ical
consequence and hey we e coun ed as ha ing ecei ed hei in e en ion
as planned
Heino e al. BMC Public Heal h (2018) 18:1352 Page 8 o 16
di e ences be ween he eason, succinc and op ou
g oups (F(4, 682) = 2.335, p=0.054, Wilk’sΛ= 0.973), al-
hough mul icollinea i y may ha e posed a p oblem o he
model (τ=0.81,ρ=0.93).
Dose dependence
I eading o messages is linea ly ela ed o wea ime, an
upwa d mo ing slope in means would ha e been ex-
pec ed. The dose dependence cu e Fig. 8is la , show-
ing no suppo o such a ela ionship be ween messages
and wea ime.
Discussion
In an a emp o imp o e measu emen o physical ac i -
i y and seden a y beha iou —key public heal h issues—
his s udy e alua ed he e ec s o wo in e en ions o
inc ease accele ome e wea imes du ing he i s wo
ec ui men wa es o he Le ’s Mo e I ial. Speci ically,
i es ed he e ec s o he because-heu is ic on accele -
ome e wea ime in olde adolescen s. We did no de-
ec inc eased wea imes among pa icipan s who
ecei ed a eason in hei daily SMS eminde s, no did
we de ec di e en wea imes be ween hose ecei ing
he eminde messages and hose who op ed ou . In all
Fig. 5 To al wea ime in minu es (dashed line o he eason condi ion, solid o succinc ). G ey band a ound he ke nel densi y plo s e e s o
95% likelihood o con aining he ue densi y plo , i he wo lines we e gene a ed by da a om he same dis ibu ion. Mean (SD) Reason:
4549.57 min (1642.14), n= 133. Mean (SD) Succinc : 4479.65 (1616.04), n= 129
Fig. 4 Discussing he SMS wi h pee s. I em s em: “I discussed he con en o he messages wi h my pee s a school.”
Heino e al. BMC Public Heal h (2018) 18:1352 Page 9 o 16
85. Ru e H, Sa ona N, Glon i K, Bibby J, Cummins S, Finegood DT, e al. The
need o a complex sys ems model o e idence o public heal h. Lance .
2017;390:2602–4.
86. Snieho a FF, A aújo-Soa es V, B own J, Kelly MP, Michie S, Wes R. Complex
sys ems and indi idual-le el app oaches o popula ion heal h: a alse
dicho omy? Lance Public Heal h. 2017;2:e396–7.
87. Laka os I. His o y o science and i s a ional econs uc ions: Sp inge ; 1971.
h p://link.sp inge .com/chap e /10.1007/978-94-010-3142-4_7. Accessed 2
Dec 2015
88. Ba holomew Eld edge LK, Ma kham CM, Rui e RA, Fe nández ME, Kok G,
Pa cel GS. Planning heal h p omo ion p og ams: an in e en ion mapping
app oach. New Je sey: John Wiley & Sons; 2016.
Heino e al. BMC Public Heal h (2018) 18:1352 Page 16 o 16