STUDY PROTOCOL Open Access
Tailo ing mo i a ional heal h messages o
smoking cessa ion using an mHeal h
ecommende sys em in eg a ed wi h an
elec onic heal h eco d: a s udy p o ocol
San iago Ho s-F aile
1,2*
, F ancine Schneide
2
, Luis Fe nandez-Luque
3,4
, F ancisco Luna-Pe ejon
1
, An on Ci i
1
,
Dimi is Spachos
5
, Panagio is Bamidis
5
and Hein de V ies
2
Abs ac
Backg ound: Smoking is one o he mos a oidable heal h isk ac o s, and ye he qui ing success a es a e low.
The usage o ailo ed heal h messages o suppo qui ing has been p o ed o inc ease qui ing success a es.
Technology can p o ide con enien means o deli e ailo ed heal h messages. Heal h ecommende sys ems a e
in o ma ion- il e ing algo i hms ha can choose he mos ele an heal h- ela ed i ems— o ins ance, mo i a ional
messages aimed a smoking cessa ion— o each use based on his o he p o ile. The goals o his s udy a e o
analyze he pe cei ed quali y o an mHeal h ecommende sys em aimed a smoking cessa ion, and o assess he
le el o engagemen wi h he messages deli e ed o use s ia his medium.
Me hods: Pa ien s pa icipa ing in a smoking cessa ion p og am will be p o ided wi h a mobile app o ecei e
ailo ed mo i a ional heal h messages selec ed by a heal h ecommende sys em, based on hei p o ile e ie ed
om an elec onic heal h eco d as he ini ial knowledge sou ce. Pa ien s’ eedback on he messages and hei
in e ac ions wi h he app will be analyzed and e alua ed ollowing an obse a ional p ospec i e me hodology o
a) assess he pe cei ed quali y o he mobile-based heal h ecommende sys em and he messages, using he p ecision
and ime- o- ead me ics and an 18-i em ques ionnai e deli e ed o all pa ien s who comple e he p og am, and b)
measu e pa ien engagemen wi h he mobile-based heal h ecommende sys em using agg ega ed da a analy ic
me ics like session equency and, o de e mine he indi idual-le el engagemen , he a e o ead messages o each
use . This pape de ails he implemen a ion and e alua ion p o ocol ha will be ollowed.
Discussion: This s udy will explo e whe he a heal h ecommende sys em algo i hm in eg a ed wi h an elec onic
heal h eco d can p edic which ailo ed mo i a ional heal h messages pa ien s would p e e and conside o be o a
good quali y, encou aging hem o engage wi h he sys em. The ou comes o his s udy will help u u e esea che s
design be e ailo ed mo i a ional message-sending ecommende sys ems o smoking cessa ion o inc ease pa ien
engagemen , educe a i ion, and, as a esul , inc ease he a es o smoking cessa ion.
T ial egis a ion: The ial was egis e ed a clinical ials.o g unde he ClinicalT ials.go iden i ie NCT03206619 on
July 2nd 2017. Re ospec i ely egis e ed.
Keywo ds: Recommende sys em, Tailo ed messages, Smoking cessa ion, Mobile app, Pa ien , mHeal h
* Co espondence: [email p o ec ed]
1
Depa men o A chi ec u e and Compu e Technology, Uni e sidad de
Se illa, ETSII, A enida Reina Me cedes S/N, 41012 Se ille, Spain
2
Depa men o Heal h P omo ion, School o Public Heal h and P ima y Ca e
(Caph i), Maas ich Uni e si y, P. Debyeplein 1, 6229, HA, Maas ich , The
Ne he lands
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.
Ho s-F aile e al. BMC Public Heal h (2018) 18:698
h ps://doi.o g/10.1186/s12889-018-5612-5
Backg ound
New echnologies such as sma phones and wea ables
can be used o suppo beha io change among pa ien s,
as many s udies ha e al eady shown [1–7]. One o he
waysinwhich echnologyisused odosoisbydesign-
ing ailo ed heal h messages a ge ed a pa ien s. Some
such pla o ms use expe sys ems [8] ha use he ules
o human expe easoning and in e esul s based on
people’s answe s o ques ions abou beha io knowledge
and mo i a ional aspec s like a i ude and sel -e icacy. Ye
ano he ype o sys em is a ecommende sys em, which
aims a sending messages ailo ed o use s’p e e ences
[9–11]. These pla o ms employ algo i hms o p edic
which message is mos simila o i s use s’p e iously p e-
e ed messages.
A ecommende sys em is a piece o so wa e ha
lea ns o p edic he bes i em o each use om a se
o i ems [12]. I ems can be ex messages, mo ies, books,
people, o any hing else ha can be ecommended. Rec-
ommende sys ems ha e been exploi ed mos ex ensi ely
in he sphe es o e-comme ce and leisu e, h ough he
ecommenda ion o , o ins ance, mo ies, books, and music
[13]. Fo example, i a sys em knows he books you ha e
liked in he pas , i aims o o ecas he books you may also
like in he u u e. This can be done using di e en ech-
niques like compa ing he ea u es o he books you ha e
liked wi h he ea u es o o he books you ha e no ead, as
shown in Fig. 1, o by conside ing books ha people wi h
simila as es as you ha e also liked.
Recommende sys ems ha e also been used in he
heal hca e domain. Heal h ecommende sys ems a e
especially aimed a p o iding eedback and ecommen-
da ions on heal h s a us and hea h beha io s, such as
li es yle, nu i ion [14], obesi y [15], diabe es [16,17],
d ug side e ec s [18], and smoking cessa ion [19]. The
ype o eedback is based on algo i hms p edic ing he
ype o message needed on he basis o p e iously mea-
su ed a iables. The equi ed inpu can be aken om
elec onic heal h eco ds as desc ibed by Wiesne e al.
[20] o may be based on a pe son’s in e es in speci ic
leisu e- ime beha io s.
Howe e , ew s udies use his echnique, and hei po-
en ial is s ill o be exploi ed [21–23]. In addi ion, heal h
ecommende sys ems need o be p ecise and accu a e
in o de o be e ec i e. Tha op imiza ion can al eady
be ound in ecommende sys ems used o ecomme ce
and leisu e. Howe e , assessing he quali y o heal h
ecommende sys ems in e ms o app ecia ion o he
ecommenda ions by pa ien s and hei beha io al e -
ec s has so a no been explo ed in-dep h. This can be
done using expe imen al designs aimed a measu ing
hei objec i e accu acy [24]o by indingou hei
use s’subjec i e opinions as desc ibed in he ResQue
amewo k [25]. The goal o his s udy is de elop and
e alua e he quali y o a heal h ecommende sys em o
ill his exis ing gap.
The heal h ecommende sys em chosen o he s udy
is aimed a suppo ing smoking cessa ion, because smoking
is di ec ly associa ed wi h a numbe o diseases: ch onic
obs uc i e pulmona y disease, as hma exace ba ion, ca a-
ac s, pulmona y ib osis, o al ca i y cance , pha ynx cance ,
la ynx cance , esophagus cance , lung cance , and bladde
cance , among o he s [26]. Consequen ly, qui ing smoking
is he mos impo an decision smoke s can ake o
op imize hei chances o educe heal h isks and in-
c ease he longe i y o hei li es [27]. Se e al s udies
explo ed he e ec s o inno a i e ailo ed beha io al
change me hods o smoking cessa ion and demons a ed
Fig. 1 Concep diag am o a simple ecommende sys em. In his example, he ecommended i ems a e books. The use ecei es sugges ions on
wha o ead nex based on he gen es and ea u es o books he o she has liked in he
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 2 o 10
ha hey can con ibu e o a highe success a es [28–30].
O he s udies showed ha he deli e y o ailo ed mo i -
a ional messages using mobile phones as a pla o m
could be an e ec i e way o enable smoking cessa ion
[31–35] as well as o impac o he aspec s o heal h,
such as p omo ing physical ac i i y and exe cise [36,37],
men al heal hca e [38], and alcohol- ela ed ha m [39],
among o he s. These ailo ed messages a e usually pieces
o ex based on he use ’s esponses o ce ain ques ions,
assessing hei a i ude, social suppo , sel -e icacy, and
ype o ac ion planning.
Ye , qui ing smoking is o en accompanied by se e al
ba ie s, such as nico ine abs inence synd ome ha may
ha e consequences like headaches, c a ings, in es inal dis-
o de s, weigh gain, insomnia, es lessness, ne ousness,
dep ession, and i i abili y, among o he s [40,41]. As a con-
sequence, he a e o smoking elapses is high and success
a es a e o en low [42–46], illus a ing a clea need o p o-
iding messages ha a e highly pe sonalized and ailo ed o
he needs o each pe son a ha poin in ime. Heal h ec-
ommende sys ems may ha e speci ic added alue he e as
hey use algo i hms o op imally choose no only he ype
o message o be sen bu also when i is sen .
In he cu en s udy, we will combine he p inciples o
ailo ing wi h wo ecommende sys ems, which cus omize
messages and p edic he bes ime o send he ailo ed
mo i a ional heal h messages o a coho o pa ien s who
a e ying o qui smoking using a mobile app. To selec
he mos ele an message opic ha can os e heal hie
habi s in each pa ien , a heal h ecommenda ion sys em
algo i hm (HRSA) accesses each pa ien ’s pe sonal heal h
eco ds and eeds he algo i hm wi h his heal h da a. Since
he messages a e deli e ed using sma phones and a e
ailo ed as pe each pa ien ’s heal h da a, we call ou sys em
an mHeal h Recommende Sys em (m-HRS).
Wendel e al. [47] sugges ed ou basic equi emen s
o educe po en ial ba ie s o using ecommende sys ems:
1. E o inpu minimiza ion o educe use s’bu den; 2.
P i acy assu ance; 3. Op imizing message use ulness; and 4.
Enjoymen .
The quali y o he sys em is s ongly associa ed o he
pa ien s’a i udes owa d he sys em and, consequen ly,
o i s usage [48]. A heal h sys em ha pa ien s eel is no
o a good quali y and one hey canno us will no be
used [49]. By inc easing he pe cei ed quali y o he sys em,
pa ien s will be encou aged o adop he sys em and use i
mo e, which will lead o be e heal h ou comes [50,51].
A i ion is a well-known p oblem in digi al heal hca e
in e en ions [52]. I is impo an o inc ease app eci-
a ion and hus aise pa ien engagemen , he eby keeping
a i ion o a minimum. This is especially ele an in
m-HRS because he mo e use engagemen wi h he sys-
em, he mo e he sys em lea ns abou he use , and, con-
sequen ly, he s onge he sys em becomes. Alkhaldi e al.
explained he impo ance o use engagemen in digi al
heal h in e en ions [53]. Al hough he e a e di e en de -
ini ions o engagemen , he one used in his s udy was
p oposed by Alkhaldi e al. in he s udy “Use s e isi ing
he digi al in e en ion”. A high engagemen in digi al
heal h in e en ions is also associa ed wi h be e pa ien
heal h ou comes [54,55]. This is because engaged pa ien s
ead mo e ailo ed mo i a ional heal h messages and
he e o e ecei e mo e such p omp s. This has a posi i e
impac on hei beha io change, such as qui ing
smoking, as some s udies ha e al eady shown [56,57].
Howe e , he no el y e ec may also in luence hei
a i ion. Consequen ly, i would be necessa y o com-
pa e how pa ien s’opinions ega ding he HRS mes-
sages e ol e h oughou he in e en ion.
In conclusion, he i s goal o his s udy is o desc ibe
he pa ien s’pe cei ed quali y o he m-HRS. The second
goal is o assess he pa ien s’le el o engagemen wi h he
messages gene a ed by he sys em. Iden i ying which sys em
ea u es and use cha ac e is ics de e mine di e ences in
use engagemen and quali y pe cep ion may help imp o e
he design o u u e m-HRS pla o ms.
Me hods/Design
Pa icipan s
Fo his obse a ional p ospec i e s udy, we will analyze
pa ien s pa icipa ing in he SmokeF eeB ain p ojec
[58] o a pe iod o 12 mon hs.
The inclusion c i e ia o he s udy a e pa ien s who
a e a ending he smoking cessa ion p og am Smoke
F ee B ain [59,60] a he Vi gen del Rocio Uni e si y
Hospi al, a e a leas 18 yea s old, a e willing o s a
ea men o qui smoking, own an And oid sma phone
and know how o use i , ha e ins alled he smoking ces-
sa ion app ecommended by he doc o , and who ha e
no p e iously had any known ad e se e ec s o he
pha macological ea men .
T ial design
Pa ien s will be p o ided wi h a pu pose-buil And oid-based
mobile applica ion ha allows hem o ecei e messages
and a e hem on hei sma phones. The app is called
“Lib e de humos”,o “Smoke ee”in English. F om he e
onwa d, we will e e o i as “ he app”.Theapp’so he
ea u es include a goal achie emen dashboa d, physical
exe cise eco ds, a elapse dia y, a elaxa ion ool,
mini-games o help pa ien s o e come c a ings h ough
dis ac ions—one o which is based on he webFi Fo All
exe gaming pla o m p o ocol [61,62]—and an in o m-
a i e sec ion wi h con en on a ious opics ela ed o
smoking cessa ion.
As no o mal me hodology has ye been used o assess
mobile heal h ecommende sys ems, we will use he
message a ings, pa ien s’app usage beha io , and an
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 3 o 10
adap ed se o ques ions o he pa ien s o compa e
hei le els o quali y app ecia ion and engagemen . This
will allow us o compa e di e ences be ween he m-HRS
pa ien s’opinions o e ime.
App ea u es and clinical in eg a ion
The p esen s udy will ocus on exploi ing ea u es o
sma phones ha allow use s o ecei e messages om a
se e and ack hei ac i i y and in e ac ion. The app
can connec wi h he smoking cessa ion uni o he
hospi al’s elec onic heal h eco d using Mi h Connec
so wa e. I in e connec s he ollowing:
The hospi al use da abase o access he pa ien s’
demog aphic da a
The hospi al clinical da a base o access he pa ien s’
clinical in o ma ion
The Ligh weigh Di ec o y Access P o ocol o
alida e he c eden ials o he heal hca e
p o essionals who access he hospi al sys em
The in o ma ion eques ed by he app is p ocessed
and selec ed om he elec onic heal h eco d da abase,
o ma ed using he ISO13606 s anda d, and sen back
by he Mi h Connec pla o m.
The e o e, all he in o ma ion ega ding a pa ien ’sp o-
ile—name, age, gende , da e o qui ing smoking, ype o
pha macological ea men —is au oma ically loaded in he
app wi hou pa ien s needing o inpu any hing apa om
a code p o ided by hei clinicians. Fu he mo e, using he
clinical s a ion— he sys em in e ace a he hospi al—
heal hca e p o essionals can moni o he pa ien s’ac i i y
using he app. Fo ins ance, i a pa ien logs a elapse, his
o he doc o s will be able o access i , p o ided hey ha e
he pa ien ’s consen o do so.
The app will allow pa ien s o a e he messages hey
ha e ecei ed using bu ons o “Like”(posi i e), “Dislike”
(nega i e), o “Don’ mind”(neu al), as shown in Fig. 2.
The messages hey will ecei e all unde one o he ol-
lowing i e opics: gene al mo i a ion, die ips, physical
exe cise ips, pe sonal pe o mance, and he bene i s o
being a non-smoke .
Each o hese opics has a pool o 150 di e en messages,
along wi h use ul ailo ed in o ma ion o he pa ien s.
These opicsandmessagesha ebeenapp o edbyasmok-
ing cessa ion psychologis as well as a pulmonologis om
he hospi al.
Thus, his design will show whe he pa ien s p e e he
messages selec ed by he m-HRS a he end o he pilo
as a consequence o he HRSA being ained o be e
ma ch he pa ien ’s p e e ences, as opposed o he mes-
sages sen a he beginning when he HRSA was no
con igu ed o lea n and imp o e messages based on he
pa ien s’ eedback.
Use -cen ic conside a ions in he heal h ecommende
sys em
Based on Wendel e al.’s s udy [47] on os e ing heal h
ecommenda ion sys ems, we ook in o accoun he
ollowing:
E o : Any e o equi ed om he pa ien ’s side
has been minimized by emo ing he en y ba ie
o ha ing o manually inpu hei p o ile de ails.
These de ails a e au oma ically loaded om he
hospi al elec onic heal h eco d.
P i acy isk: The app only s o es he pa ien ’s name
and all communica ion ollows he MD5 enc yp ion
p o ocol as equi ed o mee he hospi al’s p i acy
equi emen s. The app and he elec onic heal h
eco d exchange XML documen s ollowing he
HL7 p o ocol using he Mi h Connec engine.
Use ulness: The opics o he messages and he
messages hemsel es we e alida ed by a
psychologis and pulmonologis o make su e hey
con ain ele an scien i ic in o ma ion.
Enjoymen : The use o iendly and amilia language
combined wi h he equency and iming o he
messages makes he app eel pe sonal and less
obo ic. This ollows he gami ica ion p inciple o
“unp edic abili y and andomness”as pe he
Oc alysis gami ica ion amewo k [63].
Finally, Wendel e al.’s s udy also e e ed o he in e -
media y ha in oduces he sys em o he pa ien s. In
his case, i is he heal hca e p o essional om he hospi-
al’s smoking cessa ion uni who p esen s and endo ses
he m-HRS. This ac may also posi i ely con ibu e o
pa ien s using i .
Choosing he message opic
The da a on he message a ings and he pa ien s’in e -
ac ion wi h he app will be sen o he hospi al se e ,
whe e i will be p ocessed by an HRSA o enable i o
choose he nex message o each use as well as he
ime a which i should be sen . Figu e 3shows a
concep ualiza ion o his mHeal h Recommende Sys em
a chi ec u e. The HRSA is an algo i hm ha ollows a
hyb id app oach o imp o e i s pe o mance, as sug-
ges ed by Bu ke [64]. The algo i hm combines he ol-
lowing h ee ac o s o compu e he esul s:
Pa ien ’s demog aphic simila i y: Demog aphic
in luence will be based on he simila i y o pa ien s
in e ms o hei age, gende , employmen s a us,
da e hey s opped smoking, and sco es on he
Fage s öm [65] and Richmond es s [66].
Pe cei ed u ili y o he message opics: The u ili y o
he messages will be measu ed as ollows: Pa ien s
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 4 o 10
will be p omp ed o a e each message wi h “like”,
“dislike”,o “neu al”. In case o mul iple a ings
gi en o he same message, he old a ings will be
o e idden. Visi s o sec ions o he app on he same
opic as he messages o e- ead messages on a gi en
opic will also be conside ed.
S a emen o ini ial in e es : Pa ien s need o ill ou
a ques ionnai e consis ing o i e ques ions in which
hey s a e hei in e es in he i e message opics.
This in e es can be a ed as posi i e, neu al, o
nega i e. Pa ien s can modi y hei answe s any ime
h ough he app.
Fig. 3 Desc ip ion o he mHeal h Recommende Sys em’s a chi ec u e and low o da a in o ma ion
Fig. 2 Sc een cap u es o he message lis and he a ing mechanism o he “Lib e de humos”smoking cessa ion app. Pa ien s can p o ide
eedback on he messages hey ecei e ia he app. They can p o ide eedback on each message by indica ing “like”,“dislike”,o “indi e en ”
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 5 o 10

Once he opic has been selec ed, he sys em selec s a
p e iously unsen message om he da abase and ailo s
i o he use so ha i is mo e pe sonalized. Fo in-
s ance, he message may include he name o he use .
Sending he messages
The equency a which pa ien s will ecei e messages is
based on he conclusions o Ab oms e al. [67] consid-
e ed in combina ion wi h he ans heo e ical beha io
change model [68]. This model was chosen because he
counseling o e ed a he hospi al’s smoking cessa ion
uni is pa ially based on i , and we wan ed he messages
o be in line wi h his counseling. We s a ed conside -
ing pa ien s in he p epa a ion phase, which is when
hey can join he p og am.
P epa a ion: One message a day un il he day o
qui ing— he opic o hese messages is no
p edic ed by he HRSA bu chosen a andom, and
hey canno be a ed, so hey ha e no in luence on
he HRSA
Ac ion: Fou messages on he day o qui ing, and
one message a day o he ollowing week.
Main enance: Th ee messages a week a e ha , o
one yea .
In he main enance phase, as pe he ans heo e ical
beha io al change model, he messages will be sen a
andom days du ing he week. In o al, a pa ien who
comple es he yea long ea men will ecei e a o al o
150 messages om he m-HRS. The ime a which he
messages will be sen is also calcula ed by he HRSA,
based on he ime di e ence be ween when a message is
sen and when i is ead. The day will be di ided in 12
wo-hou pe iods du ing which he message can be sen
a any andom ime. The soone he pa ien eads he
message a e eceip , he mo e likely i is ha he nex
message will be sen in he same wo-hou pe iod. Pa-
ien s can deac i a e messages du ing a speci ic pe iod
o ime h ough a do-no -dis u b se ing. Jus like he
inal decision o he message opic, he messages’ iming
is also based on a p obabili y ec o .
Ou comes
Ou p ima y ou comes a e he pa ien s’pe cei ed qual-
i y o he m-HRS and hei engagemen wi h i . We de-
ine quali y as a se o ea u es ha makes he sys em
ecommend imely and ele an messages o each pa-
ien . As no o mal me hodology has ye been used o
assess he quali y o and engagemen wi h HRS, we will
use bo h objec i e and subjec i e quali y measu es
along wi h engagemen me ics a an agg ega ed and
indi idual le el.
Quali y
Objec i e quali y
Since he HRSA p edic s he message opic and iming
o deli e y, he objec i e pe cei ed quali y will be mea-
su ed using wo me ics: he p ecision o he message
opics, and how long a e ecei ing he message he
use s ead i .
The p ecision me ic is de ined as he ela ion be ween
he numbe o hi s pe message in he i s mon h (base-
line) and he ex en o “posi i e”,“neu al”,and“nega i e”
eedback pe message in mon h 3, mon h 6, mon h 9, and
mon h 12. Hi s a e messages wi h only posi i e eedback
(1) P ecision
p
and wi h bo h posi i e and neu al eedback
(2) P ecision
p&n
.
P ecisionP¼jmessages wi h posi i e eedback j
jall messages wi h any eedback jð1Þ
P ecisionpn ¼jmessages wi h posi i e and neu al eedback j
jall messages wi h any eedback j
ð2Þ
We expec he accep ance a e o be highe in mon h
12, as compa ed o he p e ious mon hs, as by his ime
he HRSA will ha e mo e pa ien s and enough in o ma-
ion o make accu a e decisions.
We will calcula e he p ecision o he sys em a di e en
poin s o ime, which will e lec he use s’ eedback and
hei in e ac ions wi h speci ic sec ions o he app. All his
in o ma ion is s o ed in he da abase o he m-HRS.
The ime- o- ead me ic is de ined as he ime di e -
ence be ween when he message was sen and when i
was ead by he use . We will use i o assess he e olu-
ion in he quali y o he HRSA o p edic he bes ime
a which o send a use messages—we heo ize ha he
lesse he ime- o- ead, he be e i is o he use o e-
cei e messages a ha ime.
Subjec i e quali y
The subjec i e quali y will be e alua ed h ough eigh een
ques ions assessed using a i e-poin Like scale answe ed
by all pa ien s a he end o hei smoking cessa ion
p og am. The ques ionnai e is adap ed o he con ex o
smoking cessa ion and o i wi hin he expec ed ime
ame in which he pa ien s will ha e o comple e i du ing
hei consul a ion a he hospi al. The i s wel e ques-
ions a e a selec ion o hose p oposed in he ResQue
amewo k [25] in ended o measu e he quali y o he
use ’s expe ience wi h a ecommende sys em and i s
in luence on he use ’s beha io s and in en ions. The
emaining six ques ions a e designed on he basis o he
i-Change beha io change model [69–71] o iden i y
whe he he messages also ha e an impac on he pa-
ien s’mo i a ions.
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 6 o 10
Engagemen
Engagemen a an agg ega ed le el
This will be measu ed h ough i e ac o s using he an-
aly ics so wa e Yahoo Flu y, which is in eg a ed in he
m-HRS app. As desc ibed by he so wa e de elope s
hemsel es [72], hese ac o s a e as ollows:
1. Rolling e en ion: The pe cen age o use s s ill
ac i e N days a e ins alla ion. This is a a io o he
numbe o use s whose las day o ac i i y is pas
day N o he numbe o use s who could ha e been
ac i e on day N (i.e., he sum o new use s up un il
day N).
2. Session leng h dis ibu ion: The session leng h is
de ined simply as he leng h o ime be ween he
s a app e en and he end app e en . The session
leng h de e mines engagemen in e ms o how
much ime pa ien s spend on he app pe session.
3. Session equency: F equency o Use is a measu e o
how o en each unique use used he app wi hin a
gi en ime in e al.
4. Sessions pe use : A session is de ined as one use
o he app by a pa ien . This begins when he
applica ion is launched and ends when he
applica ion is e mina ed.
5. Re u n a e: Re u n a e measu es he pe cen age o
use s who e u n o he app a a speci ic ime a e
ins alla ion. We look a his alue by coho g oup,
ha is, based on when use s i s opened he app. I
is calcula ed as he a io o he numbe o use s
ac i e on a gi en day, week, o mon h o he size o
he coho . The ins all da e is conside ed as Day 0.
Fo example, he Re u n Ra e o day 7 is he
pe cen age o use s ha opened he app on he 7 h
day a e ins alla ion. Use ac i i y on day 8, 9, and
so on does no impac his alue.
Engagemen a an indi idual le el
This will be assessed based on he a e o messages ead
by he use . I is calcula ed as he a io o he messages a
use has ead and he o al numbe o messages he sys-
em has sen o he use .
Sample size
We aim o ec ui 120 pa ien s. No powe analysis was
done o his s udy, as we ollowed a con enien sample
[73] me hod: ou ec ui ing pa ne Hospi al Vi gen del
Rocío es ima ed hey could only access 120 pa ien s
wi hin he ime limi s o he p ojec . The expec ed d op-
ou a e is 50%. This means he sys em will ha e sen a
leas 9000 messages o pa ien s be o e he s udy ends (a
minimum o 60 pa ien s ending he ea men mul iplied
by 150 messages sen h oughou he ea men ).
E hics
The Social Local and Mobile in e en ion his s udy is
based on has been app o ed by he E hical Commi ee o
he Vi gen del Rocío and he Vi gen Maca ena Uni e si y
Hospi als (app o al numbe SFB-APP_EC-2016-01). The
esea ch ac i i ies will s ic ly ollow hospi al egula ions.
Rega ding he use o da a, his p ojec will con o m
o he egula ions o he Pe sonal Da a Ac (based on
he Eu opean Da a P o ec ion Di ec i e).
Discussion
This s udy will explo e how pa ien s en olled in a smoking
cessa ion p og am pe cei e and engage wi h an m-HRS
ha sends hem pe iodic mo i a ional messages encou -
aging hem o s op smoking. We also expec he m-HRS
o help inc ease he low a i ion a e ha is ypically seen
in eHeal h ials [74].
The answe s o his s udy’s esea ch ques ions will
bene i u u e ecommende sys em-based beha io al
change in e en ions. Resea che s will be able o e-use
ou p oposed pe cei ed quali y and engagemen me ics
and exploi he knowledge de i ed in o de o design
new and imp o ed heal h ecommende sys ems ha
ake in o accoun use app ecia ion and engagemen .
Thus, he heal h messages sen o he use s will be mo e
in e es ing, ele an , and engaging o hem.
We expec he pa ien s’pe cei ed quali y o he sys em
o inc ease o e ime, ha is, o he p ecision o he
messages o inc ease and he pa ien s’ ime o ead he
messages o dec ease. Howe e , we should conside ha ,
a some poin du ing he in e en ion, pa ien s may s op
eading he messages because hey ha e unins alled he
app. This may be because hey ha e success ully qui
smoking and do no wan any smoking- ela ed e-
minde s, because hey ha e elapsed, o because hey
ne e engaged wi h he app in he i s place and ha e
decided o emo e i om hei sma phone. The e o e,
hese me ics will need o be ca e ully analyzed o a oid
simplis ic conclusions.
The use s’in-app ac i i y analy ics will p o ide an
insigh in o how pa ien s eac o he messages he HRSA
selec s o hem. This opens up se e al u u e esea ch
oppo uni ies, such as compa ing whe he he heo e ical
in e es s pa ien s claim o ha e a e in ac hei in e es s,
no ing he ime hey s op a ing messages because hey
a e no longe in e es ed in ecei ing mo e messages, and
ha ing an ini ial m-HRS-gene a ed message pa ien ac-
cep ance da a epo o u u e implemen a ion in smok-
ing cessa ion p og ams.
Howe e , his s udy may su e om he us p oblem
like all ecommende agen s. Despi e he elabo a e design
o he HRSA, should a pa ien ecei e a single message
ha is ou o con ex , i may uin he con idence in he
sys em and she may s op using he m-HRS al oge he
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 7 o 10
[75]. In o de o minimize his, he messages ha e been
designed such ha hey a e applicable o almos all ci -
cums ances and expe iences o a pa ien .
The p emise behind his s udy using an m-HRS in-
s ead o a s anda d ini ial ques ionnai e o an expe
sys em is he high on- he-go adap abili y po en ial o
pa ien s wi hou bo he ing hem wi h ques ionnai es o
upda e hei p e e ences. Ano he eason o his is he
po en ial scalabili y and obus ness he m-HRS can
achie e in he long e m when he HRSA da abase con-
ains in o ma ion on a la ge numbe o pa ien s. This
will p o ide he basis o heal h beha io change s udies
o be conduc ed a a la ge scale and o become mo e
ad anced since he ecommenda ions o pa ien s will
be selec ed om a mo e complex se o op ions.
Limi a ions and gene alizabili y
This s udy is limi ed by he design o he HRSA, which
canno dis inguish be ween longi udinal in e es changes
in he pa ien coho du ing he smoking cessa ion p ocess.
In addi ion, he HRSA is also limi ed because i does no
di e en ia e be ween wo king days and holidays.
The imposed maximum sample size and he lack o
a-p io i powe analysis may also be a limi ing ac o o
he gene alizabili y o he esul s o he s udy. Howe e ,
o minimize his, we will do powe calcula ions o mini-
mum de ec able e ec sizes o key ou comes.
The pa ien engagemen analysis is limi ed because i
is no possible o c oss-compa e he indi idual-le el en-
gagemen me ics o he agg ega ed-le el engagemen
me ics.
Since he s udy will only be conduc ed in he ci y o
Se ille, i may no be possible o he ou comes o be di -
ec ly gene alized o he Spanish popula ion as a whole.
Abb e ia ions
HRSA: Heal h ecommenda ion sys em algo i hm; m-HRS: mHeal h
Recommende Sys em
Acknowledgemen s
We app ecia e he con ibu ions o he pa ne s o he SmokeF eeB ain
p ojec who di ec ly wo ked in his s udy: he Vi gen del Rocío Uni e si y
Hospi al o i s suppo in in eg a ing he HRSA wi h hei se e s and
synch onizing i wi h hei elec onic heal h eco ds, Salumedia Tecnologías,
he A is o le Uni e si y o Thessaloniki, and he No he n G eece Neu oscience
Cen e o hei suppo in de eloping he app.
Funding
The p esen s udy was unded h ough he P ojec SmokeF eeB ain
“Mul idisciplina y ools o imp o ing he e icacy o public p e en ion
measu es agains smoking”o he Eu opean Union’s Ho izon 2020 esea ch
and inno a ion p og amme unde he g an ag eemen No 681120.
Au ho s’con ibu ions
SHF concei ed and designed he m-HRS s udy and also d a ed and e ised
he manusc ip . LFL, FS, and HV specially con ibu ed o he p o ocol design,
e alua ion me hods, and c i ically e isi ing he manusc ip . FLP, AC, DS, and
PB con ibu ed du ing he d a ing o he manusc ip de ailing echnical
and me hodological aspec s. All au ho s ha e ead and app o ed he
inal manusc ip .
E hics app o al and consen o pa icipa e
The E hical Re iew Boa d “CEI de los Hospi ales Uni e si a ios Vi gen Maca ena
yVi gendelRocío”, a ilia ed o he Vi gen Maca ena and Vi gen del Rocío
Uni e si y Hospi als, app o ed he SoLoMo s udy wi h code SFB-APP_EC-
2016-01. This E hical Re iew Boa d also e iewed he in o med consen ,
which all pa icipan s we e gi en in w i en ex , and had o ead and sign
o be included in he s udy.
Compe ing in e es s
LFL is he owne o Salumedia Tecnologías, SHF is i s adminis a o . Salumedia
Tecnologías is he company ha de eloped he app used in his s udy. The
au ho s decla e ha hey ha e no compe ing in e es s.
Publishe ’sNo e
Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in
published maps and ins i u ional a ilia ions.
Au ho de ails
1
Depa men o A chi ec u e and Compu e Technology, Uni e sidad de
Se illa, ETSII, A enida Reina Me cedes S/N, 41012 Se ille, Spain.
2
Depa men
o Heal h P omo ion, School o Public Heal h and P ima y Ca e (Caph i),
Maas ich Uni e si y, P. Debyeplein 1, 6229, HA, Maas ich , The Ne he lands.
3
Qa a Compu ing Resea ch Ins i u e, Hamad bin Khali a Uni e si y,
Educa ion Ci y, Doha, Qa a .
4
Salumedia Tecnologías, A enida República
A gen ina 24, Edi icio To e de los Remedios, Plan a 5, Módulo A, Se ille,
Spain.
5
Medical School, Facul y o Heal h Sciences, A is o le Uni e si y o
Thessaloniki, Thessaloniki, G eece.
Recei ed: 16 Oc obe 2017 Accep ed: 25 May 2018
Re e ences
1. Fjeldsoe BS, Ma shall AL, Mille YD. Beha io change in e en ions deli e ed
by mobile elephone sho -message se ice. Am J P e Med. 2009;36(2):
165–73.
2. Webb TL, Joseph J, Ya dley L, Michie S. Using he in e ne o p omo e
heal h beha io change: a sys ema ic e iew and me a-analysis o he
impac o heo e ical basis, use o beha io change echniques, and mode
o deli e y on e icacy. J Med In e ne Res. 2010;12:1.
3. Riley WT, Ri e a DE, A ienza AA, Nilsen W, Allison SM, Me mels ein R. Heal h
beha io models in he age o mobile in e en ions: a e ou heo ies up o
he ask? T ansl Beha Med. 2011;1(1):53–71.
4. Cole-Lewis H, Ke shaw T. Tex messaging as a ool o beha io change in
disease p e en ion and managemen . Epidemiol Re . 2010;32:56–69.
5. Lyons EJ, Lewis ZH, May sohn BG, Rowland JL. Beha io change echniques
implemen ed in elec onic li es yle ac i i y moni o s: a sys ema ic con en
analysis. J Med In e ne Res 2014;16:8;192.
6. Smi h DJ, G i i hs E, Poole R, Di Flo io A, Ba nes E, Kelly MJ, C addock N,
Hood K, Simpson S. Bea ing bipola : explo a o y ial o a no el in e ne -
based psychoeduca ional ea men o bipola diso de . Bipola Diso d.
2011;13(5–6):571–7.
7. Ro ondi AJ, Ande son CM, Haas GL, Eack SM, Sp ing MB, Ganguli R, Newhill
C, Rosens ock J. Web-based psychoeduca ional in e en ion o pe sons
wi h schizoph enia and hei suppo e s: one-yea ou comes. Psych Se .
2010;61(11):1099–105.
8. P ochaska JO, Velice WF, Redding C, Rossi JS, Golds ein M, DePue J, G eene
GW, Rossi SR, Sun X, Fa a JL, e al. S age-based expe sys ems o guide a
popula ion o p ima y ca e pa ien s o qui smoking, ea heal hie , p e en skin
cance , and ecei e egula mammog ams. P e Med. 2005;41(2):406–16.
9. Ali R, A zal M, Hussain M, Ali M, Siddiqi MH, Lee S, Kang BH. Mul imodal
hyb id easoning me hodology o pe sonalized wellbeing se ices. Compu
Biol Med. 2016;69:10–28.
10. Schulz DN, K eme s SP, Vandelano e C, Van Ad ichem MJ, Schneide F, Candel
MJ, de V ies H. E ec s o a web-based ailo ed mul iple-li es yle in e en ion o
adul s: a wo-yea andomized con olled ial compa ing sequen ial and
simul aneous deli e y modes. J Med In e ne Res 2014;16:1;26.
11. P ochaska JJ, Sp ing B, Nigg CR. Mul iple heal h beha io change esea ch:
an in oduc ion and o e iew. P e Med. 2008;46(3):181–8.
12. Ricci F, Rokach L, Shapi a B. In oduc ion o ecommende sys ems handbook.
2nd ed. New Yo k: Sp inge US; 2015.
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 8 o 10
13. Pa k DH, Kim HK, Choi IY, Kim JK. A li e a u e e iew and classi ica ion o
ecommende sys ems esea ch. Expe Sys Appl. 2012;39(11):10059–72.
14. Espín V, Hu ado MV, Nogue a M. Nu i ion o elde ca e: a nu i ional
seman ic ecommende sys em o he elde ly. Exp Sys. 2016;33(2):201–10.
15. Giabbanelli PJ, C u zen R. Suppo ing sel -managemen o obesi y using a
no el game a chi ec u e. Heal h In o m J. 2015;21(3):223–36.
16. Bocaneg a CLS, Ramos JLS, Rizo C, Ci i A, Fe nandez-Luque L. Heal h ecsys:
a seman ic con en -based ecommende sys em o complemen heal h
ideos. BMC Med In o m Decis Mak. 2017;17(1):63.
17. Hidalgo JI, Maqueda E, Risco-Ma ín JL, Cues a-In an e A, Colmena JM,
Nobel J. Glucmodel: a moni o ing and modeling sys em o ch onic
diseases applied o diabe es. J Biomed In o m. 2014;48:183–92.
18. Zhang W, Zou H, Luo L, Liu Q, Wu W, Xiao W. P edic ing po en ial side
e ec s o d ugs by ecommende me hods and ensemble lea ning.
Neu ocompu ing. 2016;173:979–87.
19. Ma lin BM, Adams RJ, Sadasi am R, Hous on TK. Towa ds collabo a i e
il e ing ecommende sys ems o ailo ed heal h communica ions.
Ame ican medical in o ma ics associa ion. In: AMIA annual symposium
p oceedings; 2013. p. 1600.
20. Wiesne M, P ei e D. Heal h ecommende sys ems: concep s, equi emen s,
echnical basics and challenges. In J En i on Res Public Heal h. 2014;11(3):
2580–607.
21. Sezgin E, Ozkan S. A sys ema ic li e a u e e iew on heal h ecommende
sys ems. In: E-Heal h and Bioenginee ing Con e ence (EHB) IEEE; 2013. p. 1–4.
22. Kos ko a P. G and challenges in digi al heal h. F on Public Heal h.
2015;3(134)
23. Fe nandez-Luque L, Ka lsen R, Vognild LK. Challenges and oppo uni ies o
using ecommende sys ems o pe sonalized heal h educa ion. In: MIE;
2009. p. 903–7.
24. McNee SM, Riedl J, Kons an JA. Being accu a e is no enough: how accu acy
me ics ha e hu ecommende sys ems. In: CHI’06 ex ended abs ac s on
human ac o s in compu ing sys ems. ACM; 2006. p. 1097–101.
25. Pu P, Chen L, Hu R. A use -cen ic e alua ion amewo k o ecommende
sys ems. In: P oceedings o he Fi h ACM Con e ence on Recommende
Sys ems ACM; 2011. p. 157–64.
26. U.S. Depa men o Heal h and Human Se ices: The heal h consequences
o smoking: a epo o Su geon gene al. A lan a, GA: US Depa men o
Heal h and Human Se ices, Cen e s o Disease Con ol and P e en ion,
Na ional Cen e o Ch onic Disease P e en ion and Heal h P omo ion,
O ice on Smoking and Heal h. 2004. 62.
27. Su geon Gene al: The Heal h Bene i s o Smoking Cessa ion: A epo o
Su geon Gene al. 1990, DHSS Publica ion No. (CDC) 90–8416: U.S.
Depa men o Heal h and Human Se ices (DHSS), Public heal h Se ice
(PHS), Cen e o Disease Con ol (CDC), Cen e o Ch onic Disease
P e en ion and Heal h P omo ion, O ice on Smoking and Heal h.
28. Te Poel F, Bolman C, Reubsae A, de V ies H. E icacy o a single compu e -
ailo ed e-mail o smoking cessa ion: esul s a e 6 mon hs. Heal h Educ
Res. 2009;24(6):930–40.
29. Shi man S, Pa y J, Rohay J, Di Ma ino M, Gi chell J. The e icacy o
compu e - ailo ed smoking cessa ion ma e ial as a supplemen o nico ine
pa ch he apy. D ug Alcohol Depend. 2001;64(1):35–46.
30. Bo land R, Balm o d J, Hun D. The e ec i eness o pe sonally ailo ed
compu e -gene a ed ad ice le e s o smoking cessa ion. Addic ion.
2004;99(3):369–77.
31. Liao Y, Wu Q, Tang J, Zhang F, Wang X, Qi C, He H, Long J, Kelly BC, Cohen
J. The e icacy o mobile phone-based ex message in e en ions (happy
qui ) o smoking cessa ion in China. BMC Public Heal h. 2016;16(1):833.
32. Vid ine DJ, Fle che FE, Danysh HE, Ma ani S, Vid ine JI, Can o SB, P okho o
AV. A andomized con olled ial o assess he e icacy o an in e ac i e
mobile messaging in e en ion o unde se ed smoke s: p ojec ac ion.
BMC Public Heal h. 2012;12(1):696.
33. Whi ake R, Bo land R, Bullen C, Lin RB, McRobbie H, Rodge s A. Mobile
phone-based in e en ions o smoking cessa ion. Coch ane Da abase Sys
Re . 2009;4:4.
34. F ee C, Knigh R, Rodge s A, Whi ake A, Cai ns J, Edwa ds P, Robe s I. A
andomised con olled ial o mobile phone based smoking cessa ion
suppo : x 2s op. Lance . 2011;378:49–55.
35. Rime BK, K eu e MW. Ad ancing ailo ed heal h communica ion: a pe suasion
and message e ec s pe spec i e. Aus J Commun. 2006;56(s1):184–201.
36. Filion AJ, Da ling on G, Chapu J-P, Yba a M, Haines J. Examining he
in luence o a ex message-based sleep and physical ac i i y in e en ion
among young adul smoke s in he Uni ed S a es. BMC Public Heal h.
2015;15(1):1.
37. Mülle AM, Khoo S, Mo is T. Tex messaging o exe cise p omo ion in olde
adul s om an uppe -middle-income coun y: andomized con olled ial.
J Med In e ne Res. 2016;18:1.
38. Agyapong VI, M klas K, Juhás M, Omeje J, Ohinmaa A, Du sun SM, G eenshaw
AJ. C oss-sec ional su ey e alua ing ex 4mood: mobile heal h p og am o
educe psychological ea men gap in men al heal hca e in Albe a h ough
daily suppo i e ex messages. BMC Psychia y. 2016;16(1):378.
39. Sha pe S, Shephe d M, Kool B, Whi ake R, Nosa V, Do ey E, Galea S, Reid P,
Ame a unga S. De elopmen o a ex message in e en ion aimed a
educing alcohol- ela ed ha m in pa ien s admi ed o hospi al as a esul o
inju y. BMC Public Heal h. 2015;15(1):1.
40. Shi man SM. The obacco wi hd awal synd ome. Ciga e e smoking as a
dependence p ocess. 1979;23:158–84.
41. Elge o A. Psychological and physiological changes du ing obacco-
abs inence in habi ual smoke s. J Clin Psychol. 1978;34(3):759–64.
42. Messe K, T inidad DR, Al-Delaimy WK, Pie ce JP. Smoking cessa ion a es in
he Uni ed S a es: a compa ison o young adul and olde smoke s. Am J
Public Heal h. 2008;98(2):317–22.
43. G imshaw G, S an on A. Tobacco cessa ion in e en ions o young people.
Coch ane Da abase Sys Re . 2006;4
44. Silagy C, Lancas e T, S ead L, Man D, Fowle G. Nico ine eplacemen
he apy o smoking cessa ion. Coch ane Da abase Sys Re . 2004;3:
CD000146.
45. S ead LF, Koilpillai P, Fanshawe TR, Lancas e T. Combined pha maco he apy
and beha iou al in e en ions o smoking cessa ion. Coch ane Da abase
Sys Re . 2016;10(10)
46. Rice VH, S ead LF. Nu sing in e en ions o smoking cessa ion. Coch ane
Da abase Sys Re . 2008;1
47. Wendel S, Dellae BG, Ron el ap A, an T ijp HC. Consume s’in en ion o
use heal h ecommenda ion sys ems o ecei e pe sonalized nu i ion
ad ice. BMC Heal h Se Res 2013;13:1;1.
48. Valdez AC, Zie le M, Ve be K, Fel e nig A, Holzinge A. Recommende
sys ems o heal h in o ma ics: s a e-o - he-a and u u e pe spec i es.
In: Machine Lea ning o Heal h In o ma ics. New Yo k: Sp inge ; 2016.
p. 391–414.
49. O’Dono an J, Smy h B. T us in ecommende sys ems. In: P oceedings o
he 10 h in e na ional con e ence on in elligen use in e aces. San Diego:
ACM; 2005. p. 167–74.
50. Dagge TS, Sweeney JC, Johnson LW. A hie a chical model o heal h se ice
quali y: scale de elopmen and in es iga ion o an in eg a ed model. J Se
Res. 2007;10(2):123–42.
51. Ak e S, D’Amb a J, Ray P. Use pe cei ed se ice quali y o m-heal h
se ices in de eloping coun ies. 18 h Eu opean Con e ence on In o ma ion
Sys ems; 2010. p. 1–12.
52. Ch is ensen H, Mackinnon A. The law o a i ion e isi ed. J Med In e ne
Res. 2006;8(3):20.
53. Alkhaldi G, Hamil on FL, Lau R, Webs e R, Michie S, Mu ay E. The
e ec i eness o echnology-based s a egies o p omo e engagemen wi h
digi al in e en ions: a sys ema ic e iew p o ocol. JMIR Res P o . 2015;4(2):
47.
54. Coupe MP, Alexande GL, Maddy N, Zhang N, Nowak MA, McClu e JB, Cal i
JJ, Rolnick SJ, S opponi MA, Li le RJ, e al. Engagemen and e en ion:
measu ing b ead h and dep h o pa icipan use o an online in e en ion.
J Med In e ne Res. 2010;12(4):52.
55. Donkin L, Ch is ensen H, Naismi h SL, Neal B, Hickie IB, Glozie N. A
sys ema ic e iew o he impac o adhe ence on he e ec i eness o e-
he apies. J Med In e ne Res. 2011;13(3):52.
56. Dijks a A. Wo king mechanisms o compu e - ailo ed heal h educa ion:
e idence om smoking cessa ion. Heal h Educ Res. 2005;20(5):527–39.
57. Zbikowski SM, Jack LM, McClu e JB, Dep ey M, Ja i z HS, McA ee TA, Ca z
SL, Richa ds J, Bush T, Swan GE. U iliza ion o se ices in a andomized ial
es ing phone-and web-based in e en ions o smoking cessa ion. Nico ine
Tob Res. 2011;13(5):319–27.
58. Ho s-F aile S, Benjumea FJN, He nández LC, Ruiz FO, Fe nandez-Luque L.
Design o wo combined heal h ecommende sys ems o ailo ing messages
in a smoking cessa ion app 2016 doi:a Xi p ep in a Xi :1608.07192.
59. Bamidis PD, Pa aske opoulos E, Kons an inidis E, Spachos D, Billis A.
Mul imodal e-heal h se ices o smoking cessa ion and public heal h: he
smoke eeb ain p ojec app oach. In: MedIn o. Ams e dam: IOS p ess; 2017.
Ho s-F aile e al. BMC Public Heal h (2018) 18:698 Page 9 o 10