Electronic 12-Hour Dietary Recall (e-12HR): Comparison of a Mobile Phone App for Dietary Intake Assessment With a Food Frequency Questionnaire and Four Dietary Records
Abstract
Background: One of the greatest challenges for nutritional epidemiology is improving upon traditional self-reporting methods for assessing usual dietary intake. Objective: Evaluating the relative validity of a new method known as the current-day dietary recall (current-day recall), based on a Smartphone app called 12-hour dietary recall (e-12HR), for determining the habitual intake of a series of key food/drink groups using a food frequency questionnaire (FFQ) and four dietary records (DRs) as reference methods. Methods: University students over the age of 18 recorded their consumption of certain groups of food/drink using e-12HR for 28 consecutive days. During this 28-day period, they also completed four DRs on randomly selected days. Once the monitoring period was finished, the subjects then completed a FFQ. A comparison was made between the different methods using the Spearman correlation coefficient (SCC), a cross-classification analysis, and weighted kappa (k). Results: A total of 87 participants completed the study (64% women, 56/87; 36% men, 31/87). e-12HR versus FFQ, for all food/drink groups: The average SCC was .70. Crossclassification analysis showed that the average percentage of individuals classified in the exact agreement category was 51.5%; exact agreement + adjacent was 91.8%; and no participant (0.0%) was classified in the extreme disagreement category. The average was .51. e-12HR versus the four DRs, for all food/drink groups: The average SCC was .63. Crossclassification analysis showed that the average percentage of individuals classified in the exact agreement category was 47.1%; exact agreement + adjacent was 89.2%; and no participant (0.0%) was classified in the extreme disagreement category. The average k was .47. Conclusions: Current-day recall, based on the e-12HR app, is in good agreement with the two reference methods (FFQ & four DRs), demonstrating its potential usefulness for categorizing individuals according to the usual dietary intake of certain food/drink groups.
Full text
O iginal Pape
Elec onic 12-Hou Die a y Recall (e-12HR): Compa ison o a
Mobile Phone App o Die a y In ake Assessmen Wi h a Food
F equency Ques ionnai e and Fou Die a y Reco ds
Luis Ma ía Béja 1, PhD; Ósca Ad ián Reyes2, BMed; Ma ía Dolo es Ga cía-Pe ea3, PhD
1Depa men o P e en i e Medicine and Public Heal h, School o Medicine, Uni e si y o Se ille, Se ille, Spain
2Mu ua Balea , Se ille, Spain
3Vi gen Maca ena Uni e si y Hospi al, Se ille, Spain
Co esponding Au ho :
Luis Ma ía Béja , PhD
Depa men o P e en i e Medicine and Public Heal h
School o Medicine
Uni e si y o Se ille
Ins i u e o Ana omy, 3 d Floo
Sánchez-Pizjuán A enue
Se ille, 41009
Spain
Phone: 34 954551771
Fax: 34 954556481
Email: [email p o ec ed]
Abs ac
Backg ound: One o he g ea es challenges in nu i ional epidemiology is imp o ing upon adi ional sel - epo ing me hods
o he assessmen o habi ual die a y in ake.
Objec i e: The aim o his s udy was o e alua e he ela i e alidi y o a new me hod known as he cu en -day die a y ecall
(o cu en -day ecall), based on a sma phone app called 12-hou die a y ecall, o de e mining he habi ual in ake o a se ies
o key ood and d ink g oups using a ood equency ques ionnai e (FFQ) and ou die a y eco ds as e e ence me hods.
Me hods: Uni e si y s uden s o e he age o 18 yea s eco ded hei consump ion o ce ain g oups o ood and d ink using
12-hou die a y ecall o 28 consecu i e days. Du ing his 28-day pe iod, hey also comple ed ou die a y eco ds on andomly
selec ed days. Once he moni o ing pe iod was o e , subjec s hen comple ed an FFQ. The wo me hods we e compa ed using
he Spea man co ela ion coe icien (SCC), a c oss-classi ica ion analysis, and weigh ed kappa.
Resul s: A o al o 87 pa icipan s comple ed he s udy (64% women, 56/87; 36% men, 31/87). Fo e-12HR e sus FFQ, o
all ood and d ink g oups, he a e age SCC was 0.70. C oss-classi ica ion analysis e ealed ha he a e age pe cen age o
indi iduals classi ied in he exac ag eemen ca ego y was 51.5%; exac ag eemen + adjacen was 91.8%, and no pa icipan
(0%) was classi ied in he ex eme disag eemen ca ego y. The a e age weigh ed kappa was 0.51. Fo e-12HR e sus he ou
die a y eco ds, o all ood and d ink g oups, he a e age SCC was 0.63. C oss-classi ica ion analysis e ealed ha he a e age
pe cen age o indi iduals classi ied in he exac ag eemen ca ego y was 47.1%; exac ag eemen + adjacen was 89.2%; and no
pa icipan (0%) was classi ied in he ex eme disag eemen ca ego y. The a e age weigh ed kappa was 0.47.
Conclusions: Cu en -day ecall, based on he 12-hou die a y ecall app, was ound o be in good ag eemen wi h he wo
e e ence me hods (FFQ & ou die a y eco ds), demons a ing i s po en ial use ulness o ca ego izing indi iduals acco ding o
hei habi ual die a y in ake o ce ain ood and d ink g oups.
(JMIR Mheal h Uheal h 2018;6(6):e10409) doi:10.2196/10409
KEYWORDS
die a y assessmen ; ood equency ques ionnai e; 24-hou die a y ecalls; die a y eco d; mobile phone app
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.1h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
In oduc ion
Many epidemiological in es iga ions and in e en ion s udies
do no equi e a comple e pic u e o he habi ual die (o a e age
long- e m consump ion) [1], as i can ep esen bo h an
unnecessa y wo kload o s udy pa icipan s and an a oidable
was e o he sca ce esou ces a ailable o esea ch [2]. The
essen ial concep o hese s udies is o de e mine habi ual
consump ion o speci ic key g oups o ood and d ink (he ea e
e e ed o as ‘ ood’) [3-9]. Classi ying indi iduals acco ding
o ca ego ies o habi ual consump ion o speci ic ood g oups
is su icien o iden i ying po en ial nu i ional de iciencies
[10] and o e alua ing he ela ionship be ween ela i e anking
and disease [5,8-14], and e ec i eness o pe sonalized me hods
ha a e implemen ed o p omo e changes in die a y pa e ns
ega ding selec ed ood g oups [1,5,8-10,12].
The h ee p incipal and adi ional sel - epo ing me hods o
de e mining die a y in ake can be classi ied as ollows: (1)
Sho - e m me hods: die a y eco d (DR) and 24-hou ecall.
(2) Long- e m me hods: ood equency ques ionnai e (FFQ).
These me hods p esen signi ican and well-documen ed
limi a ions [5,11,15-18]. (1) DRs ha equi e weighing ood
equi e oo much p epa a ion ime and c ea e a la ge wo kload
o he s udy pa icipan s, which can lead o de ia ions in no mal
ood in ake (especially an unde es ima ion o quan i ies), as
well as bo h low pa icipa ion and comple ion a es
[5,11,15-17,19,20]. 24-hou ecalls equi e ained pe sonnel
and depend on he sho - e m memo y o pa icipan s.
Addi ionally, a p ope pic u e o habi ual die using sho - e m
me hods equi es me hods o be epea ed mul iple imes [16,19],
which inc eases p oblems inhe en in hese p ocedu es. (2) FFQs
depend p ima ily on long- e m memo y o he in e iewed
subjec , do no accoun o in ape sonal a ia ion in eco ding
he daily consump ion o ood du ing he s udy pe iod, and do
no allow a p ecise es ima ion o se ing sizes o ood consumed.
Due o hese limi a ions, esul s ob ained by hese inexac
me hods o e bo h he sho - and long- e m could lead o inexac
conclusions and inco ec decisions [18].
Newe al e na i es o de e mining die a y in ake include audio
signal p ocessing, ine ial sensing, image p ocessing,
non-in usi e nea -in a ed scanning, and ges u e ecogni ion
in e acing [21-24]. Some au ho s main ain ha mo e esea ch
is needed o de elop hese and o he ools ha a e mo e objec i e
and p ecise and ha esou ces should be in es ed o his end
[18]. Un il hese al e na i es a e a ailable, digi al echnologies
o sel - epo ing me hods can, and mus , be de eloped and
u ilized [25], as an imp o emen in adi ional sel - epo ing
me hods. This p og ess cons i u es one o he mos impo an
challenges in he ield o nu i ional epidemiology [4,20].
Fo hese easons, he need is e iden o de elop be e me hods
ha can e en ually eplace cu en adi ional sel - epo ing
ones and p o ide be e accu acy in measu ing usual die a y
in ake o ee-li ing indi iduals. This upg ade would ep esen
a g ea boon no only o esea che s [10] bu also o socie y in
gene al, conside ing he c i ical epe cussions om bo h
epidemiological in es iga ions and in e en ional s udies
ega ding he die a y in ake o he popula ion a la ge [25].
The objec i e o his s udy was o de e mine he ela i e alidi y
o he new me hod known as cu en -day die a y ecall (o
cu en -day ecall), based on a sma phone app called 12-hou
die a y ecall (e-12HR—p e iously known as e-Epidemiology
[25]), u ilizing e e ence me hods, such as a semiquan i a i e
FFQ and ou es ima ed DRs, o e i y he compa abili y o
consump ion da a ega ding a lis o key ood g oups in he h ee
me hods.
Me hods
Rec ui men
S uden s o he Schools o Medicine and Pha macy a he
Uni e si y o Se ille (Andalusia, Spain, Sou h o Eu ope)
pa icipa ed in his s udy. Two classes om each School we e
andomly selec ed and one membe o he esea ch eam
p esen ed he p ojec . Pa icipan ec ui men ook place om
Janua y o Decembe 2017. Pa icipan s we e inco po a ed in
he s udy p og essi ely du ing he en i e ec ui men pe iod so
ha e e y day o he week and e e y season o he yea could
be ep esen ed [26]. Pa icipa ion in he s udy was incen i ized
wi h a a le ha included school ma e ials alued a 250 eu os.
S udy Sample
The s udy was p esen ed o 219 s uden s and o hose, 26 we e
no in e es ed, and 98 did no mee he inclusion c i e ia. O he
95 indi iduals who signed he in o med consen , 87 comple ed
he s udy. Inclusion c i e ia: (1) Olde han 18 yea s o age; (2)
a s uden o he Schools o Medicine o Pha macy (Uni e si y
o Se ille); and (3) possesses a sma phone wi h in e ne access
(3G/4G/Wi-Fi) and And oid ope a ing sys em.
Fi s In e iew: Pa icipan Da a Collec ion
The 95 s uden s who we e in e es ed and me he inclusion
c i e ia we e scheduled o a i s pe sonal in e iew. The
p ocedu e o he in e iew was as ollows: (1) he same esea ch
eam membe ha p esen ed he p ojec pe sonally explained
he esea ch p o ocol in de ail; (2) each pa icipan signed he
in o med consen o m; (3) each pa icipan was assigned a
pe sonal alphanume ic code; (4) each pa icipan illed ou an
ini ial ques ionnai e (on pape ) wi h da e o bi h, da e o he
in e iew, gende , and school; (5) each pa icipan downloaded
he e-12HR app o his o he pe sonal sma phone; (6) he same
membe o he esea ch eam pe sonally explained o each
pa icipan how o use he app wi h a p ac ical demons a ion
be o e w i en ins uc ions we e gi en o he pa icipan s [14,27]
o be consul ed la e , i necessa y; (7) he same esea ch eam
membe pe sonally ga e each s uden de ailed ins uc ions on
how o comple e he ou es ima ed DRs and how o es ima e
se ing sizes consumed using an explana o y pamphle ha was
also gi en o pa icipan s [12,13].
The esea che also insis ed ha pa icipan s main ain hei
habi ual die h oughou he s udy.
Comple ing he 12-Hou Die a y Recall App
The e-12HR app was de eloped o eco d daily consump ion
o a lis o 10 ood g oups. The lis could no be oo long o
minimize he wo kload on pa icipan s as well as he esea ch
cos s [28]. These ood g oups we e selec ed as hey a e
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.2h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
indica o s o heal h/disease and a e conside ed p o ec i e ac o s
( ui , ege ables, legumes o ish) o isk ac o s (so d inks,
comme cial baked oods, and p ecooked meals) o ch onic
illnesses [4,12,29]. They also p o ide consump ion pa e ns ha
ange om almos e e y day o e e y inhabi an o he
popula ion o in equen ly o he majo i y [4] (Tex box 1).
The i s ime he e-12HR app was used, he pa icipan s we e
equi ed o in oduce hei pe sonally assigned alphanume ic
code and he e-mail o he esea che who would ecei e he
da a om he app. Pa icipan s we e ins uc ed o comple e he
app a e consuming he las ood o he day [14,27]. The
e-12HR app includes a no i ica ion op ion, so ha he app can
emind he pa icipan , a he hou p e iously selec ed by he
pa icipan (be ween 20:00 h and 04:00 h), ha i is ime o
comple e he ques ionnai e. The app was s uc u ed acco ding
o ood g oups o acili a e comple ion o he e-12HR ask. Fo
each ood g oup, he pa icipan would choose he mos
app op ia e image (o images) om a se ies o colo pho og aphs
wi h 2-4 possible op ions, shown simul aneously [14,27], which
illus a ed he di e en se ing sizes o assis wi h selec ing he
numbe o s anda d se ings consumed [10,12,14,20,27]. To
u he assis wi h es ima ing se ing sizes, each pho og aph
was accompanied by explana o y ex and h ee objec s o
known/p edic able size o he s uden s [30,31] ( iducial
ma ke s): a commonly used pencil, pen, and a ma ke . Fo
example, on he i s sc een o he app, he ollowing would
appea : How many pieces o ui ha e you ea en oday?, wi h
he Ra ions bu on, and Nex bu on and a box wi h he alue
se o 0by de aul . Supposing ha he pa icipan had,
h oughou he day, consumed an apple and h ee manda ins,
hey would p oceed as ollows: (1) ap he Ra ions bu on; (2)
a new window opens wi h di e en pho og aphs o ui , an
Accep bu on, and a Cancel bu on; (3) ap once on he pho o
co esponding o he apple and ap h ee imes on he pho og aph
co esponding o he manda in; (4) ap he Accep bu on; (5)
he app e u ns o he p e ious window bu he box now has,
ins ead o he 0 alue, he co esponding numbe o s anda d
a ions o he ui selec ed wi h he pho og aphs, in his
example, a alue o 2.5 (1 × 1 s anda d a ion + 3 × 0.5 s anda d
a ion); (6) ap he Nex bu on o access he nex ood g oup
and p oceed as be o e; (7) i an e o occu s, he pa icipan can
ap he Cancel bu on ins ead o Accep in s ep 4, s a ing he
p ocess o e again (Figu e 1).
The use o ac i e images (accompanied by explana o y ex and
h ee e e ence objec s o known/p edic able size o he s uden s
[ iducial ma ke s]) is he only modi ica ion o e-12HR ega ding
he p e ious e sion (known as e-Epidemiology). Howe e , his
single, appa en ly simple, modi ica ion o he app is ac ually an
impo an e olu ion o e he p e ious e sion. In ac , he
inclusion o ac i e images was designed wi h a double pu pose:
on he one hand, o acili a e comple ion o e-12HR (by apping
he images ins ead o di ec ly in oducing he numbe o s anda d
se ings consumed); and on he o he o assis wi h es ima ing
he quan i y consumed ( his new e sion o he app di ec ly
shows 2-4 possible op ions o se ing sizes).
A e comple ing he daily ques ionnai e wi h e-12HR, he
in o ma ion is au oma ically sa ed and sen , ia 3G/4G/Wi-Fi,
o he e-mail add ess o he esea ch adminis a o (en e ed when
i s accessing he app). Once he ques ionnai e is comple ed
and sen , he pa icipan canno change hei esponses o access
he app un il he ollowing day.
The consump ion eco d o he selec ed ood g oups on he app
was pe o med o 28 consecu i e days. The ime in e al
selec ed is simila o o he compa ison/ alida ion s udies
[8,12,32-34].
The ques ionnai e and he size o he a ions used in e-12HR
a e based on a semiquan i a i e FFQ p e iously alida ed o
he popula ion o Spain [35].
Tex box 1. Ques ionnai e used in 12-hou die a y ecall (e-12HR).
1. How many pieces o ui ha e you ea en oday?
2. How many po ions o ege ables ha e you ea en oday?
3. How many po ions o legumes (len ils, ga banzos, beans, e c) ha e you ea en oday?
4. How many po ions o chicken/ u key ha e you ea en oday?
5. How many po ions o ish ha e you ea en oday?
6. How many po ions o ed mea (bee , po k, lamb) ha e you ea en oday?
7. How many se ings o so d inks ha e you had oday?
8. How many po ions o comme cially p oduced swee s (no home-made) (cookies/pas ies) ha e you ea en oday?
9. How many po ions o p epa ed/ ozen oods ha e you ea en oday (c oque es, pizza, e c)?
10. How many se ings o bee ha e you consumed oday?
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.3h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
Figu e 1. Sc een cap u e o he 12-hou die a y ecall (e-12HR).
Comple ing he Fou Es ima ed Die a y Reco ds
Du ing he 28-day pe iod ha e-12HR was in use o each
pa icipan , ou es ima ed DRs (on pape ) we e scheduled on
andomly assigned, non-consecu i e days [9,32]: h ee days
du ing he weekdays and one weekend day [9,12,13,32]. The
choice o be ween h ee and se en DRs is no mally conside ed
su icien o e alua e ood g oup in ake [9,36]. Fou es ima ed
DRs we e chosen ins ead o weighed DRs o logis ical easons
[9,12]. In he daily li e o he pa icipan s ha made up he
sample (uni e si y s uden s who spend a la ge pa o hei days
ou side he home a hei cen e o s udy), weighing ood is no
easible and ep esen s a g ea wo kload when compa ed wi h
es ima ing se ings consumed.
Each pa icipan , du ing he i s in e iew, ecei ed an
explana ion o how o use he es ima ed DRs and how o
es ima e he se ing size consumed [12,13], using a pamphle
wi h a se ies o 2-4 colo pho og aphs [10,13,14,27] (one se ies
o each ood g oup). To assis wi h es ima ing se ing sizes,
each pho og aph was accompanied by explana o y ex and h ee
e e ence objec s o known/p edic able size o he s uden s
[30,31] ( iducial ma ke s): a commonly used pencil, pen, and
a ma ke .
The DRs used we e based on a DR p e iously alida ed o
ano he Eu opean coun y [13], bu s uc u ed acco ding o he
ypical Spanish die (b eak as , lunch, an a e noon snack, and
dinne ), and p e-codi ied wi h a column ha included he same
10 ood g oups selec ed o he e-12HR. The se ing sizes we e
based on a semiquan i a i e FFQ p e iously alida ed o he
Spanish popula ion [35].
Pa icipan s we e old ha hey mus eco d he consump ion
da a on a sepa a e page o each day [7], and immedia ely a e
consuming he ood [7,13].
Second In e iew: Comple ing he Semiquan i a i e
Food F equency Ques ionnai e
A he end o he e-12HR da a collec ion pe iod, each pa icipan
was scheduled o a second pe sonal in e iew a hei
con enience. In his in e iew, he pa icipan was equi ed o
ill ou a s uc u ed, semiquan i a i e FFQ (on pape ) ha
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.4h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
included he same 10 ood g oups selec ed o he e-12HR. The
same esea ch eam membe who pe o med he i s in e iew
hen explained o each pa icipan , in his second in e iew, he
p ocess o comple ing he semiquan i a i e FFQ, and p o ided
hem wi h an explana o y pamphle o es ima e wha was
conside ed a s anda d se ing o each ood g oup. This
pamphle con ained a pho og aph o a s anda d se ing o each
ood g oup along wi h explana o y ex and h ee e e ence
objec s o a known/p edic able size o s uden s [30,31] ( iducial
ma ke s): a commonly used pencil, pen, and ma ke . The ime
pe iod conside ed by he FFQ co esponded o he 28 days o
he app. All pa icipan s comple ed he semiquan i a i e FFQ
wi hin he i s week o inishing he e-12HR app wi h he
excep ion o wo pa icipan s, who comple ed he FFQ 8 and
11 days la e .
The semiquan i a i e FFQ, as well as s anda d se ing sizes,
was based on a semiquan i a i e FFQ p e iously alida ed o
he Spanish popula ion [35].
Finally, in he second pe sonal in e iew, pa icipan s we e
asked how much ime, on a e age, was needed o comple e he
daily ques ionnai e on he app. Pa icipan s could choose
be ween a ious op ions. O he o al pa icipan s, 11% (10/87)
selec ed he op ion less han 1 minu e pe day; 36% (31/87)
selec ed he op ion app oxima ely 1 minu e pe day; 28% (24/87)
chose app oxima ely 2 minu es pe day; 17% (15/87) chose
app oxima ely 3 minu es pe day; 7% (6/87) chose
app oxima ely 4 minu es pe day; and 1% (1/87) chose he 5
minu es pe day o mo e op ion. Almos hal o he pa icipan s,
47% (41/87) indica ed ha he ask ook 1 minu e o less pe
day o comple e and 75% (65/87) s a ed ha i ook 2 minu es
o less. Fo his eason, he esea ch eam conside s ha he ime
necessa y o comple e he app is, no mally, 2 minu es pe day
o less.
Legal Conside a ions
The s udy was ca ied ou acco ding o he ules es ablished by
he Decla a ion o Helsinki and he Law on Biomedical Resea ch
[37].
All o he pe sonal in o ma ion collec ed in his s udy was
anonymized in acco dance wi h cu en Spanish legisla ion [38].
To achie e his, each pa icipan was assigned a pe sonal
alphanume ic code ha hey had o en e in o he e-12HR app
(only when i s accessing he app) and include on he di e en
pape ques ionnai es ha we e p o ided: he ini ial ques ionnai e
(pe sonal in o ma ion), on he ou es ima ed DRs, and on he
semiquan i a i e FFQ. The pe sonal alphanume ic code was
used o ma ch all o he da a pe aining o he pa icipan while
a he same ime espec ing hei anonymi y.
Da a Con e sion
Using he e-12HR app, each pa icipan eco ded he numbe
o s anda d se ing sizes consumed daily o each ood g oup
h oughou he 28-day s udy pe iod. Wi h he ou es ima ed
DRs, each pa icipan collec ed he numbe o s anda d se ing
sizes consumed daily o each ood g oup on ou di e en days
h oughou he 28-day moni o ing pe iod. On he
semiquan i a i e FFQ, each pa icipan selec ed he numbe o
s anda d se ing sizes habi ually consumed o each ood g oup
h oughou he 28-day moni o ing pe iod.
Fo each pa icipan , he da a om he e-12HR app, he ou
DRs and he FFQ had o be exp essed in he same ca ego ies o
habi ual consump ion o make compa isons: Less han once a
week; Once o wice a week, 3-4 imes a week, 5-6 imes a week;
Once o wice a day, and 3 o mo e imes a day. On he FFQ,
hese di e en op ions o habi ual consump ion we e al eady
a ailable o he pa icipan s o choose om and, as such, he
FFQ da a was no modi ied. Rega ding he e-12HR app, he
da a needed o be ans o med. As an example, one pa icipan
egis e ed an a e age daily consump ion o 0.52 s anda d
se ings o legumes o e 28 days using he app. This a e age
consump ion ep esen s 3.64 s anda d se ings pe week (0.52
× 7 = 3.64), which would be classi ied as 3–4 imes pe week
[25]. As o he ou DRs, he in o ma ion hey con ained also
needed o be con e ed [9]. As an example, one pa icipan
eco ded consuming 0, 0.3, and 1 s anda d piece o ui on he
DRs du ing he weekdays and 1.5 s anda d pieces o ui on
he DR comple ed a he weekend. This consump ion ep esen s
an a e age daily consump ion du ing weekdays o : (0 + 0.3 +
1)/3 = 0.43 s anda d pieces. As o weekly consump ion, he
con e sion was as ollows: (0.43 × 5 weekdays) + (1.5 × 2
weekend days) = 2.17 + 3 = 5.17 s anda d pieces, which would
hen be classi ied as 5-6 imes pe week.
To enable making compa isons, he h ee ools egis e ed he
consump ion o he same ood g oups, used he same s anda d
se ings as a e e ence and he in ake eco d co esponded o
he same ime pe iod, o a oid possible a ia ions in he speci ic
die s du ing di e en pe iods [20,32,39,40].
S a is ical Analysis
In his s udy, when compa ing e-12HR wi h he FFQ and he
ou DRs o each o he selec ed ood g oups, he associa ion
be ween he ca ego ies o habi ual consump ion was e alua ed
using he Spea man co ela ion coe icien (SCC) whe eas he
deg ee o ag eemen be ween he ca ego ies o habi ual
consump ion was e alua ed using c oss-classi ica ion analysis
and he weigh ed kappa index [1,41]. Fo he c oss-classi ica ion
analysis, he pe cen age o pa icipan s classi ied in he same
g oup was labeled exac ag eemen ; in he same ca ego y o
adjacen ca ego ies, exac ag eemen + adjacen ; o in opposi e
ca ego ies, ex eme disag eemen , using di e en me hods. Fo
weigh ed kappa, pa ial alues we e assigned using he s a is ical
p og am S a a: om 1.00 o an exac ag eemen o 0.00 o
ex eme disag eemen .
The compa ison c i e ia conside ed in his s udy we e: a e age
SCC≥0.5 [1,32]; a e age c oss-classi ica ion pe cen age in he
exac ag eemen ca ego y ≥50% [1], in he exac ag eemen +
adjacen ≥75% [32], and in he ex eme disag eemen ca ego y
<10%; wi h an a e age κ≥0.41 [1].
The s a is ical analysis was pe o med u ilizing STATA MP
13.1 (S a a Co p LP, College S a ion, Texas, USA), and a P
alue < .05 was conside ed s a is ically signi ican [42].
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.5h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
Resul s
O e iew
O he 95 pa icipan s who signed he in o med consen , 8 did
no comple e he s udy. Resul s o hese indi iduals we e no
included in la e s a is ical analyses. Pa icipan s who did no
comple e he s udy we e hose who pe o med he ask ewe
han 22 days, did no comple e he FFQ and did no comple e
he ou DRs.
O he 87 pa icipan s who comple ed he s udy, 78% (68/87)
comple ed he ask e e y day (28 days o moni o ing), 94%
(82/87) o he pa icipan s comple ed he app a leas 24 days,
whe eas he emaining 6% (5/87) comple ed he app o 22 days
(Table 1).
The a e age age o pa icipan s was 19.2 yea s old; 64% (56/87)
we e women, and 36% we e men (31/87), and 67% (58/87)
we e medical s uden s and 33% (29/87) we e pha macy s uden s
(Table 1). No s a is ically signi ican di e ences in he a iables
s udied we e ound among pa icipan s who comple ed he s udy
and hose who did no .
12-Hou Die a y Recall Ve sus Food F equency
Ques ionnai e
Fo all ood g oups: he a e age SCC was .70 (pe ood g oup,
om .50 [legumes] o .86 [bee ]; Table 2). C oss-classi ica ion
analysis showed ha he a e age pe cen age o indi iduals
classi ied in he exac ag eemen ca ego y was 51.5% (pe ood
g oup, om 37.9% [chicken/ u key] o 70.1% [bee ]); exac
ag eemen + adjacen was 91.8% (pe ood g oup, om 85.1%
[swee s] o 96.6% [legumes & p epa ed oods]; Table 3); and
no pa icipan s (0%) we e classi ied in he ex eme disag eemen
ca ego y. The a e age weigh ed kappa was .51 (pe ood g oup,
om .34 [chicken/ u key] o .66 [ ui & bee ]; Table 4).
12-Hou Die a y Recall Ve sus he Fou Die a y
Reco ds
Fo all ood g oups: he a e age SCC was 0.63 (pe ood g oup
om 0.46 [legumes] o 0.83 [ ui ]; Table 2). C oss-classi ica ion
analysis showed ha he a e age pe cen age o indi iduals
classi ied in he exac ag eemen ca ego y was 47.1% (pe ood
g oup, om 31.0% [chicken/ u key] o 66.7% [bee ]); exac
ag eemen + adjacen was 89.2% (pe ood g oup, om 83.9%
[so d inks] o 95.4% [legumes]; Table 3); and no pa icipan s
(0%) we e classi ied in he ca ego y o ex eme disag eemen .
The a e age weigh ed kappa was 0.47 (pe ood g oup, om
0.26 [ ish] o 0.72 [ ui ]; Table 4).
Table 1. Cha ac e is ics o s udy pa icipan s.
95% CIMean (SD)n (%)Cha ac e is ics
––a
87Pa icipan s who comple ed he s udy
Numbe o days comple ed he app
––68 (78.2)28
––5 (5.7)27
––3 (3.4)26
––5 (5.7)25
––1 (1.1)24
––5 (5.7)22
–19.2 (3.3)–Age (yea s)
Gende
74.4-53.4–56 (64.4)Female
25.6-46.6–31 (35.6)Male
Facul y
55.7-76.4–58 (66.7)Medicine
44.3-23.6–29 (33.3)Pha macy
aNo applicable.
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.6h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
Table 2. Spea man co ela ion coe icien (SCC) de i ed om 12-hou die a y ecall (e-12HR) e sus he ood equency ques ionnai e (FFQ) and
om e-12HR e sus he ou die a y eco ds (DRs).
e-12HR s DRse-12HR s FFQCompa ison
P alue
SCC
P alue
SCC
<.0010.83<.0010.84F ui
<.0010.72<.0010.80Vege ables
<.0010.46<.0010.50Legumes
<.0010.60<.0010.53Chicken/ u key
<.0010.47<.0010.65Fish
<.0010.50<.0010.69Red mea
<.0010.72<.0010.81So d inks
<.0010.74<.0010.71Swee s
<.0010.60<.0010.61P epa ed oods
<.0010.70<.0010.86Bee
N/A0.63N/Aa
0.70A e age
aN/A: no applicable.
Table 3. C oss-classi ica ion analysis de i ed om elec onic 12-hou die a y ecall (e-12HR) e sus he ood equency ques ionnai e (FFQ) and om
e-12HR e sus he ou die a y eco ds (DRs).
e-12HR s DRe-12HR s FFQCompa ison
Exac ag eemen + adjacen (%)Exac ag eemen (%)Exac ag eemen + adjacen b(%)Exac ag eemen a(%)
93.165.587.454.0F ui
90.839.192.049.4Vege ables
95.455.296.656.3Legumes
87.431.086.237.9Chicken/ u key
88.532.290.848.3Fish
88.539.195.458.6Red mea
83.957.595.452.9So d inks
89.640.285.140.2Swee s
88.544.896.647.1P epa ed oods
86.266.792.070.1Bee
89.247.191.851.5A e age
aExac ag eemen : % o cases c oss-classi ied in o he same ca ego y.
bExac ag eemen + adjacen : % o cases c oss-classi ied in o he same o adjacen ca ego y.
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.7h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
Table 4. Weigh ed kappa de i ed om elec onic 12-hou die a y ecall (e-12HR) e sus he ood equency ques ionnai e (FFQ) and om e-12HR
e sus he ou die a y eco ds (DRs).
e-12HR s DRse-12HR s FFQCompa ison
P alue
Kappa
P alue
Kappa
<.0010.72<.0010.66F ui
<.0010.51<.0010.61Vege ables
<.0010.34<.0010.38Legumes
<.0010.36<.0010.34Chicken/ u key
<.0010.26<.0010.41Fish
<.0010.34<.0010.54Red mea
<.0010.60<.0010.59So d inks
<.0010.49<.0010.48Swee s
<.0010.44<.0010.41P epa ed oods
<.0010.63<.0010.66Bee
N/A0.47N/Aa
0.51A e age
aN/A: no applicable.
Discussion
O e iew
Cu en -day ecall is a modi ied 24-hou ecall ocused on a
se ies o ood g oups and comple ed a he end o e e y day
du ing he moni o ing pe iod. This me hod has been designed
o classi y indi iduals acco ding o ca ego ies o habi ual
consump ion o selec ed ood g oups and is no in ended o be
used o de e mine o al ood consump ion o an indi idual [25].
This s udy is he i s ime ha cu en -day ecall, based on he
e-12HR app, has been compa ed wi h wo di e en e e ence
models, one long- e m ( he FFQ) and he o he sho - e m ( ou
DRs) o de e mine he u ili y o his new me hod as a ool o
es ima ing he usual die a y in ake o a se ies o ood g oups.
Re e ence Me hods
E alua ing he ue alidi y o a me hod equi es measu ing,
wi h a high deg ee o accu acy, he habi ual die o ee-li ing
indi iduals du ing a p olonged pe iod o ime, which is no
easible [1]. As a esul , he esea che s o his s udy ha e
e alua ed he ela i e alidi y o he me hod by compa ing i
wi h an al e na i e me hod o die a y assessmen , wi h i s own
limi a ions [1]. Gi en ha none o he compa ison me hods a e
pe ec [5,8,39,43], as all a e subjec i e, i is o c ucial
impo ance o hei e o s o be as independen as possible o
a oid alsely a ibu ing alidi y o he me hod in ques ion
[5,6,8,16,20,34,44]. Fo his eason, he esea ch eam
conside ed he FFQ and DRs o be possible e e ence me hods
and no 24-hou ecalls (cu en -day ecall is a modi ied 24-hou
ecall). As nei he he FFQ no he DRs can be conside ed
de ini i e e e ence me hods, he esea ch eam selec ed bo h
me hods o his alida ion s udy. Awa e ha he quan i y o
wo k o pa icipan s could be excessi e, a semiquan i a i e
FFQ was selec ed ( o be comple ed a he end o he e-12HR
moni o ing pe iod) and ou DRs (es ima ed).
P incipal Resul s
12-Hou Die a y Recall Ve sus he Food F equency
Ques ionnai e
Fo all ood g oups: he a e age SCC was 0.70 (in 10/10 ood
g oups SCC≥0.50 [1,32]). C oss-classi ica ion analysis showed
ha he a e age pe cen age o indi iduals classi ied in he
ca ego y o exac ag eemen was 51.5% (in 5/10 ood g oups
≥50% [1]); exac ag eemen + adjacen was 91.8% (10/10 ood
g oups ≥75% [32]); and no pa icipan s (0%) we e classi ied in
he ca ego y o ex eme disag eemen [1]. The a e age weigh ed
kappa was .51 (in 8/10 ood g oups κ≥0.41 [1]). These alues
indica e good ag eemen be ween he wo me hods [1,32]).
Values ob ained we e e y simila o hose om p e ious s udies
wi h he olde e sion o he e-12HR app, which did no use
pho og aphs o acili a e es ima ion o se ings consumed [45].
12-Hou Die a y Recall Ve sus he Fou Die a y Reco ds
Fo all ood g oups: he a e age SCC was 0.63 (in 8/10 ood
g oups SCC≥0.50 [1,32]). C oss-classi ica ion analysis showed
ha he a e age pe cen age o indi iduals classi ied in he
ca ego y o exac ag eemen was 47.1% (in 4/10 ood g oups
≥50% [1]); exac ag eemen + adjacen was 89.2% (10/10 ood
g oups ≥75% [32]); and no pa icipan s (0%) we e classi ied in
he ex eme disag eemen ca ego y [1]. The a e age weigh ed
kappa was 0.47 (in 6/10 ood g oups κ≥0.41 [1]). These alues
indica e good ag eemen be ween he wo me hods [1,32]). This
is he i s ime ha he e-12HR app has been compa ed wi h
he DRs and as such he e a e no p io esul s o e e ence.
I is impo an o no e ha he c oss-classi ica ion analysis and
k depend on he numbe o ca ego ies used [1]. Wi h he goal
o educing his dependence when e alua ing he deg ee o
ag eemen be ween di e en me hods, he six o iginal ca ego ies
could ha e been eo ganized in o h ee [1,8,25], ou [32,33]
o i e [46,47], as o he au ho s ha e done. No wi hs anding,
his esea ch eam p e e ed o main ain six ca ego ies o he
s a is ical analysis [45] as a g ea e numbe o ca ego ies o
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.8h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X
habi ual consump ion p o ides compac in o ma ion on he
abili y o me hods o assign indi iduals acco ding o ag eemen
wi h he dis ibu ion o die a y in ake [46]. Compa ison c i e ia
applied o his s udy we e also conside ed by Masson e al [1]
and Fo s e e al [32]. Howe e , conside ing ha hese au ho s
only used h ee and ou ca ego ies o habi ual consump ion,
espec i ely, he demands o his s udy a e much mo e
es ic i e. In his s udy, he classi ica ion o he exac ag eemen
ca ego ies su e ed especially, whe eas he c i e ia used by
Mason e al (exac ag eemen ≥50%) was de ined by h ee
ca ego ies e sus he six conside ed he e.
The esul s demons a e good ag eemen be ween e-12HR and
bo h e e ence me hods (FFQ and DRs). Ne e heless, ce ain
disag eemen was obse ed wi h he app. Rega ding he FFQ:
he a e age c oss-classi ica ion analysis showed ha 8.2% o
pa icipan s we e classi ied a a dis ance o 2–4 ca ego ies. As
o he DRs: he a e age c oss-classi ica ion analysis showed
ha 10.8% o pa icipan s we e classi ied a a dis ance o 2-4
ca ego ies. The h ee me hods (e-12HR, FFQ, and DRs) a e all
based on sel -adminis e ed and anonymized ques ionnai es and
use he same colo pho og aphs o di ec ly es ima ing se ings.
Because hese ques ionnai es we e bo h sel -adminis e ed and
anonymized (wi h he use o alphanume ic codes), hey enabled
esea che s o a oid in o ma ion biases, such as obsequiousness
bias and esea che bias in all h ee me hods. Using he same
pho og aphs also a o s he same p ecision in he h ee me hods
when es ima ing se ings. In his sense, each pa icipan se es
as hei own con ol, he eby minimizing di e ences when
conside ing se ings consumed. In any case, consump ion
equency seems o ha e a g ea e impac on die a y in ake han
on se ing size [5,9].
Howe e , di e ences be ween he h ee me hods a e ac o s
ha could po en ially con ibu e o he disc epancies obse ed.
The undamen al di e ences be ween he me hods a e cen e ed
on bo h he egula i y in collec ing da a on die a y in ake and
he o ma used in he ques ionnai es.
Regula i y in Collec ing Da a on Die a y In ake
Wi h DRs, he collec ion o da a was pe o med a e consuming
each ood g oup h oughou he ou days selec ed; wi h he
e-12HR app, da a collec ion akes place a he end o each day
du ing he 28-day moni o ing pe iod; in he FFQ, da a was
collec ed a he end o he moni o ing pe iod. These di e ences
in egula i y mean ha he accu acy o he in o ma ion depends
on bo h he memo y o he pa icipan and he daily a ia ion
in die a y in ake among ee-li ing indi iduals.
Wi h DRs, dependance on memo y is no an issue [12]; he
e-12HR app is sho - e m memo y dependen ; while FFQs
depend on long- e m memo y.
The e-12HR does no in e e e wi h daily a ia ion as i is
comple ed e e y day; howe e , an e ec is seen on he FFQ,
wi h only one poin o da a collec ed, and he DRs, whe e da a
is collec ed spo adically (only 4/28 days du ing he moni o ing
pe iod).
Fo ma Used in Ques ionnai es
The app is digi al and bo h he FFQ and DRs a e comple ed on
pape . Pape o ma s a e ypically associa ed wi h e o s such
as unanswe ed ques ions, ques ions wi h mul iple esponses
[10] (FFQ), and no egis e ing he quan i y consumed o some
o he di e en ood g oups selec ed (DRs) [48]. In addi ion,
he in o ma ion on pape ques ionnai es mus be con e ed
subsequen ly o digi al o ma o analysis, which inc eases bo h
he wo kload and cos s o he esea ch eam [17,40]. Online
FFQs and DRs, o e en sma phone apps, o e di ec solu ions
o hese limi a ions [10,13,20,30-32,40,49-53], by educing
pape was e and mailing cos s, and emo ing he need o secu e
s o age space and o ganiza ion inhe en wi h pape documen s
[50]. Addi ionally, online adminis a i e me hods can be used
o gain access o speci ic popula ion g oups ha would o he wise
be di icul o s udy due o geog aphy, o example, as hey can
be used emo ely [32]. Despi e he po en ial ad an ages o
u ilizing FFQs and DRs in digi al o ma , in he end i was
decided o use pape o ma s in his s udy. The esea ch eam
ook in o accoun ha e idence shows ha da a collec ed om
sma phone apps and bo h Web-based FFQs and DRs a e
compa able wi h da a om pape o ma s
[7,14,16,27,29,31-34,40,46,54]. The eam also conside ed he
cha ac e is ics o his s udy and he po en ial disad an ages o
de eloping FFQs and DRs in a digi al o ma ha could
ou weigh he possible bene i s. In ac , in his s udy, he
pape -based FFQ and he DRs we e e y sho and simple ( hey
only con ain 10 ood g oups), and he sample popula ion is made
up o s uden s a he Schools o Medicine and Pha macy a he
Uni e si y o Se ille, which was easily accessible o he
esea ch eam.
The simplici y o he pape -based FFQ and DRs minimized
possible e o s, he amoun o pape used, p oblems wi h s o age
space, and cos s associa ed wi h da a con e sion. These cos s
we e minimal when compa ed wi h he po en ial cos s o
de eloping an online o sma phone-based FFQ and DRs.
Easy access o he sample made i possible o comple e he pape
FFQ in pe son, wi hou he need o esea che s o pa icipan s
o a el o pay mailing cos s [25].
S eng hs and Limi a ions
The s eng hs o his s udy include ha he e-12HR is a ool ha
depends only on sho - e m memo y ( he app is comple ed a
he end o each day), i does no in e e e wi h a iabili y o
die a y in ake in daily li e ( he app is comple ed daily). In
addi ion, i has a digi al o ma (young people and adolescen s
ha e shown hei p e e ence o me hods ha use digi al
echnology e sus adi ional me hods [31,53]—imp o ing
mo i a ion and comple ion a es [48]). These cha ac e is ics
suppo g ea e accu acy in assigning he ca ego y o habi ual
consump ion in hose cases whe e he anking assigned by
cu en -day ecall is di e en om ha assigned by e e ence
me hods. Also, wi h he e-12HR app, da a collec ion is
pe o med digi ally, elimina ing he need o in es iga o s o
en e he da a manually; i is a sel - epo ing ool, no equi ing
in e iewe s; i is bo h simple and in ui i e o use, wi h
pho og aphs o assis wi h es ima ing se ings consumed; and
o e all esea ch cos s a e g ea ly educed.
JMIR Mheal h Uheal h 2018 | ol. 6 | iss. 6 | e10409 | p.9h p://mheal h.jmi .o g/2018/6/e10409/
(page numbe no o ci a ion pu poses)
Béja e alJMIR MHEALTH AND UHEALTH
XSL
•
FO
Rende X