Development of theory-based health messages: three-phase programme of formative research

Development of theory-based health messages:three-phase programme of formative research

TRACY EPTON1*, PAUL NORMAN1, PETER HARRIS2, THOMAS WEBB1,F. ALEXANDRA SNOWSILL1 and PASCHAL SHEERAN3

1Department of Psychology, University of Sheffield, Western Bank, Sheffield S10 2TP, UK 2Schoolof Psychology, University of Sussex, Falmer BN1 9QH, UK and 3Psychology Department, Universityof North Carolina, 323 Davie Hall, Chapel Hill, NC 27599-3270, USA*Corresponding author. E-mail: [email protected]

SUMMARY

Online health behaviour interventions have great potentialbut their effectiveness may be hindered by a lack of forma-tive and theoretical work. This paper describes the processof formative research to develop theoretically and empiric-ally based health messages that are culturally relevant andcan be used in an online intervention to promote healthylifestyle behaviours among new university students.Drawing on the Theory of Planned Behaviour, a three-phase programme of formative research was conductedwith prospective and current undergraduate students toidentify (i) modal salient beliefs (the most commonly heldbeliefs) about fruit and vegetable intake, physical activity,binge drinking and smoking, (ii) which beliefs predicted

intentions/behaviour and (iii) reasons underlying each ofthe beliefs that could be targeted in health messages. Phase1, conducted with 96 pre-university college students, eli-cited 56 beliefs about the behaviours. Phase 2, conductedwith 3026 incoming university students, identified 32 ofthese beliefs that predicted intentions/behaviour. Phase 3,conducted with 627 current university students, elicited 102reasons underlying the 32 beliefs to be used to constructhealth messages to bolster or challenge these beliefs. Thethree-phase programme of formative research providesresearchers with an example of how to develop health mes-sages with a strong theoretical- and empirical base for usein health behaviour change interventions.

Key words: health; theory; intervention; theory of planned behaviour

INTRODUCTION

Too few young people engage in health beha-viours that are known to reduce the risk of devel-oping serious diseases and conditions includingcancer, heart and circulatory disease, obesity andtype 2 diabetes (Parliamentary Office of Scienceand Technology, 2007; National Centre forSocial Research, 2008). For example, a recent na-tional survey in England found that only 20% ofyoung people (aged 16–24) eat five portions offruit and vegetables daily, less than 50% meetweekly physical activity guidelines, 40% exceeddaily recommended alcohol limits and 25%

smoke (National Centre for Social Research,2008). Interventions are therefore needed topromote healthy eating, physical activity andsensible alcohol consumption and to discouragesmoking in young people.

Medical Research Council guidelines for thedevelopment of complex interventions empha-size the importance of conducting formative andtheoretical work before proceeding to rando-mised controlled trials (Craig et al., 2008).Formative research ensures the development of‘targeted, culturally appropriate, health messagesthat work’ [(Witte et al., 2001), p. 49]. The use oftheory in the development of health behaviour

Health Promotion International

doi:10.1093/heapro/dau005

# The Author 2014. Published by Oxford University Press. All rights reserved.

For Permissions, please email: [email protected]

of 13

Health Promotion International Advance Access published February 5, 2014 at U

niversity of Sheffield on February 13, 2014http://heapro.oxfordjournals.org/

Dow

nloaded from

http://heapro.oxfordjournals.org/


interventions ensures that appropriate targetsfor intervention are identified (Michie andPrestwich, 2010). Moreover, theory-based healthbehaviour interventions have been found to bemore efficacious (Noar et al., 2007; Webb et al.,2010). With the importance of theory-driven, for-mative research in mind, this paper reports athree-phase programme of research, based onthe Theory of Planned Behaviour (TPB) (Ajzen,1988), to develop the content of messagesdesigned to promote four key health behaviours.

The TPB is one of the most widely appliedmodels of health behaviour and provides clearguidelines for the identification of beliefs totarget in health behaviour change interventions(Conner and Norman, 2005). According to theTPB, the proximal determinant of behaviour isintention which, in turn, is determined by (i) atti-tude (i.e. positive or negative evaluations of per-forming the behaviour), (ii) subjective norm (i.e.perceived social pressure to perform or notperform the behaviour) and (iii) perceived be-havioural control (Ajzen, 1988). The TPB hasbeen found to be predictive of health behaviour;for example, a recent meta-analysis (McEachanet al., 2011) found that the TPB explained largeamounts of variance in students’ intentions in re-lation to dietary behaviours (50.3%), physical ac-tivity (52.1%) and abstinence from smoking andbinge drinking (43.6%). The TPB was also foundto explain significant proportions of variance instudents’ prospective health behaviour for diet(21.2%), physical activity (29.7%) and abstin-ence (15.5%). Overall, these findings suggestthat the TPB provides a strong theoretical basisfor health behaviour interventions.

Health behaviour interventions are often deliv-ered as health messages that typically comprise anumber of techniques to change behaviour, suchas providing information about health conse-quences, others’ approval and environmental cues(Michie et al., 2013). Regardless of the specifictechniques used in order to change health behav-iour it is necessary for health messages to targetthe key motivational factors that underlie the be-haviour (Conner and Norman, 2005). Accordingto the TPB, interventions should target (i) behav-ioural beliefs about the perceived consequences ofperforming the behaviour, (ii) normative beliefsabout the views of others and (iii) control beliefsabout the power of factors to facilitate or inhibitperformance of the behaviour (Ajzen, 2013). Aswell as its success in predicting behaviour, theTPB has also been used to develop successful

interventions designed to promote health be-haviour including fruit and vegetable consump-tion, physical activity and alcohol consumption(Jemmott et al., 2011; Koning et al., 2011), al-though effect sizes have typically been small(Hardeman et al., 2002). Nevertheless, a review ofearly TPB-based interventions noted that manyinterventions did not explicitly target the TPBconstructs and/or underlying beliefs in the inter-vention, instead the TPB was more often used toinform the choice of dependent variables to evalu-ate the interventions (Hardeman et al., 2002).More recent health behaviour interventions basedon the TPB, including those delivered online, havereported medium-sized effects on behaviour(Webb et al., 2010).

Ajzen (Ajzen, 2013) proposes that extensiveformative research should be conducted whendeveloping interventions based on the TPB. Thisformative research should involve (i) elicitingmodal salient beliefs (i.e. the most commonlyheld beliefs about the behaviour) and (ii) asses-sing which beliefs are associated with intentionand/or behaviour (Sutton, 2002). The formativeresearch is therefore used to identify whichbeliefs should be targeted in an intervention, typ-ically through the development of persuasivemessages. However, as noted by Eagly andChaiken [(Eagly and Chaiken, 1998), p. 240] ‘themodel provides no formal guidance for choosingarguments to include in messages designed to in-fluence a specific belief’. Given the importanceof identifying key beliefs for developing healthinterventions and the potential knowledge gapbetween those developing the intervention andthe intended recipients (e.g. do academicresearchers know why students binge drink?), itis surprising that these formative procedures arerarely, if ever, undertaken when developinghealth behaviour change interventions based onthe TPB. Many TPB studies have examined thebehavioural, normative and control beliefs asso-ciated with a range of health behaviours includ-ing eating fruit and vegetables, physical activity,binge drinking and smoking (e.g. Rhodes et al.,2009; Murnaghan et al., 2010; French and Cooke,2012); however, these beliefs are rarely used asthe basis of subsequent interventions. Forexample, none of nine studies included in arecent meta-analysis of online health behaviourchange interventions that were based on the TPB(Webb et al., 2010) used recommended formativeresearch procedures (Sutton, 2002) to identifybeliefs to target in the intervention.

of 13 T. Epton et al.

at University of Sheffield on February 13, 2014

http://heapro.oxfordjournals.org/D

ownloaded from



The present research

We conducted a three-phase programme of for-mative research to develop theory- and empiric-ally based messages to target four importanthealth behaviours (fruit and vegetable intake,physical activity, binge drinking and smoking)among new university students. Phase 1 identi-fied the modal salient behavioural, normativeand control beliefs that underpin pre-universitystudents’ attitudes, subjective norms and per-ceived behavioural control concerning the fourhealth behaviours. Phase 2 identified the specificbeliefs that were associated with new students’intentions to perform each of the health beha-viours before university and their subsequentbehaviour at university. Phase 3 identified themodal salient reasons underpinning each of thebeliefs for each of the behaviours. These reasonswere used as the basis for developing messagesto promote regular exercise and fruit and vege-table intake and discourage excessive alcoholconsumption and smoking in new universitystudents.

PHASE 1: BELIEF ELICITATION

Method

Participants were pre-university students in theirfinal term at a Further Education College in theNorth of England (n ¼ 96). Participants wererandomly assigned to complete one of four ques-tionnaires in a classroom group setting. Eachquestionnaire was designed to elicit participants’beliefs about engaging in one of four healthbehaviours at university: (i) eating five portionsof fruit and vegetable daily (n ¼ 21), (ii) en-gaging in regular exercise (n ¼ 24), (iii) bingedrinking (n ¼ 26) and (iv) smoking (n ¼ 25).

Behavioural beliefs were assessed by fourquestions for each health behaviour, e.g. ‘Whatdo you believe would be the advantages of eatingfive portions of fruit and vegetables a day at uni-versity?’, ‘What do you believe would be the dis-advantages of eating five portions of fruit andvegetables a day at university?’, ‘What wouldyou like or enjoy about eating five portions offruit and vegetables a day at university?’, and‘What would you dislike or hate about eating fiveportions of fruit and vegetables a day at univer-sity?’. Normative beliefs were assessed by twoquestions for each health behaviour, e.g. ‘Whichindividuals or groups of people would approve

(i.e. think it was a good idea) of you eating fiveportions of fruit and vegetables a day at univer-sity?’ and ‘Which individuals or groups of peoplewould disapprove (i.e. think it was a bad idea) ofyou eating five portions of fruit and vegetables aday at university?’. Control beliefs were accessedby two questions for each health behaviour, e.g.‘What things (i.e. factors or circumstances)would make you more likely to engage in eatingfive portions of fruit and vegetables a day at uni-versity?’ and ‘What things (i.e. factors or circum-stances) would make you less likely to engage ineating five portions of fruit and vegetables a dayat university?’.

Results

Two independent raters coded all beliefs (90–92% agreement). Modal salient beliefs weredefined as those mentioned by at least 20% ofparticipants (see Tables 1–4). Five behavioural(e.g. fruit and vegetables are boring), two norma-tive (e.g. parents approve of eating fruit andvegetables) and one control (e.g. fruit and vege-tables are expensive) modal salient beliefs wereidentified for fruit and vegetable intake; seven,three and five behavioural, normative andcontrol modal salient beliefs were identified, re-spectively, for physical activity; eight, four andfour behavioural, normative and control modalsalient beliefs were identified for binge drinkingand seven, six and four behavioural, normativeand control modal salient beliefs were identifiedfor smoking. These beliefs were used in Phase 2to identify the key beliefs that predict intentionand/or behaviour for each health behaviour.

PHASE 2: BELIEF CORRELATION

Method

Incoming students to a large university in thenorth of England (n ¼ 5473) were asked by emailto complete an online questionnaire 3 weeksbefore starting university. The students were ran-domly allocated to receive a link to one of fourquestionnaires focusing on fruit and vegetables(n ¼ 1366), physical activity (n ¼ 1369), bingedrinking (n ¼ 1370) or smoking (n ¼ 1368).

The questionnaires assessed behavioural, nor-mative and control beliefs, as well as intentionsto perform the target behaviour. To measurebehavioural beliefs, participants were asked to

Development of theory-based health messages of 13



ownloaded from



Table 1: Fruit and vegetables: summary of results from Phases 1, 2 and 3

Phase 1: Belief elicitation Phase 2: Behaviourprediction (bs)

Phase 3: Reasons elicitation

n 21 702 244 117Belief category Intentions Behaviour

Behaviouralbelief

Advantages/like/enjoy Healthy (n ¼ 18) 0.13*** 0.06 Important health benefitsof eating fruit and vegetables

Reduced risk of disease(n ¼ 68)

Vitamins and minerals (n ¼ 46)Boost immune system (n ¼ 37)

Healthy weight (n ¼ 28)

Study better(n ¼ 6)

0.24*** 0.04

Better/varied diet(n ¼ 6)

20.02 0.06

How fruit and vegetables canhelp you study better

Energy (n ¼ 48)Improved concentration (n ¼ 35)Not being too ill to study

(n ¼ 24)Improved well-being (n ¼ 22)No sluggishness (n ¼ 20)

Disadvantages/dislike/hate

Cost (n ¼ 11) 20.15*** 20.11 Ways of eating fruit andvegetables cheaply

Buy from market/greengrocers(n ¼ 56)

Special offers (n ¼ 31)Buy frozen (n ¼ 22)

Boring (n ¼ 6) 20.29*** 20.23***

Ways to make eating fruit andvegetables interesting

Try new fruit and vegetables(n ¼ 25)

Try new recipes (n ¼ 22)Add to dishes (n ¼ 16)

Normativebeliefs

People who wouldapprove

Parents (n ¼ 7) 20.19*** 20.06 Reasons your parents wantyou to eat fruit and vegetables

Health (n ¼ 75)Healthy weight (n ¼ 24)Helps with studies (n ¼ 17)

People who woulddisapprove

Friends (n ¼ 5) 20.19*** 20.13 Reasons your friends want youto eat fruit and vegetables

Health (n ¼ 82)Energy (n ¼ 26)Healthy weight (n ¼ 16)

Control beliefs Factors that would make itmore likely

Factors that would make itless likely

Cost (n ¼ 9) 0.08* 20.02 Cost See above

Page

4o

f13

T.E

pto

net

al.

at University of Sheffield on February 13, 2014 http://heapro.oxfordjournals.org/ Downloaded from



Table 2: Physical activity: summary of results from Phases 1, 2 and 3



n 24 766 270 112Belief category Intentions Behaviour

Behaviouralbelief

Advantages/like/enjoy

Health (n ¼ 18) 0.16*** 0.11 The important healthbenefits of exercise

Reducing the risk of disease(n ¼ 37)

Improving well-being (n ¼ 18)Maintaining a healthy weight

(n ¼ 17)

Fitness (n ¼ 14) 0.16*** 0.08Make friends (n ¼ 11) 0.002 0.10Socializing (n ¼ 10) 0.09 20.07Stress relief (n ¼ 9) 0.24*** 0.07

The important healthbenefits of improvingfitness

Reducing the risk of disease(n ¼ 33)

Increased energy (n ¼ 19)Living longer (n ¼ 11)

How exercise relieves stress Distraction (n ¼ 34)Expending negative energy (n ¼ 19)Improving well-being (n ¼ 11)


Lack of time for study (n ¼ 16) 20.13*** 20.07 Why/how exercise mightnot reduce time for study

Plan it into your day (n ¼ 24)Aids concentration (n ¼ 18)More productive (n ¼ 12)

Time consuming (n ¼ 16) 20.07 0.05

Normativebeliefs

People who wouldapprove

Sporty people (n ¼ 9) 0.05 20.05 Reasons that your parentsand family think youshould exercise

Health (n ¼ 42)Improving well-being (n ¼ 11)Helps to avoid unhealthy things

(n ¼ 9)

Family (n ¼ 5) 0.10* 0.06Friends (n ¼ 7) 0.23*** 0.11

Reasons that your friendsthink you should exercise

Health (n ¼ 35)Do it together (n ¼ 32)Improving well-being (n ¼ 6)


Control beliefs Factors that wouldmake it morelikely

Access to range of facilities(n ¼ 9)

0.23*** 0.11 The best sports facilities inSheffield

The Peak District (n ¼ 15)S10 gym (n ¼ 9)University swimming pool (n ¼ 7)Cost (n ¼ 6) 0.32*** 0.14

How to exercise for free inSheffield

Running (n ¼ 31)Cycling (n ¼ 12)The Peak District (n ¼ 9)

Factors that wouldmake it less likely

Time restrictions (n ¼ 7) 0.08* 20.22 See aboveCost (n ¼ 6) 0.23*** 0.21Poor/few facilities (n ¼ 6) 20.03 0.01

Develo

pm

ent

of

theo

ry-b

ased

hea

lthm

essages

Page

5o

f13




Table 3: Binge drinking: summary of results from Phases 1, 2 and 3



n 26 726 325 137 (drinkers only)Belief category Intentions Behaviour

Behaviouralbelief

Advantages/like/enjoy Make friends/meetpeople (n ¼ 19)

0.09* 0.20* Ways to make friends and funthings to do at university thatdoesn’t involve binge drinking

Socialize (e.g. cinema, eatout) (n ¼ 83)

Sports (n ¼ 75)Join societies (n ¼ 41)

Socializing (n ¼ 17) 0.06 20.16Fun (n ¼ 10) 0.52*** 0.38***Stress relief (n ¼ 7) 0.01 0.05


Impact on studies(n ¼ 16)

20.18*** 20.17* How binge drinking has a negativeimpact on studies

Hangovers (n ¼ 107)Missing lectures (n ¼ 82)Study problems (n ¼ 49)Hangover (n ¼ 14) 0.07** 20.01

Impact on health (n ¼ 12) 20.03 0.06Cost (n ¼ 10) 20.01 0.001

Normativebeliefs

People who would approve Students (n ¼ 8) 0.09 0.06 Reasons your friends might not Health (n ¼ 62)Friends would have to look

after you (n ¼ 55)Spoil their night (n ¼ 33)

Friends (n ¼ 8) 0.51*** 0.30*** Want you to binge drink


Teachers/lecturers(n ¼ 17)

0.01 20.10 Reasons your parents think thatyou shouldn’t binge drink

Health (n ¼ 101)Bad for studies (n ¼ 78)Safety (n ¼ 78)Money (n ¼ 50)

Parents (n ¼ 11) 0.23*** 0.18*

Control beliefs Factors that would make itmore likely

Friends drinking (n ¼ 12) 0.44*** 0.35*** Reasons not to binge drink evenwhen your friends are

It’s my decision (n ¼ 66)Money (n ¼ 59)Avoid hangover (n ¼ 39)Look after friends (n ¼ 30)

Other people drinking/norms (n ¼ 9)

0.23*** 0.15*

Factors that would make itless likely

Cost (n ¼ 8) 0.19*** 20.09 Reasons why they should not beinfluenced when they see othersdrinking

Their decision (n ¼ 89)Poor role models (n ¼ 59)Studies (n ¼ 7) 20.07* 0.06

Page

6o

f13

T.E

pto

net

al.




Table 4: Smoking: summary of results from Phases 1, 2 and 3

Phase 1: Belief elicitation Phase 2: Behaviour prediction (bs) Phase 3: Reasons elicitation

N 25 606 27 114 79 179 35 47Intentions Behaviour

Belief category NS EX SMO SMO NS EX SMO

Behaviouralbelief

Advantages/like/enjoy

Make newfriends(n ¼ 11)

20.02 20.00 20.13 0.11 Ways of relievingstress atuniversity

Exercise (n ¼ 86)Talking things

over (n ¼ 28)Relaxation/

meditation(n ¼ 27)

How you couldreduce stresswithoutsmoking

Exercise(n ¼ 22)

Socializing(n ¼ 6)

Relaxation/mediation(n ¼ 5)

Stress relief(n ¼ 10)

0.12** 0.26 0.23* 20.00

Socializing(n ¼ 10)

0.19* 0.02 0.00 20.34Reasons why

smokingdoesn’t helpyou socialize

Have to go outside(n ¼ 65)

The smell (n ¼ 56)Puts off

non-smokers(n ¼ 32)


Cost (n ¼ 16) 20.12*** 20.23 20.08 0.11 What you couldspend yourmoney oninstead ofsmoking

Travel/holidays(n ¼ 66)

Nice food (n ¼ 48)Socializing

(n ¼ 40)Clothes (n ¼ 38)

Why othersdisapprove ofsmoking

Health(n ¼ 19)

The smell(n ¼ 16)

Endangersothers(n ¼ 9)

Smell (n ¼ 14) 20.01 0.18 20.15 20.11Health risks

(n ¼ 11)20.05 20.04 0.05 0.06

Others’disapproval(n ¼ 8)

20.08* 20.51* 0.05 0.30

Why othersdisapprove ofsmoking

Risk to health(n ¼ 85)

Endangers others(n ¼ 49)

The smell (n ¼ 63)Normative

beliefsPeople who would

approveSmokers

(n ¼ 16)0.04 0.05 20.14 20.03

Friends (n ¼ 8) 20.24*** 0.04 20.01 20.07People who would

disapproveNon-smokers

(n ¼ 10)20.06 20.37 0.14 20.06 Why friends

might notwant you tosmoke

Risk to health(n ¼ 75)

Concern aboutown health(n ¼ 56)

The smell (n ¼ 55)

Why yourlecturers mightnot want youto smoke

Health (n ¼ 16)The smell

(n ¼ 15)Waste of study

time (n ¼ 14)

Friends (n ¼ 7) 20.24*** 0.04 20.01 20.07Teachers/

lecturers(n ¼ 8)

20.07 20.17 20.30*** 0.15

Family/parents(n ¼ 6)

20.04 0.20 20.33*** 20.50*** Why your familymight not wantyou to smoke

Health (n ¼ 30)Money (n ¼ 18)The smell

(n ¼ 9)

Continued

Develo

pm

ent

of

theo

ry-b

ased

hea

lthm

essages

Page

7o

f13




indicate (on a seven-point scale anchored by un-likely–likely) the extent to which they thoughteach of the outcomes would be likely to occurshould they engage in the behaviour (e.g. ‘Myeating five portions of fruit and vegetables eachday at university would be healthy’). To measurenormative beliefs, participants were asked to in-dicate (on a seven-point scale anchored by thinkI should–think I should not) the extent to whichthey thought specific referents would approve ordisapprove of them engaging in the behaviour(e.g. my friends, my parents). To measurecontrol beliefs, participants were asked to indi-cate (on a seven-point scale anchored by lesslikely–more likely) the extent to which theythought different factors would influence theirbehaviour (e.g. ‘The following factors wouldmake my eating five portions of fruit and vegeta-bles per day at university . . . the high cost of fruitand vegetables’). Two questions measured inten-tions to perform the behaviour at university on aseven-point scale (definitely don’t–definitely do)(e.g. ‘Do you intend to eat five portions of fruitand vegetable a day at university?’).

To assess subsequent behaviour, participantswere emailed a link to one of four follow-upquestionnaires 1 month after starting at universitythat assessed their health behaviour, i.e. an eight-item fruit and vegetable intake measure (NationalCentre for Social Research, 2008), a three-itemphysical activity questionnaire (Booth, 2000), a7-day retrospective binge drinking diary (Gmeland Rehm, 2004) or two items on smoking(Health Survey for England, 2008).

Results

There was a 54% response rate to the baselinequestionnaires (n ¼ 2959): fruit and vegetables(n ¼ 702), physical activity (n ¼ 774), binge drink-ing (n ¼ 735) and smoking (n ¼ 747). Eighteenparticipants (1.12%) were excluded because theyhad incomplete baseline data concerning eitherbinge drinking (n ¼ 9) or physical activity (n ¼ 9).There was a 39.47% response rate to the follow-up questionnaires (n ¼ 1168): fruit and vegetables(n ¼ 244), physical activity (n ¼ 270), binge drink-ing (n ¼ 325) and smoking (n ¼ 329). For each ofthe four health behaviours, intentions and behav-iour at follow-up were regressed separately ontobehavioural beliefs, normative and control beliefs[see Tables 1–4], and intentions were correlatedwith subsequent behaviour.T

able

4:C

on

tin

ued

Ph

ase

1:B

elie

fel

icit

atio

nP

has

e2:

Beh

avio

ur

pre

dic

tio

n(b

s)P

has

e3:

Rea

son

sel

icit

atio

n

N25

606

2711

479

179

3547

Inte

nti

on

sB

ehav

iou

r

Bel

ief

cate

gory

NS

EX

SM

OS

MO

NS

EX

SM

O

Co

ntr

ol

bel

iefs

Fac

tors

that

wo

uld

mak

eit

mo

reli

kel

y

Pee

rp

ress

ure

(n¼

12)

0.05

20.

402

0.18

20.

16H

ow

toav

oid

smo

kin

gw

hen

you

see

oth

ers

smo

kin

g

Avo

idsm

ok

ers

(n¼

56)

Th

ink

of

neg

ativ

eas

pec

ts(n¼

51)

Rem

emb

erh

ealt

hri

sks

(n¼

41)

Do

som

eth

ing

else

(n¼

27)

Go

od

way

so

fre

liev

ing

stre

sso

ther

than

smo

kin

g

Exe

rcis

e(n¼

18)

Rel

ax /med

iati

on

(n¼

11)

Tal

kin

git

ove

r(n¼

6)

Str

ess

(n¼

11)

0.13

**2

0.51

*0.

24*

0.04

Oth

erp

eop

lesm

ok

ing

(n¼

8)

0.17

***

0.37

0.05

0.11

See

abo

ve

Fac

tors

that

wo

uld

mak

eit

less

lik

ely

Peo

ple

no

tsm

ok

ing

(n¼

10)

0.05

0.33

0.23

*0.

01W

hy

oth

ers

no

tsm

ok

ing

mig

ht

red

uce

tem

pta

tio

nto

smo

ke

Do

n’t

wan

tto

smo

ke

alo

ne

(n¼

21)

No

tth

en

orm

(n¼

6)N

ocu

es(n¼

5)




ownloaded from



For fruit and vegetable intake, the beliefs thatpredicted intention were as follows: the behav-ioural beliefs ‘fruit and vegetables are healthy’,‘fruit and vegetables help you study better’, ‘fruitand vegetables are expensive’ and ‘eating fruitand vegetables is boring’; the normative beliefs‘parents would approve of you eating fruit andvegetables’ and ‘friends would disapprove of youeating fruit and vegetables’; and the controlbelief ‘the cost makes eating fruit and vegetablesless likely’. Intention was found to correlate sig-nificantly with subsequent fruit and vegetableintake at university, r(241) ¼ 0.40, p , 0.001.One belief also predicted behaviour, i.e. the be-havioural belief that ‘eating fruit and vegetablesis boring’.

For physical activity, the beliefs that predictedintentions were: the behavioural beliefs ‘exerciseis healthy’, ‘exercise improves fitness’, ‘exerciseis good for stress relief’ and ‘exercise reducestime for study’; the normative beliefs ‘friendswould approve of exercise’ and ‘family wouldapprove of exercise’ and the control beliefs‘access to a wide range of good facilities makesexercise more likely’, ‘access to cheap facilitiesmakes exercise more likely’, ‘the cost make exer-cise less likely’ and ‘time restrictions makes exer-cise less likely’. Intention was found to correlatesignificantly with subsequent physical activitylevels at university, r(264) ¼ 0.20, p , 0.001. Nobeliefs predicted behaviour.

For binge drinking, the beliefs that predictedintentions were as follows: the behavioural beliefs‘binge drinking is good for making friends’, ‘bingedrinking is fun’, ‘binge drinking has a negativeimpact on studies’ and ‘binge drinking would leadto a hangover’; the normative beliefs ‘friendswould approve of binge drinking’ and ‘parentswould disapprove of binge drinking’ and thecontrol beliefs ‘friends drinking makes bingedrinking more likely’, ‘seeing other people drink-ing makes binge drinking more likely’, ‘alcoholbeing expensive makes binge drinking morelikely’ and ‘needing time to study makes bingedrinking less likely’. The same beliefs predictedbehaviour, with the exception of the behaviouralbelief ‘hangover’ and the control beliefs ‘alcoholbeing expensive’ and ‘needing time to study’.Intention was found to correlate significantly withthe frequency of binge drinking at university,r(253) ¼ 0.64, p , 0.001.

Responses to the smoking survey were ana-lysed by smoking status (non-smoker, ex-smokerand smoker). For non-smokers, the beliefs that

predicted intentions to smoke were as follows:the behavioural beliefs ‘smoking is good for re-lieving stress’, ‘smoking is good for socializing’,‘smoking is expensive’ and ‘smoking leads to dis-approval from others’; normative beliefs ‘friendswould approve of smoking’ and control beliefs‘stress is likely to increase the temptation tosmoke’ and ‘other people smoking is likely to in-crease the temptation to smoke’. For ex-smokers,one behavioural belief, ‘disapproval of others’and one control belief, ‘stress makes smokingmore likely’ predicted intentions to smoke. Themost predictive normative belief was ‘non-smokers disapprove of smoking’. For smokers,the beliefs that predicted intentions were asfollows: the behavioural belief ‘smoking relievesstress’; normative beliefs ‘your lecturers woulddisapprove of smoking’ and ‘your family disap-prove of smoking’ and control beliefs ‘stressmakes smoking more likely’ and ‘people notsmoking makes smoking less likely’. Intentionwas found to correlate significantly with numberof cigarettes smoked at university amongsmokers, r(44) ¼ 0.29, P ¼ 0.05. One normativebelief ‘your family disapprove of smoking’ pre-dicted smoking behaviour.

After removal of duplicate/similar items, 32beliefs significantly predicted intentions: fruitand vegetable intake (6 beliefs), physical activity(8 beliefs), binge drinking (6 beliefs) andsmoking (non-smokers: 6 beliefs, ex-smokers: 2beliefs and smokers: 4 beliefs). The 32 beliefswere used in Phase 3 to elicit the reasons under-lying each belief.

PHASE 3: REASON ELICITATION

Method

University students were invited by email tocomplete one of four questionnaire studies focus-ing on the key beliefs identified for each healthbehaviour: fruit and vegetables (n ¼ 117), phys-ical activity (n ¼ 112), binge drinking (n ¼ 137)and smoking (non-smokers n ¼ 179, ex-smokersn ¼ 35 and smokers n ¼ 47). Participants wereasked a series of questions to identify the reasonsunderpinning each of the 32 beliefs identified inPhase 2.

Each questionnaire asked participants to listup to three reasons underlying each belief and toindicate which of these was the most important.For example, to identify reasons underlying




ownloaded from



behavioural beliefs related to fruit and vegetableconsumption participants were asked ‘What arethe three most important health benefits of eatingfruit and vegetables?’, ‘What are the ways thatfruit and vegetables can improve your studying?’and ‘What are the ways you can make eating fruitand vegetables less boring?’. Reasons underlyingnormative beliefs related to fruit and vegetableconsumption were identified by asking partici-pants ‘What are the reasons your friends wouldwant you to eat fruit and vegetables’ and ‘Whatare the reasons your parents would want you toeat fruit and vegetables’. For reasons underlyingcontrol beliefs, participants were asked, ‘Whatare the ways to eat fruit and vegetables cheaply’.

Results

Participant’s responses to each of the questionswere collated and coded by two independentraters (71– 100% agreement) to identify themost common reasons associated with eachbelief to use in health messages. The number ofparticipants endorsing each underlying reasonwas calculated.

Overall, 102 reasons underlying beliefs wereidentified: fruit and vegetable intake (21), physic-al activity (24), binge drinking (19) and smoking(non-smokers: 20, ex-smokers: 6 and smokers:12). Responses were chosen by between 5 and107 of participants: fruit and vegetables (16–68),physical activity (9–37), binge drinking (33–101)and smoking (non-smokers 10–86; ex-smokers6–19; smokers 5–30). For example, five reasonswere identified to ‘why eating fruit and vegeta-bles can improve your studying’: increasedenergy, improved concentration, no sluggishness,not being too ill to study and increased well-being. See Tables 1–4 for full details of themodal reasons for each behaviour/belief.

DISCUSSION

A three-phase programme of formative researchwas conducted to identify (i) modal salientbeliefs associated with four health behaviours(fruit and vegetable intake, physical activity,binge drinking, smoking), (ii) the key belief(s)that predicted intentions and/or behaviour foreach health behaviour and (iii) the reasonsunderlying each belief. This process can be usedas the basis for developing health messages thathave a strong theoretical and empirical basis and

that are highly relevant for the target population.The formative research reported here is in linewith recommendations for good practice in thescience of behavioural change such as publishingintervention content (Michie and Abraham,2008). This enables ease of replication and allowssubsequent research to build on the learningaccrued during the development of the interven-tion, regardless of the eventual success of theintervention (Michie et al., 2012). The formativeresearch described in this paper provides a tem-plate for developing theory-based messages forhealth behaviour change interventions. In par-ticular, the addition of a reasons elicitation phaseprovides a way of forming messages to target keybeliefs identified in preceding formative researchbased on the TPB.

Previous attempts to develop interventionsbased on the TPB have been hampered by thelack of a clear procedure for constructing persua-sive messages to target TPB beliefs (Eagly andChaiken, 1998; Sutton, 2002). The reasons elicit-ation procedure reported in this paper providesone such procedure for developing the content ofTPB persuasive messages and, in doing so, mayalso provide a timely impetus for the developmentof TPB-based interventions. Such interventionsare of both practical and theoretical importanceas they not only seek to change health behaviour,but also to provide important tests of theory.

There are a number of issues that warrant dis-cussion in relation to the different phases of for-mative research reported in this paper. RegardingPhase 1 (i.e. belief elicitation), there are noagreed sample size guidelines for such studies.For example, studies designed to elicit the salientbeliefs underlying physical activity have usedsamples ranging in size from 7 to 120 (SymonsDowns and Hausenblas, 2005). Current recom-mendations for health services research (Franciset al., 2004) suggest a sample size of 25, which isbroadly in line with the sample sizes in the currentresearch. A similar issue arises in relation to howto select modal salient beliefs. Ajzen and Fishbein[(Ajzen and Fishbein, 1980), p. 70] suggest that‘those beliefs that exceed a certain frequency’should be chosen, but fail to indicate what repre-sents an appropriate frequency. In the present re-search, we selected beliefs that were endorsed byat least 20% of respondents in order to ensurethat a large range of beliefs were included inPhase 2. Previous physical activity belief elicit-ation studies have identified a mean of seven, fourand six behavioural, normative and control beliefs




ownloaded from



(Symons Downs and Hausenblas, 2005), which issimilar to the numbers of beliefs chosen in thecurrent research. However, there was some vari-ation in the number of modal salient beliefs iden-tified for the four health behaviours. In particular,fewer behavioural, normative and control beliefswere identified for eating fruit and vegetablesthan for the other health behaviours, reflectingdiffering levels of agreement among participants.

Regarding Phase 2 (identifying beliefs asso-ciated with intention and/or behaviour), thecurrent research only assessed expectancy com-ponents of behavioural, normative and controlbeliefs, rather than testing expectancy X valuecombinations. According to the TPB, behaviour-al beliefs about the likelihood of the behaviourleading to a particular outcome should beweighted (multiplied) by the value attached tothe outcome. However, only assessing expect-ancy components has three key advantages. First,it halves the number of items that are needed toassess the beliefs. Secondly, it avoids problemsassociated with the scaling and analysis of multi-plicative composites (Evans, 1991). Thirdly, itproduces similar, or stronger, correlations withTPB constructs as does using expectancy X valuemultiplicative terms (Gagne and Godin, 2000;Rhodes et al., 2009).

The present study examined associationsbetween the modal salient beliefs and intentionand subsequent behaviour in new university stu-dents, in line with Sutton’s (Suttons, 2002)recommendations and consistent with other TPBstudies (e.g. Rhodes et al., 2009; Booth et al., inpress). Interventions should then target thesebeliefs in order to produce changes in intentionand/or behaviour. However, according to theTPB, behavioural, normative and control beliefsdetermine attitude, subjective norm and per-ceived behavioural control which, in turn, deter-mine intention and behaviour. Thus, analternative approach taken in other TPB studies(e.g. Symons Downs and Hausenblas, 2005;Murnaghan et al., 2010) is to identify the modalsalient beliefs that are most strongly associatedwith their respective constructs. Interventionsshould then target these beliefs in order toproduce changes in attitude, subjective norm andperceived behavioural control which, in turn,should produce changes in intention and behav-iour. The latter approach is more consistent withthe theoretical structure of the TPB; however,given the longer proposed causal chain, anychanges in underlying beliefs may not carry

through to produce measurable changes in be-haviour (Sutton, 2002).

It is also noteworthy that very few of the beliefswere found to be predictive of subsequent healthbehaviour at university, despite a larger numberbeing predictive of intentions. As a result, most ofthe beliefs identified in Phase 2 to be used inPhase 3 to develop health messages were basedon significant associations with intention only.However, this pattern of results is consistent withthe theoretical structure of the TPB in which in-tention mediates the effects of attitude, subjectivenorm and perceived behavioural control on be-haviour (although a direct link is hypothesizedbetween perceived behavioural control and be-haviour). In the current studies, the intentionmeasures were found to have significant correla-tions with subsequent behaviour, suggesting thatchanging underlying behavioural, normative andcontrol beliefs may impact on intentions which, inturn, impact on subsequent behaviour.

Regarding Phase 3 (reasons elicitation), as withPhase 1, issues arise regarding the number of par-ticipants needed for such studies and cut-offpoints for selecting modal salient reasons. In thepresent study the criterion for selecting modalsalient reasons varied for each belief dependingon the number of reasons listed and numbers ofparticipants endorsing them, to ensure an appro-priate number of reasons to be used in subsequenthealth messages to change the beliefs. In line withmost interventions based on the TPB (for areview, see Hardeman et al., 2002), the present re-search sought to develop messages to changeexisting beliefs. However, although the formativeresearch ensured that the resultant messages wererelevant for the target population, it does notguarantee that the messages will be persuasive.At present, there is a lack of research on whatconstitutes a ‘strong’ message (Sutton, 2002).Nonetheless, culturally relevant messages thattarget the key beliefs that influence behaviour,such as the ones developed using this programmeof research, are likely to make strong messages.

The use of the TPB as a theoretical frameworkto develop health behaviour change interventionsis not without its own limitations. Despite thestrong predictive utility of the TPB (McEachanet al., 2011), the model has been criticized forfailing to account for other, non-cognitive, influ-ences on health behaviour (Norman and Conner,2005). In particular, past behaviour is often foundto be the strongest predictor of future behavioursuggesting the impact of habitual processes on




ownloaded from



health behaviour (Ouellette and Wood, 1998).Thus, habit strength has been found to be a strongcorrelate of a range of health behaviours. Forexample, Gardner et al. (Gardner et al., 2011)reported a medium-to-strong average correlation(rþ¼ 0.44) between habit strength and behaviourin their meta-analysis of 23 studies of nutritionand physical activity behaviour. This is verysimilar to the average intention–behaviour correl-ation (rþ ¼ 0.44) reported by McEachan et al.(McEachan et al., 2011) across 237 prospectivetests of the TPB in relation to health behaviour.Similarly, Norman (Norman, 2011) found thatboth intention and habit strength were predictiveof binge drinking in university students, suggest-ing that interventions should focus on both inten-tional (e.g. underlying beliefs) and habitual (e.g.environmental cues) influences when seeking tochange health behaviour.

The formative research reported in the currentpaper describes three phases in the developmentof health messages based on the TPB. However, itis important to note that there is a wide range ofbehaviour change techniques that can be used topromote health behaviour change in addition tothe construction of health messages targetingunderlying beliefs (Michie et al., 2013). Moreover,a process of intervention mapping (Bartholemewet al., 2011) can be used to select theory-based be-haviour change techniques to target specific deter-minants of the targeted health behaviour. Thiswork should then be followed by further activitiesfocusing on the delivery, adoption/implementa-tion and evaluation of the intervention.

CONCLUSION

The three-phase programme of research describedin this paper provides a clear, replicable procedurefor developing health message that have a strongtheoretical and empirical basis. The formative re-search includes identifying (i) the modal salientbeliefs associated with eacsh health behaviour, (ii)the key beliefs associated with intentions toperform each behaviour and (iii) the key reasons/arguments to construct health messages to targeteach belief. This systematic process provides thebasis for creating health messages that are theoret-ically and empirically based and highly relevantfor the target population. In addition, describingthe process of developing the messages promotestransparency in the reporting intervention contentand thereby adds to the science of behaviour

change (Michie and Abraham, 2008). The beliefsidentified in this formative research can be used toform the basis of persuasive messages designed tochange health behaviours in young people (Eptonet al., 2013).

ACKNOWLEDGMENTS

Thanks to Anna Spring for her assistance withdata coding.

FUNDING

The work reported in this paper was supported bythe UK National Prevention Research Initiative(NPRI) Phase 4 (grant number: MR/J0004501/1).The NPRI includes the following Funding Partners(in alphabetical order): Alzheimer’s ResearchTrust, Alzheimer’s Society, Biotechnology andBiological Sciences Research Council, BritishHeart Foundation, Cancer Research UK, ChiefScientist Office, Scottish Government HealthDirectorate, Department of Health, Diabetes UK,Economic and Social Research Council, Healthand Social Care Research and DevelopmentDivision of the Public Health Agency (HSC andR and D Division), Medical Research Council,The Stroke Association, Wellcome Trust, WelshAssembly Government and World Cancer Re-search Fund.

REFERENCES

Ajzen, I. (1988) Attitudes, Personality and Behaviour. OpenUniversity Press, Milton Keynes.

Ajzen, I. (2013) Behavioural interventions based on the theoryof planned behaviour. http://people.umass.edu/aizen/pdf/tpb.intervention.pdf (last accessed 7 February 2013).

Ajzen, I. and Fishbein, M. (1980) Understanding Attitudesand Predicting Social Behaviour. Prentice-Hall, EnglewoodCliffs, NJ.

Bartholemew, L. K., Parcel, G. S., Kok, G., Gottleib, N. H.and Fernadez, M. E. (2011) Planning Health PromotionPrograms: An Intervention Mapping Approach, 3rd edition.Wiley, Chichester.

Booth, M. L. (2000) Assessment of physical activity: aninternational perspective. Research Quarterly for Exerciseand Sport, 71, s114–s120.

Booth, A., Norman, P., Goyder, E. and Harris, P. (in press)Using the Theory of Planned Behavior to identify keybeliefs underlying chlamydia testing intentions in asample of young people living in deprived areas. Journalof Health Psychology. doi:10.1177/1359105313510335.




ownloaded from

http://people.umass.edu/aizen/pdf/tpb.intervention.pdf









Conner, M. and Norman, P. (2005) Predicting HealthBehaviour: Research and Practice with Social CognitionModels, 2nd edition. Open University Press, Buckingham.

Craig, P., Dieppe, P., Macintyre, S., Michie, S., Nazareth, I.and Petticrew, M. (2008) Developing and evaluatingcomplex interventions: the new Medical ResearchCouncil guidance. British Medical Journal, 337, 979–983.

Eagly, A. H. and Chaiken, S. (1998) Attitude structure andfunction. In: Gilbert, D. T., Fiske, S. T. and Gardner, L.(eds), The Handbook of Social Psychology, 4th edition.Vols 1 and 4. McGraw-Hill, New York, pp. 269–322.

Epton, T., Norman, P., Sheeran, P., Harris, P. R., Webb, T.L., Ciravegna, F. et al. (2013) A theory based onlinehealth behaviour intervention for new university students:Study protocol. BMC Public Health, 13, 107.

Evans, M. G. (1991) The problem of analyzing multiplicativecomposites: interactions revisited. American Psychologist,46, 6–15.

Francis, J. J., Eccles, M. P., Johnston, M., Walker, A.,Grimshaw, J., Foy, R. et al. (2004) ConstructingQuestionnaires Based on the Theory of Planned Behaviour:A Manual for Health Services Researchers. Newcastle, UK.http://pages.bangor.ac.uk/~pes004/exercise_psych/downloads/tpb_manual.pdf (last accessed 7 February 2013).

French, D. P. and Cooke, R. (2012) Using the theory ofplanned behaviour to understand binge drinking: the im-portance of beliefs for developing interventions. BritishJournal of Health Psychology, 17, 1917.

Gagne, C. and Godin, G. (2000) The theory of planned be-haviour: some measurement issues concerning belief-based variables. Journal of Applied Social Psychology,30, 2173–2193.

Gardner, B., de Bruijn, G.-J. and Lally, P. (2011) A system-atic review and meta-analysis of applications of the self-report habit index to nutrition and physical activity beha-viours. Annals of Behavioral Medicine, 42, 174–187.

Gmel, G. and Rehm, J. (2004) Measuring alcohol consump-tion. Contemporary Drug Problems, 31, 467–540.

Hardeman, W., Johnston, M., Johnston, D. W., Bonetti, D.,Wareham, N. J. and Kinmonth, A. L. (2002) Applicationof the theory of planned behaviour in behaviour changeinterventions: a systematic review. Psychology andHealth, 17, 123–158.

Jemmott, J. B., Jemmott, L. S., Leary, A. O., Ngwane, Z.,Icard, L., Bellamy, S. et al. (2011) Cognitive-behaviouralhealth-promotion intervention increases fruit and vege-table consumption and physical activity among SouthAfrican adolescents. Psychology and Health, 26, 167–185.

Koning, I. M., van den Eijnden, R. J., Verdurmen, J. E.,Engels, R. C. and Vollebergh, W. A. (2011) Long-termeffects of a parent and student intervention on alcoholuse in adolescents. American Journal of PreventativeMedicine, 40, 541–547.

McEachan, R. R. C., Conner, M., Taylor, N. J. and Lawton,R. J. (2011) Prospective prediction of health-relatedbehaviours with the theory of planned behaviour: ameta-analysis. Health Psychology Review, 5, 97–144.

Michie, S. and Abraham, C. (2008) Advancing the scienceof behaviour change: a plea for scientific reporting.Addiction, 103, 1409–1410.

Michie, S. and Prestwich, A. (2010) Are interventionstheory based? Development of an intervention theorycoding scheme. Health Psychology, 29, 1–8.

Michie, S., Brown, J., Geraghty, A. W. A., Miller, S.,Yardley, L., Gardner, B. et al. (2012) Development ofStopAdvisor: a theory-based interactive internet-basedsmoking cessation intervention. TranstheoreticalBehavioural Medicine, 2, 263–275.

Michie, S., Richardson, M., Johnston, M., Abraham, C.,Francis, J., Hardeman, W. et al. (2013) The behaviorchange technique taxonomy (v1) of 93 hierarchically clus-tered techniques: building an international consensus forthe reporting of behavior change interventions. Annals ofBehavioural Medicine, 46, 81–95.

Murnaghan, D. A., Blanchard, C. M., Rodgers, W. M.,LaRoas, J. N., MacQuarrie, C. R., MacLellan, D. L. et al.(2010) Predictors of physical activity, healthy eating andbeing smoke free in teens: a theory of planed behaviourapproach. Psychology and Health, 8, 925–941.

National Centre for Social Research (2008). Health surveyfor England 2008. http://www.ic.nhs.uk/statistics-and-data-collections/health-and-lifestyles-related-surveys/health-survey-for-england (last accessed 7 February 2013).

Noar, S. M., Benac, C. N. and Harris, M. S. (2007) Does tai-loring matter? Meta-analytic review of tailored printhealth behaviour change interventions. PsychologicalBulletin, 133, 673–693.

Norman, P. (2011) The theory of planned behaviour andbinge drinking among undergraduate students: assessingthe impact of habit strength. Addictive Behaviors, 36,502–507.

Norman, P. and Conner, M. (2005) Predicting and changinghealth behaviour: future directions. In: Conner, M. andNorman, P. (eds), Predicting Health Behaviour: Researchand Practice with Social Cognition Models, 2nd edition.Open University Press: Buckingham, pp. 324–371.

Ouellette, J. and Wood, W. (1998) Habit and intention ineveryday life: the multiple processes by which past behaviorpredicts future behavior. Psychological Bulletin, 124, 54–74.

Parliamentary Office of Science and Technology. (2007)Health behaviour. Http://www.parliament.uk/documents/post/postpn283.pdf (last accessed 7 February 2013).

Rhodes, R. E., Blanchard, C. M., Courneya, K. S. andPlotnikoff, R. C. (2009) Identifying belie-based targetsfor the promotion of leisure-time walking. HealthEducation and Behaviour, 36, 381–393.

Sutton, S. (2002) Using social cognition models to develophealth behaviour interventions. In: Rutter, D. and Quine,L. (eds), Changing Health Behaviour. Open UniversityPress: Buckingham, pp. 193–208.

Symons Downs, D. and Hausenblas, H. A. (2005)Elicitation studies and the theory of planned behaviour: asystematic review of exercise beliefs. Psychology of Sportand Exercise, 6, 1–31.

Webb, T. L., Joseph, J., Yardley, L. and Michie, S. (2010)Using the Internet to promote health behaviour change.Journal of Medical Internet Research, 12:e4.

Witte, K., Meyer, G. and Martell, D. (2001) Effective HealthRisk Messages: A Step-by-Step Guide. Sage, ThousandOaks, CA.




ownloaded from

http://pages.bangor.ac.uk/~pes004/exercise_psych/downloads/tpb_manual.pdf








http://www.ic.nhs.uk/statistics-and-data-collections/health-and-lifestyles-related-surveys/health-survey-for-england

















Http://www.parliament.uk/documents/post/postpn283.pdf









Development of theory-based health messages: three-phase programme of formative research

Documents

Transcript of Development of theory-based health messages: three-phase programme of formative research