Self-efficacy and multiple illness representations in older adults: A multilevel approach

18
This article was downloaded by: [University of Tasmania] On: 26 January 2012, At: 20:25 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Psychology & Health Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gpsh20 Self-efficacy and multiple illness representations in older adults: A multilevel approach Benjamin Schüz a , Susanne Wurm a , Lisa M. Warner a b & Jochen P. Ziegelmann b a German Centre of Gerontology, Manfred-von-Richthofen-Str. 2, 12101 Berlin, Germany b Department of Health Psychology, Freie Universität Berlin, 14195 Berlin, Germany Available online: 07 Jul 2011 To cite this article: Benjamin Schüz, Susanne Wurm, Lisa M. Warner & Jochen P. Ziegelmann (2012): Self-efficacy and multiple illness representations in older adults: A multilevel approach, Psychology & Health, 27:1, 13-29 To link to this article: http://dx.doi.org/10.1080/08870446.2010.541908 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and- conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Transcript of Self-efficacy and multiple illness representations in older adults: A multilevel approach

This article was downloaded by: [University of Tasmania]On: 26 January 2012, At: 20:25Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Psychology & HealthPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/gpsh20

Self-efficacy and multiple illnessrepresentations in older adults: Amultilevel approachBenjamin Schüz a , Susanne Wurm a , Lisa M. Warner a b & JochenP. Ziegelmann ba German Centre of Gerontology, Manfred-von-Richthofen-Str. 2,12101 Berlin, Germanyb Department of Health Psychology, Freie Universität Berlin,14195 Berlin, Germany

Available online: 07 Jul 2011

To cite this article: Benjamin Schüz, Susanne Wurm, Lisa M. Warner & Jochen P. Ziegelmann(2012): Self-efficacy and multiple illness representations in older adults: A multilevel approach,Psychology & Health, 27:1, 13-29

To link to this article: http://dx.doi.org/10.1080/08870446.2010.541908

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representationthat the contents will be complete or accurate or up to date. The accuracy of anyinstructions, formulae, and drug doses should be independently verified with primarysources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand, or costs or damages whatsoever or howsoever caused arising directly orindirectly in connection with or arising out of the use of this material.

Psychology and HealthVol. 27, No. 1, January 2012, 13–29

Self-efficacy and multiple illness representations in older adults:

A multilevel approach

Benjamin Schuza*, Susanne Wurma, Lisa M. Warnerab

and Jochen P. Ziegelmannb

aGerman Centre of Gerontology, Manfred-von-Richthofen-Str. 2, 12101 Berlin, Germany;bDepartment of Health Psychology, Freie Universitat Berlin, 14195 Berlin, Germany

(Received 14 January 2010; final version received 17 November 2010)

Objectives: The Common-Sense Model assumes that individuals formsubjective representations about their illnesses, which in turn guidecognitive and behavioural responses. This assumption is complicated inindividuals with multimorbidity, and it is an open question to which degreeillness-specific and person-level factors determine the representations ofspecific illnesses. This study examines the structure and interrelations ofillness representations in multimorbidity employing a hierarchical frame-work based on Cognitive Theory.Methods: Multiple illness representations were assessed in 305 people aged65 and older using two Brief Illness Perception Questionnaires. Multilevelmodelling was used to explore the relations between illness representationsand to explain how two illness-specific representations – personal controland treatment control – were determined by a person-level factor, self-efficacy.Results: Self-efficacy had significant main (B¼ 0.29; p5 0.01 for personalcontrol; B¼ 0.19; p5 0.05 for treatment control) and interaction effects(B¼ 0.38; p5 0.01 personal control on self-efficacy� timeline; B¼�0.31;p5 0.05 treatment control on self-efficacy� coherence).Conclusions: This study suggests that illness-specific representations ofolder people with multimorbidity are a product of both illness-specific andperson-level factors, such as self-efficacy. Strengthening individualself-efficacy may improve illness controllability regardless and on top ofillness-specific information.

Keywords: illness representations; multiple illnesses; self-regulation model;cognitive theory; self-efficacy; multilevel modelling

Introduction

The way people deal with their illnesses depends to a large extent on subjective illnessrepresentations, and these vary substantially between persons and illnesses. TheCommon-Sense Model (CSM) of health and illness (H. Leventhal, Brissette, &E.A. Leventhal, 2003; Leventhal, Meyer, & Nerenz, 1980) explains how individualsform such representations of their illnesses and how these representations in turn

*Corresponding author. Email: [email protected]

ISSN 0887–0446 print/ISSN 1476–8321 online

� 2012 Taylor & Francis

http://dx.doi.org/10.1080/08870446.2010.541908

http://www.tandfonline.com

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

guide their reactions. In this article, we aim at exploring to what extentrepresentations of specific illnesses are shaped by characteristics of the person andby specific aspects of the illnesses. We examine this in a sample of older adults withmultiple illnesses – a group at specific risk for increasing functional limitations andmortality (Fried, 2000). We employed a hierarchical framework based on CognitiveTheory (Clark & Beck, 1999) to account for the multilevel structure of globalcharacteristics of an individual and their representations of specific illnesses.

Multiple illnesses

Multimorbidity – the simultaneous occurrence of multiple illnesses within one person– is highly prevalent in older people. It has been estimated that between 61% (males)and 65% (females) of the adult population over 60 years suffer from two or more co-occurring diseases (van den Akker, Buntinx, Metsemakers, Roos, & Knotterus,1998). This accumulation of conditions poses a serious threat to health, functionalstatus and quality of life of those affected (Fortin et al., 2004). The CSM can help inlearning more about how individuals react to multiple illnesses in terms of adaptiveor maladaptive cognitive and behavioural responses. This is a particularly relevantissue, as with changing demographics, more and more individuals are likely to beaffected by multiple illnesses (Fried, 2000).

CSM of health and illness

The CSM assumes that people form subjective representations of their illnesses,which in turn determine their responses to these illnesses. It assumes thatrepresentations are organised along five dimensions: (1) Identity (i.e. the diseaselabel and individual ideas about prototypical somatic disease representations such assymptoms, location, etc.), (2) causes (i.e. causal attribution of an illness, e.g. to one’sown behaviour or genetic predispositions), (3) timeline (i.e. representations about theduration of a disease, in particular whether it is acute or chronic/cyclical), (4) cure/control (i.e. perceptions about the potential of curing the illness or controlling itscourse by personal means and/or medical treatment), and (5) consequences (i.e.beliefs about consequences of the illness). The CSM further assumes that parallel tothe cognitive representations in these five core dimensions, individuals formemotional representations of illnesses. These include emotional responses to theillness and concerns about the outcomes of the illness (Broadbent, Petrie, Main, &Weinman, 2006). Finally, the degree to which somebody is able to understand orcomprehend causes and cures of their illness constitutes representations of coherence.

The cure/control dimension has been separated into a personal control and atreatment control dimension, and current instruments such as the Revised IllnessPerceptions Questionnaire (IPQ-R; Moss-Morris et al., 2002) and the Brief IllnessPerception Questionnaire (B-IPQ; Broadbent et al., 2006) reflect this development.Personal control refers to how much a specific illness is perceived controllable byone’s own behaviour. Perceiving a specific illness to be controllable in turn is relatedto adaptive outcomes, e.g. cognitive reappraisals or initiating problem-focusedcoping, such as changing health behaviour (Hagger & Orbell, 2003). Treatmentcontrol refers to the degree a specific illness is perceived to be curable or manageableby specific treatment. These representations are related to medication adherence and

14 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

treatment fidelity, which in turn are important prerequisites for effective illnesstreatment (Horne, 2003; Horne, Weinman, & Hankins, 1999).

Hierarchical representations – Individual characteristics and illness-specific aspects

The content of illness representations is assumed to rely on informationacquired during the course of the illness (Leventhal et al., 2003; H. Leventhal,Forster, & E.A. Leventhal, 2007).

In this article, we propose that apart from such illness-specific information,general characteristics of the person affect the way individuals represent their specific(multiple) illnesses. This idea is comparable to the hierarchic structure of concrete,specific assumptions and higher order core beliefs in Cognitive Theory and CognitiveBehavioural Therapy (Clark & Beck, 1999).

Illness representations as specific assumptions

Illness representations are formed along a set of heuristics triggered by experiencingsymptoms – first, the symptoms are attributed to an illness depending on theindividual’s identity beliefs about specific illnesses. This attribution then determinesspecific representations about causes, timeline, consequences and controllability(Leventhal et al., 2007). This implies that such representations determine each other.In the hierarchical framework of Cognitive Theory, specific illness representationsform the level of concrete assumptions on specific topics.

Individual characteristics as core beliefs

Cognitive Theory assumes that a set of relatively stable, rigid and generalising corebeliefs affect the way an individual perceives, interprets and responds to specificinformation. Applied to illness representations, core beliefs would affect howsomebody translates e.g., illness symptoms into illness representations (concreteassumptions). Core beliefs result from generalised, cumulative experiences in variousfields of life, which are then applied to specific assumptions. For example, a personwho has experienced that he or she was able to successfully master novel tasks suchas learning new procedures at work, performing better in leisure sports after trainingor solving interpersonal problems, might perceive her-/himself as being capable tosolve and master further novel challenges, if she/he tries (mastery episodes in self-efficacy theory; Bandura, 1997). Such a person might then perceive illnesses ingeneral as better controllable by own behaviour than somebody with low self-efficacy.

This means that the way individuals perceive and interpret their illnesses is on theone hand dependent on illness-specific representations and on the other hand on thecontents and the directions of their higher order core self-beliefs. Following this logic,specific illness representations should be affected by the specific representations of anillness and their general core beliefs such as a sense of self-efficacy.

Previous research has shown that individuals high in self-efficacy are in fact morelikely to perceive their illnesses to be more controllable (Griva, Myers, & Newman,2000; Lau-Walker, 2004), and that self-efficacy interacts with treatment controlbeliefs (Nouwen, Law, Hussain, McGovern, & Napier, 2009). This supports the ideathat higher order core beliefs can affect individual representations of illnesses, but

Psychology and Health 15

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

this assumption has not been tested for multiple, potentially diverse illnesses withinthe same individuals.

Cognitive Theory also assumes that generalised core beliefs such as self-efficacycould transfer to expectations not only about personal control of an illness, but alsotreatment control of an illness (overgeneralisation).

Research questions

Our study aims at examining to which degree illness representations of individualswith multiple illnesses are determined by illness-specific aspects on the one andcharacteristics of the individual on the other hand. Using a hierarchic structure basedon the Cognitive Theory framework (Clark & Beck, 1999), we will examine the roleof core beliefs (self-efficacy) and concrete assumptions (representations of multiplespecific illnesses) in predicting control representations of specific illnesses inindividuals suffering from multimorbidity. In particular, we aim at answering thequestions in how far core beliefs predict representations of specific illnesses and inhow far core beliefs affect the interrelations of such specific illness representations.

The illness representation dimensions have been confirmed in a range of studies(for overviews see Hagger & Orbell, 2003; Moss-Morris et al., 2002). This is animportant prerequisite to our study – comparable dimensionality of illnessperceptions over a range of different illness contexts allows for examining illnessperceptions in potentially highly individualised clusters of multiple illnesses. Inaddition, the assumption that representations of illnesses rather than objectivefeatures are the factors that determine individual reactions (Leventhal, Nerenz, &Steele, 1984) is another important assumption for our study, because it implies thatattributes of illnesses affect individual reactions indirectly, mediated by theindividual representations of illnesses. This mediation assumption allows examina-tion of the representations of potentially different multiple illnesses within one study.

Our study focuses on personal control and treatment control. A review identifiedpersonal control representations as central for behavioural responses to illnesses(Hagger & Orbell, 2003). Similarly, treatment control perceptions are importantpredictors of adaptive behaviours in the face of chronic illnesses (Yohannes, Yalfani,Doherty, & Bundy, 2007).

The research questions and research objects are organised on different levels. Onthe one hand, there are representations of specific illnesses. As our sample consists ofindividuals suffering from multiple illnesses, representations of several of thesespecific illnesses are nested within one person. Self-efficacy, on the other hand, is afactor on the level of the individual suffering from multiple illnesses. We willtherefore examine whether such person-level variables affect the magnitude and theinteractions of illness-specific representations using multilevel modelling.

Method

Participants and procedure

Participants were recruited from the third assessment wave of the German AgeingSurvey (Wurm, Tomasik, & Tesch-Romer, 2010), a) population-representativesurvey of adults aged 40 and over with a total N of 6204. Participants wereconsidered eligible for our study if they were a) 65 years or older (n¼ 2728),

16 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

b) suffered from at least two conditions mentioned either in the CharlsonComorbidity Index (Charlson, Szatrowski, Peterson, & Gold, 1994) or theFunctional Comorbidity Index (Groll, To, Bombardier, & Wright, 2005; n¼ 644)and c) had provided explicit consent to be contacted for further studies and feasiblecontact details (n¼ 443). Of these, n¼ 309 (69.7%) participants gave informedconsent for this study. There were several reasons for not consenting and taking partin the study: In total, n¼ 134 (30.2%) could not be assessed. Of these, n¼ 61 (13.8%of eligible participants) could not be reached by telephone after several attempts,n¼ 24 (5.4%) were in inpatient care, and n¼ 49 (11.1%) declined to participate.Participants were visited at their homes by trained interviewers and completed a 30-minute interview about their personal health history (not reported here).Additionally, they filled in a questionnaire containing all assessments with a prepaidreturn envelope. Participants not returning the questionnaire received a postalreminder. In total, n¼ 305 questionnaires were available for the study.

Measures

Self-efficacy was assessed with a 5-item short version of the general self-efficacy scale(Schwarzer & Jerusalem, 1995).

Illnesses were assessed using a 24-item illness list that was informed by theillnesses included in the Charlson Comorbidity Index (Charlson et al., 1994) and theFunctional Comorbidity Index (Groll et al., 2005) and further included frequentgeriatric health problems such as hypertension, hyperlipidaemia and visual impair-ments. Participants were asked to indicate which of 24 conditions they suffered fromand were additionally asked to indicate the degree of subjective illness burden foreach ticked condition on a 4-point scale from 1 ‘no burden at all’ to 4 ‘very heavyburden.’

After filling in the illness list, participants were asked to select from this list thosetwo illnesses they considered most severe and fill in two Brief Illness PerceptionsQuestionnaire (B-IPQs; Broadbent et al., 2006) on subsequent pages in thequestionnaire using the selected illnesses as references. The B-IPQ is a shortinstrument assessing the dimensions of consequences, timeline, personal control,treatment control, identity, concern, coherence and emotional response with oneitem each. Coherence was reverse-coded for translation reasons. As a result of a pilotstudy in elderly individuals with multiple illnesses, the response format was changedto a 4-point Likert-type rating scale as compared to the original 10-point format.Additionally, potential causes of the illnesses were assessed in an open question (notreported here).

Analytical procedure

After descriptive analyses, selectivity analyses were conducted in order to comparethe total sample to the sample that provided data on multiple illness perceptions. Inorder to disentangle illness-specific and person-level influences on illness perceptions,multilevel modelling was applied. In this case, representations about specific illnesses(level-1-variables) are nested within individuals with personal characteristics(level-2-variables). Specific illness perceptions can thus be disentangled in illness-leveland person-level influences. The degree to which a specific illness perception reflects

Psychology and Health 17

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

person-level variance is analysed with the intraclass correlation coefficient (ICC),which is the proportion of level-2 variance to the total variance. Substantial ICCs(roughly 0.05 and larger) warrant multilevel analysis of data (Hox, 2002), smallerICCs suggest that most of the variance is due to differences of the level-1-units (inour case, representations of specific illnesses). In the case of a substantial ICC, it is ofimportance to identify level-2-characteristics that could explain the differencesbetween level-2-units (individuals), and to identify how these level-2-characteristicsare related to level-1-variables.

Specific illness representations can accordingly be decomposed as follows:

yij ¼ �00 þ �10Xij þ �01Zj þ �11ZjXij þ u0j þ rij:

Here, yij is an illness representation (e.g. personal control) for a specific illness i (e.g.diabetes; this constitutes the level-1-unit) within person j (level-2-unit). In thisexample, yij is predicted from another illness perception Xij (e.g. timeline) for theparticular illness with a regression coefficient �10, leading to an intercept of y whichthen can be decomposed into a mean intercept (across both illnesses per persons) �00and an illness-specific residual u0j. This is also referred to as the random interceptpart of a multilevel model.1 Apart from illness-specific cognitions, a person-levelcharacteristic Zj (level-2-variable, e.g. self-efficacy) could also be a predictor of yij(personal control) yielding a regression coefficient �01 and a level-2 residual rij.This means that the level of yij at the point where Xij equals 0 depends on the level ofZj. The level-2-predictor (self-efficacy) might even interact with the level-1-predictorXij (timeline) with a regression coefficient �11, leading to �11Zj Xij, which means thatthe degree to which Xij (timeline) predicts yij (personal control) depends on the levelof Zj (self-efficacy). See Figure 1 for a conceptual model.

The interrelations and both illness-level as well as person-level covariates ofmultiple illness representations were analysed using multilevel models with HLM6.03. Level-1-predictors (illness representations) were group-mean centred (i.e.centred within persons) and level-2-predictors were grand-mean-centred (i.e. centredacross all participants) in order to interpret both main and cross-level-interactioneffects (Enders & Tofighi, 2007). Descriptive and all other analyses were performedusing SPSS for Windows version 15.0.

Results

Participants were on average 73.3 years old (SD¼ 5.1), and 41.7% were women.They reported a mean of 5.5 (SD¼ 2.99) illnesses out of a list of 24 conditions(range 0–19).

Selectivity analysis

Selectivity analyses showed that there were no differences between participants fillingin both B-IPQs (n¼ 203), only one B-IPQ (n¼ 72) or no B-IPQ (n¼ 30) with regardto gender (�2¼ 1.66; df¼ 2; ns), age (F(2, 303)¼ 0.11; ns) and self-efficacy(F(2, 303)¼ 0.88; ns). There were significant differences with regard to the numberof illnesses (F(2, 303)¼ 18.77; p5 0.01), suggesting that participants with moreillnesses (6.24) were significantly more likely to fill in both B-IPQs than to fill in one

18 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

(4.52 illnesses) or none (3.62 illnesses; Tukey’s HSD post-hoc tests).

Similarly, participants with a higher illness burden (16.12 burden sum score) were

more likely to fill in both B-IPQs than to fill in just one (12.02 burden sum score) or

none (10.65 burden sum score; F(2, 297)¼ 8.94; p5 0.01), also based on Tukey’s

HSD post-hoc tests.

Illness prevalence and comorbidities

Table 1 shows the prevalence of the 10 most frequent illnesses from the illness list.

Table 2 shows the illnesses individuals considered most severe (first B-IPQ) and the

respective comorbidities (second B-IPQ).These diseases roughly correspond to the most frequent diseases in German

General Practices (Statistisches Bundesamt (Federal Statistical Office), 2009), with

cardiovascular, age-related and metabolic diseases being the most frequently

mentioned. However, the sample overrepresents visual impairments as well as

hearing impairment and underrepresents goitre, adipositas and hepatic diseases,

which might be due to the age and multimorbid characteristic of the sample. Table 2

demonstrates that the patterns of comorbidities are heterogeneous, but also suggests

that there are high co-prevalences of cardiovascular diseases and diabetes.

Figure 1. Hierarchical model of core beliefs (self-efficacy) on the level of the person (level-2)affecting personal control beliefs on the level of specific illnesses (level-1) and the intercepts aswell as slopes of specific illness representations predicting personal control on level-1 (cross-level interactions represented by nodes). The hierarchical model for predicting treatmentcontrol is structurally identical.

Psychology and Health 19

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

Illness representations

The first analysis step in multilevel analysis was to determine the degree to which thevariance of illness representations can be decomposed into level-1 (illness-specific)and level-2 (person-specific) variance components by examining the ICCs. The ICC ofpersonal control was �¼ 0.29 and that of treatment control was �¼ 0.19. Thissuggests that a substantial part of the variance is attributable to level-2 units, and thatignoring the multilevel structure would lead to biased estimates of standard errors.We subsequently tested whether models assuming random intercepts (i.e. theintercepts of a DV in level-1 units (illnesses) are similar within level-2 units (persons))in the prediction of level-1 variables personal control and treatment control yieldedthe best fit to the data or whether the inclusion of a level-2 predictor (self-efficacy) for

Table 2. Comorbidities (three most frequent) of the five most frequentlymentioned illnesses considered most severe by participants in the first B-IPQand the corresponding most frequent comorbidities mentioned in the secondB-IPQ.

Illness (n; % of B-IPQ 1) Comorbidities (n; %a)

Osteoarthritis (53; 26.1) Hypertension (9; 17)Diabetes (5; 9.4)Hearing impairments (4; 7.5)

Diabetes (18; 8.9) Hypertension (9; 45)Hyperlipidaemia (2; 11.1)Peripheral vascular disease (2; 11.1)

Hypertension (14; 6.9) Osteoarthritis (4; 28.6)Diabetes (2; 14.3)Congestive heart failure (2; 14.3)

Congestive heart failure (12; 5.9) Osteoarthritis (5; 41.7)Seven others with n¼ 1

COPD (10; 4.9) Hypertension (2; 20)Osteoarthritis (2; 20)Six others with n¼ 1

Note: aPercentage of those suffering from the respective illness in B-IPQ 1.

Table 1. List of the 10 most frequent illnesses from theillness list.

Illness n %

Hypertension 209 67.64Osteoarthritis 195 63.11Hyperlipidaemia 152 49.19Visual impairments 116 37.54Hearing impairment 104 33.66Arthritis 96 31.07Peripheral vascular disease 95 30.74Congestive heart failure 79 25.57Diabetes 74 23.95Glaucoma 61 19.74

20 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

both the intercepts and slopes improved the fit significantly. Deviance tests – thedeviance scores of nested models follow a �2 distribution with degrees of freedomequalling the difference of estimated parameters (Snijders & Bosker, 1999) – suggestedthat the model fit improves significantly if self-efficacy is included as level-2 predictor.Consequently, both personal control and treatment control were predicted in separatemultilevel regression models with all other illness perceptions serving as single2 level-1predictors and self-efficacy as level-2 predictors. Tables 3 and 4 show summaries ofthe multilevel regression analyses of personal control and treatment control on illnessperceptions and self-efficacy. We refrained from including more level-2 variables,such as SES in these equations, as the limited number of level-1 observations(illnesses) per level-2 unit (person) would have led to serious estimation problems.

Main effects

Personal control (Table 3) was significantly predicted by timeline (�10¼ 0.34,p5 0.001), which means that the longer an illness is perceived to last, the more

Table 3. Summary of multilevel regression analyses of personal control on illness perceptions(level-1) and self-efficacy (level-2).

Fixed effects CoefficientStandarderrors

Randomeffects

Standarddeviation

Variancecomponent

Level 2Intercept personal controlon self-efficacy

0.29** 0.08

Level 1– Consequences �0.07 0.06 Level-1 residual 0.39 0.16**

Intercept 0.64 0.42– Timeline 0.34** 0.04 Level-1 residual 0.43 0.18**

Intercept 0.59 0.35– Treatment control 0.40** 0.04 Level-1 residual 0.46 0.21**

Intercept 0.56 0.31– Identity �0.07 0.06 Level-1 residual 0.39 0.16**

Intercept 0.65 0.42– Concern �0.09 0.06 Level-1 residual 0.40 0.16**

Intercept 0.65 0.42– Coherence �0.08 0.06 Level-1 residual 0.40 0.16**

Intercept 0.65 0.42– Emotional response �0.15** 0.06 Level-1 residual 0.40 0.16**

Intercept 0.64 0.41

Cross-level interactions– Consequences�self-efficacy

0.14 0.13

– Timeline� self-efficacy 0.38** 0.12– Treatment control�self-efficacy

�0.06 0.09

– Identity� self-efficacy 0.05 0.11– Concern� self-efficacy 0.05 0.13– Coherence� self-efficacy �0.23 0.12– Emotional response�self-efficacy

�0.08 0.13

Note: **p5 0.01.

Psychology and Health 21

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

control over its outcomes individuals perceive. Treatment control also predictedpersonal control (�10¼ 0.40, p5 0.001), indicating that the more effective thetreatment for a specific illness is perceived to be, the more personal control over thecourse of the illness is perceived. Emotional response negatively predicted personalcontrol (�10¼�0.15, p5 0.001), which indicates that the more a specific illnesscauses emotional distress, the less controllable it is perceived to be.

Treatment control (Table 4) was predicted by timeline (�10¼ 0.26, p5 0.001),which means that illnesses with a longer temporal horizon are perceived to be bettercontrollable. Personal control also predicted treatment control (�10¼ 0.66,p5 0.001), indicating that if an illness is perceived controllable by one’s ownbehaviour, it is also more likely to be perceived controllable by medical treatment.

The intercepts of both personal control and treatment control were predicted byindividual levels of self-efficacy (�01 for personal control¼ 0.29, p5 0.01 and �01 fortreatment control¼ 0.19, p5 0.05). This means that higher levels of self-efficacy goalong with higher levels of perceived personal and treatment control in every

Table 4. Summary of multilevel regression analyses of treatment control on illnessperceptions (level-1) and self-efficacy (level-2).

Fixed effects CoefficientStandarderrors

Randomeffects

Standarddeviation

Variancecomponent

Level 2Intercept treatmentcontrol on self-efficacy

0.19* 0.09

Level 1– Consequences �0.09 0.08 Level-1 residual 0.39 0.16**

Intercept 0.83 0.70– Timeline 0.26** 0.06 Level-1 residual 0.42 0.18**

Intercept 0.81 0.66– Personal control 0.66** 0.06 Level-1 residual 0.49 0.25**

Intercept 0.72 0.51– Identity �0.02 0.08 Level-1 residual 0.40 0.16**

Intercept 0.83 0.70– Concern 0.10 0.08 Level-1 residual 0.41 0.16**

Intercept 0.83 0.69– Coherence �0.11 0.08 Level-1 residual 0.41 0.16**

Intercept 0.83 0.69– Emotional response �0.06 0.07 Level-1 residual 0.39 0.16**

Intercept 0.84 0.70

Cross-level interactions– Consequences�self-efficacy

0.12 0.17

– Timeline� self-efficacy �0.01 0.16– Personal control�self-efficacy

�0.15 0.12

– Identity� self-efficacy 0.19 0.15– Concern� self-efficacy �0.32 0.16– Coherence� self-efficacy �0.31* 0.16– Emotional response�self-efficacy

�0.07 0.16

Notes: **p5 0.01 and *p5 0.05.

22 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

prediction when the predicting variable is zero. Or, more simply, the more self-efficacious somebody is, the more personal control and treatment efficacy concern-ing a specific illness he or she will perceive.

Cross-level interactions

Apart from the main effects of self-efficacy and timeline in predicting personalcontrol (Figure 2), there was a significant cross-level interaction of self-efficacy andtimeline (�11¼ 0.38, p5 0.01). Simple slopes analyses (Aiken & West, 1991) suggestthat in self-efficacious persons (level-2), timeline (level-1) is more predictive ofwhether an illness is perceived personally controllable (level-1) than in less self-efficacious persons.

In addition, there was a marginally significant cross-level interaction of self-efficacy with (in)coherence (�11¼�0.23, p¼ 0.07). Simple slopes analyses suggestedthat in more self-efficacious people (level-2), illnesses are perceived equallycontrollable (level-1) irrespective of whether they are perceived coherent or not(level-1). In less self-efficacious people, illnesses perceived incoherent are more likelyto be also perceived uncontrollable.

For treatment control (Figure 3), there was a significant cross-level interaction ofself-efficacy with (in)coherence (�11¼�0.31, p5 0.05). Here, simple slopes analysessuggested that more self-efficacious individuals (level-2) are more likely to perceive aspecific illness to be controllable by treatment (level-1), irrespective of whether it isperceived incoherent or not. Less self-efficacious people are more likely to perceivean illness to be less controllable by treatment, if they also perceive it to be incoherent.

There also was a marginally significant cross-level interaction of self-efficacy andconcern (�11¼�0.32, p¼ 0.08). Simple slopes analyses suggested that more self-efficacious people (level-2) perceive an illness to be controllable by treatment (level-1)irrespective of possible concerns this illness causes (level-1). In less self-efficaciouspeople, illnesses causing concerns are also less likely to be perceived controllable bytreatments.

Figure 2. Cross-level interaction of self-efficacy (level-2) and timeline (level-1) in predictingpersonal control (level-1) at higher (þ1 SD), mean and lower (�1 SD) levels of self-efficacy.

Psychology and Health 23

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

Discussion

In this study, we examined the degree to which control representations of specificillnesses are shaped by specific representations of these illnesses and by globalcharacteristics of the person in a sample of older adults with multiple illnesses. Weapplied a hierarchical framework based on Cognitive Theory (Clark & Beck, 1999) tostructure the relation between specific illness representations and individual corebeliefs. In particular, we examined the prediction of treatment and controlrepresentations of specific illnesses by other representations of the same illness andby self-efficacy core beliefs on the level of the person suffering from multiple illnesses.

Illness representations in multimorbid individuals

We found that illness representations tend to be clustered within persons, indicatedby substantial and significant ICCs. This suggests that substantial parts of thevariance in specific illness representations are due to variation on the level of theperson suffering from multiple illnesses. This means that in addition to the specificcontents of illness representations, individual characteristics shape the way peoplethink about specific illnesses. This can be interpreted along the hierarchical structurein Cognitive Theory: Specific representations of multiple illnesses within one personare very likely to be both dependent on information specific to the illness (e.g.someone might have short timeline representations about his/her broken leg, butchronic timeline representations about his/her diabetes), but also by individual corebeliefs affecting these specific beliefs (e.g. a stable disposition to interpret all illnessesas long-lasting). This could lead to a person with the disposition to rate the timelinesof a broken leg and diabetes as more similarly long-lasting than a person without thisdisposition.

Predicting personal and treatment control representations – The illness level

We found that illness-specific representations of personal control and treatmentcontrol were predicted by timeline. This means that if a specific illness is perceived to

Figure 3. Cross-level interaction of self-efficacy (level-2) and reverse-coded coherence(incoherence; level-1) in predicting treatment control (level-1) at higher (þ1 SD), mean andlower (�1 SD) levels of self-efficacy.

24 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

be long-lasting, it is also more likely to be perceived controllable. At first sight, thisfinding might seem counter-intuitive because longer timelines of illnesses often areassociated with increased distress (Hagger & Orbell, 2003; Llewellyn, McGurk, &Weinman, 2007). However, acute illnesses with limited timeline often are the resultsof uncontrollable events such as accidents or infections, whereas longer lasting,especially chronic illnesses involve illness management strategies, which are subjectto personal control. Illnesses which were associated with strong emotional responseswere perceived less controllable, which might be an indicator of illness severity. Thefinding that illness representations about treatment control are significantlypredicted by representations about personal control and vice versa is in line withself-efficacy theory (Bandura, 1997), which would classify treatment such asmedication as a means to exert control over outcomes of illnesses.

Self-efficacy and specific personal/treatment control representations

In addition to the specific illness-level representations, self-efficacy on the level of theperson suffering from multiple illnesses predicted the intercepts of both personalcontrol and treatment control. More self-efficacious persons were more likely toperceive their specific illnesses to be under personal and treatment control and viceversa. There were also cross-level interactions of self-efficacy and specific illnessrepresentations: self-efficacy moderated the degree to which the expected timeline ofan illness predicted the controllability of this illness, and it also moderated the degreeto which the coherence of an illness was predictive for treatment control.

The finding that self-efficacy core beliefs predict illness-specific representations ofpersonal control is especially interesting, as it suggests that apart from representationsabout the controllability of a specific illness, person-level core beliefs about agencyand control might be important for how specific illnesses are perceived. This is anessential finding, as it suggests that strengthening an individual’s self-efficacy beliefsmight have implications for how controllable somebody perceives a specific illness,over and above other illness-related cognitions.

Self-efficacy core beliefs also predicted illness-specific treatment control repre-sentations indicating that more self-efficacious people were also more likely toperceive a specific illness to be better controllable by treatments. Beliefs about theefficacy of treatment strongly affect treatment adherence (Yohannes et al., 2007),and our findings suggest that apart from illness-specific representations abouttimeline or personal control, self-efficacy as a person-level characteristic positivelyaffects the degree to which treatment is perceived to help with a specific illness.Nevertheless, beliefs about the effectiveness of personal means to control a specificillness seem to be more important predictors of treatment control than self-efficacyon the level of the individual. However, the significant cross-level effect of self-efficacy on treatment control suggests that strengthening a person’s general controlbeliefs might also affect beliefs about very specific and potentially diverse illnesses.This can be due to several pathways: the efficacy of treatments depends on whethera treatment is administered in the frequency, regularity and intensity as prescribed.Individuals who feel self-efficacious are more likely to feel able to adhere toprescribed treatments (Griva et al., 2000) and might thus perceive a higherpotential for treatments to control their illness. On the other hand, the hierarchicalstructure of core beliefs and specific assumptions outlined in Cognitive Theory

Psychology and Health 25

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

suggests that core beliefs such as self-efficacy can, due to the overgeneralisation ofcore beliefs, also affect other control-related specific representations such astreatment control beliefs.

We applied this hierarchical framework to examine the idea that in individualssuffering from multiple illnesses, core beliefs on the level of the individual such asself-efficacy affect multiple representations of specific illnesses nested within thisperson, both in terms of extent and interactions. Our data show that individual-level variables affect illness-specific representations. This goes beyond previousresearch reporting a relation between self-efficacy and control representations(Griva et al., 2000; Lau-Walker, 2006) in that we show that this relation pertainsto different illnesses within one person. Our study shows for the first time thatthese personal characteristics could transfer across different illnesses nested withinone person. Self-efficacy core beliefs interacted with illness-specific timelinerepresentations indicating that more self-efficacious individuals rated an illness tobe more controllable if it had an extended timeline. As mentioned above, thetimeframe of an illness affects its potential for personal and treatment control.More self-efficacious people might be better able to realise this potential and thusincrease cognitions about personal control in illnesses with longer time frames. Inaddition, an extended time perspective on an illness might suggest that thisparticular illness is non-fatal and might thus be perceived less severe and bettercontrollable.

In predicting specific treatment control representations, self-efficacy corebeliefs interacted with illness coherence representations. Individuals high in self-efficacy perceived an illness to be better controllable by treatment, regardless ofwhether this illness made sense to them or not. Self-efficacy acted as a bufferingfactor in this respect, as it attenuated adverse effects of incoherence on treatmentefficacy beliefs. This is important, as it suggests that supporting self-efficacymight also improve the belief that a particular illness can be controlled bytreatment, even if this illness does not make sense to the individual. This in turnmight improve treatment fidelity and medication adherence (French, Cooper, &Weinman, 2006).

Our study suggests that illness representations of individuals with multipleillnesses can be conceptualised in a hierarchy of core beliefs and specificrepresentations. Core beliefs (here: self-efficacy) might play a significant role inindividuals’ representations of their illnesses. This is of particular importance in thecontext of multiple illnesses, a frequent condition in older adults (Fried, 2000; vanden Akker et al., 1998), because it is essential to understand which factors predictadaptive illness-related cognitions. In complex and individually different multi-morbidity patterns, it might be hard to distinguish and eventually change specificillness-related beliefs.

Limitations

There are some conceptual and methodological limitations to our study. Due to theimpaired health status of our participants, we refrained from using the full illnessperception questionnaire (Moss-Morris et al., 2002) with probably more reliableassessments of illness perceptions. We also refrained from analysing the individualcauses of illnesses, as these are assessed in an open format in the Brief-IPQ

26 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

(Broadbent et al., 2006), and could thus not be included in the multilevel analyses.

Similarly, we did not assess medication beliefs specific to the illnesses. A wide body

of research shows that beliefs about treatment and medicines are important

predictors of adaptive behavioural outcomes (Horne, 2003). Future research

therefore should examine specific medication beliefs and potential interactions in

people with multiple illnesses. Furthermore, the assessment of only two illnesses per

participant might be an oversimplification of the complex picture of multimorbidity

in advanced age – for example, individuals aged 75 and over suffer, on average, from

5 conditions simultaneously (van den Akker et al., 1998). However, this would have

meant to employ five Brief-IPQs and more per person, which was not feasible in the

study protocol. The analyses were based on cross-sectional data, which impairs the

strong causality assumptions implied in the regression approach, particularly with

regard to the direction of the examined predictions. However, given that core beliefs

are relatively stable (Clark & Beck, 1999) and that illness representations are prone

to change over time (Leventhal et al., 2003), our directed analyses might be justified.

Future studies should examine changes in multiple illness perceptions over time in

cross-lagged designs to establish the causal direction of predictions and to examine

differences and potential interactions of multiple illness perceptions within one

person in predicting individual coping attempts. The focus on the two illnesses

deemed most severe by the participants might have led to limited variance in the

illness representations, particularly identity, timeline, consequences and the control

representations, as illnesses tend to cause greater distress if these representations are

prominent (Hagger & Orbell, 2003). Due to restrictions on the number of level-1

units (illnesses), we were not able to include more predictors on the level of the

person, such as demographic variables. Similarly, our focus on self-efficacy as solecore belief on the level of the person might be an oversimplification of the picture, as

a range of other individual characteristics could affect the way illness representations

influence each other. Future studies might want to employ more predictors on the

level of the person, such as optimism (Llewellyn, Weinman, McGurk, & Humphris,

2008) or personality factors such as neuroticism (Millar, Purushotham, McLatchie,

George, & Murray, 2005). For further applications in multimorbidity, especially with

regard to predicting outcomes from multiple illness representations, it might be

necessary to reconceptualise the CSM or add further dimensions, such as the

subjective prioritisation of one illness over another.

Conclusion

Notwithstanding these limitations, we think that our study makes an important

contribution to the literature on illness perceptions in that it is the first study to

examine multiple illness perceptions by accounting for the hierarchical structure of

multiple illness representations within one person. It is unique in showing the

importance of person-level factors, in particular self-efficacy, for illness-specific

personal control and treatment control representations. This suggests that interven-

tions targeted at improving the perceptions relevant for adaptive reactions to

illnesses should aim at strengthening individual general self-efficacy beliefs over a

range of life domains, as this has implications for how controllable illnesses are

perceived – regardless and on top of illness-specific information.

Psychology and Health 27

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

Acknowledgements

This study (PREFER) was funded by the German Federal Ministry of Education andResearch (Grant no. 01ET0702). The German Ageing Survey was funded by the GermanFederal Ministry for Family, Senior Citizens, Women and Youth (Grant No. 301-1720-2/2).The content is solely the responsibility of the authors.

Notes

1. Note that the slopes of Xij in predicting yij were not allowed to vary across level-1-units(illnesses), as the number of level-1-units per level-2-unit (individuals) was restricted to 2,which does not allow us to estimate random slopes.

2. Simple regressions were used, as every level-2 unit of analysis consisted of only n¼ 2illnesses at level 1. More predictors could have led to instable parameter estimates.

References

Aiken, L.S., & West, S.G. (1991). Multiple regression: Testing and interpreting interactions.

Thousand Oaks, CA: Sage.Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: WH Freeman/

Times Books/Henry Holt and Co.

Broadbent, E., Petrie, K.J., Main, J., & Weinman, J. (2006). The brief illness perceptionquestionnaire. Journal of Psychosomatic Research, 60, 631–637.

Charlson, M.E., Szatrowski, T.P., Peterson, J., & Gold, J. (1994). Validation of a combined

comorbidity index. Journal of Clinical Epidemiology, 47, 1245–1251.Clark, D.A., & Beck, A.T. (1999). Scientific foundations of cognitive theory and depression.

New York: Wiley.Enders, C.K., & Tofighi, D. (2007). Centering predictor variables in cross-sectional multilevel

models: A new look at an old issue. Psychological Methods, 12, 121–138.Fortin, M., Lapointe, L., Hudon, C., Vanasse, A., Ntetu, A.L., & Maltais, D. (2004).

Multimorbidity and quality of life in primary care: A systematic review. Health andQuality of Life Outcomes, 2, 51.

French, D.P., Cooper, A., & Weinman, J. (2006). Illness perceptions predict attendance at

cardiac rehabilitation following acute myocardial infarction: A systematic review withmeta-analysis. Journal of Psychosomatic Research, 61, 757–767.

Fried, L.P. (2000). Epidemiology of aging. Epidemiologic Reviews, 22, 95–106.Griva, K., Myers, L.B., & Newman, S. (2000). Illness perceptions and self efficacy beliefs in

adolescents and young adults with insulin dependent diabetes mellitus. Psychology andHealth, 15, 733–750.

Groll, D.L., To, T., Bombardier, C., & Wright, J.G. (2005). The development of a

comorbidity index with physical function as the outcome. Journal of ClinicalEpidemiology, 58, 595–602.

Hagger, M.S., & Orbell, S. (2003). A meta-analytic review of the common-sense model ofillness representations. Psychology and Health, 18, 141–184.

Horne, R. (2003). Treatment perceptions and self-regulation. In L.D. Cameron &H. Leventhal (Eds.), The self-regulation of health and illness behaviour (pp. 138–153).

New York, NY: Routledge.Horne, R., Weinman, J., & Hankins, M. (1999). The beliefs about medicines questionnaire:

The development and evaluation of a new method for assessing the cognitive

representation of medication. Psychology and Health, 14(1), 1–24.Hox, J.J. (2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Erlbaum.

Lau-Walker, M. (2004). Relationship between illness representation and self-efficacy. Journalof Advanced Nursing, 48, 216–225.

28 B. Schuz et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12

Lau-Walker, M. (2006). Predicting self-efficacy using illness perception components: A patientsurvey. British Journal of Health Psychology, 11, 643–661.

Leventhal, H., Brissette, I., & Leventhal, E.A. (2003). The common-sense model of self-regulation of health and illness. In L.D. Cameron & H. Leventhal (Eds.), The self-

regulation of health and illness behaviour (pp. 42–65). Howard: Routledge.Leventhal, H., Forster, R., & Leventhal, E.A. (2007). Self-regulation of health threats, affect,

and the self: Lessons from older adults. In C.M. Aldwin, C.L. Park, & A. Spiro III (Eds.),

Handbook of health psychology and aging (pp. 341–366). Oxford: Guilford Press.Leventhal, H., Meyer, D., & Nerenz, D.R. (1980). The common sense representation of illness

danger. In S. Rachman (Ed.), Medical psychology (pp. 7–30). New York, NY: Pergamon

Press.Leventhal, H., Nerenz, D.R., & Steele, D.J. (1984). Illness representations and coping with

health threats. In A. Baum, S.E. Taylor, & J.E. Singer (Eds.), Handbook of psychology and

health (Vol. 4, pp. 219–252). Hillsdale, NJ: Erlbaum.Llewellyn, C.D., McGurk, M., & Weinman, J. (2007). Illness and treatment beliefs in head and

neck cancer: Is Leventhal’s common sense model a useful framework for determiningchanges in outcomes over time? Journal of Psychosomatic Research, 63, 17–26.

Llewellyn, C.D., Weinman, J., McGurk, M., & Humphris, G. (2008). Can we predict whichhead and neck cancer survivors develop fears of recurrence? Journal of PsychosomaticResearch, 65, 525–532.

Millar, K., Purushotham, A.D., McLatchie, E., George, W.D., & Murray, G.D. (2005). A1-year prospective study of individual variation in distress, and illness perceptions, aftertreatment for breast cancer. Journal of Psychosomatic Research, 58, 335–342.

Moss-Morris, R., Weinman, J., Petrie, K.J., Horne, R., Cameron, L.D., & Buick, D. (2002).The revised illness perception questionnaire (IPQ-R). Psychology and Health, 17(1), 1–16.

Nouwen, A., Law, G.U., Hussain, S., McGovern, S., & Napier, H. (2009). Comparison of therole of self-efficacy and illness representations in relation to dietary self-care and diabetes

distress in adolescents with type 1 diabetes. Psychology and Health, 24, 1071–1084.Schwarzer, R., & Jerusalem, M. (1995). Generalized self-efficacy scale. In J. Weinman,

S. Wright, & M. Johnston (Eds.), Measures in health psychology: A user’s portfolio. Causal

and control beliefs (pp. 35–37). Windsor, UK: NFER-NELSON.Snijders, T.A.B., & Bosker, R. (1999). Multilevel analysis: An introduction to basic and

advanced multilevel modeling. New York, NY: Sage.

Statistisches Bundesamt (Federal Statistical Office). (2009). Haufigste Diagnosen in Prozentder Behandlungsfalle [Most common diagnoses in percent of cases]. Retrieved from http://www.gbe-bund.de/oowa921-install/servlet/oowa/aw92/dboowasys921.xwdevkit/xwd_init?

gbe.isgbetol/xs_start_neu/33750472/567647van den Akker, M., Buntinx, F., Metsemakers, J.F.M., Roos, S., & Knotterus, J.A. (1998).

Multimorbidity in general practice: Prevalence, incidence, and determinants ofco-occurring chronic and recurrent diseases. Journal of Clinical Epidemiology, 51, 367–375.

Wurm, S., Tomasik, M.J., & Tesch-Romer, C. (2010). On the importance of a positive view onaging for physical exercise among middle-aged and older adults: Cross-sectional andlongitudinal findings. Psychology and Health, 25, 25–42.

Yohannes, A.M., Yalfani, A., Doherty, P., & Bundy, C. (2007). Predictors of drop-out froman outpatient cardiac rehabilitation programme. Clinical Rehabilitation, 21, 222–229.

Psychology and Health 29

Dow

nloa

ded

by [

Uni

vers

ity o

f T

asm

ania

] at

20:

25 2

6 Ja

nuar

y 20

12