15
Assessment The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index Katarzyna Zawisza, 1 Aleksander Galas, 2 Beata Tobiasz-Adamczyk, 1 * Somnath Chatterji, 3 Josep Maria Haro, 4,5 Marta Miret, 4,6,7 Seppo Koskinen, 8 Mick Power 9 and Matilde Leonardi 10 1 Department of Medical Sociology, Chair of Epidemiology and Preventive Medicine, Jagiellonian University Medical College, Krakow, Poland 2 Department of Epidemiology, Chair of Epidemiology and Preventive Medicine, Jagiellonian University Medical College, Krakow, Poland 3 Department of Health Statistics and Information Systems, World Health Organization, Geneva, Switzerland 4 Instituto de Salud Carlos III, Centro de Investigación en Red de Salud Mental, CIBERSAM, Madrid, Spain 5 Parc Sanitari Sant Joan de Déu, Universitat de Barcelona, Sant Boi de Llobregat, Barcelona, Spain 6 Department of Psychiatry, Hospital Universitario de La Princesa, Instituto de Investigación Sanitaria Princesa (IP), Madrid, Spain 7 Department of Psychiatry, Universidad Autónoma de Madrid, Madrid, Spain 8 National Institute for Health and Welfare, Helsinki, Finland 9 Section of Clinical and Health Psychology, University of Edinburgh, Edinburgh, UK 10 Fondazione IRCCS, Neurological Institute Carlo Besta, Milano, Italy The aim of the study was to create a simplied, easy implementable multidimensional instrument to assess all relevant elements of the structure and function of social network within individuals across different European countries and to provide the tool for health professionals and policy makers. The analysis was based on the sample of 10 446 non-institutionalized adult population from Finland, Poland and Spain. The Social Network Questionnaire Collaborative Research on Ageing in Europe Social Network Index (COURAGE-SNI) was part of the COURAGE questionnaire. The indicators of the functioning of social network ties (close relations), frequency of direct contact and general support were evaluated. Functions were assess within the main structural components as spouse, parents, children, grandchildren, other relatives, friends, coworkers and neighbours. The exploratory factor analysis revealed ve main latent components of social network with one component com- posed of hierarchical part. The conrmatory factor analysis provided an acceptable t for the model. The general- ize partial credit model was used to calculate factor scores for ve components of the COURAGE-SNI considering the social networks of spouse/partner, parents, other family members , neighboursand friends and co- workers. The scores for every component were recalculated so as to provide the social network saturation ranged from 0 (the lowest) to 100% (the highest possible). Finally, the COURAGE-SNI score was obtained as the sum of weighted information calculated by the item response theory procedure for every aforementioned component. In summary, the COURAGE-SNI showed good reliability and content validity and seems to be a promising tool for the assessment of the social network phenomenon across European countries. Copyright © 2013 John Wiley & Sons, Ltd. Key Practitioner Message: The Courage-SNI is a new tool to assess the construct of social network in population studies. The Courage-SNI is an instrument useful to identify high risk groups or populations whose social network is poorer. Keywords: Social Networks, Measurement, Scale, Validity, Reliability, Item response theory *Correspondence to: Beata Tobiasz-Adamczyk, Ph.D., Department of Medical Sociology, Chair of Epidemiology and Preventive Medicine, Jagiellonian University Medical College, Kopernika 7a St, 31034 Krakow, Poland. E-mail: [email protected] Clinical Psychology and Psychotherapy Clin. Psychol. Psychother. 21, 227241 (2014) Published online 12 August 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpp.1860 Copyright © 2013 John Wiley & Sons, Ltd.

The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

  • Upload
    matilde

  • View
    217

  • Download
    3

Embed Size (px)

Citation preview

Page 1: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Assessment

The Validity of the Instrument to Evaluate SocialNetwork in theAgeing Population: The CollaborativeResearch on Ageing in Europe Social Network Index

Katarzyna Zawisza,1 Aleksander Galas,2 Beata Tobiasz-Adamczyk,1*Somnath Chatterji,3 Josep Maria Haro,4,5 Marta Miret,4,6,7 Seppo Koskinen,8Mick Power9 and Matilde Leonardi101Department of Medical Sociology, Chair of Epidemiology and Preventive Medicine, Jagiellonian University MedicalCollege, Krakow, Poland2Department of Epidemiology, Chair of Epidemiology and Preventive Medicine, Jagiellonian University MedicalCollege, Krakow, Poland3Department of Health Statistics and Information Systems, World Health Organization, Geneva, Switzerland4 Instituto de Salud Carlos III, Centro de Investigación en Red de Salud Mental, CIBERSAM, Madrid, Spain5Parc Sanitari Sant Joan de Déu, Universitat de Barcelona, Sant Boi de Llobregat, Barcelona, Spain6Department of Psychiatry,HospitalUniversitario de LaPrincesa, Instituto de Investigación Sanitaria Princesa (IP),Madrid, Spain7Department of Psychiatry, Universidad Autónoma de Madrid, Madrid, Spain8National Institute for Health and Welfare, Helsinki, Finland9 Section of Clinical and Health Psychology, University of Edinburgh, Edinburgh, UK10 Fondazione IRCCS, Neurological Institute Carlo Besta, Milano, Italy

The aim of the study was to create a simplified, easy implementable multidimensional instrument toassess all relevant elements of the structure and function of social network within individuals acrossdifferent European countries and to provide the tool for health professionals and policy makers.The analysis was based on the sample of 10 446 non-institutionalized adult population from Finland,Poland and Spain. The Social Network Questionnaire Collaborative Research on Ageing in Europe SocialNetwork Index (COURAGE-SNI) was part of the COURAGE questionnaire. The indicators of thefunctioning of social network ties (close relations), frequency of direct contact and general support wereevaluated. Functions were assess within the main structural components as spouse, parents, children,grandchildren, other relatives, friends, coworkers and neighbours.The exploratory factor analysis revealedfivemain latent components of social networkwith one component com-posed of hierarchical part. The confirmatory factor analysis provided an acceptablefit for themodel. The general-izepartial creditmodelwasused to calculate factor scores forfive components of theCOURAGE-SNI consideringthe social networks of ‘spouse/partner’, ‘parents’, ‘other family members’, ‘neighbours’ and ‘friends and co-workers’. The scores for every componentwere recalculated so as to provide the social network saturation rangedfrom 0 (the lowest) to 100% (the highest possible). Finally, the COURAGE-SNI score was obtained as the sum ofweighted information calculated by the item response theory procedure for every aforementioned component.In summary, the COURAGE-SNI showed good reliability and content validity and seems to be a promisingtool for the assessment of the social network phenomenon across European countries. Copyright © 2013 JohnWiley & Sons, Ltd.

Key Practitioner Message:• The Courage-SNI is a new tool to assess the construct of social network in population studies.• TheCourage-SNI is an instrumentuseful to identifyhigh riskgroupsorpopulationswhose social network ispoorer.

Keywords: Social Networks, Measurement, Scale, Validity, Reliability, Item response theory

*Correspondence to: Beata Tobiasz-Adamczyk, Ph.D., Department ofMedical Sociology, Chair of Epidemiology and Preventive Medicine,Jagiellonian University Medical College, Kopernika 7a St, 31034Krakow, Poland.E-mail: [email protected]

Clinical Psychology and PsychotherapyClin. Psychol. Psychother. 21, 227–241 (2014)Published online 12 August 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpp.1860

Copyright © 2013 John Wiley & Sons, Ltd.

Page 2: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

INTRODUCTION

The development of the concept of ‘social network’ hasopened up a new area in research that focused on relation-ships between different dimensions of social network andhealth (Berkman & Glass, 2000). Several studies haveshown the impact of social network, not only on mortality(Bowling & Grundy, 1998; Iwasaki et al., 2002; Kawachiet al., 1996; Olsen, 1993; Rutledge, Matthews, Lui, Stone, &Cauley, 2003; Shye, Mullooly, Freeborn, & Pope, 1995), butalso on health-related quality of life and on mental health(Berkman, 1995; Finch, Okun, Barrera, Zautra, & Reich,1989; López García, Banegas, Graciani Pérez-Regadera,Herruzo Cabrera, & Rodríguez-Artalejo, 2005; Smith &Christakis, 2008; Mendes de Leon et al., 1999).Different definitions of social network have been

developed, on the basis of various theoretical frameworksto identify social relationships that surround a person,their characteristics and the individual's perception ofthem (Victor, Scambler, Bond, & Bowling, 2000). Socialnetworks refer to the structural aspects of socialrelationships—they are the channels through whichpragmatic help as well as emotional and psychologicalsupports can be exchanged between individuals (Achat,Kawachi, Levine, Berkey, & Coakley, 1998) as set of nodesthat are tied to one another by the types of relationsbetween them (Moren-Cross & Lin, 2006).Depending on the definition used, different aspects of

social network can be taken into account: objective andsubjective; also the interrelations between social network,social support and social capital need explanation beforeplanning the research (Due, Holstein, Lund, Modvig, &Avlund, 1999).The interest in the role of social network in health

outcomes is not free from different theoretical and method-ological ambiguities, difficulties and controversies. Bowling(1997) has mentioned that existing research on socialnetwork and health outcomes suffers from methodologicalproblems, such as imprecise definitions andmethodologicallimitations due to lack of well validated indicators for thismultidimensional concept.The review of existing theoretical frameworks as well as

empirical experience in the development of the instrumentsmeasuring social network reveals several significant chal-lenges to be faced by researchers. The analysis of methodsused to measure social network performed by Wenger(1997a, 1997b) showed the development of aspects of themethodological approach. Those highlighted are as follows:

1. Network mapping—based on the list of all peopleknown to an individual. This method shows theaffective or elective nature of the network ties.

2. Proxy measures—based on network indicators, suchas the availability of adult children or siblings orfriends within 15min travel time or 20miles or in the

same community; households composition; theavailability of confidant or having friends in the sameneighbourhood.

3. Measuring partial and purposive networks—developed as a way of overcoming the problemsexperienced with the two previous methods ofmeasurement, based on the list of specific people whofill specific roles, perform special actions or help inparticular ways in relation to respondents.

4. Measurement of network structures (less developed)—based on the correlation between the network type,demographic and outcome variables (Wenger, 1997a,1997b).

Regarding the general structure of social network,characteristics of individual ties include the following:frequency of contact (number of face-to-face contactsand/or contacts by phone or mail), multiplexity (thenumber of types of transactions or support flowingthrough a set of ties), duration (the length of time that anindividual knows another) and reciprocity (the extent towhich exchanges or transactions are even or reciprocal)(pp. 847-848, Berkman & Glass, 2000).In order to assess/analyse social networks, the following

considerations have to be made: range or size (number ofnetwork members), density (the extent to which the mem-bers are connected to each other), boundedness (the degreeto which they are defined on the basis of traditional groupstructures such as kin, work and neighbourhood) andhomogeneity (the extent to which individuals are similarto each other in a network) (p. 847, Berkman & Glass, 2000).From the methodological perspective, Bowling (1997)

reported that many surveys relied on single-item questionsonly, such as marital status, frequency of contacts withothers and existence of a confidant. Such approach wasused by Berkman and Syme (1979), who measured thefrequency of social relations and activities: visits withfriends and relatives, involvement in formal organizations(church and voluntary associations), active and socialleisure and number of hours per day spent on solitaryactivity. The same approachwas also employed by Seeman,Kaplan, Knudsen, Cohen, and Guralnik (1987) and in 1994,Sugisawa, Liang, & Liu, who followed Berkman and Syme'sline of research by measuring three indicators of social net-works: social contacts, social participation and maritalstatus.Litwin (2001) proposed to assess supportive network

and social network based on data on interpersonal tiesthat people of all ages maintain in varying contexts. Shyeet al. (1995) gathered information on marital status,network size and frequency of social interactions. Net-work size was measured by the number of people withwhom the respondent reported informal social contactsacross all available categories of social relationships(family, kin, friends, neighbours and co-workers).

228 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 3: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Frequency of social interaction was a summary measureof the degree of social contact the respondent reportedwith relatives (other than those lived with), friends andco-workers and frequency of attending church andmeetings of organizations the person belonged to.Several approaches have been used to define the indica-

tors of social network, such asmarital status or the structureof hierarchy in mapping of social relations (Antonucci,1986), as well as an evaluation of the individual's socialrelations structure (Ajrouch, Blandon, & Antonucci, 2005).In the study by López García et al. (2005), social networkwas assessed by asking four questions: ‘what is yourmaritalstatus?’, ‘with whom are you currently living?’, ‘do you seefamily members other than those who are living with you?’and ‘do you see friends or neighbours?’. This informationwas subsequently used to make a classification consistentwith the network typology developed by Wenger (1997a,1997b). In another study performed by Gallegos-Carrilloet al. (2009), the structure of the social network wasmeasured by estimating the availability of contacts offeringresources within the network and the size of the socialnetwork (based on specific elements of social network, suchas marital status, living alone, size of the network of closerelatives and size of the network of available friends). Scottand Hofmeyer (2007) studied social relationships in fami-lies, communities, teams, organizations and othercollectives in relation to social capital—identifying thenature and the extent of the impact of social relationshipstaking into account the definition of social capital as powergradients in society across which networks may or may notprovide links.Many instruments intended to measure social support

have been developed. However, according to Bowling,some of these have poor, and majority have only modestagreement with conceptual definitions (Bowling, 1997).These are the Social Relationship Scale (McFarlane, 1981;reliability correlations 0.62–0.99 among 73 students), theSocial Support Questionnaire (Sarason, 1983; reliability:inter-item correlations 0.35–0.71, for satisfaction scores0.21–0.71 among 103 students), the Research and Develop-ment social health battery (Research and DevelopmentCorporation, 1978; reliability: only five of 45 correlationsexceeding 0.40), the Medical Outcomes Study Social Sup-port Survey (Sherbourne and Stewart, 1991; 1-y test-retestreliability 0.72–0.76 for each subscale), the Duke—Univer-sity of North Carolina Functional Social Support Question-naire (Broadhead, 1988; two-week test-retest reliability from0.50 to 0.77), the Duke Social Support and Stress Scale,(1989; two-week test-retest correlations 0.40–0.76), the inter-view schedule for social interaction (Henderson, 1980; 18-day test-retest correlations 0.71–0.76), the Inventory ofSocially Supportive Behaviours (1981), the Perceived SocialSupport from Family and Friends (Procido and Heller,1983), the family environment scale (Moons and Moos,1981), the interpersonal support evaluation list (Cohen,

1985), the network typology and the network assessmentinstrument (Wenger, 1989, 1994) (McDowell, 2006).It is worth mentioning that in the last decade the

Lubben Social Network Scale (Lubben, 2003) , the scalewas well developed conceptually with the scale scoringthat was equally weighted sum of the six items, withscores ranking from 0 to 30. Additionally, the LSNS-6 fam-ily subscale was created from the three LSNS-6 items thatask about relatives and LSNS-6 friends subscale from thethree items that ask about friends (Lubben et al., 2006).In sum, it is increasing evidence that existing instruments

are unable to meet methodological expectations associatedwith good psychometric properties and that this givesopportunity to improve social network research. Moreover,there are no instruments designed to measure the constructof social network in population studies.

AIM

The aim of this paper is to create a simplified, easyimplementable multidimensional instrument to assess allrelevant elements of the structure and function of socialnetwork for adults through different age groups acrossvarious European countries and to provide the tool forhealth professionals and policy makers. The approach usedin the presented network analysis is based on the egocentricor personal network—amethod that samples individuals orentities and examines their network's characteristics (thecomposition and structure) extending from the focal person.

MATERIALS AND METHODS

The Collaborative Research on Ageing in Europe SocialNetwork Index (COURAGE-SNI) was developed andevaluated in the context of the Collaborative Research ofAgeing in Europe—the COURAGE project. A cross-sectionalgeneral population survey was performed in threeEuropean countries (Finland, Poland and Spain). Aprobabilistic representative sample of non-institutional-ized adult population was selected on the basis ofmultistage clustered design, with the oversampling ofadults 50–79 years and >80 years.In total, 10 800 individuals took part in the study (1976

from Finland, 4071 from Poland and 4753 from Spain),with a response rate of 53.4% for Finland, 66.5% forPoland and 69.9% for Spain. The 354 individuals wereexcluded from the social network analysis, mainly dueto the limited cognitive function of interviewee (131 inPoland, 170 in Spain and 42 in Finland) requiring the useof proxy interviews. There were also 11 cases (in Finland)with missing values for all questions concerning the socialnetwork part of the COURAGE protocol. Finally, the totalof 10 446 reports (96.7% of the study sample) wereavailable for the analysis.All data were collected during the face-to-face interviews

performed by trained interviewers assisted by the

229The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 4: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

computer-assisted personal interviewing system. Data onsocial network were collected in one part of the interview.Moreover, in the random sample of 519 participants, the

second interview was performed within 2-week timeinterval from the first one to assess the test-retest reliability.All of the questions concerning social network were used.

The Development of the Collaborative Research onAgeing in Europe Social Network Index

ConceptualizationThe COURAGE-SNI was developed in terms of the

available concepts and empirical experience of informalsocial networks. Social networks were defined as amultidimensional set of networks consisting of eightindependent, but partially interrelated, specific networksinvolving the relations with spouse, children, grandchildren,parents, other relatives, friends, neighbours and co-workers,taking into account the degree of boundedness to whichnetworks are defined on the basis of traditional groupstructures such as kin, neighbourhood, friendship and work.Four dimensions of each specific social network were distin-guished: structure, frequency of direct contacts, ties andgeneral support coming from specific networks (ability toreceive help on any matter). Next, the structural andfunctional aspects of the network were defined, namely, thenumber of specific networks (size) and the ties (closerelations), help (general social support) and the frequencyof face-to-face contacts (involvement).

It was expected that the theoretical model wouldenable us to classify and stratify different categories ofpersons from the highest social networks (high quantityand quality of social ties, high frequency of face-to-facecontacts and high level of social support) to persons withpoor networks (lack of social ties, lack or very poor socialsupport and low frequency or no contacts) and evenindividuals with no social network (lack of everystructural component within the network).

OperationalizationThe Social Network Questionnaire COURAGE-SNI was

part of the COURAGE protocol. In order to assess specificfunctions provided by each structural component of theSocial Network, the following questions were developed:(1) to assess ties with the spouse or partner, we used thequestion ‘How close is your relationship with your spouseor partner ’ (very close, quite close, not very close and notat all close), with other members of the social network,‘With how many people with the following groups (par-ents, children, grandchildren, other relatives, neighbours,friends, co-workers) would you say you have a closerelationship?’ (with nobody, with some people from thisgroup, with all the persons of this group); (2) to assessgeneral support, we used the question ‘How easy can

you get help from the following (spouse or partner,parents, children, grandchildren, other relatives, neigh-bours, friends, co-workers) if you should need it?’ (veryeasy, easy, possible, difficult and very difficult); and (3)to assess personal contacts within the network, we usedthe question ‘How often in the last 12months have youhad face-to-face contacts with… (spouse or partner, par-ents, children, grandchildren, other relatives, neighbours,friends, co-workers - outside working place)’ (never, onceor few times per year, once or few times per month, onceor few times per week and daily) (Appendix).

Construct validityConstruct validity refers to whether a scale measures or cor-relates with the theoretical construct that it is intended tomeasure. Construct validity considers structural componentdefined as ‘the extent to which structural relations betweentest items parallel the structural relations of other manifesta-tions of the trait being measured’ (Loevinger, 1957). There-fore, as a next step, it was decided to use factor analysis touncover the latent structure of a set of variables (items)created to assess social networks. For this purpose, two in-dependent random subsamples were created: the develop-mental subsample of 70% (7308) respondents forexploratory analysis and the validation subsample of 30%(3138) respondents for confirmatory analysis.

Exploratory factor analysis. To reveal the domains of socialnetworks and verify the consistency between the constructdeveloped theoretically and the ability of the items to mea-sure the underlying trait, we performed a principal factoranalysis. Because of the fact that the questionnaire used or-dinal scale items, polychoric correlations were used. Al-though the initial sample size was high (7308 eligibleindividuals), the listwise deletion procedure dramaticallydecreased the data available for the analyses to 187 respon-dents—thosewho reported to have all members of all inves-tigated structural components of social network (i.e., parent(s), spouse/partner, children and so on). Thus, the pairwisedeletion procedure was chosen.Subsequently, the principal factor analysis and the

principal component extraction method were used.Several methods were taken into consideration inextracting the number of factors including the Scree Test,the parallel analysis, the Velicer's Minimum AveragePartial Test, the very simple structure method and theeigenvalues criterion. To simplify and clarify the datastructure, we also tested some rotations (as suggested byBasto & Pereira, 2012) (including varimax, quartimax,oblimin group, equamax, parsimax, factor parsimony,simplimax, bentler, tandem, geomin and infomax) tochoose the solution that has the more logical interpreta-tion. Finally, the orthogonal infomax rotation has beenobserved to show acceptable fit between the theoreticalconcept and the data structure.

230 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 5: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Internal consistency reliabilityThe ordinal coefficient alpha was assessed to evaluate thedegree to which the items that build the COURAGE-SNIare inter-correlated (Zumbo et al., 2007). The averageinter-item correlations, which are a straightforward mea-sure of internal consistency (Clark &Whatson, 1995), werealso presented. The second ‘subsample’ was used for thispart of the analysis.

Confirmatory factor analysis. In order to assess how thedeveloped model fits the data, a confirmatory factoranalysis (CFA) has been performed and the absolute fitindices were estimated. The SPSS R-Menu for ordinalfactor analysis was used to estimate these measures(Basto & Pereira, 2012).

Test-retest reliability. To examine test-retest reliability of theSNI and its five components, we used the two-way mixedeffects single measure as the intraclass correlation coeffi-cient (ICC). Additionally, for every question, the kappacoefficient and the level of agreement were calculated.According to Landis and Koch, the value of ICC or kappacoefficients between 0.6 and 0.8 was classified as substan-tial, values greater than 0.8 were rated almost perfect,whereas values ranging from 0.4 to 0.6 were classified asmoderate, from 0.2 to 0.4 as fair and below 0.2 as slightor poor (Streiner & Norman, 1998). The suggested criteriafor the level of agreement are different. They are describedas excellent for the values greater than 0.9, as good for therange 0.75–0.89, as moderate for 0.60–0.74 and as poor forvalues lower than 0.6 (Saelens et al., 2006).

Item response theory modelling. The item response theory(IRT) modelling was developed to evaluate the ability ofthe questionnaire and to assess the latent trait (dimensions).There are severalmodels that can be used, depending on theitem(s) structure. The COURAGE protocol contained threetypes of polytomous items (with 3, 4 or 5 ordered responsecategories, see below and in the Appendix) used for thedevelopment of the COURAGE-SNI. For this reason, thegeneralized partial credit model (GPCM) developed byMuraki (Muraki, 1992) had to be applied (for the individualitems as well as for the whole COURAGE-SNI). TheGPCM is a parametric IRT model, which is especiallyrecommended for the analysis of items with differentnumbers of ordered response categories providing thelocation (threshold) and slope (discrimination) parame-ters as results. Typically, in test theory, the larger valueof the location parameter indicates the measurementconstruct (theta) that the respondent has endorsed.Subsequently, this concept was used in a posteriori dataanalysis; the value of location parameter describes thelevel of latent construct of social network present in the

individual. The second parameter indicates the extent towhich the item is related to the underlying construct. InIRT, models reliability is described as a continuous functionconditional on values of theta (the measured construct,social network in our case) and it is often depicted by iteminformation curves (IICs). Thus, IICs indicate the range oftheta (i.e., social network) where an item is best at discrimi-nating among individuals. Next, IICswere used to calculatethe total information available from the items used tocreate the specific structural component of social networkdeveloped by the exploratory factor analysis (EFA).Five structural components have been developed in theCOURAGE-SNI and they finally provide the total avail-able information concerning the social network (depictedby the total test information) (Edelen & Reeve, 2007;Streiner & Norman, 1998; Embretson & Reise, 2009).

Collaborative Research on Ageing in Europe Social NetworkIndex scoringItem response theory was subsequently used to create

the final scoring system for the COURAGE-SNI. TheGPCM model was used to calculate factor scores, whichwere assigned as the modes of the posterior distributionevaluated at the maximum likelihood estimations by theempirical Bayes estimates. The calculated factor scoreswere subsequently used to assess the level of social net-work across individuals. To normalize the distributionand to unify the range of variability across developedcomponents of social network, the zero unitarizationmethod was used, according to the following formula:

global score SNSð Þ ¼ factor score-min factor scoresð Þmax factor scoresð Þ �min factor scoresð Þ � 100:

Thus, the so called ‘social network saturation’ (SNS) vary-ing from 0% to 100% was obtained for every structuralcomponent and every individual.

It was observed that the different structural componentsof social network developed for the COURAGE-SNI pro-vide different amounts of information regarding the assess-ment of the total SNS. Subsequently, the total informationprovided by the IRT for every structural component wasused to calculate weights and to obtain the weighted totalSNS (WTSNS), by the following formula:

WTSNS ¼∑n

i¼1TCIi�SNSi

∑n

i¼1TCIi

where

WTSNS – weighted total social network saturation

i – one of the components in COURAGE-SNI (spouse/partner,

231The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 6: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

parents, other family members, fiends/co-workers and neighbours)

n – the total number of available components of the COURAGE-SNI for a person (max=5)

SNS – social network saturation

TCI – total component information by IRT

Finally, the COURAGE-SNI total score defined by the afore-mentioned formula (WTSNS) provides the conceptually con-sistent measure of social network ranging from 0% to 100%.

RESULTS

In order to verify comparability of the developmental andvalidation subsamples, the basic socio-demographiccharacteristics and existing members of the socialnetworks were analysed and compared (Table 1). Ingeneral, there were no significant differences betweenthese two subsamples, with the exception of the numberof people who reported that they have neighbours;nevertheless, the percentage difference was slight andthe p-value was relatively high.

The next step was the EFA, performed as describedabove in the methods section. The results of factor extrac-tion are presented in Figure 1. Majority of methods

Table 1. Basic characteristics in the total and the developmental and validation Collaborative Research on Ageing in Europe SocialNetwork Index samples

Variable Total

Sample (n=10 446)

t/χ2 pDevelopmental (7308) Validation (3138)

≥50 years: n (%) 7979 (76.4) 5569 (76.2) 2410 (76.8) 0.43 0.51Female: n (%) 5975 (57.2) 4192 (57.4) 1783 (56.8) 0.26 0.61Country

Finland 1923 (18.4) 1314 (18.0) 609 (19.4) 3.24 0.20Poland 3940 (37.7) 2760 (37.8) 1180 (37.6)Spain 4583 (43.9) 3234 (44.3) 1349 (43.0)Rural setting: n (%) 2752 (26.3) 1949 (26.7) 803 (25.6) 1.32 0.25Currently employed: n (%) 3653 (38.1) 2581 (38.6) 1072 (37.1) 1.94 0.16Married or in partnership: n (%) 6185 (59.2) 4364 (59.7) 1821 (58.0) 2.58 0.11Years of education, mean±SD 11.35 (5.27) 11.35 (5.22) 11.35 (5.40) -0.02 0.99Number of people in the household (HH), mean± SD 2.42 (1.29) 2.42 (1.28) 2.42 (1.30) -0.10 0.99COURAGE-SNI total score, mean± SD 68.5 (13.3) 68.6 (13.3) 68.1 (13.3) �1.67 0.10Existing members of social network(by self-report)

Spouse 6711 (64.2) 4731 (64.7) 1980 (63.1) 2.57 0.11Parents 4368 (41.8) 3079 (42.1) 1289 (41.1) 1.03 0.31Children 8353 (80.0) 5834 (79.8) 2519 (80.3) 0.27 0.60Grandchildren 5277 (50.5) 3691 (50.5) 1586 (50.5) 0.00 0.97Other relatives 10 040 (96.1) 7037 (96.2) 3003 (95.7) 2.17 0.14Friends 9640 (92.3) 6760 (92.5) 2880 (91.8) 1.49 0.22Co-workers 5181 (49.6) 3614 (49.5) 1567 (50.0) 0.21 0.64Neighbours 10 226 (97.9) 7168 (98.1) 3058 (97.5) 4.28 0.04

SD= standard deviation. COURAGE-SNI =Collaborative Research on Ageing in Europe Social Network Index.Unpaired t-test for years of education and number of people in the HH score. χ2 test for the other variables.

Figure 1. A Scree plot and factor extraction for the Collabora-tive Research on Ageing in Europe Social Network Index

232 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 7: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

suggested 7-factors criterion, so it was decided to retainthis criterion for further analysis.After the initial extraction, several rotations (see above

in the methods section) were tested to obtain better fitwith the theoretical concept. As was already observed, theresults of the analyses agreed with the multilevel constructcreated by the theory. Moreover, structural modellingsuggested a hierarchical structure within the domain of‘other family members’ (Figure 2). Factor loadings rangedbetween 0.43 and 0.89 with the only one loading of 0.25for ‘face-to-face contacts with other relatives’. In this com-ponent called ‘other family members’, the aforementioneditem was loaded with the higher value on the factor‘parents’; however, it was decided to keep this item in itsprimary position taking into account the theoretical concept

of social networks. Item communalities in themodel rangedbetween 0.58 and 0.90 (Figure 2).Confirmatory factor analysis performed on the second

random subsample provided an acceptable fit of themodel: goodness of fit index= 0.97; adjusted goodness offit index= 0.93, the root mean square residual = 0.056 andthe root mean square partial correlation controllingfactors = 0.21. Moreover, the fit of the model to thecorrelation matrix assessed by the residual matrix showedacceptable adjustment (residuals <0.05 = 65%).To assess the internal consistency measures, we used

the average correlation and ordinal coefficient alpha.Fairly good internal consistency across layers, definedon the basis of EFA (varying from 0.61 to 0.86), wasobserved (Table 1).

.42

.25

.75

.79

.77

.60

.57

.65

.88

.89

.70

.76

.82

.83

.61

.46

.77

.59

.70

.65

.75

.48

.60

.52

.43

.61.42

.43

.69

.69.47

.53.67

.5

.4

.5

OTHER FAMILY MEMBERS

children

grandchildren

other relatives

face-to-face

children

grandchildren

other relatives

children

grandchildren

other relatives

close

help

close

help

face-to-face

SPOUSE

close

help

face-to-face

PARENTS

close

help

face-to-faceNEIGHBOURS

C-close

C-help

C-face-to-face

F-close

F-help

F-face-to-face FRIENDS &CO-WORKERS

A1=0.72AICC=0.35

A1=0.77AICC=0.21

A1=0.64AICC=0.29

A1=0.61AICC=0.29

A1=0.86AICC=0.29

A1=0.80AICC=0.29

A1=0.76AICC=0.33

Figure 2. Structural model of social network revealed after exploratory factor analysis (A-ordinal coefficient alpha; A1-tested for theseparate layer; AIIC-the average inter-item correlation)

233The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 8: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Subsequently, the ability to discriminate betweenrespondents at the appropriate level of the trait ofsocial network for every item was assessed by the IRTmodels.For the five social network components (spouse/part-

ner, parents, other family members, neighbours andfriends/co-workers), discrimination and difficulty pa-rameters were estimated. The widest range of the slopeestimates was observed for the structural component of

the spouse/partner (from 0.8 to 3.0) and the narrowestfor parents (from 0.6 to 1.1). The considerable range of lo-cation parameters was noticed for all of the five compo-nents (the average range was 5.22—from 4.5 for‘neighbours’ to 7.3 for ‘other family members’). The anal-ysis of the location parameters as well as the IICs revealedthat the majority of the response categories was endorsedby the respondents with lower than the average socialnetwork (Tables 2–3).

Table 3. Item discrimination and item difficulty parameters based on the generalized partial credit models

Difficulty 1 Difficulty 2 Difficulty 3 Difficulty 4 Discrimination

Spouse/partner

Close �2.85 �2.13 �1.09 — 3.04Face-to-face �6.27 �5.26 �4.33 �3.23 0.86Help �2.86 �2.55 �2.11 �1.00 2.37Parents

Close �2.43 �2.26 — — 0.80Face-to-face �3.34 �1.63 �0.77 0.07 0.57Help �1.32 �1.01 �1.55 �0.40 1.14Children

Close �3.86 �5.61 — — 0.43Face-to-face �3.62 �2.16 �1.44 �0.82 0.86Help �2.37 �2.59 �2.29 �0.90 0.97Grandchildren

Close �2.34 �4.18 — — 0.49Face-to-face �2.69 �1.73 �0.76 0.54 1.22Help �1.19 �2.33 �1.37 �0.39 0.84Other relatives

Close �3.12 0.05 — — 0.88Face-to-face �3.24 �0.87 0.67 1.71 0.89Help �1.81 �2.20 �1.06 0.90 1.00Co-workers

Close �2.49 0.70 — — 0.99Face-to-face �3.08 �2.51 �0.36 �0.04 0.51Help �1.81 �1.64 �0.42 1.23 2.16Friends

Close �4.05 �0.20 — — 0.86Face-to-face �3.84 �2.51 �0.36 1.53 0.76Help �2.54 �2.51 �1.07 0.79 1.50Neighbours

Close �1.23 0.59 — — 1.94Face-to-face �1.14 �2.99 �1.58 0.17 0.69Help �2.36 �2.58 �1.12 1.50 0.76

Table 2. Internal consistency measures for the whole 7-factor solution model

OFM help OFM close Friends and co-workers OFM Face-to-face Partner Spouse Neighbours

Average correlation0.214 0.213 0.245 0.538 0.272 0.350 0.294Ordinal coefficient alpha0.816 0.875 0.683 0.870 0.696 0.717 0.642

OFM=other family members.

234 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 9: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Variations in the measurement precision across thecontinuum of social network are presented in Figure 3.The analysis of IICs showed that within the fivecomponents of the social network structure, some itemsprovide more information, whereas others provide lessinformation, however, spread out over a larger range ofthe social network. Besides, test information functions forthe layers indicated that its peaks are not far from the leftend of the axis representing theta (Figure 4).The ICCs for test-retest reliability were in the range from

0.35 for the component of ‘friends/co-workers’ to 0.54 for‘parents’. The value of the ICC for the whole SNI was 0.36.In the case of questions concerning help and face-to-facecontacts, kappa coefficients were in the range from 0.35to 0.79. The kappa coefficients for questions aboutcloseness were at least fair apart from the question regard-ing friends, where the value of the coefficient was 0.15.Simultaneously, the percentage agreement for everyquestion was greater than 0.84 (Tables 4–5).

DISCUSSION

The COURAGE project was designed to develop ameasure for health and health-related outcomes and toassess the role of different health determinants, with socialnetworks among them. In the development of thequestionnaire to evaluate the quality of the social networkfor the population studies, a theoretical conception of theconstruct was elucidated and it was decided to use thestructural model of social network. This model has beenmore frequently suggested across literature and wassuccessfully used by Lubben et al. (2006).The structural (morphologic) characteristic of the

network was developed on the basis of eight pre-definedstructural components (spouse/partner, parents, children,grandchildren, other relatives, friends, neighbours and co--workers). Every component (if present) is supposed to pro-vide social network functionalities. The model is consistentwith the analytic concept of the social networks that en-compasses the structure of linkages between individualsthrough which several functionalities may flow. The mostimportant of these were considered in the COURAGE-SNI protocol. The questions ‘how close’ and ‘how manypeople are close’ refer mainly the level of the social ties,whereas subsequent questions regarding help express gen-eral support, given all the types of support: emotional, in-strumental, appraisal and informational. The third type ofquestions about the frequency of face-to-face contacts re-veals the intensity of contacts and particularly the level ofinvolvement, thus constituting additional informationabout the function of the social network.Face validity was achieved by a detailed investigation of

the available theory. One of the limitations was thenecessity of creating an easily implementable relatively

short tool to investigate the construct of social networkin population studies. Subsequently, we had to limit thenumber of items used, although those finally acceptedfor the COURAGE-SNI protocol showed relatively goodcapacity to evaluate the structure and major functions ofsocial network. As the next step, the EFA showed thatthe data collected correspond with the developed theory.It was decided to use the principal factor analysis insteadof principal component analysis. According to Gorsuch(1997), the aim of factor analysis is to reveal any latentvariables that cause the manifest variables to co-vary.During factor extraction, the shared variance of a variableis partitioned from its unique variance and error varianceto reveal the underlying factor structure; finally, onlyshared variance appears in the solution. Principal compo-nents analysis does not discriminate between shared andunique variances and subsequently was not used.The EFA revealed seven latent layers of social network

with a part of the hierarchical model within the construct.Although somewhat different from our primary expecta-tions, we believe that the revealed model supports ourtheoretical concept.Network types created by Wenger (1997a, 1997b) were

used in England, Germany and the Netherlands. Other net-work analyses showed that all social network typologies ex-hibit several similarities especially in regard to composition,along with some other characteristics (e.g., demographic).Mendes de Leon et al. (1999) developed the notion of socialnetwork in terms of the structure of ties with children, tieswith other relatives, tieswith friends and tieswith confidant(see also Glass, Mendes de Leon, Seeman, & Berkman,1997). CFA was used to construct the model of scales forfour separate domains, each scale combining informationon the number of ties with frequent visual contact, frequentnon-visual contact and the geographic proximity of ties. TheChildren Ties Scale also included information on(emotional) closeness and reciprocity (help given andreceived). The interscale correlations were generally low (r10), except for the correlation between the friends and therelatives scale (r=0.33). Finally, to examine the effect ofmultiple social relationships, a Total Social Network Scalewas constructed by adding the score for each of Domain-Specific Scales into a summary score. This strategy has beenadopted for many scales. These scales were developed andscored according to classical test theory (CCT), and typi-cally the sums of answers were used. However, CCT hassome limitations. The major problem concerns the fact thatthe result of reliability and validity in the CTT applies onlyto the specific group for whom the test was validated.Secondly, CTT assumes that each item contributes equallyto the final score, when usually some items are more impor-tant than others. This is partially solved by assigningdifferent weights to item, but there is no effective way toaccount for this fact while building the scale. CTT assumesthat all items are measured on the same interval scale, when

235The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 10: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

SPOUSEItems:1-close; 2-face-to-face; 3-help;

PARENTSItems:1-close; 2-face-to-face; 3-help;

OTHER RELATIVESItems:1-children-close; 2-children-face-to-face; 3-children-help; 4-grandchildren-close; 5-grandchildren-face-to-face; 6-grandchildren-help; 7-other relatives-close; 8-other relatives-face-to-face; 9-other relative-help

CO-WORKERS AND FRIENDSItems:1-co-workers-close; 2-co-workers–face-to-face; 3-co-workers-help; 4-friends-close; 5-friends-face-to-face; 6-friends-help;

NEIGHBOURS Items:1-close; 2-face-to-face; 3-help;

Figure 3. Total information functions for the defined layers

236 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 11: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

SPOUSE PARENTS

OTHER RELATIVES CO-WORKERS AND FRIENDS

NEIGHBOURS TOTAL TEST INFORMATION

Figure 4. Total information functions for the defined layers and the total test information

237The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 12: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

in fact items are usually ordinal or nominal, unless the‘psychological’ distance between responses of the itemsvary and different from one item to the other. IRT definesa scale for the underlying latent variable (the construct ofsocial networks in our case) that is measured by a set ofitems, where all these items are calibrated with respectto the same scale (Streiner & Norman, 1998; Edelen &Reeve, 2007).The analysis of the characteristics of test information

functions and the total test information shows that the curvepeaks are shifted to the left side of the social networkcontinuum. This suggests that the COURAGE-SNImay be

particularly useful to identify individuals whose socialnetwork is poorer and thus suggests the benefit for the toolif is used to identify high risk groups or populations.The COURAGE-SNI makes it possible for us to assess

social network in five general structural components. Socialnetwork is assessed as the SNS ranges from 0% to 100%,providing the opportunity to compare different groupsdirectly or creating a scoring system. Moreover, theWTSNS(the COURAGE-SNI total score) proposed for the assessmentof the social networks is another advantage due to consider-ing the proportional amount of information regarding socialnetwork provided by every specific structural component.

Table 4. Intraclass correlation coefficients for the Social Network Index and the five social network components

Test mean (SD) Retest mean (SD) ICC (95% CI) n

Parents 72.5 (21.8) 69.9 (23.0) 0.54 (0.43–0.64) 187Spouse 87.3 (21.9) 84.3 (25.8) 0.42 (0.33–0.51) 323Other family 67.0 (14.7) 64.4 (14.4) 0.48 (0.40–0.56) 341Friends/co-workers 61.6 (17.1) 53.2 (16.4) 0.35 (0.26–0.43) 411Neighbours 63.5 (22.0) 59.8 (20.3) 0.43 (0.35–0.50) 482Social network index 67.6 (13.8) 53.4 (19.3) 0.36 (0.28–0.43) 518

SD= standard deviation. ICC= Intraclass correlation coefficients. CI = confidence interval.

Table 5. Kappa coefficients and percentage agreement for items

Agreement (%) Expected agreement (%) Kappa (95% CI) n

CloseSpouse 97.3 93.9 0.56 (0.42–0.65) 325Parents 91.4 87.6 0.31 (0.19–0.49) 187Children 95.5 93.2 0.34 (0.22–0.53) 387Grandchildren 92.3 90.1 0.23 (012–0.34) 258Other relative 86.6 81.7 0.27 (0.22–0.33) 464Co-workers 87.4 82.9 0.26 (0.23–0.33) 204Friends 87.5 85.3 0.15 (0.07–0.22) 432Neighbours 84.0 76.5 0.32 (0.27–0.37) 470Face-to-face

Spouse 99.3 97.4 0.73 (0.70–0.87) 324Parents 97.0 85.5 0.79 (0.73–0.82) 197Children 97.0 89.2 0.72 (0.64–0.73) 407Grandchildren 95.7 88.7 0.62 (0.52–0.65) 269Other relative 94.6 88.2 0.54 (0.53–0.61) 488Co-workers 87.4 80.4 0.36 (0.24–0.40) 206Friends 94.2 90.8 0.37 (0.32–0.42) 455Neighbours 92.4 88.2 0.36 (0.36–0.40) 492Help

Spouse 95.9 93.4 0.38 (0.26–0.51) 325Parents 91.8 86.9 0.38 (0.27–0.43) 192Children 95.1 91.6 0.42 (0.40–0.47) 391Grandchildren 93.5 89.2 0.40 (0.35–0.50) 179Other relative 93.5 90.0 0.35 (0.34–0.40) 422Co-workers 94.3 90.5 0.40 (0.25–0.49) 206Friends 94.7 93.1 0.24 (0.21–0.31) 445Neighbours 95.0 91.7 0.40 (0.39–0.41) 492

CI = confidence interval.

238 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 13: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

SUMMARY

The COURAGE in Europe project made it possible todevelop a tool called the COURAGE-SNI that was consis-tent with a conceptual definition of social network andallows us to assess the construct of social network. Thequantitative model provides an opportunity to differentiateindividuals across different ages with respect to the qualityof their social network. The COURAGE-SNI showed goodreliability and content validity and seems to be a promisingtool for the assessment of the social network phenomenonacross European countries. Some of the validity measuressuch as convergent validity, criterion validity and predictivevalidity require further evaluation.

ACKNOWLEDGEMENTS

The research leading to these results has received fundingfrom the European Community's Seventh FrameworkProgramme (FP7/2007–2013) under grant agreementnumber 223071 (COURAGE in Europe), from the Institutode Salud Carlos III-FIS research grant number PS09/00295and PS09/01845, and from the Spanish Ministry of Scienceand Innovation's ACI-Promociona (ACI2009-1010) andthe Polish Ministry for Science and Higher Educationgrant for an international co-financed project (number1277/7PR/UE/2009/7, 2009–2012). The study wassupported by the Centro de Investigación Biomédica enRed de Salud Mental (CIBERSAM), Instituto de SaludCarlos III.

DISCLOSURE

The views expressed in this paper are those of the authorsand do not necessarily represent the views or policies ofthe World Health Organization.

DISCLOSURE OF CONFLICTS OF INTEREST

None.

REFERENCESAchat, H., Kawachi, I., Levine, S., Berkey, C., & Coakley, E.

(1998). Social networks, stress and health-related quality of life.Quality of Life Research, 7, 735-750.

Antonucci, T. C. (1986). Measure social support networks: Hierar-chical mapping techniques. Generations, X, 10-12.

Ajrouch, K. J., Blandon, A. Y., & Antonucci, T. C. (2005). Socialnetworks among men and women: The effects of age and so-cioeconomic status. Journal of Gerontology: Social Sciences, 60B(6), 311-317.

Barrera, M., Sandler, I.N., & Ramsay, T.B. (1981). Preliminarydevelopment of a scale of social support: Studies on collegestudents. American Journal of Community Psychology, 9(4),435–447.

Basto, M., & Pereira, J. M. (2012). An SPSS R-Menu for ordinalfactor analysis. Journal of Statistical Software, 46(4), 1–29.

Berkman, L. F. (1995). The role of social relations in health promo-tion. Psychosomatic Medicine, 57, 245-254.

Berkman, L. F., & Glass, T. (2000). Social integration, social net-works, social support, and health. In: L. F. Berkman, & I.Kawachi (Eds.), Social epidemiology (pp. 137-173). OxfordUniversity Press.

Berkman, L. F., & Syme, S. L. (1979). Social networks, hostresistance, and mortality: a nine-year follow-up study ofAlameda County residents. American Journal of Epidemiology,109(2), 186-204.

Bowling, A., & Grundy, E. (1998). The association between socialnetworks and mortality in later life. Reviews in ClinicalGerontology, 8, 353-361.

Bowling, A. (1997). Measuring social networks and socialsupport. In: A. Bowling (ed.), Measuring health. A review ofquality of life measurement scales (2nd ed., pp. s.90-s.110).Open University Press, Buckingham, Philadelphia.

Broadhead, W.E., Gehlbach, S.H., De Gruy, F.V., & Kaplan, B.H.(1988). The Duke-UNC Functional Social Support Question-naire: Measurement of social support in family medicine pa-tients. Medical Care, 26, 709–723.

Broadhead, W.E., Gehlbach, S.H., de Gruy, F.V., Kaplan, B.H.(1989) Functional versus structural social support and healthcare utilization in a family medicine outpatient practice. Medi-cal Care, 27, 221–233.

Clark, L.A., & Whatson, D. (1995) Constructing validity: Basic is-sues in objective instrument development. 7(3), 309-319.

Cohen, S., Mermelstein, R., Kamarck, T., & Hoberman, H. (1985).Measuring the functional components of social support. Socialsupport: Theory, research and applications, 73–94.

Donald, C.A., Ware Jr, J.E., Brook, R.H., & Davies-Avery, A.(1978). Conceptualization and measurement of health foradults in the health insurance study: Vol. IV, Social health.Santa Monica, CA: Rand Corporation.

Due, P., Holstein, B., Lund, R., Modvig, J., & Avlund, K. (1999).Social relations: Network, support and relational strain. SocialScience & Medicine, 48, 661-673.

Edelen, M. O., & Reeve, B. B. (2007). Applying item response the-ory (IRT) modeling to questionnaire development, evaluation,and refinement. Quality of Life Research, 16, 5–18. doi: 10.1007/s11136-007-9198-0

Embretson, S. E., & Reise, S. P. (2009). Item response theory forpsychologists. Psychology Press Taylor & Francis Group.

Finch, J. F., Okun, M. A., Barrera, M., Zautra, A. J., & Reich, J. W.(1989). Positive and negative social ties among older adults:Measurement models and the prediction of psychological dis-tress and well-being. American Journal of Community Psychology,17(5), 585-605.

Gallegos-Carrillo, K., Mudgal, J., Sánchez-Garcia, S., Wagner,F. A., Gallo, J. J., Salmerón, J., & Garcia-Pena, C. (2009).Social networks and health-related quality of life: A popula-tion based study among older adults. Salud Pública deMéxico 51(1), 6-13.

Glass, T. A., Mendes de Leon, C. F., Seeman, T. E., & Berkman, L. F.(1997). Beyond single indicators of social networks a lisrel anal-ysis of social ties among the elderly. Social Science &Medicine, 44(10), 1503-1517

239The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 14: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Gorsuch, R. L. (1997). New procedure for extension analysis inexploratory factor analysis. Educational and PsychologicalMeasurement, 57, 725-740.

Henderson, S., Duncan-Jones, P., Byrne, D.G., & Scott, R. (1980).Measuring social relationships. The interview schedule for so-cial interaction. Psychological Medicine, 10(4), 723–734.

Iwasaki M., Otani T., Sunaga R., Miyazaki H., Xiao L., Wang N.,Yosiaki S., & Suzuki S. (2002). Social networks and mortalitybased on the Komo-Ise cohort study in Japan. InternationalJournal of Epidemiology, 31, 1208-1218.

Kawachi, I., Colditz, G. A., Ascherio, A., Rimm, E. B.,Giovannucci, E., Stampfer, M. J., & Willett, W. C. (1996). Aprospective study of social networks in relation to totalmortality and cardiovascular disease in men in the USA.Journal of Epidemiology and Community Health, 50, 245-251.

Loevinger, J. (1957). Objective tests as instruments of psychologicaltheory. Psychological Reports, 3: 635-694.

Litwin, H. (2001). Social networks type and morale in old age.The Gerontologist, 41(4), 516-524.

LópezGarcía, E., Banegas, J. R., Graciani Pérez-Regadera,A.,HerruzoCabrera, R., & Rodríguez-Artalejo, F. (2005). Social network andhealth-related quality of life in older adults: A population-basedstudy in Spain. Quality of Life Research, 14, 511-520.

Lubben, J., Blozik, E., Gillmann, G., Iliffe, S., von Renteln Kruse,W., Beck, J. C., & Stuck, A. E. (2006). Performance of anabbreviated version of the Lubben Social Network Scaleamong three European community-dwelling older adultpopulations. The Gerontologist, 46(4), 503-513.

Lubben, J., Gironda, M. (2003). Centrality of social ties to thehealth and well-being of older adults. In Social Work andHealth Care in an Aging Society. Eds. Berkman, B., Harootyan,L. Springer Publishing Company.

McDowell, I. (2006). Social Health. In: McDowell, I. (Ed.).Measuring health. A guide to rating scales and questionnaires(pp. s 150-s 184). New York: Oxford University Press.

McFarlane, A.H., Neale, K.A., Norman, G.R., Roy, R.G., &Streiner, D.L. (1981). Methodological Issues in Developing a Scaleto Measure Social Support Schizophr Bull, 7, 90–100.

Mendes de Leon, C. F., Glass, T. A., Beckett, L. A., Seeman, T. E.,Evans, D. A., & Berkman, L. F. (1999). Social networks anddisability transitions across eight intervals of yearly data inthe new haven EPESE. Journal of Gerontology, 54B(3), 162-172.

Moos, R.H., &Moos, B.S. (1981).Manual for the family environmentscale. Palo Alto, Calif.: Consulting Psychologists Press, 50, 33–56.

Moren-Cross, J. L., & Lin, N. (2006). Social networks and health.In: R. H. Binstock, & L. K. George (Eds.), Handbook of agingand the social sciences (6th ed., pp. 111-126). Elsevier.

Muraki, E. (1992). A generalized partial creditmodel: Application ofan EM algorithm. Applied Psychological Measurement, 16, 159–176.

Olsen, O. (1993). Impact of social network on cardiovascularmortality in middle aged Danish men. Journal of Epidemiologyand Community Health, 47, 176-180.

Procidano, M.E., & Heller, K. (1983). Measures of perceived socialsupport from friends and from family: Three validation stud-ies. American Journal of Community Psychology, 11, 1–24.

Rutledge, T., Matthews, K., Lui, L.-Y., Stone, K. L., & Cauley,J. A. (2003). Social networks and marital status predictmortality in older women: Prospective evidence from thestudy of osteoporotic fractures (SOF). PsychosomaticMedicine, 65, 688-694.

Saelens, B. E., Frank, L. D., Auffrey, C., Whitaker, R. C., Burdette,H. L., & Colabianchi, N. (2006). Measuring physicalenvironments of parks and playgrounds: EAPRS instrumentdevelopment and inter-rater reliability. Journal of PhysicalActivity & Health, 3(Suppl 1), S190–S207.

Sarason, I.G., Levine, H.M., Basham, R.B., Sarason, B.R. (1983)Assessing social support: The Social Support Questionnaire.Journal of Personality and Social Psychology, 44(1), 127–139.

Scott, C., Hofmeyer, A. (2007). Networks and social capital: Arelational approach to primary healthcare reform.Health ResearchPolicy and Systems 5, 9.

Seeman, T. E., Kaplan, G. A., Knudsen, L., Cohen, R., & Guralnik,J. (1987). Social network ties and mortality among the elderlyin the Alameda Country study. American Journal ofEpidemiology, 126(4), 714-723

Sherbourne, C.D., & Stewart, A.L. (1991). The MOS social sup-port survey. Social science & medicine, 32(6), 705–714.

Shye, D., Mullooly, J. P., Freeborn, D. K., & Pope, C. R. (1995).Gender differences in the relationship between social networksupport and mortality: A longitudinal study of an elderly co-hort. Social Science & Medicine, 41(7), 935-947.

Smith, K. P., & Christakis, N. A. (2008). Social networks andhealth. Annual Review of Sociology, 34, 405-429.

Streiner, D. L., & Norman, G. R. (1998). Health measurementscales. A practical guide to their development and use. OxfordUniversity Press.

Sugisawa, H., Liang, J., & Liu, X. (1994). Social networks, socialsupport, and mortality among older people in Japan. Journalof Gerontology Social Sciences, 49(1), 3-13.

Victor, Ch., Scambler, S., Bond, J., & Bowling, A. (2000). Beingalone in later life: Loneliness, social isolation and living alone.Reviews in Clinical Gerontology, 10, 407-417.

Wenger, G.C. (1989). Support networks in old age: constructing atypology. Growing old in the twentieth century, 166–185.

Wenger, G.C. (1994). Support Networks of Older People: A Guidefor Practitioners. Bangor: CSPRD, University of Wales.

Wenger, G. C. (1997a). Review of findings on supportnetworks of older Europeans. Journal of Cross-CulturalGerontology, 12, 1-21.

Wenger, G. C. (1997b). Social network and the prediction ofelderly people at risk. Aging & Mental Health, 1(4), 311-320.

Zumbo, B. D., Gadermann, A. M., & Zeisser, C. (2007).Ordinal versions of coefficients alpha and theta for likertrating instruments. Journal of Modern Applied StatisticalMethods, 6, 21-29.

240 K. Zawisza et al.

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)

Page 15: The Validity of the Instrument to Evaluate Social Network in the Ageing Population: The Collaborative Research on Ageing in Europe Social Network Index

Appendix 1 The courage social network questionnaire

Filter questions

Do you have…1. Spouse or partner 1. YE S 2. NO If ‘No’, omit 101, 201, 3012. Parents 1.YE S 2. NO If ‘No’, omit 102, 202, 3023. Children 1.YE S 2. NO If ‘No’, omit 103, 203, 3034. Grandchildren 1.YE S 2. NO If ‘No’, omit 104, 204, 3045. Other relatives 1. YE S 2. NO If ‘No’, omit 105, 205, 3056. Co-workers 1.YE S 2. NO If ‘No’, omit 106, 206, 3067. Friends 1.YE S 2. NO If ‘No’, omit 107, 207, 3078. Neighbours 1.YE S 2. NO If ‘No’, omit 108, 208, 308101 How close is your relationship with your spouse or partner? 1. Very close

2. Quite close3. Not very close4. Not at all close

With how many people from the following groups would you say you have a close relationship?

With nobody With some people from this group With all the persons of this group

102 Parents 1 2 3103 Children 1 2 3104 Grandchildren 1 2 3105 Other relatives 1 2 3106 Co-workers 1 2 3107 Friends 1 2 3108 Neighbours 1 2 3

How often in the last 12months have you had face-to-face contacts with the following:

Never Once or few times per year Once or few times per month Once or few times per week Daily

201 Spouse or Partner 1 2 3 4 5202 Parents 1 2 3 4 5203 Children 1 2 3 4 5204 Grandchildren 1 2 3 4 5205 Other relatives 1 2 3 4 5206 Co-workers

(outside working place)1 2 3 4 5

207 Friends 1 2 3 4 5208 Neighbours 1 2 3 4 5209 Acquaintances 1 2 3 4 5

How easy can you obtain help from the following if you should need it?

Very easy Easy Possible Difficult Very difficult NA

301 Spouse or partner 1 2 3 4 5 —302 Parents 1 2 3 4 5 —303 Children 1 2 3 4 5 9304 Grandchildren 1 2 3 4 5 9305 Other relatives 1 2 3 4 5 —306 Co-workers 1 2 3 4 5 —307 Friends 1 2 3 4 5 —308 Neighbours 1 2 3 4 5 —

NA, not applicable

241The Validity of the COURAGE Social Network Index

Copyright © 2013 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 21, 227–241 (2014)