www.elsevier.com/locate/yebeh

Epilepsy & Behavior 10 (2007) 595–603

Development and psychometric properties of the Thinkingabout Epilepsy questionnaire assessing children’s

knowledge and attitudes about epilepsy

Alexandra L.C. Martiniuk a,*, Kathy N. Speechley b,c, Mary Secco d, M. Karen Campbell b,c

a The George Institute for International Health at the University of Sydney, Sydney, M201 Missenden Road, Camperdown, NSW 2050, Australiab Departments of Epidemiology and Biostatistics and Pediatrics, University of Western Ontario, Canada

c Children’s Health Research Institute, London, ON, Canadad Canadian Epilepsy Alliance, ON, Canada

Received 2 June 2006; revised 25 January 2007; accepted 26 January 2007Available online 7 March 2007

Abstract

Objective. Epilepsy is the most common neurological disorder in the world, yet it is still widely misunderstood. A lack of knowledgeand negative attitudes about epilepsy are largely blamed for the stigma felt by people with epilepsy and their families. Recent calls forresearch into stigma have been made by the World Health Organization and international epilepsy organizations. Our objective is todescribe the development, structure, and psychometric properties of the Thinking about Epilepsy questionnaire.

Methods. A 36-item questionnaire was designed to assess Grade 5 (ages 9–11) students’ knowledge of and attitudes about epilepsyand to evaluate changes in knowledge and attitudes following an epilepsy education program. The questionnaire contains 18 knowledge,10 attitude, and 8 demographic questions.

Results. Psychometric properties of the Thinking about Epilepsy questionnaire were ascertained using data from 783 Grade 5 stu-dents. Three items (one knowledge item and two attitude items) were removed prior to the factor analysis due to their low extractioncommunalities. Factor analysis revealed a bidimensional structure (knowledge and attitudes) with five knowledge factors and two atti-tude factors. The questionnaire was found to have good internal consistency reliability (Cronbach’s a = 0.74 for knowledge and 0.82 forattitudes). Both the knowledge and attitude measures were deemed to have acceptable face, content, and construct validity.

Conclusion. The final 33-item Thinking about Epilepsy questionnaire demonstrates adequate reliability for the knowledge measure,good reliability for the attitude measure, and excellent validity for both measures. The Thinking about Epilepsy questionnaire offers aviable option for assessing elementary school students’ knowledge and attitudes regarding epilepsy in general or in conjunction with itsaffiliated Thinking about Epilepsy education program.� 2007 Elsevier Inc. All rights reserved.

Keywords: Epilepsy; Stigma; Children; Knowledge; Attitudes; Questionnaire development; Validity; Reliability

1. Introduction

The lifetime prevalence of epilepsy worldwide is estimat-ed to be 100 million individuals [1]. Although epilepsy issaid to be the most common neurological disorder, itremains misunderstood and unaccepted, particularly in

1525-5050/$ - see front matter � 2007 Elsevier Inc. All rights reserved.

doi:10.1016/j.yebeh.2007.01.011

* Corresponding author. Fax: +61 2 9657 0301.E-mail address: amartiniuk@george.org.au (A.L.C. Martiniuk).

developing countries [1–3]. In many countries, epilepsy isbelieved to be contagious or a sign of a curse [3]. The stig-ma associated with epilepsy may affect an individual’semployment opportunities, housing, and access to medicalcare [3,4]. Stigmatization is said to begin when labeled indi-viduals are set into their own category, which leads to rejec-tion and discrimination [5]. Measuring knowledge andattitudes associated with a stigmatizing condition is thefirst step toward understanding stigma and is necessary

596 A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603

to inform programs and policies that aim to alleviatestigma and prevent discrimination [6].

There is a need to educate the public about epilepsy. TheWorld Health Organization (WHO), International Leagueagainst Epilepsy (ILAE), International Bureau for Epilepsy(IBE) [7], and Canadian Epilepsy Alliance (CEA) [8] haveset a priority to educate the public about epilepsy todecrease stigma. In response, the Epilepsy Support Centre(ESC) in London, Ontario, Canada, together with familiesaffected by epilepsy, primary school teachers, and research-ers at the University of Western Ontario (UWO), designeda Grade 5 epilepsy education program and accompanyingquestionnaire to facilitate evaluation of the program.Because behaviors (such as stigmatizing behaviors towardsomeone with epilepsy) are difficult to measure, Fishbeinand Ajzen’s Theory of Planned Behavior offered insightinto prebehavior components that are more measurable[9]. Prebehavior components include knowledge, attitudes,and behavioral intent [9].

Prior to implementation of an epilepsy education pro-gram, a questionnaire was required to assess baseline knowl-edge and attitudes and to evaluate changes following such aprogram. None existed at that time that was reliable and val-id. Although there are several reliable and valid surveys thatassess knowledge and attitudes of epilepsy in adults [10–22],neither of the two surveys for children available at the timewas appropriate. One survey, developed by the EpilepsyFoundation of America, was intended for the older agegroup of adolescents [23,24]. The second was developed in1985 to assess epilepsy knowledge in elementary school chil-dren. It measures knowledge only, not attitudes, and has notbeen assessed for reliability and validity [25]. A new Elemen-tary School Epilepsy Survey (ESES), to assess epilepsyknowledge and attitudes of children (aged 10–14), was indevelopment at the same time as the Thinking about Epilep-sy questionnaire described here. Its measurement propertieshave been reported [26]. The objective of this article is todescribe the development, structure, and psychometricproperties of the Thinking about Epilepsy questionnaire.

2. Methods

A self-administered questionnaire was developed to assess students’baseline knowledge of and attitudes toward epilepsy, as well as changesin knowledge and attitudes following an epilepsy education program.The Thinking about Epilepsy program and the randomized controlled tri-al conducted to evaluate it are described elsewhere [27]. Briefly, the Think-ing about Epilepsy program was the first educational program in Canadato target elementary school children and be aligned with several of theexisting school curriculum objectives. Bandura’s Social Learning Theoryinformed the development of this 30-minute, scripted program about epi-lepsy [28]. To impart its messages, the program uses posters, photographs,a television commercial, a puppet show, and role play. The program wasevaluated using a stratified cluster randomized control trial in 783 Grade 5students. There was a significant increase in epilepsy knowledge and posi-tive attitudes, compared with the control group, 1 month after delivery ofthe program [27].

Elementary schools are good locations at which to offer such an educa-tion program in that children spend the majority of their time at school [29],and schools are motivated hosts given their safety and liability concerns

about epilepsy [30]. The decision to focus on students in Grade 5 was basedon developmental appropriateness and alignment with the educational cur-riculum. The child development theories of Piaget and Eriksson point to7- to 12-year-olds as the most appropriate target group for an epilepsy edu-cation program. They are learning to think logically about events andobjects and becoming aware of other people’s perspectives [31]. They arefocused on their school and family communities and enjoy developing con-crete skills. They are also developing their self-worth, so this is a particular-ly crucial age group at which to target antidiscrimination interventions [32].At this age, students begin to understand the meaning of friendship. Theylearn that qualities such as loyalty, honesty, trustworthiness, and being agood listener are the characteristics that make a good friend [33]. They alsobecome able to describe themselves using a variety of adjectives and areable to express their emotions with words, an essential developmental skillfor completing a questionnaire about attitudes [34]. Grade 5 students inCanada are 9 to 11 years old. Placing the education program here allowedus to align it with established curriculum guidelines and, thus, increase theodds that the program would be sustained in the future. In Ontario, chil-dren in Grade 5 learn about first-aid and bullying through their PhysicalEducation curriculum and about the brain and nervous system in theScience and Technology curriculum [35].

Development of the Thinking about Epilepsy questionnaire (AppendixA) and assessment of its psychometric properties followed the process rec-ommended by Aday [36]. Analyses were conducted using SPSS Version10.0.1 [37]. This study was approved by the Research Ethics Board forNon-Medical Research Involving Human Subjects at the University ofWestern Ontario. Approval was obtained from the school principals andteachers via a letter of information and formal consent forms were givento parents/guardians.

2.1. Questionnaire development

2.1.1. Identifying target concepts

The concepts selected for teaching, and therefore assessment by theepilepsy questionnaire, were those for which there is a persistent lack ofknowledge or toward which negative attitudes are known to exist. Deci-sions were informed by the existing literature regarding the public’s knowl-edge of and attitudes toward epilepsy, as well as those concepts deemed tobe important by the WHO, ILAE, IBE, and CEA and stakeholders inLondon, Ontario (ESC, UWO, and families affected by epilepsy). The con-cepts assessed by the Thinking about Epilepsy questionnaire were groupedinto several components. The five components for knowledge were: role ofthe brain (location, parts, and function), epilepsy (what epilepsy is, typesof epilepsy, and their manifestations), diagnosis (EEGs and treatment),first-aid (for tonic–clonic seizures), and causes of epilepsy. The three com-ponents for attitudes were: contagiousness (sharing a drink and sittingbeside), friendship (no-one’s fault, play with, can tell who has epilepsyby looks, hard to make friends), and that epilepsy should not limitachievement (intelligence, school and sports). The questionnaire alsoincludes demographic questions on: students’ school, gender, languagespoken at home, and familiarity with epilepsy. Socioeconomic status(SES) was assessed at the school level using data from the school boards.School boards rank each school on SES by combining data from theCanadian Census and internally collected school board data on: income,unemployment, lone parent, education, immigration, mobility, and schoolboard-ranked literacy levels. Schools are ranked from 1 (high) to 4 (low).

2.1.2. Questions from previous epilepsy questionnaires

Nine of the 36 questions in the Thinking about Epilepsy survey weretaken directly or revised from questions used in previously published stud-ies investigating epilepsy knowledge and attitudes [10–21]. The remainingitems were newly developed.

2.1.3. Refining the questionnaire

Readability of the questions was assessed for grade level prior to pre-testing, using a graphic technique developed by Fry, and rated to be at areading level of Grade 3 [38,39]. We conducted a pilot study in five Grade

A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603 597

5 classes (103 students) to assess the format of the questionnaire. As rec-ommended by Litwin, the pilot assessed: item numbering, font size, read-ability, length, skip patterns, flow of questions, item wording, and thecultural sensitivity of items [40]. Several changes were made based onthe pilot including: lowering the reading level of the instructions, catego-rizing the open response option about the TV commercial, and standard-izing the process of reading the questionnaire aloud to participants in allschools.

2.1.4. Scoring the questionnaire

The 36-item questionnaire has several sections. Questions 1–8 collectinformation on respondents’ sociodemographic characteristics. Questions9–26 focus on epilepsy knowledge. Epilepsy attitudes are assessed in ques-tions 27–36. The first set of knowledge questions used a true/false/don’tknow response format. ‘‘Don’t know’’ responses were included to mini-mize the perceived threat of the questions and to decrease the tendencyof the respondent to guess. These were later recoded as incorrect respons-es. Scores ranged from 0 to 11, with 11 signifying maximum knowledge.The second set of knowledge questions (No. 21–26) employed multipleresponse options using the concept of ‘‘sleepers,’’ which are fictionaloptions included to help estimate whether respondents are overreportingtheir knowledge [36]. The lowest possible score was assigned ‘‘0’’ for com-plete lack of knowledge. Scores could range from 0 to 3 for question 21, 0to 6 for question 22, 0 to 12 for question 23, 0 to 3 for question 24, 0 to 7for question 25, and 0 to 4 for question 26. Individual item scores wereadded to obtain a summary knowledge score [41]. The total knowledgescore for the multiple response items could range from 0 to 46, with 46signifying maximum knowledge. Therefore, the highest possible totalknowledge score was 57 points.

Questions 27–36 asked about attitudes regarding epilepsy. These ques-tions used a 5-point Likert scale scored from 1 to 5, chosen because scaleswith five to seven options are more valid than those with only two or threeoptions [36]. The number of attitudinal questions was kept to a minimumbecause the quality of responses tends to decline as the number of ques-tions increases, and this is particularly evident for questions about atti-tudes [36]. Positive questions (Nos. 29 and 30) were those in which‘‘strongly agree’’ was the most positive attitude. Reverse scoring was usedafter data entry for negative questions (Nos. 27, 28, 31, 32, 33, 34, 35, and36), where ‘‘strongly disagree’’ was the most positive attitude. Higherscores reflect more positive attitudes overall. The total attitude score couldrange from 10 to 50.

2.2. Psychometric properties of the questionnaire

2.2.1. Factor analysis

Prior to assessing the questionnaire’s reliability and validity, a factoranalysis was conducted to determine whether the questionnaire assessedtwo overall measures, knowledge and attitudes, as hypothesized, and todetermine how well the factor structure revealed fit with the original con-cepts identified as potential components of knowledge and attitudes [42].First, a scree test was used to determine how many measures to interpretfrom the full questionnaire (not including the demographic questions).Then, to better understand the relationships among items, correlationmatrices were obtained for the two domains emerging from the scree test.Variance accounted for by each of the items within these measures wasassessed using extraction communalities. We decided a priori that anyitems with extraction communalities below 0.2 at this stage would bedropped from further analyses [43]. Because we were interested in under-standing the factors making up the overall conceptual components(knowledge and attitudes), scree plots were then calculated for two mea-sures that emerged from the first scree plot, and the initial factors wereextracted using the maximum likelihood method of factor extraction forboth measures separately [24]. Factor extraction provided loading valuesto observe which items were involved in each factor pattern and to whatdegree. Following initial factor extraction, a varimax rotation with Kaisernormalization was used to transform the original matrices for easier inter-pretation [43]. The post-rotation factor loadings measure the importance

of an item within each measure and, thus, were used to inform the itemsmaking up each factor. Items were considered acceptable and retained ifthe loading coefficient was 0.3 or greater [44]. Items loading on more thanone factor were ascribed to the factor for which the loading coefficient washighest [45]. Following the distinction of the total number of measures toretain using a scree plot of all items on the questionnaire, and then furtherelucidating the multiple factors making up each measure, the theoreticalbackground and observation of high loading items then assisted in theinterpretation and naming of the final measures and their subfactors[43,45].

2.2.2. Reliability

Following factor analysis, Cronbach a values were calculated to assessthe internal consistency of the overall (knowledge and attitude) measuresand the subfactors for each measure. Test–retest reliability was not includ-ed as a measure of reliability, as it may underestimate the degree of reli-ability of measurements over time by interpreting true change asmeasurement instability [46]. As recommended for scale development, acutoff of 0.2 was used for inclusion of items based on item-to-total corre-lations within each domain [47]. In comparisons of groups, a values of 0.6to 0.7 are considered satisfactory and values of 0.7 to 0.9 ideal [48].

2.2.3. Validity

Face, content, and construct validity was assessed to determine howwell the questionnaire measures what it aims to measure. Face and contentvalidity was assessed qualitatively by consulting 11 experts regarding theappropriateness of the measure including the executive director, a socialworker, an epilepsy educator, and parent representative (from the EpilepsySupport Center), school board curriculum chairs (two school boardresearch officers, physical education curriculum chair, science and technol-ogy curriculum chair), as well as university faculty (two epidemiologistsand a biostatistician). Content validity was also sought through a compre-hensive literature review, as well as a review and application of existinginstrumentation where applicable. A table outlining program components,objectives, and rationale for each component and questions to assess eachcontent area was also used to ensure content validity. Last, we assessedconstruct validity, given the absence of any current ‘‘gold standard’’ forthe measurement of epilepsy knowledge and attitudes for children [49].Our method involved checking for congruent responses to particularknowledge and attitude items using the Spearman rank correlation coeffi-cient [40]. Correlations of 0.4–0.8 have been suggested as representingacceptable validity [41]. Three a priori predictions were made based onhypothesized theory that, if upheld, would offer support of constructvalidity: (1) subjects who had not heard about epilepsy would have overalllower knowledge scores than those who had heard about epilepsy; (2) sub-jects who knew someone with epilepsy would have higher overall knowl-edge scores than those who did not know anyone with epilepsy; and (3)subjects would score poorly on the question asking how epilepsy is diag-nosed on the pre-intervention questionnaire because EEG testing wouldnot likely be a topic Grade 5 students would know about prior to the epi-lepsy education program. Two predictions were made for the attitudedomain: (1) subjects agreeing or strongly agreeing with ‘‘you can catch epi-lepsy’’ would agree or strongly agree with ‘‘children with epilepsy proba-bly should not go to school’’; and (2) subjects agreeing or stronglyagreeing with ‘‘you can catch epilepsy’’ would agree or strongly agree with‘‘I would be afraid to sit beside someone with epilepsy.’’

3. Results

3.1. Study population

The total sample consisted of 810 Grade 5 students (ages9–11) from two school boards in southwestern Ontario.Twenty-seven students did not provide consent to partici-pate; therefore, all analyses were based on a sample of783 students who completed the questionnaire. The SES

598 A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603

of participating schools was well distributed (24% ofrespondents were from high SES schools, 20% from medi-um SES schools, 36% from low SES schools, and 20% fromvery low SES schools). There was an equal balance of maleand female respondents and just over half had heard of epi-lepsy at baseline (�60%). The majority spoke English astheir first language (73%). On average, the questionnairetook about 22 minutes to complete.

3.2. Factor analysis

The overall structure of the questionnaire was investi-gated using a scree plot (Fig. 1). The ‘‘elbow’’ point onthe scree plot revealed a three-measure structure overall[41,43]. We hypothesize that three measures emerged dueto the phenomenon where items with similar response for-mats load together (true/false/don’t know knowledge itemsas one measure, multiple-response knowledge items as asecond factor, and attitude items as a third factor) [42].To further investigate this, a sensitivity analysis was doneto examine the factor loadings after forcing three- and thentwo-measure results in SPSS. When a three-measure struc-ture was forced, knowledge questions with three responseoptions loaded together (questions 9–20 ‘‘true/false/don’tknow’’ as well as question 21 ‘‘shuts off/sending mixed mes-sages/is asleep’’ and question 24 ‘‘respirologist/neurologist/cardiologist’’); knowledge questions with more than threeresponse options loaded together; and attitude questionsloaded together. In comparison, when a two-measurestructure was forced, knowledge questions with threeresponse options loaded together and knowledge questionswith multiple responses loaded with the multiple-responseattitude questions. We believe this supports the phenome-non whereby questions with similar format complexityload together regardless of content [42].

Correlation matrices for the knowledge and attitudeitems showed that none of the correlations was beyondthe 0.9 range that would have indicated item redundancy.

Fig. 1. Scree plot of all Thinking about Epilepsy questionnaire items.

The initial and extracted communalities for the knowledgeand attitude domains demonstrated low extraction com-munalities for the knowledge item ‘‘people can have epilep-sy because of’’ and for the attitude items ‘‘it is someone’sfault if they have epilepsy’’ and ‘‘you cannot play sportsif you have epilepsy’’. These three items were subsequentlyremoved from further analyses.

Based on the interpretation of the overall scree plotdescribed above, we proceeded with analyses as conceptu-ally hypothesized, assuming a two-measure structure over-all (knowledge, attitudes). To assess the factors underlyingthis bidimensional structure, we conducted factor analyseswithin these two measures. Using a varimax rotation withKaiser normalization and the default option in SPSS toextract factors for items with eigenvalues greater thanone, five subfactors fell out for the knowledge measureand two for the attitude measure. Tables 1 and 2 list thefinal factor loadings and items included within each factorfor the knowledge and attitude measures, respectively.

Overall, the two overall measures (knowledge and atti-tudes) and their seven subfactors identified through factoranalysis fit relatively well with the original five knowledgeand three attitude concepts articulated during question-naire development. The seven items loading on the firstdomain appear to address our original target concept epi-

lepsy (what it is, types, causes): cannot be a seizure if notshaking, can always tell when someone has epilepsy, onlykids get epilepsy, there is only one type of seizure, seizuresaffect the whole body, seizures start in the brain, and a per-son is in control if he or she is having a seizure. The twoitems loading on the second domain both discuss whatbystanders may be worried about or what they might doin the event of seeing someone having a seizure (can swal-low tongue during a seizure and you need to hold a persondown to stop the shaking during a seizure). This domainwas named epilepsy manifestation. The two items loadingon the third domain discuss first-aid for a seizure and whento call an ambulance. This domain was named first-aid.The three items loading on the fourth domain are aboutmedicines for seizures and how to diagnose epilepsy (some-one with epilepsy must always take medicine, medicine canprevent some seizures, and EEGs help to diagnose epilep-sy). This domain was called diagnosis and treatment. Last,three items loaded on the fifth domain (what is happeningto the brain during a seizure, special doctor who looks afterthe brain is called, and how doctors look at brain waves).This domain was called role of the brain (Fig. 2). As men-tioned previously, one knowledge item (cause of epilepsy)was dropped from the questionnaire because it demonstrat-ed a low extraction communality (0.07) and failed to reachthe 0.3 cutoff in the rotated factor matrix for knowledge(see Tables 3 and 4).

Two domains were observed within the epilepsy attitudemeasure. Four items loaded on the first domain (wouldplay with, share drink, can catch, and sit beside), and allappear to be related to whether epilepsy is contagious ornot. For simplicity, this factor was called contagiousness.

Table 1Rotated factor matrix for knowledge

Note: Question 25 (People can have epilepsy because of . . .) was dropped because it loaded <0.20 (therefore, the knowledge domain was based on 17 itemsin total).

Table 2Rotated factor matrix for attitudes

Note: Questions 27 and 31 (‘‘It is someone’s fault if they have epilepsy’’ and ‘‘You cannot play sports if you have epilepsy’’) were dropped because theyloaded <0.20 (therefore, the attitude domain was based on 8 items in total).

A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603 599

This fell out as a separate factor, although contagiousnesswas not originally specified as a separate target concept.Four separate items loaded on the second domain (shouldnot go to school, can tell by looks, not smart, and hard tomake friends) and appear related to our a priori targetconcept that ‘‘epilepsy should not limit achievement’’(Fig. 3).

3.3. Reliability

Cronbach’s a for the total knowledge measure (includesall five subdomains) was 0.65, with item-to-total correla-tions ranging from 0.21 to 0.43. The a values for the fiveknowledge factors were 0.74, 0.57, 0.47, 0.52, and 0.26,respectively. Cronbach’s a for the total attitude measure(includes both subdomains) was 0.82, with item-to-totalcorrelations ranging from 0.36 to 0.68. The a values forthe two attitude factors making up the attitude domainwere 0.77 and 0.65.

3.3.1. Validity

Three a priori predictions were made to assess constructvalidity of the knowledge measure. First, mean knowledgescores were higher for those who had heard about epilepsythan for those who had not (27 points vs 23 points both witha SD of 5, P < 0.001). Second, mean knowledge scores werehigher for those who knew someone with epilepsy than forthose who did not (28 points vs 25 points, both with a SD of5, P < 0.001). Third, as hypothesized, few students knew thatan EEG is used to diagnose epilepsy, prior to receiving theeducation program (182 students, or 22%, answered correctlythat EEGs are used to diagnose epilepsy, and 630 students, or78%, answered incorrectly or said that they did not know).

Two predictions were made to assess construct validityof the attitude measure. First, it was expected that believingyou can catch epilepsy would be correlated with believingthat people with epilepsy should not go to school. TheSpearman correlation was 0.33 (P < 0.001, two-tailed),slightly lower than the minimum value suggested as

ruziesagnivahebnacnosrepA:41Q gnikahstonsiehs/ehfineveeeruziesagnivahsienoemosnehwlletsyawlanacuoY:51Q

yspelipeevahnac,stludaton,nerdlihcylnO:9QyspelipEeruziesfoepytenoylnosierehT:21Q

ydobelohwehttceffaseruziesemoS:31Q )sesuac,sepyt,sititahw(niarbehtnitratsseruzieS:11Q

eruziesagnivahsinosrepanehW:01Q ydobreh/sihfolortnocnisiehs/eh

eugnotriehtwollawsnacelpoeP:61Q yspelipEeruziesagniruderuziesgnikahsagnivahsienoemosnehW:71Q )noitatsefinam(nwodreh/mihdlohotdeenuoy

egdelwonK

…dluowIeruziesafoesuacebgnikahsenoemoswasIfI:32QyspelipehtiwenoemosnehW:22Q dia-tsriF…dellacebdluohsecnalubmanaeruziesagnivahsi

llewgnileeferayehtnehwneveenicidemekattsumyspelipehtiwenoemoS:02QtnemtaerT&sisongaiDseruziesemostneverpnacenicideM:91Q

yspelipeesongaidsplehGEEnA:81Q

ivahsinosrepanehW:12Q …niarbreh/siheruziesagnhtretfaskoolohwrotcodlaicepsehT:42Q niarbehtfoeloR…niarbe

…ybsevawniarbtakoolsrotcoD:62Q

Fig. 2. Questionnaire items and factors making up the knowledge measure.

Table 3Item-to-total correlations for the knowledge measure

Item Item-to-total correlation

Q9: Only children, not adults, can have epilepsy. 0.32Q10: When a person is having a seizure he/she is in control of his/her body. 0.35Q11: Seizures start in the brain. 0.40Q12: There is only one type of seizure. 0.36Q13: Some seizures affect the whole body. 0.35Q14: A person can be having a seizure even if he/she is not shaking. 0.36Q15: You can always tell when someone is having a seizure. 0.39Q16: People can swallow their tongue during a seizure. 0.26Q17: When someone is having a shaking seizure you need to hold him/her down. 0.43Q18: An EEG helps diagnose epilepsy. 0.29Q19: Medicine can help prevent some seizures. 0.29Q20: Someone with epilepsy must take medicine even when they are feeling well. 0.37Q21: When a person is having a seizure his/her brain . . . 0.22Q22: When someone with epilepsy has a seizure an ambulance should be called. 0.27Q23: If I saw someone shaking because of a seizure I would . . . 0.30Q24: The special doctor who looks after the brain is called . . . 0.21Q26: Doctors look at brain waves by . . . 0.26

Table 4Item-to-total correlations for the attitude measure

Item Item-to-total correlation

Q28: You can catch epilepsy. 0.57Q29: I would play with someone who has epilepsy. 0.66Q30: I would share a drink with someone who has epilepsy. 0.46Q32: Most people with epilepsy are not very smart. 0.52Q33: You can tell by looks who has epilepsy. 0.48Q34: Children with epilepsy probably should not go to school. 0.64Q35: It must be hard to make friends when you have epilepsy. 0.36Q36: I would be afraid to sit beside someone who has epilepsy. 0.68

600 A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603

yspelipe sah ohw enoemos htiw yalp dluow I :92Qenoemos htiw knird a erahs dluow I :03Q ssensuoigatnoC yspelipe sah ohw

yspelipe hctac nac uoY :82Q tis ot diarfa eb dluow I :63Q yspelipe sah ohw enoemos ediseb

sedutittA

loohcs ot og ton dluohs ylbaborp yspelipe htiw nerdlihC :43Q timil ton dluohs yspelipE yspelipe sah ohw skool yb llet nac uoY :33Q

era yspelipe htiw elpoep tsoM :23Q tnemeveihca trams yrev tonekam ot drah eb tsum tI :53Q yspelipe evah uoy nehw sdneirf

Fig. 3. Questionnaire items and factors making up the attitude measure.

A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603 601

supportive of the prediction [41]. Second, we expected thatbelieving you can catch epilepsy would be correlated withbeing afraid to sit beside someone who has epilepsy. Thisprediction was supported with a Spearman correlation of0.42 (P < 0.001, two-tailed).

4. Discussion

This article described the development and structure ofthe Thinking about Epilepsy questionnaire, designed toassess knowledge and attitudes of Grade 5 students regard-ing epilepsy, and assessed its psychometric properties. Ouranalyses supported a two-measure structure overall(knowledge, attitudes), with five factors within knowledgeand two factors within the attitude measure.

Reliability analyses demonstrated an acceptable Cron-bach a for the overall knowledge measure (0.7) and accept-able item-to-total correlations (0.2–0.4). However, a valuesfor the five knowledge subfactors were not ideal, rangingfrom 0.3 to 0.7. A strong a value was obtained for the over-all attitude measure (0.8), with good item-to-total correla-tions (0.4–0.7). a values for the two attitude subfactorswere good, 0.8 and 0.7. The low a values observed forthe knowledge subfactors may be due to the fact that fewitems are included in each factor or reflect a low correlationbetween items for each factor. Low a values can also indi-cate a factor that is more multidimensional. In this case, webelieve that the low knowledge subfactor a values reflectthat some knowledge factors have an attitudinalcomponent.

The criteria set to assess validity were surpassed for theknowledge measure and met for the attitude measure. Theprediction of a correlation between the items ‘‘you cancatch epilepsy’’ and ‘‘those with epilepsy should not attendschool’’ was weaker than hypothesized. We speculate thismay be because the items measure slightly differentconcepts, as there may be several reasons for answering‘‘people with epilepsy should not go to school’’ outside ofthe belief that ‘‘you can catch epilepsy.’’

There is one other questionnaire designed to assess epi-lepsy knowledge and attitudes in elementary school stu-dents. This measure, the Elementary School EpilepsySurvey (ESES), was developed at the same time as theThinking about Epilepsy questionnaire [26]. It wasdesigned to target children enrolled in grades 4–6 (ages

10–12) in Halifax, Canada. Its measurement propertieswere assessed in 155 Grade 4, 5, and 6 students, of whom36 were Grade 5 students. The participation rate was35% for the 447 eligible students. The ESES knowledgesubscale demonstrated low internal consistency reliabilitywith a Cronbach a of 0.50, whereas the attitudes subscalehad an acceptable value of 0.81. In terms of discriminantvalidity, the ESES knowledge scores did not significantlydiffer between those who had heard/read about epilepsyand those who had not, but ESES attitude scores were sig-nificantly better for the former group. The authors suggestthat continued item refinement is important to address theissue that the initial version of the ESES may have tappedinto different specific areas of knowledge about epilepsy.

The Thinking about Epilepsy questionnaire, as a newmeasure of knowledge and attitudes of students 10–12 yearsold, has two key methodological strengths important toconsider when interpreting the findings of acceptable psy-chometric properties reported here. The first is the largesample size on which the measurement analyses were per-formed. Insufficient sample sizes are a common methodo-logical flaw in principal component and factor analysis.According to a review of more than 1000 articles using prin-cipal component or factor analyses to evaluate question-naires, about half failed to obtain appropriate samplesizes [50]. There are a variety of popular ‘‘rules’’ for deter-mining appropriate sample size: the commonly used ‘‘ruleof 10’’ that there should be at least 10 cases for each itemon the instrument [47]; the ‘‘rule of 100,’’ where many moresubjects than 100 are recommended when few variables loadon many factors [51]; the ‘‘rule of [greater than] 150’’ [52];‘‘the rule of 200’’ [45]; ‘‘the rule of 300’’ [53]; and the scaleproposed by Comfrey and Lee [44], where a sample of 100is poor, 200 fair, 500 good, and 1000 excellent. The mainconcern with an insufficient sample size is that the final solu-tion (component/factor structure) is more likely to beunstable. Unstable solutions may mean that erroneous fac-tors are extracted or that items are assigned to the wrongfactors [42]. In this case, the addition of more or new data(i.e., applying the questionnaire to a new population) maylead to a different structure or cause items to switch fromone factor to another [54]. The sample size used to assessthe Thinking about Epilepsy questionnaire was sufficientto exceed each ‘‘rule’’ with a total sample of 783 cases,lending confidence to the estimates we reported.

602 A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603

The second key methodological strength is the excellentresponse rate. We were successful in obtaining completedsurveys from 97% of the target sample. Obtaining a repre-sentative sample such as this minimizes the likelihood thatnonresponders may be systematically different thanresponders, thus strengthening generalizability of the find-ings and potentially increasing the stability of the findingsobserved. Future research could assess the stability of theproposed structure for the Thinking about Epilepsy ques-tionnaire in other populations.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.yebeh.2007.01.011.

References

[1] World Health Organization (WHO). Mental and neurologicaldisorders. Fact sheet No 265 [monograph on the Internet]; 2001[cited 2006 Nov 17]. <http://www.who.int/mediacentre/factsheets/fs265/en/>.

[2] World Health Organization (WHO). Epilepsy: aetiology, epidemiol-ogy and prognosis. Fact sheet No 165 [monograph on the Internet];2001 [cited 2006 Nov 17]. Available at: <http://www.who.int/media-centre/factsheets/fs165/en/>.

[3] Keusch GT, Wilentz J, Kleinman A. Stigma and global health:developing a research agenda. Lancet 2006;367:525–7.

[4] Link BG, Phelan JC. Stigma and its public health implications.Lancet 2006;367:528–9.

[5] Goffman E. Stigma: notes on the management of spoiled identi-ty. New York: Simon & Schuster; 1986.

[6] Weiss MG, Ramakrishna J. Stigma interventions and research forinternational health. Lancet 2006;367:536–8.

[7] World Health Organization (WHO). Global campaign againstepilepsy: out of the shadows. [monograph on the internet]; 1997[cited 2006 Nov 17]. <http://www.who.int/mental_health/manage-ment/globalepilepsycampaign/en/index.html/>.

[8] Canadian Epilepsy Alliance [homepage on the internet]. National surveyresults: people with epilepsy speak out. 2000. Vol 2, Issue 2 [cited 2006Nov 17]. <http://www.epilepsymatters.com/english/survey3.html/>.

[9] Fishbein M, Ajzen I. Belief, attitude, intention and behavior: anintroduction to theory and research. Addison–Wesley: Reading,MA; 1975.

[10] DeIorio CA, Kobau R, Holden EW, et al. Developing a measure toassess attitudes toward epilepsy in the US population. Epilepsy Behav2004;5:965–75.

[11] Kobau R, DiIorio CA, Anderson LA, Price PH. Further validationand reliability testing of the attitudes and beliefs about living withepilepsy (ABLE) components of the CDC epilepsy program instru-ment on stigma. Epilepsy Behav 2006;8:552–9.

[12] Bener A, al-Marzooqi FH, Sztriha L. Public awareness and attitudestoward epilepsy in the United Arab Emirates. Seizure 1998;7:219–22.

[13] Gambhir SK, Kumar V, Singhi PD, Goel RC. Public awareness,understanding and attitudes toward epilepsy. Indian J Med Res1995;102:34–8.

[14] Gutteling JM, Seydel ER, Wiegman O. Previous experiences withepilepsy and effectiveness of information to change public perceptionof epilepsy. Epilepsia 1986;27:739–45.

[15] Jensen R, Dam M. Public attitudes toward epilepsy in Denmark.Epilepsia 1992;33:459–63.

[16] Lai CW, Huang XS, Lai YH, Zhang ZQ, Liu GJ, Yang MZ. Surveyof public awareness, understanding and attitudes toward epilepsy inHenan Province, China. Epilepsia 1990;31:182–7.

[17] Mirnics Z, Czikora G, Zavecz T, Halasz P. Changes in publicattitudes toward epilepsy in Hungary: results of surveys conducted in1994 and 2000. Epilepsia 2001;42:86–93.

[18] Nyame PK, Biritwum RB. Epilepsy: knowledge, attitude and practicein literate urban population, Accra, Ghana. West Afr J Med1997;16:139–45.

[19] Radhakrishnan K, Pandian JD, Santhoshkumar T, et al. Prevalence,knowledge, attitude and practice of epilepsy in Kerala, South India.Epilepsia 2000;41:1027–35.

[20] Rwiza HT, Matuja WB, Kilonzo GP, et al. Knowledge, attitude andpractice toward epilepsy among rural Tanzanian residents. Epilepsia1993;34:1017–23.

[21] Santos IC, Mirnics Z. Public awareness and attitudes toward epilepsy indifferent social segments in Brazil. Arq Neuropsiquiatr 1998;56:32–8.

[22] Young BG, Derry P, Hutchinson I, et al. An epilepsy questionnairestudy of knowledge and attitudes in Canadian college students.Epilepsia 2002;43:652–8.

[23] Austin JK, Shafer PO, Deering JB. Epilepsy familiarity, knowledge,and perceptions of stigma: report from a survey of adolescents in thegeneral population. Epilepsy Behav 2002;3:368–75.

[24] Austin JK, MacLeod J, Dunn DW, Shen J, Perkins SM. Measuringstigma in children with epilepsy and their parents: instrumentdevelopment and testing. Epilepsy Behav 2004;5:472–82.

[25] Mudge PR, Turner W. Health education on epilepsy. Aust FamPhysician 1987;16:1356–65.

[26] Lowe-Pearce C, Camfield CS. Elementary School Epilepsy Survey(ESES): a new measure of elementary school students’ knowledge andattitudes about epilepsy. Epilepsy Behav 2005;7:687–96.

[27] Martiniuk ALC, Speechley KN, Campbell MK, Donner A. A clusterrandomized trial to evaluate an epilepsy education program inOntario, Canada, submitted for publication.

[28] Bandura A. Social learning theory. New York: General LearningPress; 1971.

[29] Juster FT, Ono H, Stafford FP. Changing times of American youth:1981–2003. [monograph on the internet]. Institute for Social ResearchUniversity of Michigan; Ann Arbor, MI; November 2004. [cited 2006Dec 20]. <http://www.umich.edu/news/index.html?Releases/2004/Nov04/r111704a/>.

[30] Olson AL, Seidler AB, Goodman D, Gaelic S, Nordgren R. Schoolprofessionals’ perceptions about the impact of chronic illness in theclassroom. Arch Pediatr Adolesc Med 2004;158:53–8.

[31] Maier HW, Erikson EH, Piaget J, Sears RR. Three theories of childdevelopment: the contributions of Erik H. Erikson, Jean Pia-get. New York: Harper & Row; 1969.

[32] Gordon N, Sillanpaa M. Epilepsy and prejudice with particularrelevance to childhood. Dev Med Child Neurol 1997;39:777–81.

[33] Dembo MH. Applying educational psychology. 5th ed. NewYork: Longman; 1994.

[34] Eller CL, Mulroy MT, for the National Network for Child Care(NNCC). Developmentally appropriate programming for school-agechildren. Storrs, CT: National Extension Service Children Youth andFamily Educational Research Network; 1993.

[35] Ontario Ministry of Education. [homepage on the internet] Schoolcurriculum guidelines. [cited 2006 Dec 20]. <http://www.edu.gov.on.ca/eng/curriculum/elementary/grade5.html/>.

[36] Aday LA. Designing and conducting health surveys. San Francis-co: Jossey–Bass; 1996.

[37] SPSS for Windows, Rel. 10.0.1 2000. Chicago: SPSS Inc.[38] Fry EB. A readability formula for short passages. J Reading

1990;33:594.[39] Bruinsma R. Readability of B.C. high school textbooks: Fry and

SMOG formulae. Vancouver, BC: Educational Research Institute ofBritish Columbia; 1978.

A.L.C. Martiniuk et al. / Epilepsy & Behavior 10 (2007) 595–603 603

[40] Litwin MS. How to measure survey reliability and validity. Lon-don: SAGE; 1995.

[41] Streiner DL, Norman GR. Health measurement scales: a practicalguide to their development and use. 2nd ed. Oxford: Oxford MedicalPublisher; 2001.

[42] Tabachnick BG, Fidell LS. Using multivariate statistics. 3rd ed. NewYork: Harper Collins; 1996.

[43] Pett MA, Lackey NR, Sullivan JJ. Making sense of factor analysis:the use of factor analysis for instrument development in healthcare research. Thousand Oaks, CA: Sage; 2003.

[44] Comrey AL, Lee HB. A first course in factor analysis. Hillsdale,NJ: Lawrence Erlbaum; 1992.

[45] Gorsuch RL. Factor analysis. 2nd ed. Hillsdale, NJ: LawrenceErlbaum; 1983.

[46] Carmines EG, Zeller RA. Reliability and validity assessment.Newbury Park, CA: Sage; 1979.

[47] Nunnally JC, Bernstein IH. Psychometric theory. New York: Mc-Graw–Hill; 1994.

[48] Bland JM, Altman G. Statistics notes: Cronbach’s alpha. Br Med J1997;314:572.

[49] Rossi PH, Freeman HE, Lipsey MW. Evaluation: a systematicapproach. 6th ed. Thousand Oaks, CA: Sage; 1999.

[50] Costello AB, Osborne JW. Exploring best practices in factor analysis:four mistakes applied researchers make. Paper presented at theannual meeting of the American Educational Research Association;Chicago, IL; 2003.

[51] Hatcher L. A step-by-step approach to using the SAS system forfactor analysis and structural equation modeling. Cary, NC: SASInstitute; 1994.

[52] Hutcheson G, Sofroniou N. The multivariate social scientist: intro-ductory statistics using generalized linear models. Thousand Oaks,CA: Sage; 1999.

[53] Norusis MJ. SPSS 13.0 guide to data analysis. NJ: Prentice–Hall: Upper Saddle River; 2005.

[54] Guadagnoli E, Velicer WF. Relation of sample size to the stabilityof component patterns. Psychol Bull 1988;103:265–75.