Communily Dcnl Oral Epidemiol 1996: 24: 346-50Piinlcil in Denmark . .•(// righis reservcil

t © Muiiksiia(n-({ 1996

Communify Dentistryand Oral Epidemiology

ISSN 0301-5661

Who is dentally anxious?Concordance between measures otdental anxiety

D. Locker\ D. Shapiro^ andA. LiddelP^Faculty of Dentistry, tjniversity of Toronto,^Department of Psychology, fVlemorial iJniversityof Newfoundland, St John's, Canada

Locker D, Shapiro D, Liddell A: Who is dentally anxious? Concordance betweenmeasures of dental anxiety. Community Dent Oral Fpidemiol 1996; 24:346-50. © Munksgaard, 1996

Abstraet - Studies of the prevalence of dental anxiety in general population sam-ples have produced estimates which range from a low of 2.6% to a high of20.4%. It is not clear whether these reflect real differences among populations orwhether they are the result ofthe use of different measures and different cut-off points. We undertook a large scale mail survey of dental anxiety in a randomsample of the adult population living in Metropolitan Toronto designed to assessthe performance of and agreement between three measures. These were Corah'sDAS, the single item used by Milgrom and colleagues in Seattle and the ten-point fear scale used by Gatchel. These measures and their published cut-offpoints produced prevalence estimates of 10.9'>̂ i, 21>.A"/n and 8.2'yii respectively.While there was a significant association between scores on pairs of measures theagreement between them was far from perfect. Kappa values ranged I'rom 0.37to 0.56, indicating only fair to moderate agreement beyond chance. There wasevidence to indicate that the dentally anxious subjects identified by each measurediffered according to certain behavioural and other eharaeteristies. The results ofthe study suggest the need to levisit the issue of measurement in studies ofdental anxiety.

Key words: dental anxiety: prevalence: measures

David Locker, Faculty of Dentitry, University ofToronto, 124, Edward Street, Toronto, OntarioM5G 1G6, Canada

Accepted for publication 26 February 1996

Given the significance of dental anxiety(1-4), investigators have invested con-siderable effort in studies of its preva-lence and etiology. In order to facilitatethis work, many have developed mea-sures or scales for the identification ofdentally anxious subjects and assess-ment of tbeir level of dental anxiety.These range I'rom single item indicatorssuch as those used by MtLGROM el al.(1) and GATCHEL (5), to more compre-hensive multi-item scales such as theDental Fear Survey (6) and the Struc-tured Interview for Assessing DentalFear (7).

While there are some advantages tohaving a number of measures and scalesfrom which to chose, it also has its dis-advantages. For example, estimates ofthe prevalence of dental anxiety based

on these scales vary I'rom a low of 2.6%(8) to a high of 20.4'/o (1). Consequent-ly, since investigators have used dif-ferent seales and measures and differentcut-offs in identifying who is and is notdentally anxious, it is not clear whetherthese estimates refiect real differencesamong populations or whether they aremethodological in origin.

Only a few investigators have com-pared the performance of differentmeasures by using them in the samestudy. HAKi-tsiiRG et al. (9), in study ofthe Swedish population, used the Den-tal Anxiety Sealc and the Gatchel 10-point Fear Scale and found that theprevalence of high dental anxiety was6.7*̂ 11 according to the former and 5.4'^according to the latter. 'VA.sstiND (10)also used the same measures and re-

ported prevalence rates of A.2% and7.1% for a population of Norwegianadults. Moore et al (11) used CorahtsDental Anxiety Scale (DAS) (12), thesingle item used by MILGROM et al. (1)in the Seattle study, and the final itemfrom the Dental Fear Survey (6). Theprevalence of dental anxiety obtainedby each measure was 10.2'^, 9.1"Ai and10.3'VIi respectively. Correlations be-tween scores on pairs of tbcsc measuresranged from 0.68 to 0.78. Same subjectanalyses also indicated a close associa-tion between the measures. For exam-ple, 78'yii of subjects responding "terri-fied" or "very afraid" to the Seattleitem had DAS scores of 15 or more.

However, although these investiga-tors assessed the level of agreement be-tween these different measures in terms

Measures of dentitl anxiety 347

of prevalence, none assessed the level ofagreement in terms of cases. This is im-portant because the statistical conceptsof association and agreement are verydifferent. Consec^uently, even whenprevalence estimates are close, it is pos-sible tbat these measures identify dif-ferent individuals as being dentally anx-ious. If tbis is the case, then the preva-lence of dental anxiety may beunderestimated and tbe results of stud-ies of such issues as etiology may differaccording to which sub-set of the den-tally anxious population is identified.

Accordingly, we included three com-monly-used measures of dental anxietyin a large scale study of Canadianadults in order to assess their associa-tion and agreement. The aims of thispaper are as follows: 1) to identify theextent of similarities and differences inestimates of the prevalence of dentalanxiety produced by different measures;2) to assess the degree of agreement be-tween measures and 3) to determine ifthere are differences in the characteris-tics of dentally anxious subjects iden-tified by each measure.

Methods

The target population for the study wasall adults 18 years and over living in theCity of Etobicoke, one of five munici-palities which comprise MetropolitanToronto. Census data for 1991 showedthat the city contained 256,390 personsaged 18 years and over. The samplingframe was the list of registered voterscovering tbe city. This list is compiledby a household enumeration processand, as a result, has been estimated byStatistics Canada to contain the nameand address of 91"/« of those eligible tovote.

A two-stage random start systematicsampling procedure was used in whichprimary sampling units were pollingsub-divisions and secondary units werenamed persons. The sampling fractionswere selected to give a final sample of6360 subjects.

Data were collected by means of afour wave-mail survey. The question-naire used in the study was comprehen-sive and included sections on aversiveexperienees, dental anxiety, dental visit-ing behaviours, attitudes towards den-tists and fear of pain. A complete ver-sion of the questionnaire was used in

tbe lirst and third mailings (the secondconsisted of a reminder post-card),while the fourth mailing contained ashortened version of the questionnairein the hope of stiiiuilating a responsefrom hard-core non responders.

The three measures of dental anxietyused in the study were chosen becausethey have recently been employed instudies of North American populationsand because they have clear cut-olTpoints for classifying subjects as den-tally anxious or not. The first was Cor-ah's DAS, a four item measure givingrise to scores of 4 to 20. Subjects withscores of 13 or above are considered tobe dentally anxious (13). The secondwas the single item I'rom the Seattlestudy in which subjects are asked "Howdo you rate your feelings toward dentaltreatment". The response options wereas follows: not at all afraid (1), a littleafraid (2), somewhat afraid (3), veryafraid (4) and terrified (5). Subjectswith response options 3 to 5 were con-sidered to be dentally anxious. Thethird was Gatchel's 10-point fear .scale(5). Subjects were asked to rate theirfear of dentistry on this scale in which1 indicates no fear, 5 moderate fear and10 extreme fear. Subjects seoring 8 to10 were considered to be highly anxiousabout dental treatment.

Other measures included on the ques-tionnaire were the Dental Fear Survey(6) which includes questions on anxiety-invoking stimuli and sclf-perccivcdphysiological responses. Subjeets werealso asked about negative dental experi-enees in tbe form of treatment that waspainful, frightening or embarrassing.Other questions addressed the fear ofpain and whether or not subjects con-sidered themselves generally anxiousand fearful.

The relationships between the mea-sures of dental anxiety were assessedusing correlation coeffieients, chi-square tests and oneway analysis ofvariance. Agreement between pairs ofmeasures was assessed using tbe kappastatistic. This gives an indication of theextent of the agreement beyond thatwhich would be expected by chance(14).

Results

Response and representativeness of suh-jeets - Of 6360 mailed ciuestionnaires.

1254 were returned by the Post Officeand a further 45 were returned becausetbe subject bad died, leaving 5061 sub-jects presumed to be alive, living at thelisted address and. therefore, eligible forthe study. Completed questionnaireswere returned by 3055 persons, giving aresponse rate of 60.4%. The long ver-sion ofthe questionnaire was completedby 2729 subjects and the short versionby 326. This paper is based on datafrom the former.

A comparison of the sociodetno-graphic characteristics of tbe subjectsI'rom whom complete data were ob-tained and tbe target population fromwhich they were drawn showed that,while study subjects were broadly re-presentative of the target popula-tion, males, persons less than 30 yearsof age and those with lower levels of ed-ucational attainment were under-sam-pled.

Prevalenee of dental att.xiety - Table1 shows the percentage of subjects cate-gorized as dentally anxious by each ofthe three measures. There was a three-fold difference in the prevalence rate ac-cording to the measure used. This var-ied from 8.2% for the Gatchel FS to23.4'y;i for the Seattle item. The DASgave a prevalence of \0.9%.

Table 1 also shows tbat all three mea-sures revealed significant differences indental anxiety by gender with womenhaving rates almost twice as high asmen. The pattern according to age wasalso similar for the three measures withthe oldest age group less likely to beanxious than tbe younger age groups.Of some interest was the fact that alltbree revealed tbat those aged 18 to 29years had a rate lower than those aged30 to 49 years.

Assoeiation hetween the three mea-sures - The Spearman rank correlationcoefficients between tbe three measureswere all high and significant (P<0.001)as follows: DAS vs Seattle - 0.78; DASvs Gatchel - 0.77; Seattle vs Gatchel -0.74.These elose associations betweenscores on the three measures were alsorevealed in analyses using oneway anal-ysis of variance and the chi-square test.For each pair of measures tbe associa-tion was signilicant at the /'<0.0001level.

Agt-eentettt betweeti the tneasures -Although the associations between thethree measures were highly significant.

3 4 8 LOCKER ET AL.

Table I. Prevalence of dental anxiety according to three measures

DASSeattle Gatchel

item FS

All subjects:

Sex:MalesFemales

Age:18-29 years30-49 years50-69 years70 years and over

10.9 23.4 8.2

7 3***13.7

10.1**13.39.57.5

17.5****28.1

23.1***26.223.714.9

5.5****9.9

8.2*9.67.74.0

*P<0.05; **P<().()\-, ***P<0.001: ****P<O.OOOIP values from chi-square tests

they were far from perfect. For exam-ple, n.l% of those responding "terri-fied" to the Seattle item had DAS seoresof 12 or less and were not categorizedas anxious on this measure. Similarly,22.2% of these subjeets were not classi-fied as dentally anxious according totheir scores on the Gatchel dental fearscale.

This less-than-perfect agreement isfurther illustrated in Table 2 whichshows cross-tabulations between eachofthe three pairs of measures. The pub-lished cut-off points were used to reduceeach measure to a dichotomy. Although

Table 2. Agreement between measures ofdental anxiety (n's only)

A. DAS with Seattle Item

DAS y?^No

Seattle

Yes

366k=0.48

B. Gatchel FS with Seattle Item

Gatchel FS f̂̂^No

"All agreement=82.3%:

C. DAS with Gatchel

DAS yf**No

% agreement=92.4yi);

Seattle

Yes

188440

k=0.37

FS

Item

No

352002

Item

No

301991

Gatchel FS

Yes

15062

k=0.56

No

1382294

Totals vary due to missing values on somemeasures

the percent agreement was high betweenpairs of measures, ranging from 82.3'̂ !!to 92.4'^, the kappa statistics rangedfrom 0.37 to 0.56, suggesting thatagreement corrected for chance wasonly fair to moderate (14)

Given the much higher prevalencerate obtained by the Seattle item, lowlevels of agreement with the other twomeasures are to be expected. However,the agreement between the DAS andthe Gatchel item, which producedbroadly comparable prevalence rates, isnot much better. Between them, the twomeasures classified 350 people as beingdentally anxious but agreement wasrcaehed on only 150. In addition, ifonly those subjects responding "veryafraid" or 'terrified' to the Seattle itemare considered dentally anxious, theprevalence drops to l.AVn, very close tothe 8.2'y;, obtained by the Gatchel FS.In this case the agreement between themeasures is better (k = 0.62), but tbeyconcur on only 134 of the 280 casesidentified. Similarly, if a DAS cut-olTpoint of 12 is used (11), the prevalenceon this measure inereases to 15.4'V!i andthe agreement with the Seattle item im-proves (k=0.59), but only to a levelwhere they agree on 49.9% of casesidentified. However, using this cutoffpoint, agreement with the Gatchel FSdeclines (k=0.50) and the two measuresagree on only 35.4'^ of the eases iden-tified.

When all three measures were exam-ined together, a total of 679 subjects, or25.6%i overall, were classified as dentallyanxious,but only 145 were so classifiedby all tbree measures. Anotber 161 sub-jects were judged to be dentally anxiousby two of the measures and 376 by one.The Venn diagram in Figure 1 shows in

diagrammatic form the extent of over-lap between the three measures.

Characteristics of dentctllv an.xitntssubjects - Dentally anxious subjectsidentified by each measure were com-pared on a number of variables knownto be associated with dental anxiety.These variables reflect antecedent fac-tors, evoking stimuli and anxiety re-sponses in the form of behavioural,physiologieal and cognitive outcomes.

The data summarized in Table 3 sug-gest that the dentally anxious subjectsidentified by eacb of tbe measures weresimilar with respect to antecedent fac-tors and evoking stimuli, but differentwith respect to anxiety responses. Ex-treme behavioural, physiological andemotional responses were less commonamong those identified by the Seattlemeasure than the other two.

Discussion

When used with the same population,three commonly-used measures of den-tal anxiety produeed very differentprevalence rates, ranging from 8.2'/<i to23.4'!/]. While they varied widely, therates produced by the individual mea-sures were similar to those reported inprevious studies of North Americanpopulations. Gatchel's (5) study of arandom sample of adults in Dallasfound that 1 \ .2% were dentally anxiousand MlLGROM et at. (I) reported a rateof 20.4'>;i in their Seattle survey of asample of the general population. Norecent study of North American adultshas used the DAS. However, Locker atal (15) reported a rate of 8.4'V;i for asample of Canadians aged 50 years andover. In this study, the rate for this agegroup was 9.0'^. This suggests that eachmeasure produces fairly stable estimateswhen used across populations. Aninteresting exception to this observationis to be found in a study by MOORE et

al. (II) conducted in Denmark, ln theirstudy, the Seattle item produced aprevalence estimate 10.2'^, only half therates reported by MILGROM et al. {1)

and observed bere.

There are two reasons why the ratesobtained by the three measures differ bysuch a wide margin. The most obviousconcerns the cut-off points each mea-sure uses to distinguish between tbosewho are and are not dentally anxious.The Gatchel fear scale uses the most

Mea.utres of dental an.xiety 349

Seattle

DAS

Gatchel FSFig 1. Overlap between the three measures of dental anxietv.

stringent cut-off point and tbe Seattleitctn the most liberal.-The former istben likely to include only the mostsevere cases while the latter probably in-cludes those with both moderate andhigh levels of anxiety about dentaltreatment.

Tbe second reason concerns differ-ences in the explicit or implicit con-structs underlying these measures. Thewording of three of the DAS items re-flects botb feelings and physiological re-sponses (16). LINDSAY & .IACK.SON (17)

have argued that this wording meansthat the DAS may miss or underestimatethe anxiety levels of individuals who donot respond physically to dental treat-ment. The construct underlying theSeattle item appears to be behaviouralrather than physiological in terms of

anxiety response. According to MtL-GROM et al. (1), individuals who reportbeing "somewbat afraid" of dental treat-ment should be included in the high feargroup sinee clinical experience indicatesthat these individuals show behaviouralpatterns, such as cancellation of dentalvisits or avoidance of dental care, typicalof dentally anxious individuals.

Another interesting observation wasthat the agreement between the mea-sures of dental anxiety was at best fairto moderate. This was to be expectedwhen measures resulted in very dif-ferent prevalence rates but it remainedtbe case when measures producingbroadly similar prevalence rates werecompared. In this situation, the mea-sures agreed on less than half the casesthey identified.

Table 3. Characteristics of dentally anxious subjects identified by eaeh measure

DASSeattle

ItemGatchel

FS

"All with prior aversive dental experiencesMean number of fear evoking stimuli% with fear of pain"Ao who have avoided dental care% with extreme physiological response to treatment% having had a panic attack in opcratory"All generally fearful

94.57.6

83.081.664.234.710.9

91.65.3

82.969.745.020.812.4

94.18.0

87.479.071.633.512.2

In tbe absence of any "gold stan-dard" against which the performance ofthe measures ean be judged, this raisesthe question of who should be consid-ered to be dentally anxious. Is it the 145subjects so classified by all three mea-sures or the 679 classified by at leastone? The former gives a prevalence rateof 5.A% and the latter a rate of 25.6%.

As Sct4uuRS & HOOGSTRATEN (16) in-

dicated in their review of dental anxietyscales, many of tbose used in epidemio-logical research are ad hoc in the sensethat they are not explicitly linked to aconceptual construct or theoreticalframework. Moreover, the validity ofmany measures bas not been adequatelyassessed, perbaps because of the diffi-culty of identifying an external referentagainst which it may be tested. Tbe is-sue of measurement in dental anxietyresearch may need to be revisited andconsideration given to whether it is suf-ficient for an individual simply to re-port high levels of fear to be classifiedas dentally anxious or whether evidencein tbe form of behavioural, physiologi-cal or cognitive characteristics shouldbe required as well. As LINDSAY &

.IACK.SON (17) note, new questionnairesto measure dental anxiety and associ-ated responses are needed which en-compass new knowledge about contrib-uting factors. Since dental anxiety is acontinuum, where to place the cut-offpoint is a crucial issue in researeh whichaims to compare subjects fearful of den-tistry witb tbose wbo are not.

We also attempted to assess if therewere dilTerenees in the characteristics ofthe subjects identified as anxious byeach of tbe measures. On some charac-teristics the groups were remarkablysimilar and on others, particularly be-havioural and physiological responses,there were significant differences. Atthis point it is difficult to judge whetherthis is simply a reflection of differencesin levels of severity of dental anxiety orwhether it means that some measuresidentify qualitatively different sub-groups of the dentally anxious popula-tion. Nevertheless, it does suggest thatfindings regarding dental anxiety, its eti-ology and its outcomes may be infiu-eneed by the measure used to identifydentally anxious subjects.

One immediate solution to the prob-lems of measurement revealed in tbispaper is to use more than one measure

350 LOCKER ET AL.

of dental anxiety in each study and tolook for corroborating evidence thatthose who say they are fearful of dentaltreatment indeed are. SCHUURS ANDHOOGSTRATEN (16) recommend this ap-proach beeause all dental anxiety mea-sures and questionnaires have limita-tions and do not completely cover theconcept. It would also aid in our under-standing of dental anxiety if investiga-tors were consistent in their use of mea-sures and the eut-off points used toidentify dentally anxious subjects. Truevariations in prevalence rates and asso-ciated characteristics across popula-tions could then be identified wbieb arecurrently masked by the use of differentscales and indices.

References

I. MILGROM P, FISI:T L, MHLNICK S.WiiiNSTEiN P. The prevalence and prac-tice management consequences of dentalfear in a major U.S. city. J Am DentA.tsoc 1988; 116: 641-647.

2. BiiRGGKiiN U. 1V1EYNI;KT G. Dental fearand avoidance: causes, symptoms andconsequences. J Atn Dent Assoc 1984;109: 247-251.

3. LOCKER D, LIDDELL A. Clinieal corre-lates of dental anxiety in older adults.Comtnunitv Dent Oral Epidetniol 1992;20: 371-375.

4. Liu R, BARTY J, WEINMAN J. Effects ofbackground stress and anxiety on post-operative recovery. Anaesthesia 1994; 49:382-386.

5. GATCHEL R. The prevalence of dentalfear and avoidance: expanded adult andrecent adolescent surveys. J Atn DentA.S.SOC 1989; 118: 591-593.

6. KLEINKNECHT R, KLEPAC R , ALI:XAN-DER L. Origins and characteristics offear of dentistry ./ Am Detit A.s-.s'oc 1973:20: 437-443,

7. VRANA S, MCNEIL D, MCGLYNN D, Astructured interview for assessing dentalfear. J Behav Ther E.\p Psvchiat 1986;17: 175-178.

8. AGRAS D, SYLVESTER D. OLIVEAU D. Theepidemiology of common fears and pho-bias. Comprehensive Psyehiatrv 1969; It):151-156.

9. HAKEUERG M . BERGREN U, CARLSSONSG. Prevalence of dental anxiety in anadult population in a major urban area

in Sweden. Cotittnitnilv Dent OralEpidetniol 1992; 20: 97-101.

10. VASSEND O. Anxiety, pain and discom-fort associated with dental tieatmcnl.Behav Res Ther 1993; 31: 659-666.

It. MOORE R, BIRN H , KIRKEGAARD E, BRO-DAGAARD I, ScHEUTZ F. Prevalence andeharaeteristies of dental anxiety inDanish adults. Commuiiitv Dent OntlEpidetniol 1993; 21: 292-296.

12. CORAH N . Development of a denial anx-iety scale. J Dent Res 1969: 48: 596.

13. CORAH N , GALE E, ILLIG S. Assessmentof a dental anxiety scale. ./ Atn DentAssoe 1978; 97: 816-819,

14. ALTMAN D, Practical Statistics for Medi-cal Research. London: Chapman andHall, 1991.

15. LOCKER D, LIDDELL A. Correlates ofdental anxiety among older adults. JDetU Res 1991; 70: 198-203.

16. SCHUURS A, HOOGSTRATEN .1. Appraisalof dental anxiety and fear question-naires: A review. Conimiinitv Dettt OrittEpidetniol 1993; 21: 329 339.

17. LINDSAY S, .IACKSON C. Fear of routinedental treatment in adults: Its nature andmanagement. Psvehologv and Health1993; 5.- 135-154.'