Vous êtes sur la page 1sur 8

Cotnmunity Detit Oral Epidemiol 1997: 25: 284-90 Copyright © Miinksgaani 1997

Printed iti Denmark . Alt rights reserved


Communify Dentistiy
and Oral Epidemiology
ISSN 0301-5661

Derivation and validation of a short- Gary D, Slade


Department of Dental Ecology, University of
North Carolina, USA

form oral health impact profile


Slade GD; Derivation and validation of a short-fortn oral health impact profile.
Comtnunity Dent Oral Epidetniol 1997; 25; 284-90. © Munksgaard, 1997

Abstract - Growing recognition that quality of life is an important outcome of


dental care has created a need for a range of instrutnents to measure oral health-
related quality of life. This study aimed to derive a subset of items from the Oral
Health Impact Profile (OHIP-49) - a 49-itetn questionnaire that tneasures peo-
ple's perceptions of the impact of oral conditions on their well-being. Secondary
analysis was conducted using data from an epidemiologic study of 1217 people
aged 60-1- years in South Austraha. Internal reliability analysis, factor analysis
and regression analysis were undertaken to derive a subset (OHIP-14) question-
naire and its validity was evaluated by assessing associations with sociodetno-
graphic and clinical oral status variables. Internal reliabihty of the OHIP-14
was evaluated using Cronbach's coefficient a. Regression analysis yielded an opti-
mal set of 14 questions. The OHIP-14 accounted for 94% of variance in the OHIP-
49; had high reliability (a=0.88); contained questions from each ofthe seven
conceptual dimensions of the OHIP-49; and had a good distribution of preva-
lence for individual questions. OHIP-14 scores and OHIP-49 scores displayed the
Key words: epidemiology; orai health:
same pattern of variation among sociodemographic groups of older adults. In psychometrics; quality of life
a multivariate analysis of dentate people, eight oral status and sociodemographic
variables were associated {P<0.05) with both the OHIP-49 and the OHIP-14. Gary D. Slade, Department of Dental Ecoiogy,
University of North Caroiina, CB#7450,
While it will be important to replicate these findings in other populations, the Chapei Hili, NC 27599-7450, USA
findings suggest that the OHIP-14 has good reliability, validity and precision.
Accepted for pubiication 4 August, t996

The extension of people's life span and cial impact of oral disorders on their tures?" Qtiestions in the disability di-
the enhancement of their quality of life well-being. The development, reliability mension refer to itnpacts on everyday
are two central goals of the Healthy and validity of the OHIP have been activities, such as; "Have you had to
People 2000 initiative (1). This emphasis described previously (4). The 49 ques- interrupt meals because of problems
on quality of life is consistent with the tions in the OHIP capture seven con- with your teeth, mouth or dentures?"
concept that health is a resource and ceptually formulated dimensions that while handicap represents the extent of
not simply the absence of disease. are based on Locker's theoretical mod- disadvantage caused by oral health,
Increasingly, quality-of-life assessment el of oral health (Fig. I) (5). The seven such as; "Have you been unable to en-
is being regarded as an essential compo- dimensions are: functional limitation, joy other people's company because of
nent for as.sessing outcomes of health physieal pain, psychological discom- problems with your teeth, tnouth or
care, including outcomes for public fort, physical disability, psychological dentures?" Separate subscales can be
health programs (2). Until a decade disability, social disability and handi- calculated for each dimension of the
ago, there was a virtual absence of in- cap. The hierarchy captures outcomes OHIP, while overall levels of social im-
dices to measure quality of life as it re- that have an increasingly disruptive pact have been computed using all 49
lates to oral health. However, there is itnpact on people's lives. For exatnple, questions. Epidemiological studies that
now an impressive range of instruments the functional litnitation qttestions have used the OHIP have found that
that assess the impact of oral conditions capture impacts that would be ap- missing teeth, untreated decay, perio-
on well-being and quality of life (3). parent primarily to the individual, dontal attachment loss and barriers to
The Oral Health Impact Profile such as; "Have you had difficulty dental care are associated with increas-
(OHIP) is one such instrument that chewing any foods because of prob- ing levels of impact on well-being (6-
measures people's perception of the so- lems with your teeth, tnouth or den-
Deriyalion attd yalidation of short-form OHIP 285

DISEASE had the strongest associations with the response was selected to identify ques-
long-form scales (13). In other in- tions that caused respotidents the great-
stances, such as the final revision of the est problems with interpretation or
Sickness Itnpact Profile, additional cri- completion. Theti, three statistical pro-
IMPAIRMENT teria were used, including the frequency cedures were used with the intention of
with which items were reported and ex- deriving a subset of approxitnately 10-
perience with adtninistration of the in- 15 questions that could capture as
FUNCTIONAL DISCOMFORT strutnent in research settings (14). much information as possible from the
LIMITATION &PAIN The purpose of this report is to de- 49-item OHIP questionnaire.
scribe the derivation of a short-form 1. Internal reliability attalysis used
OHIP questionnaire and to report ini- Cronbach's coefficient a to deter-
DISABILITY tial findings concerning its reliability mine whether retnoval of individual
Physical and validity. questions would increase the instru-
Psychological tnent's reliability.
Social 2. Prineipctl components factor analysis
Methods was undertaken to identify a set of
I This paper reports findings frotn a sec- underlying factors contributing to
HANDICAP ondary analysis of data collected in OHIP responses. This was followed
Eig. 1. Locker's conceptual model of oral 1991/92 during a cross-sectional study by promax rotation of the factors
health. of older adults in South Australia (6). that accounted for the greatest
Subjects in that study were a stratified amoutit of variation, and cotnputa-
randotn satnple of cotntnunity-dwelling tion of factor loadings for each ques-
While the OHIP is intended to pro- persons aged 60+ years living in Ade- tioti to idetitify atiy that exceeded
vide comprehensive data about percep- laide, the state capital, and Mt Gatn- 0.4, which was used as a threshold
tions of impact on well-being, some re- bier, a rural city in the south-east of the for moderate to high loadings. For
search settings do not pertnit use ofthe state. Some 1650 people took part in a the factor analysis, cotntnunalities
full battery of 49 questions in the in- face-to-face interview that asked about were set to the variable's squared
strutnent. For example, health services dental visits, tnedical conditions and multiple correlation with all other
researchers frequently require a more sociodemographic characteristics. The variables.
succinct instrument to assess the per- data for this report come from self- 3. Least-sc/uarcs regression was used
ceived impact of oral health as one of administered OHIP questionnaires, with the total OHIP score (obtaitied
several outcotnes of dental care. which were cotnpleted by 1217 inter- by summing the coded Likert-type
Although there is a general psychomet- viewed people, and frotn clinical oral responses from all questions) as the
ric principle that reliability of ati index examinations that were conducted dependent variable and each ques-
decreases as the nutnber of items de- among 716 interviewed people who tion was an independent variable. A
creases (9) other criteria provide a ra- were dentate. default stepwise procedure was used
tionale lor developitig short-fortn in- For each of the 49 OHIP questions, in which individual questions were
dices. STEWART et al. have described people were asked how frequently they considered sequentially for their
conceptual, statistical and pragtnatic re- had experienced the impact in the pre- contribution to total R~, atid the
quirements for health status questioti- ceding 12 months. Responses were first 14 items tnaking the largest ad-
naires (10); 1. they should represent tnade on a Likert-type scale and coded dition to R- were selected. A second
multiple health concepts and a range of 4="very often", 3 = "fairly often", 2 = procedure itnposed greater control
health states pertaining to general func- "occasionally", l="hardly ever" and over the stepwise procedure; again,
tioning atid well-being; 2. they should O="nevcr". For this report, descriptive itetns making the greatest contribu-
have good psychometric properties (re- statistics were created by computing the tion to total R- were added sequen-
liability, validity and precision); and 3. mean of the coded response for each tially, except that no tnore than two
for clinic settings, they should be simple itetn which is described below as the se- items frotn each coticeptuai dimen-
and easy to use. verity score for each itetn. In addition, sion were pertnitted to enter the
Several statistical techniques have a threshold of occasionally, fairly often model. This "controlled" regression
been used to identify short forms of or very often was used to dichototnize procedure was also condueted until
health questionnaires including ititernal responses, thereby indicating people 14 itetns, two from each ditnension,
reliability analysis, regression atialysis who had experienced at least sotne itn- were selected.
and factor analysis (11). For example, pact. This is described below as the When the final set of questions was
the General Health Questionnaire was prevalence of each itetn. identified, the vahdity of the short form
shortened frotn 60 itetns to 28 itetns The first step in deriving a short (OHIP-14) was investigated by deter-
using factor analysis (12). The Medical fortn begat! by elimitiating items that tnining whether its associations with
Outcomes Study (MOS) set of 116 core applied only to denture wearers and sociodetnographic and clinical variables
items was shortened to versions that items where 5%) or more of responses were sitnilar to the associations between
contained 20 itetns (SF-20) and 36 were left blank or marked "don't the full OHIP (OHIP-49) and those
itetns (SF-36) by selecting items that know". This percentage of non- same variables. This was used to pro-
286 SLADE

vide evidence of discriminant validity produce a standardized score for each lence frotn 11.8% (Q34) to 1.6% (Q48)
(15). Summary scores for the OHIP-14 subject. For the 14 items, new weights and severity ranged from 0.40 (Q34) to
and OHIP-49 were obtaining by count- were computed from the same data set 0.06 (Q48). This set of 14 questions was
ing the number of items reported occa- of popvtlatioti judgments about the rela- dominated by the two disability dimen-
sionally, fairly often or very often by tive unpleasantness of impacts. The sions and the handicap dimension.
each person. Differences between sub- weights represented the proportion of In the default stepwise regression
groups of each sociodemographic vari- people who judged the impact within procedure, in which the sutn of coded
able were evaluated using analysis of each ditnension as more unpleasant responses to all OHIP questions was the
variance. than the other impact in that dimen- dependent variable, the first seven ques-
Additional assessment of validity sion. Coded responses to each question tions selected (Q23, QI5, Q35, Q20, Q6,
used multivariate, least-squares regres- were multiplied by the weights and the Q9, Q2) had a total R- of 0.92,
sion to identify oral status and socio- products were added to produce seven although they did not include questions
demographic factors that were signifi- subscale scores that were standardized from the physical disability, soeial dis-
cantly related to both the OHIP-49 and (1.0±1.0) and sumtned. ability or handicap ditnensions (Table
the OHIP-14 scores for dentate per- 1). After including the next seven ques-
sons. Models were constructed separ- tions, the total R-^ increased to 0.96,
ately for each depetident variable by iti- Results and the set of 14 questions included at
corporating all clinical and sociodetno- Three of the OHIP-49 questions (Q17, least one from each dimension. The
graphic variables used previously in this Q18 and Q30) referred to denture- prevalence of the 14 questiotis ranged
study. The clinical variables were; related disorders, while seven other from 34.1% (Q12) to 2.2%, (Q43) and
number of missing teeth (anterior and questions were left blank or marked severity scores ranged frotn 0.91 (Q12)
posterior); number of unreplaced "don't know" by at least 5% of respon- to 0.10 (Q43) (Table 1). Although not
spaces (anterior atid posterior); number dents (Table 1). Consequently, those 10 shown in Table 1, Cronbach's a for this
of decayed surfaces (coronal and root); questions were excluded from the set set of 14 questions was 0.88.
number of teeth with occlusal attrition for consideration in the short fortn. For In the controlled stepwise regression
to dentin; number of retained roots; the remaining 39 questions, the preva- procedure, in which entry into the mod-
maximutn periodontal recession; maxi- lence of reported impacts (at the thresh- el was litnited to two questions frotn
mutn periodontal probing depth; and old of occasionally, fairly often or very each ditnension, the 14 itetns ranged in
maximum periodontal attachtnent loss. often) ranged from 73.7% (Q7) to 1.6% prevalence frotn 33.1%) (Q15) to 1.6%
Sociodemographic variables were; age; (Q48) and the severity scores ranged (Q48) while severity ranged frotn 0.97
sex; country of birth; period since last frotn 1.96 (Q7) to 0.06 (Q48) (Table 1). to 0.06 (Q48) (Table 1). The total R^ for
dental visit; usual reason for dental vis- Internal reliability for the 39 ques- this set of 14 questions was 0.94. Eleven
its; site of last dental visit; perceived tions was very high (a=0.94) and could of the questions were identical to the
need for dental treatment; difficulty be improved only to the third decimal questions obtained using the default
paying a AUD 100 dental fee; living place by deletion of individual ques- method of selection. Cronbach's a for
alone; educational attaintnent and tions. Furthermore, reliability was this set of 14 questions was 0.88.
household income. Variables which moderate or high among questions The set of 14 questions from the con-
were not statistically significant (F-test, within the seven conceptual dimensions trolled selection procedure was used to
P>0.05) were then eliminated from the (0.66 to 0.89) and could not be im- compute OHIP-14 scores, indicating
model until a final model of significant proved substantially by deletion of that a mean of 1.64 questions were re-
first-order tertns was obtaitied. questions. Correlations between the ported occasionally, fairly often or very
The validation process was replicated sum of coded responses to all OHIP often (Table 2). The corresponding
by computing new summary scores that questiotis atid the individual coded re- mean for the OHIP-49 score was 7.30.
took advantage ofthe full range of cod- sponses were all statistically signifieant In bivariate analysis, both sutntnary
ed responses to OHIP questions. The (P<0.0\) and Pearson's correlation co- seores varied consistently among sub-
method has beeti described previously efficients ranged from 0.27 (Q12) to groups of older adults in South Austral-
(4), and it permitted calculation of stan- 0.74 (Q36). ia. Age group and sex differences in
dardized OHIP scores. For all 49 ques- Factor atialysis revealed one princi- OHIP-14 scores were stnall and statis-
tions, coded responses (ranging frotn 0 pal cotnponent with an eigenvalue of tically tioti-significant, and the differ-
to 4) within each dimension were 15.1 that accounted for 69.2"/i of varia- ences were similarly small and non-
multiplied by previously developed tion, and an additional three principal significant using the OHIP-49. How-
weights, obtained using Thurstone's components that had eigetivalues ratig- ever, for subgroups defined by dentition
method of paired comparisons. The ing frotn 1.5 to 1.9. Thirty-two of the status, site of last dental visit, perceived
weights reflected population judgments OHIP questions had factor loadings need for dental treattnent and financial
about the relative unpleasantness of that exceeded 0.4 for the first rotated hardship, OHIP-14 scores differed by as
each impact. The sutn of the products factor. Factor loadings for the ques- mueh as a factor of two, and the differ-
within each ditnension represented sub- tions with the 14 highest loadings (0.64 ences were statistically significant
scale scores which were then standard- or more) are presented in Table 1 (col- {P<O.Ol). Those diffetences were of a
ized to a common mean and standard umn headed "Factor loading"). The sitnilar magtiitude and were statistically
deviation (1.0±1.0), and summed to high-loading questions ranged in preva- significant using the OHIP-49 (Table 2).
Dcriyation and yalidatiott of short-Jorm OHIP 287

Table 1. Prevalence, means. factor analysis and regression analysis for 49 OHIP items
Sequential R- for;
Prevalence; % reporting item Severity:
% blank/ occasionally, fairly often or item Factor Default Controlled
Conceptual dimension and item don't know very often mean loading' selection- selection-^

Eiitictiotuil litititation
Ql Difficulty chewing 1.2 32.8 1.03
Q2 Trouble pronouncing words 1.3 11.7 0.38 0.016 0.016
Q3 Noticed tooth that doesn't look right 9.5 18.2 0.60 *
18.3 0.58 * *
Q4 Appearance affected 4.6
22.1 * *
Q5 Breath stale 6.1 0.70
Q6 Taste worse 2.7 10.5 0.34 0.032 0.032
Q7 Food catching 0.7 73.7 1.96
10.8 * *
Q8 Digestion worse 5.4 0.37
29.0 0.88 * *
Q17 Dentures not fitting 3.1
Physieal poiti
Q9 Painful aching 0.5 23.8 0.74 0.024 0.024
QIO Sore jaw 1.6 16.5 0.49
4.5 * *
Qll Headaches 7.8 0.15
Ql2 Sensitive teeth 2.5 34.1 0.91 0.004
Ql3 Toothache 3.4 22.2 0.71
QI4 Painful gums 2.3 29.4 0.85 0.008
Q15 Uncomfortable to cat 0.8 33.1 0.97 0.14 0.14
Q16 Sore spots 1.1 44.6 1.18
22.2 0.71 *
Q18 Discomfort (dentures) 2.5
Psychological discotnfort
Qi9 Worried ' 2.7 24.6 0.82 0.007
Q20 Self-conscious 1.0 17.2 0.55 0.057 0.057
Q21 Miserable 1.6 14.5 0.50
Q22 Appearance 1.2 15.6 0.49
Q23 Tense 1.2 12.1 0.42 0.56 0.56
Physical disability
Q24 Speech unclear 2.8 12.7 0.39
Q25 Others misunderstood 7.8 10.7 0.33 *
Q26 Less flavor in food 3.7 9.0 0.31 0.64
Q27 Unable to brush teeth 2.1 8.6 0.29
Q28 Avoid eating 0.7 27.3 0.80
Q29 Diet unsatisfactory 1.2 4.2 0.20 0.003
Q30 Unable to eat (dentures) 1.4 11.8 0.43 *
Q31 Avoid smiling 1.6 8.6 0.29
Q32 Interrupt meals 0.8 15.1 0.48 0.004 0.004

Psychological disability
Q33 Sleep interrupted 1.2 7.7 0.29
Q34 Upset 0.7 11.8 0.40 0.69
Q35 Difticult to relax 0.7 8.1 0.31 0.75 0.087 0.087
Q36 Depressed 0.7 7.4 0.27 0.76
Q37 Concentration affected 2.1 4.2 0.19 0.78
Q38 Been embarrassed 0.7 13.2 0.44 0.005 0.006

Soeial disability
Q39 Avoid going out 0.2 3.7 0.12 0.74
Q40 Less tolerant of others 2.2 4.3 0.16 0.73
Q41 Trouble getting on with others 1.6 1.7 0.09 0.75
Q42 Irritable with others t.5 4.8 0.18 0.70 0.002
Q43 Difficulty doing jobs 1.2 2.2 0.10 0.68 0.007 0.007

Hatidleap
5.3 0.19 *
Q44 Health worsened 4.7
Q45 Financial loss 2.5 6.7 0.20
Q46 Unable to enjoy people's company 0.8 4.3 0.17 0.77
Q47 Life unsatisfying 1.2 7.6 0.26 0.71 0.011 0.011
Q48 Unable to function 0.7 1.6 0.06 0.65 0.001
Q49 Utiable to work 1.3 1.8 0.10 0.64

' 14 questions with highest factor loadings on first rotated factor.


- 14 questions selected to maximize increase in total R^.
-^ 14 questions selected to maximize increase in total R-, with no more than two items per dimension permitted to enter the model.
* Question excluded because it applies only to denture wearers or because of high non-response.
288 SLADE

Table 2. Mean nuinber of OHIP-49 and QHIP-14 items for South Australians aged 60-1- yrs

No. of items reported occasionally,


fairly often or very often
OHIP-49 OHIP-14
Group n mean (SD) mean (SD)
All persons 1217 7.30 (7.75) 1.64 (2.44)
Age 60-69 yrs 576 7.74 (7.83) 1.75 (2.53)
70-79 yrs 473 6.99 (7.66) L56 (2.37)
80-t- yrs 168 6.67 (7.68) 1.51 (2.30)
P-value 0.16 0.33
Sex Male 646 7.54 (7.87) 1,68 (2.51)
Female 571 7.02 (7.60) 1.60 (2.36)
P-valuc 0.24 0.57
Dentition Dentate 905 6.89 (7.37) 1.44 (2.28)
Edentulous 312 8.47 (8.66) 2.20 (2.77)
P-value <O.OI <0.01
Site of last dental visit Private dentist 962 6.43 (6.81) 1.37 (2.12)
Public clinic 245 10.67 (10.01) 2,69 (3.23)
P-value <0.01 <0.01
Perceived need for dental treatment Yes 325 10.82 (9.72) 2.68 (3.11)
No 850 5.93 (6.34) 1.23 (1.96)
P-value <0.01 <0.01
Finaticial hardship; difficulty paying None/hardly any 808 6.47 (6.64) L39 (2.09)
AUD 100 dental fee A littlc/a lot 390 9.12 (9.49) 2.19 (2.99)
P-value <O.OI <0.01

Clinical and sociodemographic fac- lection procedure, along with the (P=0.05), although not significant
tors that were associated with OHIP-49 weights that refiect population judg- usitig the standardized OHIP-14 {P=
and OHIP-14 scores for dentate people ments about the unpleasantness of each 0.07). Both indices yielded significant
are summarized in the multivariate re- pair of items within ditnensions. When differences (P<0.01) between other
gression analysis (Table 3). There were the weights were used to compute sub- subgroups in Table 2. For the multivari-
three clinical factors and four noti- scales and a standardized OHIP-14 ate analysis, 10 explanatory variables
clinical factors that met the criteria for score, the results were generally compa- were significant in a model using the
model building (P<0.05) when OHIP- rable to the results obtained with the standardized OHIP-49 score as the de-
49 was the dependent variable. The standardized OHIP-49. In the bivariate pendent variable, and all but one of
same factors remained signifieant when analysis, there were non-significant dif- thetn (financial hardship, P=0.12) were
the OHIP-14 was the dependent vari- ferences amotig age groups using both significant when the standardized
able, and in addition country of birth the statidardized OHIP-49 {P=O.IO) OHIP-14 score was the dependent vari-
was significant (^=0.04) in that model. and the standardized OHIP-14 (P= able.
Table 4 presents the 14-OHIP ques- 0.33). However, sex differences were
tions obtained from the controlled se- significant using standardized OHIP-49
Discussion
Of the three statistical methods that
Table 3. Multivariate least-squares regression models for 716 dentate persons aged 60-1- yrs were explored in order to derive a sub-
set of OHIP questions, internal reliabil-
Dependent variable;
no. of items reported occasionally.
ity analysis and factor analysis yielded
fairly often or very often the least satisfactory results. This can
OHIP-49 OHIP-14 be attributed to the tnoderate or high
correlation that each item had with the
Explanatory variable F P F P
overall OHIP score. Cronbach's a, used
No. of missing anterior teeth 31.1 <0.01 30.0 <O.OI in internal reliability analysis, is a sutn-
No. of missing unreplaced anterior teeth 11.8 <0.01 9.0 <0.01 mary statistic which captures the extent
No. of teeth with attrition 15.7 <0.01 II.O <0.01
No. of reported medical conditions
of agreetnent between all possible sub-
12.5 <0.01 5.7 0.01
Attended public dental clinic* 14.2 <0.01 11.2 <0.01 sets of questions, and conseqtiently
Perceived need for dental treatment* 32.8 <0.01 29.8 <0.01 high intercorrelations tneant that the
Age (yrs) 12.0 <0.01 5.3 0.02 OHIP had very high a values which
Born outside Australia 3.0 0.09 4.4 0.04 could not be substantially itnproved by
R'=0.19 R2= =0.17
deletion of items. High intercorrelation
*Dummy variables coded 0=no, l=ycs. also appeared to be the teason that only
Deriyalion and yalidatiott of short-form OHIP 289

Table 4. Questions and weights for the OHIP-14* health concepts with a range of preva-
Weiaht
lences for the short-form OHIP, the
Dimension Question
method of factor analysis was foutid to
Functional Have you had trouble protiouticing any ^vords because of problems with be unsatisfactory.
limitation your teeth, mouth or dentures? 0.51
Regression analysis resulted in selec-
Have you felt that your seti.se of taste has worsened because of problems
with your teeth, mouth or dentures? 0.49 tion of itetns that had a greater range
in prevalence atid severity compared
Physical Have you had paltiful aching in your mouth? 0.34
pain Have you found it uticottifortable to eat any foods because of problems
with the range obtained with factor
with your teeth, mouth or dentures? 0.66 analysis. With the default regression
Psychological Have you been self-cotiscious because of your teeth, mouth or dentures? 0.45 procedure, an R- value of 0.92 could be
discomfort Have you felt teii.se because of problems with your teeth, mouth or obtained with only seven variables,
dentures? 0.55 although those itetns (Table 2, R- of
Physical Has your diet been unsatisfactory because of problems with your teeth, 0.016 or greater) excluded the physical
disability mouth or dentures? 0.52 disability, social disability and handicap
Have you had to interrupt ttieols because of problems with your teeth, ditnensions. However, when 14 vari-
mouth or dentures? 0.48 ables were selected by the default tneth-
Psychological Have you found it difftatlt to relax because of problems with your teeth, od, at least one item from each concep-
disability mouth or dentures? 0.60
tual dimension was included, with a
Have you been a bit etiiborras.sed bec-dwx of problems with your teeth, small increase of R- to 0.96. By control-
mouth or dentures? 0.40
litig the process, it was possible to retaiti
Social Have you been a bit irritable with other people because of problems with
0.62 two items frotn each ditnension with
disability your teeth, mouth or dentures?
Have you had difficulty doittg your usual jobs because of problems with only a small reduction in R- to 0.94.
0.38 At first appearance, there is little to
your teeth, mouth or dentures?
Handicap Have you felt that life in general was le.ss .satisfying because of problems distinguish between the subsets of 14
0.59 items obtaitied usitig the default selec-
with your teeth, motith or dentures?
Have you been totally utiable tofittictioti because of problems wilh your tion procedure atid the controlled selec-
teeth, mouth or dentures? 0.41
tion procedure. The subsets have 11
* Responses are made on a 5-point scale, coded O = ncvcr, l=haidly ever. 2 = occasionally 3 = itetns in cotntnon, both account for
fairly often, 4=very often. Within each dimension, coded responses can be multiplied by more than 9O'Mi of variation in total
weights to yield a subscale score. OHIP scores, and both have excellent
internal reliability (a=0.88). However,
closer inspection reveals that the con-
one principal component dominated the pattern of responses indicating that trolled regressioti procedure yielded
the factor analysis. This interpretation little distitiction was tnade between di- more questions that had low prevalence
is based on the finding that there was a tnensions of oral ill-health that manifest (six questions reported by lO'/o or less
very rapid reduction in eigenvalues for as dysfunction, pain, disability or hand- of people) cotnpared with the default
the second and subsequent principal icap. While this result could be used as procedure (three questions with preva-
cotnponents, corresponding to the "el- justification for the selection of the 14 lence of 10% or less).
bow" in the scree plot of eigenvalues. highest-loading factors to represent a The prevalence of items within a bat-
This elbow represents a useful criterion unidimensional index of social itnpact, tery of questions becotnes a critical fac-
for the selection of principal compo- the factor loadings in Table 1 indicated tor when the precision of the index is
nents for subsequent rotatioti (16). that a large nutnber of conceptually itn- evaluated within different populations.
While some researchers advocate use of portant items would be eliminated, in- Precision refers to the ability of an
all principal components that have ei- cluding all of the pain, discomfort and itidex to discritnitiate betweeti sub-
genvalues greater than one (four of functional limitation itetns. In addition, groups. KESSLER & MROCZEK regard the
which were found in this analysis), that only items with relatively low preva- triad of precision, reliability and validi-
criterion was judged to be too arbitrary lences of 11.8% or less were selected ty as critical psychometric properties
in this itistance. with this method. These litnitations of for health status tneasures and they ar-
ATCHISON & DoLAN have reported a factor analysis as a tool for developing gue that it is itnportant to retain less
similar phenotnenon with the Geriatric indices of health status have been ob- frequently reported items in order to
Oral Health Assessment Index, where served by others. KAPLAN ct al. found diseritninate between subgroups frotn
only one principal component emerged that factors and items that contribute populations with high levels of disease
frotn their analysis of the 12 questions little to explaining variance in occur- or disability (for exatnple, some clinical
in that index (17). The results with the rence or frequency are considered unitn- populatiotis) (11). The short Ibrtns of
OHIP are consistent with that finding, portant in factor analysis (15). How- the MOS instrument illustrate this prin-
and indicate that there is one single ever, they reject this substitution of ciple; while the SF-20 had good reliabil-
construct underlying the responses to "variation in frequency for variation in ity and validity, it failed to show further
OHIP questions tnade by these older social itnportance" when developing decrements in a satnple of hospital pa-
adults. The construct could be interpre- health status indicators. Since it was tients who scored at the lowest levels of
ted to represent "oral ill-health", with deetned itnportant to capture multiple health in sotne subscales (13). This led
290 SLADE

to the development of the SF-36 which satne conclusion was reached - the 14
References
included questions about more severe questions were effective in detecting the
levels of disability. same associations with clinical and 1. US Dept. of Health and Human Ser-
vices. Healthy people 2000. Natiotial
Precision has also been cited as an sociodemographie factors that were ob- health protitotioti atid disease prevetitioti
important requirement for oral health served using the 49 questions. objectives. Boston; Jones and Bartlett,
indices, since different subgroups may An important caveat in this study re- 1992.
have vastly different oral conditions, lates to the use of a single source of 2. Hennessy CH, Moriarty DG, Zack MM,
and therefore experience impacts on Scherr PA, Brackbill R. Measuring
data for both derivation of a subset and health-related quality of life for public
qualitatively different aspects of their validation of that subset of questions. health stirveillancc. Publie Health Rep
lives (18). In the current study, preva- Consequently, it will be important to in- 1994; 109: 665-72.
lence and severity scores have been de- vestigate the reliability and validity of 3. Gift HC, Atchison KA. Oral health,
rived from a sample of independently the OHIP-14 in other populations. health, and health-related quality of life.
Med Care 1995; 33: NS57-77.
living older adults who generally are Furthermore, not all research settings 4. Slade GD, Spencer AJ. Development
free of severe oral disease. However, will be concerned with cross-sectional and evaluation of the oral health impact
questions with low prevalence in this associations, and therefore it will be ne- profile. Cottitnutiity Detit Health 1994;
population would be reported more cessary to evaluate the perfortnance of //.3-I1.
frequently in subgroups of people with 5. Locker D. Measuring oral health; a con-
the OHIP-14 using other research de- ceptual framework. Cotntiiutiity Detit
more severe oral disease, and therefore signs, such as longitudinal studies and Health 1988; 5: 5-13.
those questions diseritninate tnore ef- experimental trials. Nonetheless, these 6. Slade GD, Spencer AJ. Social impact of
fectively within those subgroups. For cross-sectional findings provide encour- oral disease among older aldults. Aust
example, when the OHIP-49 was ad- aging results, suggesting that the OHIP- Detit J 1994; 39: 358-64.
ministered to a sample of dental pa- 7. Locker D, Slade G. Association between
14 has good statistical properties and clinical and subjective indicators of oral
tients aged 21-49 years with HIV in- validity, with the obvious benefit that health in an older adult population. Ger-
fection, their levels of social impact the data could be collected with less odcmtology 1994; //.• 108-14.
were more than twice as high as the fieldwork effort and respondent burden. 8. Hunt RJ, Slade GD, Strauss R. Racial
levels reported by general dental pa- variations in social impact among older
In conclusion, a controlled regression community-dwelling adults. J Pttblic
tients (19). procedure permitted identificatioti of a Health Dent 1995; 55: 205-9.
The findings frotn the OHIP-14 indi- subset of 14 questions about the social 9. Nunnally JC. Psyehotttetrie theory. New
cate that the items obtained from the York; McGraw-Hill, 1967; 192-3'
itnpact of oral disease that accounted
10. Stewart AL, Hays RD, Ware JD. The
controlled selection method were as ef- for 94% of variation in total OHIP MOS short-form general health survey
fective as the OHIP-49 items in de- scores and which had an internal reli- Med Care 1988; 26: llA-'il.
tecting differences among subgroups of ability coefficient (a) of 0.88. Sutnmary 11. Kessler RC, Mroczek DK. Measuring
older South Australians. In some re- scores based on the OHIP-14 displayed the effects of medical interventions. Med
Cate 1995; 33: AS 109-19.
spects, this should not be surprising, the satne pattern of variation among 12. Goldberg DP Hillicr VF A scaled ver-
since the 14 items were selected speci- sociodemographic groups that was ob- sion of the general health questionnaire.
fically because they accounted for most served using the OHIP-49, and both the Psyehol Med 1979; 9: 139-45.
of the variance in the summed OHIP OHIP-14 and the OHIP-49 resulted in 13. Hays RD, Shcrbournc CD, Mazel RM.
responses. However, both the OHIP-14 sitnilar tnultivariate models relating The RAND 36-item health survey I.O.
Health Eeon 1993; 2: 217-27.
and OHIP-49 (Tables 2 and 3) were oral status and sociodetnographic vari- 14. Bcrgner M, Bobhitt RA, Carter WB,
computed by counting the number of ables to social impact. The OHIP-14 Gilson BS. The sickness impact profile;
items reported oecasionally, fairly often contains questions that retain the origi- development and final revision of a
or very often, and therefore they rely on nal conceptual ditnensions contained in health status measure. Med Care 1981;
19: 787-805.
a threshold of reported itnpact. While the OHIP, and those questions have a 15. Kaplan RM, Bush JW, Berry CC.
this measure is simple to compute and good distribution of prevalences, sug- Health status; types of validity and the
interpret, it fails to take advantage of gesting the instrument should be useful index of well-being. Health Serv Res
the full range of responses to each ques- for quantifying levels of impact on well- 1976; //.• 478-507.
tion. Furthertnore, the use of this being in settings where only a limited 16. Cattell RB. 77;i;' .scieittifie use of faetor
analysis in behavioral and life sciences.
threshold probably contributes to "false number of questions can be adminis- New York; Plenum Press, 1978.
positive" reports (for example, an im- tered. 17. Atchison KA, Dolan TA. Development
pact that is rare, but which is incorrect- of the geriatric oral health assessment
ly reported at the "occasional thresh- index. / Dent Edite 1990; 54: 680-7.
old"). For these reasons, standardized 18. Locker D, Miller Y. Evaluation of sub-
jective oral health status indicators. /
scores are a preferred method for exam- Acknowledgtnents - The author acknow-
Public Health Dent 1994; 54: 167-76.
ining associations betweeti explanatory ledges with gratitude the comments provided
19. Coatcs E, Slade GD, Goss AN, Gorkic
by Dr Ronald Hunt and Dr Dan Shugars
factors and reported impact. Standard- during the preparation of this manuscript.
E. Oral conditions and their social im-
ized scores also utilize the weights re- pact among HIV dental patients. Aust
Original data for this analysis were collected
Detit J 1996; 41: 33-6.
fiecting lay judgments about the severi- with financial assistance from the Australian
ty of impacts. Nonetheless, when stan- National Health and Medical Research
Council and the US National Institute of
dardized scores were computed the Dental Research (Grant No. R0I-DE09588).

Vous aimerez peut-être aussi