On the basis of the initial multitrait scaling analyses, we identified three items (GOHAI3, GOHAI5, GOHAI8)On the basis of the initial multitrait scaling analyses, we identified three items (GOHAI3, GOHAI5, GOHAI8) that correlated poorly with all of the domains we originally hypothesized, so we eliminated these items. We created a separate denture subscale, recognizing that denture functioning represents a conceptual dimension separate from that of natural teeth; this also further improved scaling properties. Results indicated that the psychosocial and opportunity items covered four dimensions: 1) distress, 2) self-consciousness and worry, 3) role function, and 4) opportunity. Because most of the items in the latter construct loaded more strongly on other scales and because of skepticism about the usefulness of opportunity as an oral health construct, we deleted these three items (OHIP29, OHIP45, OHIP47). Thus, we were left with three remaining psychosocial constructs: distress, self-consciousness and worry, and role function. Additional analyses found that some items had poor loadings on the hypothesized dimensions. Accordingly, we moved the perception items (OHIP44, OHIP3, GOHAI7) from the perception dimension into the worry dimension of psychosocial items, where they had higher loadings.
We then examined the impairment items, using exploratory factor analysis, because of concerns about the multidimensionality of this domain. Indeed, we found four subdimensions: 1) mouth pain, 2) flavor, digestion, and breath, 3) tooth pain, and 4) denture discomfort. On the basis of these results, we retained all of these items but further altered our conceptual model to include five dimensions: 1) physical function, 2) impairment and disease, and three dimensions of psychosocial function: 3) role function, 4) distress, and 5) worry (Appendix 1
Using the remaining items, we standardized the item scores so that the mean of each variable was 50 and the standard deviation, 10. We scored the scales by taking the mean of all the items, after reversing the response categories where necessary so that higher scores indicated poorer oral quality of life. Thus, we created five scales to correspond with the above dimensions, a separate scale of the three denture-related items, and a summary scale comprising all items.
To develop a shorter version of the measure, we used data from the two veteran samples analyzed together to conduct forward stepwise regressions on each scale. This process allowed us to determine which items explained the most variance in each scale score. We selected items that explained either 80% of the variance or the first five items, whichever was greater. This resulted in five scales, each with five items. All of the scales had excellent internal consistency reliability, ranging from .78 (impairment) to .92 (distress), with the other scales also having excellent reliability ().
Table 1 Correlations of Scales (5 Items in Each) With Clinical Variables, Scale Internal Consistency Reliability, and Variance Explained by Each Scale, Among Participants in the Veterans Health Study and the Dental Longitudinal Study, 1993–1995 (N = 827) (more ...)
Next, we examined the correlations of each scale with clinical indices (). The strongest correlation observed was between physical function and number of teeth (r = −0.38). Coronal caries was moderately associated with worry (r = 0.23) and impairment (r = 0.18), whereas periodontal status was moderately associated with physical function (r = 0.21) and worry (r = 0.21). Root caries had the smallest correlations overall with OQOL.
We also examined mean scores on each of the quality of life dimensions by scores on the CPITN and found that individuals with greater treatment need had significantly worse OQOL ().
Meana Oral Quality of Life Scores by Varying Levels of Periodontal Disease Among Participants in the Veterans Health Study and the Dental Longitudinal Study, 1993–1995 (N = 827)
We then examined the proportion of variance explained in each oral quality of life dimension among different subgroups based on number of teeth (not shown). The impairment, physical function, worry, and role function scales explained the least variance among patients with no teeth and the most among patients having 1 to 10 teeth. The patterns observed for the distress scale were different: the most variance was explained among those with either no teeth or 1 to 10 teeth, with the least (but still 97% of the variance explained) among those with 11 to 24 teeth.
Next, we administered these five scales and the three denture-specific items to the sample of community dental patients. Using multitrait analysis, we sought to reduce the number of items further by eliminating items contributing least to each scale's internal consistency reliability and retaining items that conceptually best represented the spirit of the subscale. We eliminated items whose deletion least affected the internal consistency reliability of the scales (Cronbach α), and at the same time, sought to retain the items that we considered, from a conceptual standpoint, best represented the spirit of the subscale. We did this on two levels. First, we developed one 12-item measure (Appendix 2
) that includes 3-item subscales for each of 3 scales in the psychosocial dimension (distress, worry, and role) and single items assessing dimensions entitled physical, denture,
(Cronbach α of the 12 items = .90). We also developed a second, briefer 6-item measure that includes single items assessing each dimension (distress, worry, role, physical, denture, pain) (Cronbach α for the scale = .80).
We then took these two brief measures, refined on the community dental patient sample, and returned to our original data set of 827 veterans to examine the association of the two brief scales with clinical indices. Both summary scales were significantly correlated overall with number of teeth (r = −0.35 and −0.23, for the 6- and 12-item scales, respectively), coronal decay (r = 0.09 and 0.14), periodontal status (r = 0.19 and 0.20), and root caries (r = 0.14 and 0.12) (). Most items were significantly correlated with number of teeth, coronal decay, and periodontal status, but fewer were significantly correlated with root caries. Most items were associated with periodontal treatment need ().
Correlationsa of Items From the Two New Brief Scales and Overall Summary Scales With Clinical Variables Among Participants in the Veterans Health Study and the Dental Longitudinal Study, 1993–1995 (N = 827)
Mean Oral Quality of Life Scoresa by Varying Levels of Periodontal Treatment Need Among Participants in the Veterans Health Study and the Dental Longitudinal Study, 1993–1995 (N = 827)