Dimensions of oral health related quality of life measured by EQ-5D+ and OHIP-14

Design

The Burden of oral disease study was conducted using a 2-stage sampling design whereby dentists were randomly sampled from the South Australian Dental Register, randomised into one of seven equal-sized study groups (n = 100) and sent a mailed self-complete dentist questionnaire along with up to five self-complete patient questionnaires depending on the study group. The dentist questionnaire collected data on dentist and practice details and patient oral health details. A pilot study was conducted which collected five patients per dentist in order to establish the feasibility of the 2-stage methodology. Since the optimum number of patients to sample from dentists was not known, dentists in the main study were randomised into six groups in order to assess the sample size-related efficiency and response properties of recording data on from 1 to 5 patients and distributing between 0 to 5 patient questionnaires. Note that dentists in the group that had no patient questionnaires to distribute recorded details of 5 patients in their dentist questionnaire, while dentists in all other groups recorded the same number of patients in their dentists questionnaire as they distributed patient questionnaires. Within the questionnaire dentists were provided with a practitioner logbook in which to record for the first 1 to 5 adult patients (depending on study group assignment of dentist) of a random clinical day the diagnosis of the oral disease or condition treated and the treatment they performed. At the conclusion of treatment each practitioner (except dentists in the study group that had no patient questionnaires to distribute) passed on a survey kit to their sampled patient(s) containing a cover letter and explanation sheet, and a patient questionnaire. Sampled patients completing the patient questionnaire recorded basic socio-demographic characteristics and data concerning the nature, severity and duration of their symptoms. The patient questionnaires were identified using the practitioner identification number allowing linkage between the practitioner logbook data and patient questionnaire data, but maintaining the anonymity of each patient to the investigators. While the primary rationale for this 2-stage methodology was to allow linkage of dentist-assessed oral health status to patient perceptions of quality of life this paper reports solely on the patient perception data. The research project was reviewed and approved by the Human Research Ethics Committee of the University of Adelaide.

Sampling and data collection

Data were collected during 2001–2 with a primary approach letter sent initially to each dentist, followed a week later by the survey materials, with a reminder card two weeks later, and up to four follow-up mailings of survey materials to dentists who had not yet responded in order to ensure higher response rates [13].

Data items

Dentists recorded the details of the dental conditions that patients had, and patients recorded their experience of those dental conditions. Diagnosis of dental conditions was collected from dentists using an open-ended question in the dentist questionnaire. In the patient questionnaire, patients were asked if the dental conditions had caused problems in each of six health state dimensions. The six health state dimensions were: mobility (e.g, walking about), self-care (e.g, washing, dressing), usual activities (e.g., work, study, housework, family or leisure), pain/discomfort, anxiety/depression and cognition (e.g, memory, concentration, coherence, IQ). They were measured using the European Quality of Life indicator or EuroQol (EQ-5D+) instrument [6]. The EuroQol measures each of these six dimensions according to a 3-level response grading from 1 = no problems, 2 = some / moderate problems and 3 = extreme problems. Patients were also asked to rate their experience of dental problems in the last year using the OHIP-14 [5], which uses 14 items to capture measures of the seven dimensions of functional limitation, physical pain, psychological discomfort, physical disability, psychological disability, social disability and handicap. For each of the 14 OHIP questions subjects were asked how frequently they had experienced impact in the preceding 12 months using a 5-point scale coded 4 = very often, 3 = fairly often, 2 = occasionally, 1 = hardly ever and 0 = never.

Data analysis

The characteristics of responding patients were compared descriptively with published data on dental patients and the Australian population. The distributions of responses to the EQ-5D+ and OHIP items were examined, and the items were analysed by factor analysis and cluster analysis. Factor analysis was used to examine the battery of quality-of-life items for underlying component factors. Standard errors and confidence intervals were reported adjusted for the effect of clustering within the primary sampling unit of dentist.

Principal components factor analyses were performed using varimax rotation [14]. A range of n-factor solutions were performed and assessed. While selecting the number of factors involves the reasonableness of the solution and knowledge of the subject matter [15, 16], retaining factors with eigenvalues greater than 1.0 is commonly used, based on heuristic and practical grounds. Sampling adequacy, or the degree that the subset of variables used represents a potentially larger domain, was assessed by Kaiser's measure of sampling adequacy [15]. Communality measures the common factor variance of a variable. A communality of 0.3 or less indicates that a variable may be unreliable [17], while a value greater than 0.3 indicates a large percentage of the sample variance of each variable is accounted for by the factors [16].

Hierarchical clustering of variables was performed using SAS PROC VARCLUS [14], an oblique multiple-group component analysis. Associated with each cluster is a linear combination of the variables in the cluster, the first principal component. The first principal component is a weighted average of the variables that explains as much variance as possible. Clusters are chosen to maximise the variation accounted for by the first principal component of each cluster. SAS PROC TREE was used to construct a dendrogram to present the results of the hierarchical clustering as a tree structure [14].

Response

Characteristics of patients

Distribution of items

Factor structure

Cluster structure

Figure 1 shows that the first branch in the dendrogram split the EuroQol items mobility (EQ1), self-care (EQ2) and usual activities (EQ3) from the OHIP items and the remaining EuroQol items. Moving from the bottom towards the top of the dendrogram the next major branch split the OHIP psychological discomfort subscale items (OH5 and OH6) together with one psychological disability subscale item (OH10) and the functional limitation subscale items (OH1 and OH2). At the next major branch the EuroQol anxiety/depression (EQ5) and cognition problems (EQ6) items were split from the remaining items. At the next major branch the EuroQol pain/discomfort item (EQ4) split together with the OHIP physical pain subscale items (OH3 and OH4). At the next major branch the OHIP item OH14 (unable to function) split from the remaining items. At the next major branch the OHIP physical disability items (OH7 and OH8) split from the OHIP social disability items (OH11 and OH12) together with the psychological disability item, OH9 (difficulty relaxing).

OHIP items tended to cluster together in subscales, either in the same branch or as adjacent items in a sub-branch, with the main exception being the psychological disability subscale being split between social disability and psychological discomfort (highlighted by the dotted line connecting OH9 at the top of the dendrogram to OH10 in the lower portion of the dendrogram). Handicap subscale items were not widely separated in the branching structure (on either side of OH7 and OH8, indicated by the dotted line connecting OH13 to OH14 in the upper portion of the dendrogram) and were in the same major branch as the disability subscale items. The dotted line connecting OH5 and OH6 indicates that these OHIP psychological discomfort subscale items were separated on either side of the OH10 item.

Response

Response rates to the survey were adequate for both the dentist and patient questionnaires [21]. Comparison of respondents against estimates for private general practice and the Australian population indicated a slightly higher percentage of female patients compared to the population consistent with higher reported visiting rates by females [20], but both place of birth and time since last visit were similar. While dental insurance was higher, the percentage of check-up visits was lower among patients indicating a higher percentage of dental problems for patients compared to the population. The respondents tend to be representative in terms of demographics and dental visiting but have higher levels of dental problems. The method of sampling patients showed that response rates tended to be higher among dentists who had to sample fewer patients consistent with a lower response burden, but selection of an optimal collection methodology requires consideration of efficiency of collection as well as response rates.

Factor structure

While EuroQol items tended to load on separate factors to OHIP items, there was some overlap in dimensions for the EuroQol pain/discomfort item and the OHIP physical pain subscale. In general, OHIP items loaded highly on each of the first two factors. Since it is often recommended to sample 4 or more variables to identify each factor [22], it might be argued that using the full OHIP-49 rather than the shorter OHIP-14 would provide a better factor analytic solution, however previous analysis of a wider range of OHIP items drawn from the full OHIP-49 battery [2] showed that one principal component accounted for a large percentage of variance, with another three principal components having eigenvalues greater than one [5]. High intercorrelation of items was suggested as the reason that one principal component dominated the factor analysis [5], and interpreted as indicating that one single underlying construct could be interpreted to represent "oral ill-health".

Some technical aspects of factor analysis might be raised in relation to why the OHIP items were not separated. While interval level of measurement is implied by the use of correlation or covariance matrices as the basic input into factor analysis many variables such as measures of attitudes and opinions that do not have a clearly established metric base may be given numeric values without distorting the underlying properties [15]. Additionally, the assumption that variables have a multivariate normal distribution is only required in the case of maximum likelihood (ML) solutions. However, as a descriptive method ML factor analysis does not require a multivariate normal distribution, although the validity of Bartlett's test for the number of factors does require such an assumption [14]. The factor analysis model itself does not require such an assumption [15], with the solution presented here being based on a principal components factor analysis. Another consideration is that factor analysis may not be preferred as a scale validation method for instruments that contain causal indicators such as disease-related groups of symptoms, as these factors are likely to vary between different patient subgroups [23]. While this phenomenon may have particular relevance to quality of life measurement in terms of being able to measure both good and poor quality of life rather than just poor quality of life [24], factors in general should be interpreted with caution. Reification of factors may occur, and researchers are cautioned against attributing reality and uniqueness to factors (i.e., giving a factor a name does not give it reality). However, factors that recur from different samples and conditions point to an underlying variable [25].

Cluster structure

The purpose of cluster analysis is to place objects (i.e., variables) into groups (i.e., clusters) suggested by the data, not defined a priori, so that the variables in a given cluster tend to be similar to each other. Factor analysis is often used to classify variables into groups (i.e., factors) but the resultant groups are overlapping. While factor analysis may be preferred when overlapping groups are desired, cluster analysis can be used for both hierarchical and disjoint clustering of variables [14]. The results of the cluster analysis showed that there was some separation of the generic items from the specific items, with the EuroQol mobility, self-care and usual activities items branching off from the root of the tree structure. However, the EuroQol items anxiety/depression and cognition formed a cluster adjacent to the EuroQol pain item that clustered with the OHIP pain subscale items, and the OHIP functional limitation subscale items indicating that the separation was partial and there was a degree of overlap in the domains measured by both instruments.

Most of the items from the conceptually derived OHIP subscales formed clusters providing empirical support for those conceptual subscales. The positioning of the OHIP handicap and disability items in adjacent clusters of a major sub-branch of the tree structure fits conceptual expectations, as does the positioning of the limitation and discomfort items in adjacent clusters of another major sub-branch of the tree structure. The main departure from expected groupings was that one item from the OHIP psychological disability subscale clustered with items from the OHIP psychological discomfort subscale instead. Replication would be required to determine whether this represents incorrect assignment of the item to the conceptual subscale or the response preferences of this particular study group. The nature and magnitude of impacts may vary according to the cultural background of the population measured, and there has been some questioning of whether existing oral health status measures are sufficiently robust in their conceptual underpinning [26].