Theoretical framework

The validation process was conceptually grounded in the framework exposed by Locker in 1988, which shows different effects on Quality of Life based on changes arising in the oral cavity. This model has been used to elaborate several instruments of oral Quality of Life, as well as previous validations of the GOHAI scale ^13-15.

Participants SABE Colombia: survey on health, well-being, and aging in Colombia-Study

The study was performed on the subjects participating in the “Encuesta Salud, Bienestar y Envejecimiento” (Survey on Health, Well-Being, and Aging in Colombia) - SABE Colombia 2015, a national survey which aims to gather information about the aging process among 60 years old and older Colombians. The survey had a field collection time of one year, between 2015-2016. The SABE survey, as well as the GOHAI scale validation study, were endorsed by the Universidad del Valle’s Ethics Committee (Cod. 083/014 and 093/015), and all participants signed an informed consent form. The survey respondents were 23,694 subjects, among them, 4,689 were excluded due to cognitive impairment identified by the Minimental Test. In total, 19,005 elderly people responded to the GOHAI scale. A pre-test of the GOHAI was tested among 63 participants ¹⁶.

Forward- backward translation procedure

The GOHAI scale is a self-reporting instrument made up by 3 dimensions that assess the physical function, the psychosocial function, pain and discomfort. The instrument consists of 12 questions replicated in a Likert scale that confers each answer a score ranging from 1 to 5; the options used by the scale being Always, Often, Sometimes, Seldom and Never. The total score corresponds to the whole sum of each question and deems the oral health as adequate when the score ranks between 57 and 60, moderate between 51 to 56 and low below or equal to 50. The questions corresponding to numerals 1, 2, 4, 6, 8, 9, 10, 11 and 12 exhibit Likert categories from 1 to 5, the others 3, 5 and 7 exhibit the Likert scale inversely ¹⁷.

The translation process was carried out by two translators external to the whole research process, who met the following criteria: competent in the study languages (English and Spanish), being acquainted or immersed in the culture where the validated scale would be applied and having basic training on health measuring; training understood as having some kind of prior experience on translating instruments or health-issued documents. The objective was to check for changes in the phrasing, semantic and idiomatic equivalency. Upon receiving the translation, an expert committee was formed by a professional on dentistry, doctor on public health, oral rehabilitator, epidemiologist, Master on Gerontology, Master on epidemiology. This committee performed some joint modifications to the initial version, and an adjusted version of the GOHAI instrument was obtained according to experts.

The same translators provided a positive report in the face of the new version, which was the one tested at the pilot test and employed in the fullness of the SABE survey back translated the instrument again.

Pilot test

It was carried out at three municipalities: Bogotá (Census code: 11001), Ubaté (Census code: 25843) and Soledad (Census code: 8758) to 63 elderly persons, these three were chosen because they are culturally diverse in the country and would help to understand the aspects to be dealt with in different regions. In addition, two are small towns and one the capital city, an important aspect when testing the instrument. 3-5 blocks were selected in each municipality and all the elderly who lived in the blocks were interviewed.

The results showed that the questions were understood by interviewees and interviewers alike. Cronbach’s Alpha was 0.74, reflecting an acceptable internal consistency, on account of which the study was decided to be continued with that GOHAI version. It was decided to not perform factorial analysis at this stage due to an insufficient sample size.

Sample

In studies validating a scale, the process that requires a larger sample is the factorial analysis. Some authors define their sample’s size based on the number of items in the scale to be validated; considering between 10 and 20 subjects per item as an appropriate alternative. Other authors recommend a sample size over 500 subjects as good and over 1,000 ones as excellent ^18,19. Thus, the sample for the GOHAI study would have 240 people as it has 12 questions. However, this study took thrice that size because the answers per item showed an asymmetrical distribution, thus considering a total of 720 elderly subjects. Finally, the sample was quintupled in order to enable comparisons among some sub-groups of subjects, as such a sample of 3,600 subjects was planned for both the working sample (sample WS) and the confirmatory sample (sample CS).

Both samples (WS and CS) were obtained randomly from the 18,863 subjects who responded the GOHAI scale, after eliminating duplicates, atypical data and GOHAI scores below 12 or above 60. Then, two samples (WS and CS), about 3,600 subjects each, were selected using proportional sampling fractions according to the gender, age-groups, and dwelling area (urban vs rural) variables among the 18.863 subjects selected from the whole survey. All the analyses were performed in both samples, in this manner at the end of the study, it counted with a sample of 7,200 subjects, divided into two groups of approximately 3,600 subjects each one.

The Figure 1 describes the sampling and validation processes.

Figure 1 Sampling and validation processes.  

Variables

The following existing variables in the SABE Colombia Survey were used: age (60-64, 65-69, 70-74, 75-79 and 80 and above), gender (male-female), educational level (8 sub-groups), dwelling area (rural-urban), socioeconomic level (1-2, 3-4 and 5-6), healthcare affiliation (5 sub-groups), self-perception about dental prostheses need (yes/no), overall oral health self-perception (three sub-groups) and skin color (three sub-groups), with the latter inquired using the pallet of colors from the PEARL in Latin-America project, which displays people’s face pigmentation as a proxy of ethnicity identification ²⁰.

Appearance and content validity

They were appraised through expert analysis by asking themselves if the GOHAI scale truly measures Quality of Life regarding Oral Health, and whether the contents integrate the constructs that would be affected upon the appearance of a favorable or unfavorable oral health condition.

Statistical analysis

All analyses were carried out on working (WS) and confirmatory (CS) samples. After sampling based on frequency weights, according to the socio-demographic variables already described, work sample consisted of 3.628 registries and confirmatory sample of 3,572; response frequency for both samples and the GOHAI’s translated version are shown in the supplementary Table. Kruskal-Wallis’ and Mann-Whitney’s tests were used to determine if significant differences existed between the samples in relation to the study variables.

Reliability

Internal consistency was assessed through Cronbach’s Alpha in order to measure homogeneity among items. Using the same interviewees, it was carried out another reliability aspect defined as stability measurement over time, by replicating the GOHAI scale on two different opportunities. 75 subjects were chosen to this part of the validation. The first application was done in the first visit to the elderly whit the application of the SABE survey, the second between 5 and 7 days after the first application. For the test re-test analysis Kendall’s correlation coefficient was used. The coefficient’s interpretation ranges between -1 and +1 pointing to negative or positive associations respectively, while zero (0) means no correlation.

Validity

With the purpose of determining the construct’s validity, and stablishing dimensions of variables to be identified, exploratory factorial analysis was used by means of oblique Promax rotation. Factor analysis assumptions were assessed by Bartlett's test for sphericity and the Kaiser-Meyer-Olkin’s measure of sampling adequacy ^21,22.

Upon being a self-reported scale on a personal construct, which has no gold standard, the GOHAI scale assesses its criteria validity through discriminant and convergence analysis. The discriminant aspect sheds lights on the scale’s ability to differentiate between groups where it must do so, and its inability to differentiate the groups where it must not; this was analyzed though the relationship between the GOHAI total score and socio-demographic variables such as area, age, skin color, educational level, socioeconomic level, healthcare affiliation, self-perception of oral prostheses need and gender. The relationship with the oral health self-perception assessment was used for convergent validity, which inquires how the target construct measured by the studied scale converges to or relates with other scales assessing similar constructs. Kruskal-Wallis’ or Mann-Whitney’s tests were utilized to stablish differences between medium-sized groups depending on the dichotomy or categorical nature of the variables; these tests have been widely used throughout GOHAI validation literature ^2,23.

Construct’s validity

Construct analysis for both WS and CS yielded a Kaiser-Meyer-Olkin’s measure above 0.85, which is considered as remarkable, and a significant Bartlett's test (p <0.05); both indicating appropriate conditions in order to performing factorial analyses. The factorial structure suggested in both samples was two-factored, considering eigenvalues above 1.0. It is worth clarifying that for both WS and CF a third factor emerged very closely to an eigenvalue of 1.0: 0.96 for WS and 0.98 for CS. Thus, Promax-type rotations were performed for two- and three-factorial structures but the factorial loadings on the GOHAI items were different between both samples. In contrast, the one-factorial structure was consistent over both samples (WS and CS).

Discriminant validity

Discriminant validity for WS and CS yielded significant differences between the means of variables age, skin color, educational level, socioeconomic level, healthcare affiliation, and self-perception of prostheses need (p <0.05). Discriminant differences were not significant for gender (p >0.05); for dwelling area was not significant in WS (p >0.05), but it was significant in the CS (p <0.05).

Convergent validity

Convergent validity showed significant results in both WS and CS, between the overall oral health self-perception scale and the total GOHAI score. Table 1 describes results from discriminant and convergent validity.

Internal consistency

Cronbach’s Alpha was 0.80 for both WS and CS, thus demonstrating high correlation and homogeneity among the GOHAI items. The item-scale correlations ranked between 0.49-0.70 for WS and 0.46-0.72 for CS, with an exception for questions 3 and 4 where both samples correlation ranked between 0.35 and 0.4.

Test-retest reliability

The results showed a sound correlation between both applications over time; thus, the Kendal tau-B coefficient indicated a significant correlation of 0.85. The scale in both X and Y axis is given in points of the GOHAI score. Figure 2 shows the correlation.

Figure 2 Test- re-test scores scatter graph of the Colombian GOHAI scale. 

Validity of the construct

The construct’s validity based on eigenvalues suggested a two-factorial structure; however, Quality of Life in terms of oral health is shown as a single factor structure (i.e. a dimension). In the original version presenting the GOHAI scale ¹⁰, only a factor emerged from the factorial analysis and no sub-scales were shown, whereby GOHAI was defined to correspond to a single factor or dimension. The most important difference described through developing this part of the study was the rotation procedure used during the factorial analysis with two and three factors. All previous GOHAI validations used an orthogonal rotation (i.e. Varimax) assuming the phenomena under study to be inwardly independent (non-correlated). Conversely, for the GOHAI Colombian version oblique rotation (i.e. Promax) was used, which allows phenomena, dimensions or factors to be inwardly correlated. The correlation in oral health events is important because we can not separate conditions such as pain or dissatisfaction even more knowing that everything will affect several aspects of the quality of life ²⁴. This decision was made taking into account the Oral Quality of Life index’s result, an aspect which depends on other conditions inherent to each person and supported by the theoretical model on which the research was carried out ¹⁵.

Regarding the number of factors reported by other studies validating the GOHAI scale, the Spanish version ²⁵ found three factors with a small-sized sample of about 100 participants; these results points that the factorial structure of the original study was not replicated ¹⁰. Likewise, the Greek validation ²⁶ also found three factors (or dimensions), using a similar sample size, with three items explained by two factors simultaneously; thus, the authors recommend that structure to be corroborated with a larger sample size. On the contrary, the Mexican validation ¹³ within a much larger sample (n= 695) was able to conclude that the factorial structure corresponded to a single factor, by using the same version of the GOHAI questionnaire as the Spanish validation study. The Portuguese validation ²⁷ indicates a factorial structure similar to that found by the Mexican study. The Swiss validation ²⁸ suggests a three-factorial structure; nonetheless, after analyzing and verifying the structure the authors concluded that the best way to employ GOHAI is an uni-factorial structure, given the conceptual complexity of assessing oral Quality of Life. The explanation for the factorial structure used by the Swiss study is because the initial and original GOHAI scale was created with the belief that oral health was just a construct which takes dimensions into account within its structure, but no sub-scales ¹⁰.

In the current validation study, the structures displayed for scenarios of two and three factors are not consistent with the theoretical structure which underlies the GOHAI scale; verifying that participants do not discriminate between physical, pain-discomfort and psychosocial conceptual constructs, which could be explained by the interrelation of oral Quality of Life’s impact on each of these conceptual dimensions.

The findings above would corroborate the lack of replicability of factorial structure’s of two and three factors between both samples (WS and CS), and supports the choice of a uni-factorial structure for the Colombian GOHAI, hence matching the original scale on a single dimension where Quality of Life within Oral Health is considered a single conceptual construct ¹⁰.

Discriminant and convergent validity

The Colombian GOHAI scale allows for discriminating among subjects of different characteristics in terms of age groups, skin color, educational level, socioeconomic level, healthcare affiliation and self-perception of oral prostheses need, which are variables conceptually related with the quality of oral health. In contrast, no discriminant results were found for gender (in both WS and CS samples) neither for dwelling areas in the WS sample.

The discriminant validity results of the Colombian GOHAI scale agree with previous findings in the literature. Differences in GOHAI scores according to age groups have been found in the validation of the French version ³⁰, opposite to the Spanish, Greek, Swiss and Chinese versions ^25,26,28,29, which showed no differences in relation to age; however, this finding was corroborated in both working and confirmatory samples in this study.

Educational level differences were assessed by studies on the Mexican, American and French versions ^13,19,30) and showed significant differences, such as the current study has found. In contrast, validation studies of the Greek and Chinese versions did not find educational level differences in GOHAI scores ^26,29. Socioeconomic level and healthcare affiliation are variables assessed differently among countries, which are not properly specified in other validations or not inquired about in most of them.

Furthermore, in the current study, the “colors pallet”, for the assessment of the skin color, was used in discriminant validity analyses. This variable has not been used in other validations. In Colombian settings, skin color is related with discrimination issues and the socioeconomic status ²⁰, which could explain the differences of GOHAI scores among different ethnic groups.

The lack of discriminant validity of GOHAI for gender is in line with previous findings from Mexican, American, Spanish, Swiss, Chinese and French validation studies ^13,19,28-30 which did not report significant differences between men and women in GOHAI scores, according to the quality of oral health conceptualizations.

Dwelling area was previously evaluated only in the French version of the GOHAI ³⁰, where it yielded no significant differences; but France is a country with more homogeneous level of development when comparing rural with urban areas. In this manner, more researches are needed on the differences of the GOHAI performance between urban and rural areas in Latin-America, and the role of socioeconomic inequities in such differences.

In relation to the convergent validity, the correlation tests included overall health, oral health self-perception and the self-perception of treatment necessity, showing significant correlations among the constructs.

It is worth clarifying that differences in the results, of the current validation study, are product of transcultural conditions among countries which make these issues incomparable, aside the population’s inherent characteristics and the validation study’s design.

Internal consistency

In terms of internal consistency, most Cronbach’s alpha values, from previous GOHAI validations, range between 0.8 and 0.9. The Colombian GOHAI attained an internal consistency equal to 0.80, which is consistent with the American, Chinese, Tamil, Persian and Malaysian validations ^19,29,31-33. The Mexican, Portuguese and Romanian versions ^13,27,34 achieved lower internal consistency values, in contrast with the Spanish, Greek, Swiss, French, Dutch, Arab, Japanese and German versions ^{25,26,28,30,35-38}, which report higher internal consistency values.

When withdrawing the GOHAI item 4, the Cronbach’s alpha would come to be 0.81, meaning that internal consistency would change in a single digit, an aspect which does not imply conceptual changes in the internal consistency tests; therefore, the exclusion of item 4 was not accepted.

Test-Retest Reliability

The authors from the Greek validation ²⁶ set the test’s time of re-application one month after it was first applied; the Dutch ³⁵ set re-application time between one and two weeks, and the Chinese, Tamil, Arab, German and Chilean validations ^{29-31,36,38,39)} set the survey’s re-application time to one week after the first application. The Malaysia-validated version ³³ employed a re-application time between 1 and 14 days; the French took 3 weeks between applications ³⁰. As seen in the literature, each author set the time he/she considered appropriate based on his/her experience and previous approaches to the GOHAI scale. In this manner, the expert committee and researchers of the current study defined the re-application time between 5 and 7 days, for the Colombian validation, a period throughout which no changes in Quality of Life are believed to take place on elderly people’s oral health status.

The results as displayed by the GOHAI Colombian version validation, show the correlation measurement to be quite good by yielding a 0.85 Kendall’s coefficient, with a perfect relationship being 1.0, indicating that the Colombian GOHAI has equal or higher test-retest reliability in comparison with the validation of GOHAI versions in other countries; only the Greek, French, Tamil and Dutch versions ^26,30,31,35 showed slightly higher correlations.