Self-perceived Oral Health among Brazilian Elderly Individuals

OBJECTIVE: To identify factors associated with self-perceived oral health among Brazilian elderly individuals. A probabilistic sample of 5,349 elderly individuals, aged between 65 and 74 years, was analyzed, interviewed and subsequently grouped into dentate and edentate individuals. The dependent variable was self-perceived oral health condition, while independent variables were as follows: place of residence, individual characteristics, health-related behavior, objective health conditions and subjective health conditions. Descriptive and hierarchical multiple linear regression analyses were performed. RESULTS: In both groups, self-perceived oral health was considered positive, despite the poor oral health conditions found among the elderly. In the fi nal model, place of residence and individual characteristics contributed little to explain the variability of self-perception. The use of dental services was associated with the outcome in dentate individuals, while objective and subjective conditions were associated in both groups. Among dentate and edentate individuals, R 2 for the external environment was 0.00; with the inclusion of individual characteristics, 0.05 and 0.02, respectively; with the inclusion of health-related behavior, 0.06 and 0.03, respectively; with the inclusion of objective health conditions, 0.11 and 0.04 respectively; and with the inclusion of subjective health conditions, 0.50 and 0.43, respectively. It was possible to explain 50% of the variability of self-perception in dentate individuals, and 43% in edentate ones. CONCLUSIONS: The main factors associated with self-perceived oral health in both groups were excellent self-perceived appearance, followed by positive self-perceived mastication. The third associated factor was reporting no need for dental treatment in dentate individuals, and self-perceived speech in edentate ones.


INTRODUCTION
Self-perceived health is the interpretation of the health status and experiences in the context of daily life.It is based on information and knowledge about health and disease, which change according to social and cultural norms and experiences. 8The evaluation of self-perceived oral health and oral health condition are essential, because behavior is regulated by the perception of this condition a Ministério da Saúde.Secretaria de Atenção à Saúde.Departamento de Atenção Básica.Condições de saúde bucal da população brasileira, 2002-2003: resultados principais.Brasília, DF; 2004.(Projeto SB Brasil 2003).
and the importance it is given.Self-perceived health favors the indirect participation of the community in making political and social decisions, contributing to an approach aimed at quality of life.In Dentistry, routine evaluation of self-perceived oral health is important to encourage adherence to healthy types of behavior. 3mong the elderly, the main reason for not seeking dental services is the lack of perception of need. 23elf-perceived oral health has a multidimensional aspect. 2,3,6,10,15,17,20,21,23As one of the main components of quality of life, oral health consists in an individual's subjective experience of their functional, social and psychological well-being. 14Self-perceived oral health provides more information about how a certain disease affects an individual's life, rather than the objective measurements of this disease. 7In Brazil, public dental care for the elderly needs to be improved and the identifi cation of their self-perception of oral health could be the fi rst step towards the development of programs that include educational actions aimed at self-care, in addition to preventive and rehabilitating actions. 23e present study aimed to identify factors associated with self-perceived oral health in elderly Brazilians.

METHODS
Data from the Epidemiological Survey of the Oral Health Conditions of the Brazilian Population (Projeto SB Brasil), conducted by the Brazilian Health Ministry in 2002 and 2003, a according to what is recommended by the World Health Organization (WHO), in 1997. 24robabilistic, cluster sampling was used, with a random selection of individuals.Interviews and tests were conducted in homes or schools.More information about the methodology is available in the fi nal report of this research project.a A total of 5,349 elderly individuals, aged between 65 and 74 years, were examined and interviewed in their homes.
An adaptation 15 of the theoretical model proposed by Gift et al 6 in 1998 was used.Analyses were made in two groups: dentate elderly individuals, who had at least one remaining tooth, and edentate elderly individuals, who had not teeth left.The dependent variable studied -self-perceived oral health condition -was obtained through the following question: "How would you rate your oral health?"(very poor, poor, fair, good, excellent).The independent variables were divided into fi ve subgroups: external environment (place of residence), individual characteristics (age in years, sex, self-reported ethnicity; predisposition: years of education, access to information about how to prevent oral problems; availability of resources: per capita household income in reais), health-related behavior (use of dental services), objective health conditions (presence of soft tissue change, number of permanent teeth present, number of decayed permanent teeth present, number of fi lled permanent teeth, periodontal condition and need for prosthesis), subjective health conditions (selfperceived: gum and toothache in the last six months; appearance of teeth and gums; mastication; speech, in terms of teeth and gums; relationship, according to oral health and need for dental care).Among the objective conditions, only the presence of soft tissue changes and need for prosthesis were considered in the analyses of both groups, while the remaining aspects were considered in dentate individuals exclusively.
A descriptive analysis was performed in both groups to identify the presence of statistically signifi cant associations between self-perceived oral health and variables of interest.The contributions of independent variables to the variance of the dependent variable were estimated with multiple linear regression, used in the majority of studies on this theme. 1,2,6,17,21,23This type of regression is indicated to estimate associations between quantitative variables.In the present study, the dependent variable was considered quantitative and discrete, although being ordinal and categorical.The "macro-region", "place of residence", "sex", "ethnicity", "information about how to prevent oral problems", "soft tissue change", "periodontal condition", "self-perceived need for dental care" variables are nominal and categorical and were turned into artifi cial, indicator or dummy variables. 18The "age", "level of education", "income" and "numbers of individuals per room, permanent teeth, decayed teeth and fi lled teeth" variables are quantitative and were categorized in the descriptive analysis exclusively.The "use of dental services", "need for prosthesis" and "self-perceived pain, appearance, mastication, speech and relationship" variables were considered quantitative and discrete, although being ordinal categorical like the dependent variable.
In the multiple linear regression, the following assumptions are recommended: linearity of the phenomenon measured; constant variance of error terms (homocedasticity); independence of error terms (null covariance) or independence of residual random variables; normality of distribution of error terms; 9 non-verification of multicollinearity. 5nearity is assessed by the graphic residue analysis, 4 dispersion diagram or correlation coefficient.The diagnosis of homocedasticity is made from the graphic residue analysis or Levene's test. 9Verifi cation of null covariance is performed using the Durbin-Watson test or the graphic residue analysis. 5The validity of the normality assumption of error terms distribution is confi rmed with the graphic of normal probability for residues and the Kolmogorov-Smirnov test of adherence to normality, corrected with Lilliefors or Shapiro-Wilks. 18Explanatory variables must be linearly independent 5 and, in case multicollinearity is observed, the necessary corrective actions should be determined. 9he most common measures to assess the collinearity of two or more variables are as follows: value of tolerance and its inverse, variance infl ation factor (VIF=1/ tolerance).A reference value which is frequently used is 0.10, corresponding to a VIF higher than 10. 9,18 Multiple hierarchical analyses were conducted in both groups to determine the relative contributions of each variable proposed in the theoretical model. 15The magnitudes of association between the dependent variable and factors of interest, in each group of variables, were estimated using R 2 , with a 0.05 signifi cance level.The respective estimated parameters and standard-errors were obtained using multiple linear regression.The SPSS 11.0 software was used.The design effect correction to analyze data coming from complex samples was not made, once the data needed for such correction were not available.This adjustment was necessary because the Projeto SB Brasil included a complex cluster sample and estimates that do not consider the cluster sampling organization tend towards overestimation and loss of accuracy of estimates. 19e Projeto SB Brasil was approved by the Comissão Nacional de Ética em Pesquisa (National Research Ethics Committee -offi cial opinion 581/2000).
There was linearity in both groups and a violation of homocedasticity.Based on the analysis of null covariance, there was a violation in the graphic residue analysis, although not through the Durbin-Watson test (DW=1.9;p≤0.05).A violation of the hypothesis of normality of error terms distribution was found in the normal probability graph (Q-Q Plot) and Kolmogorov-Smirnov tests, corrected with Lilliefors or Shapiro-Wilks, (p=0.000).Results of multicollinearity tests showed that such assumption was not violated (tolerance>0.10and VIF<10).
In the multiple linear regression analysis, the R 2 for model 1, external environment, was found to be 0.00 between dentates and edentates; for model 2 (external environment and individual characteristics), 0.05 and 0.02, respectively; for model 3 (external environment, individual characteristics and health-related behavior), 0.06 and 0.03; for model 4 (external environment, individual characteristics, health-related behavior and objective health conditions), 0.11 and 0.04; for model 5 (external environment, individual characteristics, health-related behavior, objective health conditions and subjective health conditions), 0.50 and 0.43 (Tables 3 and 4).

DISCUSSION
Self-perceived appearance was the factor most strongly associated with self-perceived oral health and no previous studies that had investigated such association were found, thus revealing the need for future studies on this theme.It is suggested that the identifi cation of factors associated with self-perceived appearance could more accurately clarify self-perceived oral health among elderly Brazilians.
As well as other studies, the majority of elderly individuals perceived their oral health as positive, 1,3,6,10,11,17,21,2 2,23 even with unsatisfactory objective oral health conditions.Literature data are consistent in this sense, 6,21,23 suggesting that objective conditions are poor predictors of such self-perception. 17,21The weak association between objective conditions and self-perceived oral health in both groups had been previously observed, probably because diseases detected by professional tests can be asymptomatic and unknown to patients, 21,23 who only consider themselves to be ill when suffering from acute manifestations of oral diseases. 21Another explanation for such weak associations would be the assumption that objective indicators measure diseases, whereas subjective ones assess health and human experiences. 13e hierarchical approach of multiple linear regression enabled the assessment of the importance of each group of variables in dentate and edentate individuals' self-perceived oral health.R 2 varied slightly when the three fi rst models that hierarchically included variables related to the external environment; individual and behavioral characteristics were assessed; these variables explained approximately 6% of the variability in  a R$ 1.00 was equivalent to US$ 3.00 at the time of this study variability and, in addition to rising from 3% to 4% in edentates.This suggests that objective conditions had a greater impact on the self-perception of elderly individuals with at least one remaining tooth than in edentate ones.Subjective conditions were the main factors that explain the variability of self-perceived oral health.In both groups, the inclusion of such variables into the model explained 50% of variability of self-perceived oral health in dentates and 43% in edentates.
Studies suggest that negative self-perceived oral health increases with age. 6,21In contrast, in the present study and other ones, age is not associated with selfperception. 2,17,23There was no association between sex and self-perception, as in the majority of studies, 2,6,23 although more frequent positive self-perception in women than in men has been reported. 21Ethnicity was not associated with self-perception in both groups, thus contrasting with studies that show that white individuals perceived their own health in a more positive way than non-white ones. 2,17Level of education and access to information about how to prevent oral problems were not associated with the outcome.Studies show that, the greater an individual's level of education 2,6,17 and access to information about how to prevent oral problems, the more frequent the positive self-perceived oral health. 2,6ncome was initially associated in both groups, although such association was not found in the fi nal model.The literature is controversial in terms of income.Among elderly Brazilians, this association was not probably observed due to the homogeneity of the population, most of whom had a low income.The variables related to individual characteristics did not contribute to the understanding of variability of self-perceived oral health in elderly Brazilians.
Use of dental services was associated with self-perceived oral health in dentates.Another study showed that individuals who had used such services for less than one year before perceived their oral health as positive. 2jective conditions have limited relevance in selfperceived oral health. 21A higher number of permanent teeth present is associated with positive self-perceived oral health. 1,2,6,16,17,21However, there was an inverse association in the present study, so that the lower the number of teeth present, the more frequent the positive self-perceived oral health.Despite the apparent incoherence of results, elderly individuals seem to rate their oral health more positively when they are free from decayed teeth and probably from diseases, rather than when few teeth are maintained in precarious conditions, in insuffi cient number and without access to prostheses to guarantee effi cient and comfortable mastication. 15nother study found that the higher number of decayed teeth present in the mouth of individuals was associated with negative self-perceived oral health. 1,6,21Yet another study observed that a higher number of fi lled teeth was associated with positive self-perceived oral health. 21esults in the literature are controversial about periodontal disease.Loss of periodontal attachment 6 and the presence of periodontal pockets 21,23 were associated with negative self-perceived oral health. 6The association between periodontal condition and self-perceived oral health was neither found in the present study nor in another study. 1 Associations between self-perceived global oral health and specifi c self-rated functional or psychological limitations were observed in another study. 12The association between absence of sensitivity to pain and positive self-perceived oral health found in the present study, in both groups, was reported in another study. 2ositive self-perceived appearance was associated with positive self-perceived oral health. 1,17Other studies indicated the importance of self-perception of the need for dental care in the variability of self-perceived oral health. 1,6,10,16The frequency of positive self-perceived oral health was higher among those who reported not needing treatment in both groups of study.
Results explained 50% of the variability of selfperceived oral health in dentates and 43% in edentates and also showed the effectiveness of the proposed model.This model surpassed the explanation of such variability in other studies performed in the United States: 14% in workers of an insurance company aged 18 years or more; 21 43% in the search performed in Medicare users aged less than 65 years, in 1988; 1 37% in Medicare users aged 65 years or more in Los Angeles, California; 17 38% in residents of Baltimore; 32% in users of health services for indigenous people; 39% in residents of San Antonio; 2 and 35% in elderly individuals aged 65 years or more. 6Not all variables associated with self-perception, described in the literature and observed in the proposed model, were analyzed in the Projeto SB Brasil, preventing better understanding of the variability of self-perceived oral health.More in-depth studies will enable the factors that infl uence it to be better understood.The process that associates self-perception and the variables analyzed is dynamic, as shown in the proposed model.The Projeto SB Brasil was a cross-sectional study, thus not enabling a temporal relationship among the observed associations to be established.
In both groups, two assumptions were violated: homocedasticity and normality of error terms.In practical terms, the violation of the assumption of homocedasticity causes predictions to be better on some levels than on others, so that such violation frequently renders statistical tests more conservative or sensitive.The most common violation is the assumption of normality of error terms, 9 which means that the differences between the estimated model and the data observed are not frequently zero or close to zero anymore, i.e. that the differences above zero do not only occur occasionally.
The assumption of null covariance was considered not to have been violated, once the graphic analysis is considered subjective 5 and such violation was not confi rmed by the Durbin Watson test.In the multiple linear regression analyses where the assumption of multicollinearity is violated, there is a reduction in the accuracy of regression coeffi cients estimators. 5In the present study, this assumption was not violated.Despite the limitations resulting from the violation of two out of the fi ve recommended assumptions, the choice made was to compare the results with studies in which the analyses of all recommended assumptions were not described. 1,2,6,17,21,23 terms of the variability of self-perceived oral heath, self-perception was considered positive, although poor oral conditions in the elderly were found.The use of the linear regression model to analyze self-perceived oral health is common, though questionable, once it is not a quantitative variable.Self-perceived appearance rated as excellent, followed by positive self-perceived mastication and reporting no need for dental care, were the factors that most contributed to explain the variability of self-perceived oral health in both groups.
Objective health conditions helped to explain this, especially in dentates.The higher the number of permanent teeth present, the more negative the self-perceived oral health; moreover, the higher the number of fi lled permanent teeth, the more positive this self-perception.Factors associated with the external environment, individual characteristics and behavior contributed little to the understanding of variability of self-perception in both groups.The use of dental services infl uenced the oral health self-perception in dentates.Better understanding of such associations is necessary to guarantee quality of life for elderly individuals.

Table 1 .
Description of dentate elderly individuals, according to self-perceived oral health, external environment, individual characteristics, health-related behavior, objective health conditions and subjective health conditions.Brazil, 2002-2003.
a R$ 1.00 was equivalent to US$ 3.00 at the time of this study dentates and 3% of the variability of self-perceived oral health in edentates.The inclusion of objective conditions into the model contributed little to the understanding of self-perceived variability, although the R 2 practically doubled in dentates, explaining 11% of

Table 2 .
Description of edentate elderly individuals, according to self-perceived oral health, external environment, individual characteristics, health-related behavior, objective health conditions and subjective health conditions.Brazil, 2002-2003.

Table 3 .
Multiple linear regression analysis of self-perceived oral health in dentate elderly individuals.Brazil, 2002-2003.

Table 4 .
Multiple linear regression analysis of self-perceived oral health in edentate elderly individuals.Brazil, 2002-2003.