Characterization of the self-perception of oral health in the Brazilian adult population

Resumo O objetivo deste artigo é realizar uma análise dos fatores que determinam a autopercepção de saúde bucal dos brasileiros sob base metodológica multidimensional. Estudo transversal, com dados provenientes de inquérito em âmbito nacional. Foi realizada entrevista domiciliar com uma amostra de 60.202 adultos. Considerou-se a autopercepção de saúde bucal como variável desfecho e características sociodemográficas, de autocuidado e condição de saúde bucal, de utilização de serviços odontológicos, de condição de saúde geral e de trabalho como variáveis independentes. Empregou-se o teste de redução de dimensionalidade e as variáveis que apresentaram relação passaram pelo teste de regressão logística. A autopercepção negativa de saúde bucal apresentou-se relacionada à dificuldade para se alimentar, avaliação negativa do atendimento recebido durante a última consulta odontológica, autopercepção ruim da condição de saúde geral, não utilização de fio dental, perda dental superior e motivo da última consulta com o cirurgião dentista. A utilização de base metodológica multidimensional foi capaz de (re)desenhar modelos explicativos para a condição percebida de saúde bucal de adultos brasileiros, devendo, os seus resultados serem considerados na implementação, avaliação e qualificação da rede de saúde bucal. Palavras-chave Saúde bucal, Percepção, Autoavaliação, Estudos transversais, Mineração de dados Abstract This article aims to perform an analysis of the factors that determine the self-perception of oral health of Brazilians, based on a multidimensional methodology basis. This is a cross-sectional study with data from a national survey. A household interview was conducted with a sample of 60,202 adults. Self-perception of oral health was considered the outcome variable and sociodemographic characteristics, self-care and oral health condition, use of dental services, general health and work condition as independent variables. The dimensionality reduction test was used and the variables that showed a relationship were submitted to logistic regression. The negative oral health condition was related to difficulty feeding, negative evaluation of the last dental appointment, negative self-perception of general health condition, not flossing, upper dental loss, and reason for the last dental appointment. The use of a multidimensional methodological basis was able to design explanatory models for the self-perception of oral health of Brazilian adults, and these results should be considered in the implementation, evaluation, and qualification of the oral health network.


introduction
The personal assessment of living and becoming ill is characterized by the interpretation that an individual performs on his/her own health 1,2 . Several factors have been pointed out in the literature as capable of influencing this self-assessment: personal characteristics, especially the access to information 3 , demographic aspects [4][5][6][7][8][9][10][11][12] , level of acculturation 4,6 , beliefs, values, health practices 3,6,13 , physical and cognitive ability, and emotional relationships 14,15 .
Due to the subjective nature of the object, classical epidemiological studies are not able to apprehend the elements that are related to the evaluation of the health status of the individuals, since they are mainly devoted to aspects of human biology, clinical, and laboratorial methods. Although biologically based information is essential for the quantification of the population health conditions, since they subsidize the planning, organization, and monitoring of health services, personal measures are increasingly recommended by the World Health Organization 16 . Authors affirm that through these measures the individual conscience on health status is reflected 17 , constituting also an important contribution to the orientation of political and social decisions that have as a goal the quality of life 3,6 and to increase the search, adherence, and motivation of individuals for treatment and self-care 4 .
In the oral health field in Brazil, despite the importance of assessing patients' self-perception on broad and effective bases, most studies have focused on isolated conditions such as clinical (use of dental prosthesis, cavities, and periodontal disease), socioeconomic (income and level of education), socio-demographic (gender, age, and ethnicity), and behavioral factors (consumption of tobacco, alcohol, and other drugs), as well as specific habits of oral hygiene [3][4][5][6][7][8][9][10][11][12]18 .
The empirical evaluation of self-perception of oral health should incorporate as many critical, social, and biological indicators as possible, aiming at the formation of a synthesis measure, influenced in a micro and macro dimensional way, with repercussions on people's quality of life 6 .
Few studies on the self-perception of oral health consider representative samples of the entire country 11 , and this fact precludes the access of formulators and implementers of public policies to a true and near panorama of the evaluation of the health conditions of individuals and collectivities.
Thus, the objective of the present study was to perform a more comprehensive analysis of the factors that determine self-perception of oral health of Brazilians, based on a multidimensional methodological basis.
It is believed that the results of this study will contribute to the effectiveness of evaluation and planning of public oral health actions, making it possible to resize and qualify them. methods This is a cross-sectional quantitative study with data from the population-based survey -National Health Survey (NHS), proposed by the Ministry of Health and conducted in 2013 by the Brazilian Institute of Geography and Statistics (IBGE) 19 .

the National health survey
The research is in a household level and the sampling plan used was based on a cluster sampling in three stages, with the sectors or groups of census tracts being the primary sampling units, the households the secondary, and the selected elderly inhabitants, the tertiary units 19 .
The sample size was defined considering the level of precision desired for the estimates of some indicators of interest, resulting in a sample of 60,202 adults interviewed 19 . Details on the sampling and weighting process are available in the PNS report 19 .
Previously calibrated researchers collected the data. The information was obtained through individual interviews and stored on handheld computers. Only individuals older than 18 years old participated in the research. The interview was directed by three forms: the domicile, referring to the characteristics of the house; the residents of the house; and the individual, answered by a drawn resident of the house 19 . For the present study it was considered exclusively the data of the drawn resident from the last two forms.
The survey was approved by the National Commission for the Protection of Human Subjects of the Ministry of Health. The data used are in the public domain from the National Health Survey. Being the data available without link: http://www.pns.icict.fiocruz.br/index.php? pa=resultados.

treatment and data analysis
The outcome variable "self-perception of oral health" is the result of the question: In general, how do you evaluate your oral health (teeth and gums)? Possible answers were: very good, good, regular, poor and very poor. For the purpose of analysis in the present study, the responses were grouped into positive (very good and good) and negative (regular, poor and very poor).
In the exploration phase of the data, 47 variables of interest were cataloged in order to compose the independent variables, related to: sociodemographic characteristics (eight variables); oral self-care (five variables); oral health condition (five variables); use of dental services (five variables); general health condition (eighteen variables); job characteristics (six variables).
In order to verify the existence of one or more chronic diseases, regardless of the disease installed, a new variable grouping all the diseases listed above was created, entitled "presence of one or more chronic diseases." The variable 'number of natural teeth present in the mouth' was also developed. For that, it was subtracted the number of upper and lower teeth lost from all natural teeth (n = 32). Subsequently, the variable was categorized, according to a previous study, into more than 10 natural teeth present and 10 natural teeth present or less 8 .
After the complete database treatment, dimensionality reduction was performed using the Correlation-based Feature Selection (CFS) algorithm using the 10-fold cross validation technique. This data mining test is recommended to be used in a large set of data and variables as proposed in the present study. It evaluated, according to response patterns, the entire data set and the 47 variables at a single time to look for variables highly related to the outcome variable and not related to each other. In this way, it does not only consider the utility of variables individually, but also the correlation level between them, thereby eliminating any and all confounding relationships. The variables with nonexistent or weak and/or redundant relations, which do not bring information gain to the model, are all eliminated by the algorithm, without needing any researcher's influence.
Thus, it is possible to validate the pure and strict relations of the independent variables to the outcome variable with much more precision than other tests commonly used in the literature.
After, the variables related to 'Self-perception of oral health' were evaluated through logistic regression to measure the magnitude of the associations.
The model had an explanatory capacity of 73%. All analyzes were performed using the software WEKA (Waikato Environment for Knowledge Analysis) environment.
The WEKA software does not report the individual confidence intervals and p values for each OR; it considers the default values of 95% confidence interval and p value < 0.05. In this type of analysis, the possibility of influencing the confidence interval due to the sample size is eliminated in the steps preceding the regression analysis.

Results
The descriptive analysis shows that about 34% of the individuals interviewed evaluated their oral health negatively (Table 1). In the results of the attribute selection analysis, it was verified that the variables most strongly related to self-perception of oral health were: difficulty feeding, upper dental loss, self-perception of the general health condition, evaluation of the last dental appointment, reason for last dental appointment and flossing. No variables related to sociodemographic and job characteristics were related to self-perception of health. Table 1 shows the distribution of adults, considering the variables used in the study that did not present a strong relation with the dependent variable. Table 2 also describes the sample according to the independent variables that presented a relation with the self-perception of oral health. Table 3 shows the reasons for the chance of reporting negative oral health status according to independent variables.
The variable that most influenced the evaluation of oral health was the difficulty in feeding, where individuals who reported difficulty (11%) presented 5.81 more chance of having a poor perception of health in detriment to those without difficulty. The upper tooth loss was also linked to self-perception of health. The partial loss of upper teeth increases the probability of having a negative perception of oral health in 1.49, while the total loss of teeth was shown to be a protective factor (OR = 0.86).
Individuals who negatively evaluated their overall health (49%), presented 1.91 more chance of evaluating their oral health as bad. Regarding to the use of dental services, the search for dental consultations for reasons other than prevention (treatment or others) and negative evaluation of care increases the chances of the individual to have a negative perception of oral health.

Discussion
The variables most strongly related to the self-perception of oral health found in the study were: difficulty feeding, upper dental loss, self-perception of the general health condition, evaluation of the last dental appointment, reason for last dental appointment and flossing. Thus, it can be inferred that subjective conditions, which evaluate human experiences and health, are much more capable of explaining the variability of self-perception of oral health than objective measures, which mostly measure the context of the presence of diseases 11,20,21 . According to studies 5,8,20,22,23 , individuals who report more dysfunctions, symptoms, and oral incapacities evaluate their oral health in a more negative way, in line with the findings of the present study, which demonstrate that difficulty feeding increases by 5.8 times the chance of the individual exhibiting a negative self-perception of oral health. These results may reflect the relation between masticatory limitation and the worsening of the psychosocial conditions by the embarrassment to feed in front of other people, or on physical or functional conditions, especially as the nuisance when swallowing food 5 .
The literature shows that the number of teeth present in the mouth is an important determi- self-perception of oral health, but only the general context of health. This finding indicates that general health situations transcends chronic conditions and reinforces that the complexity in the framework of health perceptions is much more involved with subjective issues than with objective clinical demands. The reason and the evaluation of the last dental appointment were the only variables of the group of "use of dental services" that were related to the dependent variable. Visiting the dentist for check-up or prevention and positively evaluate appointment were self-referred factors of oral health protection. It is suggested that the explanation for this finding lies in the fact that routine visits to the dentist 8,20 and good dental service 27,28 can minimize dental loss and improve oral condition, as well as empower patients with healthy living habits and the health-disease process, making them more confident about their status of oral health.
A similar condition can be observed regarding flossing. Individuals who floss regularly have a better self-perception of oral health 6 . The study emphasizes that the lack of knowledge and moti-vation to adopt preventive and care practices aggravates the oral health condition and compromises the self-perception of oral health 4 .
Moreover, it should be noted that the literature is rather vast and inconclusive in relation to factors that truly affect self-perception of oral health [3][4][5][6]8,[10][11][12][13][18][19][20][21][22][23] . The only pattern that emerges from these studies is that most of the factors associated to sociodemographic, work characterize, lifestyle, health, and use of dental services by the individuals are related to the self-perception of oral health, however they do not contribute to the understanding of its variability. Given the diversity and the low power of association of the methods used to cross the factors of interest, the information obtained becomes little advantageous for the assertive decision making.
Through the feature selection test applied in the present study, it became possible to consider the pluralism of the social and health segments involved in the large database of the National Health Survey and to eliminate possible confounding or redundant factors, unraveling the variables that are truly important for the conformation of the self-perception of oral health. This condition becomes crucial for contributions that aim to increase the knowledge about self-perception of oral health in the Brazilian adult population, which can help supporting decision-making processes and redirection of oral health practices and resources.

limitations of the study
As a limitation of the study, we highlight the subjectivity imbricated in the evaluation of self-perception of oral health, since it is susceptible to changes throughout life, day, week as a result of contextual conditions, psychological state experienced by the individual, as well as involving values and feelings not always expressed 29 . However, the subjectivity does not disqualify the relevance to guide policy and to plan decisions in health 3,6 .
In addition, because this is a cross-sectional study, it was not possible to establish a temporal relationship between the associations found.

Conclusion
The use of a multidimensional methodological basis was able to (re)design explanatory models for the self-perception of oral health of Brazilian adults and its results should be considered in the implementation, evaluation, and qualification of the national oral health network.
It is noteworthy that the determinants of the self-perception of oral health of Brazilians were difficulty feeding, use of dental floss, upper dental loss, self-perception of the general health condition, and reason and evaluation last dental appointment; and that factors related to sociodemographic and work characteristics, lifestyle, and presence of chronic diseases contributed little to the understanding of the variability of the self-perception of oral health of the adult population.
Collaborations D Bordin delineated the study, interpreted the data and wrote the article. CB Fadel delineated the study and carried out the writing of the article. CB Santos conducted the statistical analysis, created the index for analysis and performed the interpretation of the data. CAS Garbin, SAS Moimaz and NA Saliba contributed writing and critical content review.