Study population
This retrospective study was based on data from the national epidemiological survey of oral health, Project SB Brasil (Projeto Saúde Bucal Brasil – Project Oral Health Brazil), which was carried out in 2010. It was a cross-sectional study that included elderly Brazilian subjects aged 65–74 years. The design of SB Brasil, a multistage cluster study, has been previously described [13, 14]. The sample domains were the state capitals and non-capital municipalities in the five regions of Brazil. The sample units were census zones and households in the state capitals, and census zones and households in municipalities of the non-capitals. A total of 30 sectors in each capital and 30 non-capital municipalities of each region were randomly selected. The sample considered fields clustered according to density of the total population and the internal variability of the indices. The final sample was representative of each state capital and federal district in Brazil. From these criteria, a total of 7,619 older adults from municipalities in the state capital and non-capital cities from all five geographic regions of Brazil participated in the study.
The SB Brasil 2010 Project was approved by the Research Ethics Committee of the Ministry of Health and was registered with the National Commission of Research Ethics under number 15.498.
Sources of data
Data on complete denture need as well as and socioeconomic and sociodemographic variables were obtained from Project SB Brasil 2010 and databases of the National Demographic Census (www.ibge.gov.br and www.pnud.org.br), carried out in 2010 by the Brazilian Institute of Geography and Statistics.
Dependent variable
The indices used in SB Brasil 2010 to assess the oral health status of older adults met the recommendations of the 4th Edition of the WHO Instruction Manual for Basic Epidemiological Surveys in Oral Health [13], and took into account experience accumulated in previous surveys in several regions of Brazil, particularly from 1980. The indices of SB Brasil included the following: upper denture need (none/one denture element/more than one prosthesis element/combination of dentures/complete denture) and lower denture need (none/one denture element/more than one prosthesis element/combination of dentures/complete denture). The category selected for this study was “complete denture”, and the dependent variable in this study was “complete denture need”, i.e., requiring complete dentures in one or both dental arches.
Independent variables
The individual variables were age (65–69 years/70–74 years), sex (male/female), race (white/mixed), and socioeconomic status (high/low), obtained from a questionnaire administered by public health service dentists to participants in their homes. Race was assessed through self-report according to the classification proposed by the Brazilian Institute of Geography and Statistics.
A socioeconomic status variable was constructed using the cluster analysis technique, and included: 1) the number of consumer goods (e.g., televisions, refrigerators, stereos, microwaves, telephones, cell phones, washing machines, dishwashers, personal computers, and cars), ranging from 0 to 11; 2) family income (monthly family income in Brazilian currency, reais); 3) years of education; and 4) household crowding (number of people living in the residence of the older adult divided by the number of rooms serving as bedrooms in the residence) [13].
The city-level sociodemographic variables were: population size (municipalities up to 20,000 inhabitants, 20,001–100,000 inhabitants, 100,001–500,000 inhabitants, and more than 500,000 inhabitants); type of municipality (capital or non-capital); region of Brazil (north, northeast, center-west, southeast, south); and Human Development Index (HDI: up to 0.762; 0.763 and over). The HDI is a composite measure encompassing information on income, education, and life expectancy, and is used to assess the social development of an area and to compare quality of life on an international basis [15]. The HDI was taken from the “Human Development Atlas” provided by the Brazilian agency of the United for Nations Development Program (www.pnud.org.br). The same method was used for all municipalities in Brazil [8]. HDI scores range from 0 to 1, and the higher the value, the better the social development. In the present study, the HDI variable was dichotomized according to the median value.
The city-level variables were population size, type of municipality, and region of Brazil and were extracted from the National Demographic Census, carried out in 2010 by the Brazilian Institute of Geography and Statistics.
Data analysis
A descriptive analysis of the dependent, individual, and contextual independent variables was performed. A Poisson robust regression model was used for bivariate analysis to obtain crude prevalence ratios (PRs) and 95 % confidence intervals (CIs). In multivariate cluster analysis, a hierarchical technique was first used to select the number of clusters and profile cluster centers to serve as initial cluster seeds (centroid elements) in the nonhierarchical procedure. Following this, hierarchical process analyses produced dendrograms that resulted in the choice of two socioeconomic clusters. The between groups linkage clustering algorithm was used in the agglomerative hierarchical method, with Euclidean simple used as a distance measure, as this was used in the nonhierarchical part of the analysis. Ward’s clustering algorithm was used to check the stability of clusters at a 2nd time stage (after initial analysis).
The nonhierarchical technique was selected because of the large sample size. From the final cluster model, based on standardized values (by Z score), the presence of two clusters was observed: one composed of individuals with high socioeconomic status, and the other composed of individuals with low socioeconomic status.
A two-level multilevel mixed-effect Poisson regression analysis was performed to verify the effect of individual characteristics and also the context’s influence on the outcome. In this case, the context was represented by one level of aggregation, taking into account the Brazilian administrative organization.
Exploratory analysis was performed to evaluate the effect of the level through calculation of PRs with respective 95 % CIs, with the better situation as the reference category. In addition, two-level multivariable modeling was constructed for the outcome using significant explanatory variables. The analysis started with a random intercepts model (null model) in order to verify whether the contextual effects were significant. The final model was executed when the variables were added and the model remained significant. Only HDI was used as a contextual variable in the final model because it was collinear with the other contextual variables, and also included the respective municipality.