This study was carried out in São Paulo, a metropolis with nearly 11 million inhabitants, the second largest city in Latin America, and the Capital of the most populous and industrialized Brazilian state. During the last decades, São Paulo experienced a relevant improvement in life expectancy and health indicators [22
]. A major reform of the national health system in 1988 has boosted initiatives in dental public health and provision of dental care [23
From September to November, 2008, the local health authority performed an oral health survey following international diagnostic criteria standardized by the World Health Organization [24
]. All students aged 12 (179,674) and 15 years (184,537) in the city were eligible to participate in the study.
A total of 4,249 12-year-old and 1,565 15-year-old schoolchildren were examined, and their parents or guardians answered a questionnaire on socioeconomic and demographic conditions. The selection of participants followed a multistage, probabilistic sampling design aimed at allowing statistical inference on the outcomes of oral health with regard to the city as a whole and to each one of its 25 areas, which were geographically divided in 2005 by the local health authority for administrative purposes. These areas were the strata for the multistage selection of sample units, and schools were the primary sampling survey units for the random selection of schoolchildren. Each participating child was assigned a sampling weight corresponding to the inverse of its probability of selection.
As the oral health survey investigated several dental outcomes (dental caries, periodontal conditions, fluorosis, and malocclusion), sample size was calculated to exceed the minimum required for each outcome, based on the prevalence levels reported by a previous municipal oral health survey. Sample-size calculation considered a sample error ranging from 0.05 (prevalence of fluorosis) to 0.20 (dental caries index), a type I error of 5%, and a design effect of 1.25 for 12-year-old students and 1.50 for 15-year-old adolescents.
Refusals to participate were compensated by adding an addition of 30% participants, thus totaling to 4,249 12-year-old adolescents. Furthermore, aiming to allow stratified analysis and to increase statistical power, the original sample was enlarged by adding 1,565 15-year-old adolescents. The refusals were not replaced.
Dental examinations were carried out at the schoolyard, using natural light, periodontal probes (CPI probes), and plane mouth mirrors. Seventy-four specifically trained dentists performed the dental examinations; kappa
statistics assessing inter examiner reliability previous to the fieldwork ranged from 0.70 (95% CI = 0.57 - 0.82) for dental fluorosis to 0.95 (95% CI = 0.94 - 0.96) for dental caries, which is satisfactory for this type of assessment [25
Dental-pain period prevalence - the main outcome variable of this study - was assessed by the direct answer to the question "have you had toothache during the last six months?" Dental pain was originally recorded according to three categories - no, mild dental pain, and severe dental pain. We created a new binary variable by grouping mild and severe dental pain into one category.
Explanatory variables assessed individual and contextual covariates. At the area level, the HDI presented the socioeconomic status. This index is a composite measurement encompassing information on income, education, and longevity, and calculated by governmental agencies [15
] based on the most recent source of information on population, observing criteria established by the United Nations Development Program [15
]. For analytical purposes, the HDI was categorically assessed, considering the median as the cutoff point.
At the individual level, demographic status was stratified by sex, age, and five categories of skin color/race group: Amerindians, Asian descendants, light- and dark-skinned blacks, and whites [27
Socioeconomic position was assessed by the per capita family income, educational level of the parents, and type of school. Family income was divided into tertiles according to their frequency distribution in Reais (Brazilian currency), with cutoffs at half and a quarter of the Brazilian Minimum wage (BMW) per capita. The minimum wage is a standard for measuring income in Brazil, which broadly corresponded to 200 US dollars during the period of data collection. The classification of educational level of parents had cutoffs at 8 and 11 years of formal schooling, which in Brazil, corresponds to completion of primary and high school. As public schools do not collect tuition fee, the enrolment of children in private schools was used as a surrogate of improved socioeconomic status in epidemiologic studies on child health. Finally, the evaluation of dental status used the prevalence of untreated caries (having at least one tooth with untreated caries) and endodontic treatment need (having at least one tooth with indication for endodontic treatment) as covariates of dental pain.
Statistical analyses used Stata 10.0 (2007, Stata Corporation; College Station, Texas, USA). Data analysis considered the organization of the sample into strata and primary survey units as well as individual sampling weights estimated in the draft of complex survey data.
Maps of the city of São Paulo assessed the overlap of areas ranking higher dental-pain prevalence and poorer human development. The assessment of covariates for dental-pain prevalence used Poisson regression analysis; the prevalence ratio (PR) with 95% confidence intervals and p values were the outputs of the analysis.
Poisson multilevel regression analysis used the scheme of fixed effects/random intercept [28
], considering two levels of data organization: the examined schoolchildren and areas of the city. The hierarchical, multilevel analysis observed a conceptual framework to appraise covariates of dental pain, according to the model described by Victora et al
]. The HDI of residential areas was considered as the most distal determinant of dental pain. At the individual level, demographic characteristics were selected as the first block, thus allowing the assessment of all remaining covariates to be adjusted for the distribution of participants by sex, age, and ethnic group. Income, education, and type of school comprised the second block, thus allowing proximal covariates on the third block (dental status) to be adjusted for the differences in the socioeconomic status in the sample (Figure ). All associations were adjusted for covariates positioned in the same and in the upper levels of the hierarchical model. Prevalence ratio for the Human Development Index was also estimated after controlled for all individual-level variables. Interaction between HDI and per capita
family income was also assessed.
Theoretical model of the relationship between contextual and individual characteristics on adolescents' dental pain.
The study followed the national and international standards of ethics in research involving human participants; the study protocol was approved by the Research Ethics Committee of the sponsoring institution - Sao Paulo Health Authority (protocol No. 048/08 - March 18th 2008) - and written informed consent was obtained from parents and guardians of the participating adolescents.