Multivariate method to identify inequalities in oral healthcare access.

OBJECTIVE: The aim of this study is to apply a multivariate method for municipality's classification according access to oral health in adults.
MATERIALS AND METHODS: This is a cross-sectional epidemiological study. Were used multivariate classification called nonhierarchical cluster analysis K-means. The strategy brings together municipalities' similarity as access to oral health, where the most similar are next and the most different from getting further away. In addition, it allows reducing the intragroup variance and maximizing intergroup variance. It was assumed the number of four groups.
RESULTS: Among adults, 3,185 (52.63%) visited the dentist less than a year, and 357 (5.90%) have never been to the dentist. Homogeneous groups showed differences in the time since the last visit to the dentist for adults. The analysis of variance by the F statistic rejected the hypothesis that the variances are equal for the variables related to the time since the last visit to the dentist.
CONCLUSION: It was possible to identify the inequalities in the access to oral healthcare services for adults and locate spatially municipalities whose subjects take longer to visit the dentist.

INTRODUCTION

Dental healthcare services differ greatly among countries regarding organization, accessibility, availability, and cost.[1234567891011121314151617181920] In some countries, full dental health services are readily available through private or public systems.[124] Countries that offer universal coverage of health services providing a healthcare package to all citizens, without suffering financial hardship when paying for them.[4]

The aim of this study was to apply a multivariate method to classify the access to oral health in adults.

MATERIALS AND METHODS

Data were extracted from the epidemiological survey of the Oral Health Conditions of the Population of the State of São Paulo (SBSP-2015) with 161 municipalities in 2015.[21] The data are available at Figshare public data repository – Licence CC BY 4.0 (DOI: 10.6084/m9.figshare. 5286025.v1). Subjects were chosen by conglomerate/cluster sampling with probabilities proportional to the population size, taking into consideration the sample weight and effect of design on the respective stages of the draw.[21] The sample size was calculated using the mean values of dental caries; prevalence of periodontal conditions; prevalence of use and need for dental prosthesis; with the respective standard deviations; acceptable error margins (ε); design effects (deff = 2), and non-response rates of 30%. Finally, a sample of 6051 adults aged 35–44 years from State of São Paulo, Brazil, was obtained.[21] Training and calibration processes of the dental teams were conducted by the gold-standard examiner with level of interrater agreement statistic Kappa of over κ > 0.76.[21]

The municipalities were classified through the nonhierarchical K-means multivariate grouping technique.[181922] K-means is one of the simplest unsupervised learning algorithms that solve the well-known clustering problem.[22] The procedure follows a simple and easy way to classify a given data set through a certain number of clusters (assume k clusters) fixed a priori. The main idea is to define k centroids, one for each cluster. These centroids should be placed in a cunning way because of different location causes different result. The next step is to take each point belonging to a given data set and associate it to the nearest centroid. At this point, we need to recalculate k new centroids as barycenters of the clusters resulting from the previous step. After we have these k new centroids, a new binding has to be done between the same data set points and the nearest new centroid. Finally, this algorithm aims at minimizing an objective function, in this case, a squared error function. The number of 4 groups (A, B, C, and D) was admitted, and the name of the municipality was defined as variable identifying the cluster.[1819] The variable “last dental visit” was used as input variable for the formation of homogeneous groups. Previous studies have used thematic maps to visualize the results of multivariate classification and visualization of homogeneous groups (clusters).[181923]

The F statistic was used to test the hypothesis that the sample variances are equal (H0) and with a level of statistical significance (α = 0.05).[19] The correlation compared the variability among the means of the formed groups.[19]

Research Ethics Committee of Dentistry College of Piracicaba approved the study with number 111/2015.

RESULTS

There was a decreasing order in the number of municipalities belonging to each group, and Group “A” showed 62 municipalities and Group “D” 10 municipalities [Figure 1].

Figure 1

Distribution of municipalities according to the time since the last visit to the dentist, São Paulo, 2016

Among adults, 3185 (52.63%) have been to the dentist less than a year, and 357 (5.90%) have never been to the dentist. Group A represented municipalities whose individuals took an average of more time or never visited the dentist. Groups B and C represent the municipalities where adults visited the dentist less time. The hypothesis of equal variances for researched variables was rejected and the individuals who took the most time to visit the dentist [Table 1].

Table 1

Profile of groups according to means of time since the last visit to the dentist, São Paulo, 2016

Boxplot summarized robust measures of central tendency and dispersion, with Group A presenting the worst median and Group B the largest. Among the municipalities that visited the dentist more than 3 years, there was an inequality in the distribution of the outcome, mainly for Groups A and C [Figure 2].

Figure 2

Boxplot of the groups according to the average time since the last visit to the dentist, São Paulo, 2016

The paid service was the most used by adults, motivated by demands for treatment and well evaluated by the user, and 1046 (18.75%) were motivated by toothache and 617 (11.04%) for tooth extraction [Table 2].

Table 2

Profile of the groups of municipalities according to means of the variables of access to the oral health service by adults, São Paulo, 2016

The use of the public oral healthcare service was inversely proportional to fee for service/private health insurance and directly motivated by pain, extraction, treatment, and evaluated positively. Adults who fee for service or private health insurance were directly related to the reason for negative review and evaluation of the service. Dental extraction was directly related to the use of the publicoral health care service [Table 3].

Table 3

Pearson's bivariate correlation coefficient between the variables of access to the oral health service by adults, São Paulo, Brazil, 2016

DISCUSSION

The multivariate classification process employed was able to identify significant and important differences in access to oral health services.

It was observed that the demands for public dental services were high; however, it was the private sector that responded by the greater coverage of these services. The improvement in the average income of the Brazilian population may influence the higher demand for the paid dental service.[234924] In this study, a higher prevalence of consultations was found in private practices and corroborates previous studies; however, the prevalence values were divergent and can be explained by the methodological differences between the studies. Another important result is the correlation of the negative evaluation of the private services or by plan and can be explained by the fact that individuals with better socioeconomic conditions tend to have better schooling and evaluate more critically the service received.[9]

The results also provide information on the geographical behavior of the main untreated oral conditions in Brazil.[2024] Adults who never attended a dentist had toothache or went to the dentist to extract a tooth indicate a serious epidemiological picture of the Brazilian adult population.[37] The cohort effect resulting from the past exposures to certain risk factors may reflect on the oral condition of the adults interviewed.[25]

The positive evaluation obtained a high prevalence and was similar to that found in a previous study.[7] On the other hand, the low prevalence of individuals who evaluated negatively the received service found in this study is based on the acceptance in the cultural question of acceptance of the oral health condition as a natural phenomenon of the aging process.

This study has a cross-sectional design, and causality has not been studied. The “time since the last visit to the dentist” is a variable that depends on the respondent's memory, and memory bias may have occurred.

CONCLUSION

The classification method identified the spatial pattern of distribution and inequality of access to oral health in adults. In addition, was possible visualize, when, where and why the adult individuals living in São Paulo State are seeking to improve the oral health condition and provide subsidies for planning in oral health according to demands of the population.

Financial support and sponsorship

This study was financially supported by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).

Conflicts of interest

There are no conflicts of interest.