Research design and scenario
An ecological study[7] was carried out in Ribeirão Preto, a city in the interior of the state of São Paulo (SP). Located 314 kilometers (km) from the capital, Ribeirão Preto has an area of approximately 650 square kilometers (km²) and a high population density of 995.3 inhabitants/km². It also had an estimated population of 711,825 inhabitants in 2020, of which 99.7% live in urban áreas[8].
The ecological analysis units of the study are the census sectors in the municipality. The census sector is the smallest territorial unit formed by a continuous area, located in an urban or rural area, with a defined size, number of households and number of residents, and is used for the main Brazilian statistical surveys[9].
Figure 1 shows the geographical location of the municipality of Ribeirão Preto, which is divided into 972 census sectors, however, for this study it was decided to use only the urban census sectors in the municipality, corresponding to 956 units of analysis. The cartographic base of the census sectors in Ribeirão Preto was obtained from the website of the Brazilian Institute of Geography and Statistics (IBGE)[9] free of charge.
Regarding tuberculosis, the city is considered one of the 22 priority municipalities in the state of São Paulo due to the high number of cases; in the last nine years, there was an average of 196 cases/year, with an incidence rate of around 35/100,000 inhabitants[10].
The active search for respiratory symptoms with sputum smear collection and X-ray order is performed by all basic health units (Ribeirão Preto has 44 health units in Primary Care), however, the treatment and monitoring of the disease is carried out in the five specialized district infectology clinics. In addition, the directly observed treatment is offered to all patients diagnosed with tuberculosis and is carried out by the outpatient teams[10].
Population
The study population consisted of all confirmed cases of pulmonary tuberculosis reported in the Tuberculosis Control System (TBWeb) from 2006 to 2017, made available by the Municipal Tuberculosis Control Program of the Ribeirão Preto Municipal Secretariat.
It is worth mentioning that in Brazil, the notification of tuberculosis is carried out through the Information System for Notifiable Diseases (SINAN), however the state of São Paulo uses its own unique system for notification and monitoring of people with tuberculosis, but which also interfaces with SINAN. The system, called TBWeb, started to be used effectively from the year 2006 and the notifications are made online, the main advantage being the exclusivity of the medical record of each person notified with tuberculosis and automatic communication in cases of transfer and hospitalization[11].
It was adopted as a selection criterion that the notification was carried out between 2006 and 2017, with only one registration per person, the most current registration being selected if there was more than one entry in the system and residents in an urban area of the city of Ribeirão Preto. It is noteworthy that only pulmonary tuberculosis records were considered, so that extrapulmonary or concomitant forms (pulmonary and extrapulmonary forms together) were excluded.
Analysis plan
Time series analysis
Initially, monthly time series of tuberculosis cases were constructed, referring to the period from January 2006 to December 2017. To verify the behavior of the time series over the study period and also its trend, the decomposition method called Seasonal Trend Decomposition using Loess (STL) was used, which is based on a locally weighted regression[12]. This analysis was performed using RStudio software through the forecast package[13].
Identification of clusters
The georeferencing of pulmonary tuberculosis cases was performed using the Google Earth Pro® software in order to obtain the geographical coordinates (latitude and longitude) of the residential addresses of the notified cases.
In order to identify areas at higher risk for pulmonary tuberculosis, the spatial analysis technique called scanning statistics, developed by Kulldorff and Nagarwalla[14], was used.
The identification of clusters was carried out by placing a circle of variable radius around the centroid of each unit of analysis (census sector) and the number of observed and expected cases is calculated. This procedure is performed until all centroids are tested and, when the value observed in the area enclosed by the circle is greater or less than expected, it is called a cluster[15].
It is considered as a null hypothesis that there is no high or low risk cluster, that is, the entire population has the same probability of contracting pulmonary tuberculosis, regardless of its location; while the alternative hypothesis assumes the existence of clusters that are areas in which the population would be more or less likely to contract the disease[14].
Unlike the purely spatial scan that is based on circles, in the space-time scan, cylinders are created around each centroid, in which the base of the circle remains the same, and additionally, the height of that cylinder reflects the period of time considered in the cluster, so as to move through time and space simultaneously[15]. Thus, incorporating time as a variable of interest, it is possible to verify the existence of clusters in a given area, and that in a specific period of time, there was a greater or lesser proportion of cases when compared to the other areas analyzed[15].
Still referring to the analysis of cluster detection, the SVTT technique was also performed, which differs from the other analyses presented by calculating the temporal trend of the clusters[4].
This analysis uses the same circles as the purely spatial scan, however, the SVTT does not seek to identify clusters with a high or low number of occurrences of the event, but verifies whether the time trend of the cases is increasing or decreasing over time[15].
The time trend is calculated inside and outside the scan circle, in which we call the internal temporal trend (ITT) the change in the time trend of the event within a cluster, and the external temporal trend (ETT) as the trend of all other areas that do not belong to the cluster in question. Therefore, what is statistically significant in this analysis are the temporal trends and not the formation of a cluster as in spatial and space-time scanning[4,16].
In SVTT, it is considered as a null hypothesis when there is no difference in the temporal trends in the analyzed areas and we have as an alternative hypothesis that the temporal trends are different.
The parameters used in the analysis of purely spatial scanning, space-time scanning, and SVTT were: discrete Poisson model, no geographic overlap of clusters, circular-shaped clusters, 999 replications in the Monte Carlo simulation, and the size of the exposed population was 8%, value stipulated by the Gini coefficient in which the number of cases is compared to the data of the base population and the expected number of cases in each census sector is proportional to the size of the population at risk[15,17].
In addition, the relative risk (RR) and 95% confidence interval (95% CI) of each cluster was calculated, allowing the comparison of information in different areas, with the exception of SVTT because, as explained above, it is emphasized that what is significant in this analysis are ITT and ETT, so that the RR of the identified cluster may not be within the CI. Clusters with p <0.05 were considered statistically significant.
The analyses were performed using SaTScan software version 9.3, and thematic maps were created using ArcGis software version 10.5.
Descriptive and association analysis
In order to identify factors associated with the epidemiological situation that occurred in the municipality, exploratory analyses (absolute and relative frequencies) were carried out, and then, the association of these variables with the fact of living in a risk area identified through the Pearson's chi-square analysis (X²) and for the variables that were statistically significant (p <0.05), Odds Ratio (OR) and 95% CI were calculated using the IBM SPSS version 25 software.