Principal findings
We studied standardized municipal TB cure rates in an area of urban social inequality in Brazil.
Few studies in Brazil have addressed the spatial distribution of endemic diseases such as TB in urban areas. Information on the spatial and temporal spread of these diseases allows understanding the occurrence of these events in the territory. In addition, the description and visualization of the events´ spatial distribution facilitate the identification of their association with local characteristics such as socioeconomic conditions.
Tuberculosis rates showed strong positive spatial autocorrelation and association with socioeconomic and demographic variables.
Mapping of cases is a convenient tool for spatial characterization of TB, the results of which promote a better understanding of TB distribution and of the areas with greatest risk of infection.
The results showed that primary healthcare coverage was related to higher TB cure rates, besides better contact tracing and higher rates of supervised treatment. The authors believe that PHC allows achieving shorter waiting times for treatment and better access to health services and TB preventive measures.
Comparison with other studies
The TB cure rate of 71.57% was lower than in other studies conducted in other regions of Brazil, such as 90.9% TB cure in the state of Maranhão [
40]. However, it is higher than in other studies; Lima et al. (2020) [
41] found a median cure rate of 29.8% among cities in Northeast Brazil. Two other studies identified decreases in TB cure rates in some cities in the state of Sergipe [
4] and in Fortaleza, Ceará [
5], in recent years.
Spatial analysis showed a significant spatial association with TB cure. In addition, the TB cure probability map shows that patients in the South and West Zones of Rio de Janeiro were more likely to achieve TB cure, while those in the North Zone were less likely to achieve cure. This result can be useful for public health policy purposes since it is possible to prioritize this region to improve TB cure in the city. The regions with the highest likelihood of cure may have been associated with higher coverage of the family health strategy, such as in the city’s West Zone, where PHC coverage has exceeded 90% since 2010, in addition to Rocinha with 100% coverage since 2012. Meanwhile, for the rest of the South Zone and the Tijuca neighborhood, the higher TB cure rates are likely due to easier access to other health services, better mean socioeconomic status, and retention of qualified physicians.
As expected, spatial analysis showed an association between TB cure and better socioeconomic conditions, including schooling and income [
11].
There was an apparent paradox between the lower probability of cure in the elderly when analyzing the age variable at the individual level and the higher likelihood of cure with higher elderly rates at the census tract level. This may be explained by the fact that at the individual level, older patients tend to adhere less to treatment due to intrinsic factors, with more adverse effects, increased drug-drug interactions, forgetting to take medication, and lower immunity. On the other hand, patients living in census tracts with higher elderly rates have greater likelihood of cure, probably because this variable represents more structured communities in socioeconomic terms, leading to longer life expectancy, and in temporal terms because they are communities that have been settled longer and have better social support networks, favoring better treatment and thus higher likelihood of cure. Interestingly, elderly rate was the variable with the highest likelihood of cure (OR 9.39) and widest variation (95% CI 1.03–85.26), evidencing the importance of the social context for TB cure.
Several previous studies have evaluated the association between TB incidence and TB mortality and socioeconomic factors [
2‐
4,
6,
7,
9,
41]. One study constructed a personalized social risk indicator for TB. The authors found an association between lower income, poverty, education, and overcrowding and TB mortality [
7].
Household crowding is associated with higher interpersonal contact, thereby increasing the likelihood of
M. tuberculosis transmission [
3]. Uppal et al. (2021) concluded that the overall parameter representing the relative risk of progression to active disease among individuals in crowded homes compared to non-crowded homes was the most influential factor in driving costs and effectiveness [
2]. Silva (2016) found that highest density of cases was strongly associated with higher population density but not with lower income or level of literacy [
6]. Nevertheless, our study did not show a significant association between the number of individuals living in the household and TB cure rate.
Despite the association between race/color and cure in the bivariate analysis, no association was found in the final spatial analysis model. This race/color variable may have been subject to data entry problems. The spatial model demonstrated that the higher the educational level, the higher the likelihood of cure, which is consistent with reports in the literature [
5,
40,
42].
HIV continues to be a major contributor to morbidity and mortality around the globe and remains a public health priority in Latin America [
43]. Positive HIV serology and alcohol abuse were associated with lower likelihood of TB cure. This finding has been reported by other authors such as Uppal et al. (2021) [
2]. Besides the expected lower immunity, these patients may show less adherence, due to drug interaction and greater occurrence of the medication’s side effects [
44‐
47].
The Stop TB Strategy of the World Health Organization (WHO) recommends household contact investigations (HCI) for active screening of TB disease among contacts of smear- positive TB cases [
1]. Saunders et al. (2018) found higher risk of progression to TB disease among close contacts of pulmonary TB cases, but the diagnostic accuracy to predict each outcome is poor [
41]. Another study concluded that proactive social policies and active contact tracing to identify missed cases may help reduce the TB burden in this setting [
5]. Our study found that household contact investigation was associated with higher TB cure.
It is important that governments take responsibility for ensuring universal health coverage as a key element in achieving global goals [
5,
41]. In one study, primary care coverage was inversely associated with TB mortality in children [
9]. Ross et al. (2018) concluded that greater population coverage of Family Health Program teams (PHC) was associated with lower TB and HIV mortality [
11]. In our study, longer time between deployment of family health teams and diagnosis of the disease was associated with higher odds of TB cure, except for the category of 41 months or more, which showed the worst probability of cure among the categories. The data were insufficient to explain this phenomenon, since one would expect that patients covered by PHC for 41 months or more would be more likely to achieve cure. One possible explanation is that the first family health teams in the city of Rio de Janeiro were deployed in socioeconomically vulnerable areas, and despite efforts by the Municipal Health Department, some of these teams remained incomplete for a long time due to shortage of medical staffing.
Strengths and limitations
Our study used comprehensive modeling based on a theoretical model relating lack of TB cure to environmental factors, access to health services, social determinants, and individual factors.
Considering that TB is closely related to socioeconomic factors, we incorporated these variables using data from Brazil’s 2010 population census. However, in some cases our results reflect limited data with imprecise measures because we applied these variables at the census level but the vital records at the individual level.
TB is closely related to housing conditions such as distance between dwellings, number of persons living in the same household, areas of social vulnerability, and sanitation, among others [
3]. The spatial element is thus closely related to tuberculosis. Therefore, spatial analysis is essential in TB statistical analysis. However, most studies evaluating the spatial distribution of TB cases use data from polygon areas. The approach to estimation with small areas is a challenge in spatial models [
11], in which the global spatial autocorrelation index should be analyzed. This can be an important barrier when the prevalence is low in the population, generating many polygons with zero cases. Therefore, it is challenging to use small polygons, closer to the cases occurred, such as polygons of the city’s census tracts, neighborhoods, or regions in a city. Another common and well-known phenomenon in these situations is the oscillation of small numbers, that is, in a small population, a random case ends up generating a high incidence rate.
Our study used a spatial analysis methodology based on geostatistics. We used the points of spatial coordinates represented by the latitude and longitude of addresses for TB cases as the units of analysis, thereby addressing the biases in spatial analysis using data aggregated in polygons.
Nevertheless, the benefits of resolving data aggregation bias in polygons should be weighed against the risk of potentially identifying individuals when analyses of exceptionally rare outcomes are conducted in extremely small areas.
It was also possible to evaluate clusters of new TB cases in the period in different areas of the city, using the Kernel point density method [
22]. This information is useful for identifying areas of greatest vulnerability and population density. The use of generalized additive model (GAM) allowed identifying areas at greater risk of lack of TB cure in the city, including level of confidence and statistical significance. Such information is useful for health system administrators to prioritize areas in the city for intensifying measures to improve TB cure rates.
Finally, GAM allowed the incorporation of the time and space components into the modeling and thus the spatial analysis of PHC performance over time in TB cure. It was also possible to establish the minimum time for deployment of primary care teams to improve the results in TB cure.
The study was subject to various limitations. First, the study design does not allow establishing a causal relationship, and external validity may not be reached. Furthermore, we used data from secondary sources; the data may not be complete, introducing some bias in the study.
Second, some residential addresses were either missing, incomplete, or impossible to geocode, with 24.22% of the reported cases that could not be geocoded. The geocode loss rate in this study is worse than the rate in other studies in Brazil [
6]. Silva et al. (2016) reached 94.6% geocoding in their sample of 387 cases [
6].
This may be explained by the fact that the address is not validated prior to data entry, compromising the record’s quality. Furthermore, there are numerous favelas in the city of Rio de Janeiro, and there are few official records of street names and zip codes in these communities, further compromising the reports´ quality and making geocoding more difficult. Such cases probably occurred in areas with the worst socioeconomic conditions.
This may represent a classification bias, since coverage by primary care did not occur homogeneously in the city, but prioritized areas of greatest social vulnerability and may have presented greater loss of geocoding, compromising the data on primary care coverage.
Third, to avoid selection bias, we attempted to include all reported TB cases in the city during the study period. However, approximately 1600 records were not used, due to lack of information on treatment completion, which may have produced a selection bias if these cases were not randomly distributed in the city.
Finally, only 3 years of TB cases were analyzed. For the study’s results and conclusions to be more robust, a longer historical series would be necessary. This would allow assessing the impact of the consolidation effect in the primary care model on TB cure in the city of Rio de Janeiro.
Implications and future research
Our findings are important for informing Brazilian policy and orienting further research on primary care-based TB treatment. The study allowed evaluating the relationship between PHC coverage and TB cure, based on spatial and temporal distribution. It was also possible to identify risk areas for failure to cure TB in the city of Rio de Janeiro, providing a comprehensive model of TB cure utilizing spatial and temporal components in the analysis.
TB interventions, such as active case tracing and mobile testing units can be resource-intensive and are utilized most effectively when prioritized to high-burden areas
Future studies to assess PHC performance in TB treatment should be implemented with careful consideration of how to address perceived barriers, especially studying a longer temporal dataset in the study. The results may also interest policymakers facing similar decisions in other countries.
Future work may assess whether factors such as treatment-seeking behavior and case-reporting completeness can be used to improve modelling of TB incidence from case notifications. Enhancing the quality of residential address data entry will be useful for all spatial analysis studies using geocoding, which is a huge challenge in these studies.