Box 2. Summary of limitations reported by the selected studies.

Author(s)/ Year	Limitations
Peres et al.21 (2011)	Lack of temporal data on the potential social determinants of homicide decline made it impossible to infer its causes. There were no discussions on the limitations resulting from the social exclusion/inclusion index.
Vilela et al. (2008)22	The limitations considered were intra-aggregate heterogeneity, inter-group mobility, and the underreporting of infant deaths.
Antunes et al. (2008)23	Different ways of measuring variables in statistical censuses in Barcelona and São Paulo. As an ecological study, it does not consider the relevant variation in individual socioeconomic characteristics. Another limitation is the relatively simple analytical scheme, which disregards non-linear relationships between mortality and socioeconomic status.
Silveira and Junger24 (2018)	Use of secondary data is a limiting factor in this study, as is the possibility of ecological studies.
Machado et al.37 (2019)	The short period analyzed.
Schuck-Paim et al.38 (2019)	Despite the synthetic control method used to detect the benefits of pneumococcal conjugate vaccine introduction and explicitly designed to minimize confounding, the ecological study design may have disregarded other uncontrolled factors that can affect the estimates.
Medeiros et al 28 (2012)	Use of secondary data is a limiting factor in this study. Information may not be completely reliable and represents population averages as it is an ecological study.
Ribeiro et al.27 (2007)	Ecological study that had to consider “ecological fallacy.” No individual assessment of socioeconomic status was performed in the study, and the smallest unit analyzed (state) was too large to represent a neighborhood effect.
Drachler et al.34 (2014)	Does not mention limitations.
Silva et al.29 (2008)	The associations found may be stronger due to spatial aggregation (neighborhood). The districts of Recife still have significant social heterogeneity, with wealth areas existing alongside pockets of poverty. Moreover, this study characterized mortality using an indicator created by other authors. Research shows that synthetic indicators do not capture the different nuances of social reality.
Bassanesi et al.30 (2008)	Does not mention limitations.
Campos et al.42 (2000)	Does not mention limitations.
Oliveira et al.43 (2010)	Considering the nature of the aggregate measure, a neighborhood classified with the highest deprivation does not always have the worst rates on all variables analyzed. The high variation in population composition (between 2,500 and 60,000 people) across districts was not considered. Regarding statistical analysis, no mortality smoothing techniques was performed, as it was not possible to assess the effect of deprivation on mortality. Use of the 2000 census to obtain socioeconomic indicators may result in limitations in understanding previous years.
Belon and Barros33 (2011)	As a unit of analysis and area of residence, a limitation of this study is that its results do not necessarily reflect the situation of those belonging to each socioeconomic stratum.
Teixeira et al.25 (2002)	Does not mention limitations.
Macedo et al.26 (2001)	The stratification adopted in the study, although performed by aggregating similar zones, has several limitations due to the particular heterogeneity of the urban area of Salvador. Problems related to the quality of information were also studied.
Bastos et al.32 (2009)	An important limitation of ecological studies is that the relationship between two variables does not necessarily reflect the situation under study. Administrative regions may have caused degrees of heterogeneity due to the specific characteristics of each neighborhood.
Faria and Santana (2016)35	Use of secondary data can be considered a limiting factor in this study.
Lotufo and Benseñor44 (2009)	Does not mention limitations.
Araújo et al.45 (2005)	Lack of data on living conditions disaggregated by neighborhood prevented the generation of a weighted indicator for classification according to more specific sociodemographic variables. Moreover, the quality of violent death records restricted a more comprehensive understanding.
Alves et al.41 (2020)	Limiting factors comprise the use of secondary data and the fact that deaths due to more severe forms of the outcome were not verified.
Guimarães et al.40 (2013)	An ecological design that needed to measure the variables as proxies: income does not directly interfere with colorectal cancer. It can promote conditions to decrease exposure to risk factors, such as diet (primary prevention), and establish early diagnosis (secondary prevention).
Alarcão et al.39 (2020)	Use of secondary data and collinear variables (schooling, income, and employment), which may impair the strength of the association, are limitations in this study.
Bonfim et al.36 (2020)	Given the difference in coverage of the Mortality Information System throughout Brazil, the use of secondary data is a possible limitation in this study.