A Local View of Informal Urban Environments: a Mobile Phone-Based Neighborhood Audit of Street-Level Factors in a Brazilian Informal Community

Abstract

Street-level environment characteristics influence the health behaviors and safety of urban residents, and may particularly threaten health within informal communities. However, available data on how such characteristics vary within and among informal communities is limited. We sought to adapt street audit strategies designed to characterize the physical environment for use in a large informal community, Rio das Pedras (RdP) located in Rio de Janeiro, Brazil. A smartphone-based systematic observation protocol was used to gather street-level information for a high-density convenience sample of street segments (N = 630, estimated as 86% of all street segments in the community). We adapted items related to physical disorder and physical deterioration. Measures selected to illustrate the approach include the presence of the following: (1) low-hanging or tangled wires, (2) litter, (3) structural evidence of sinking, and (4) an unpleasant odor. Intercept-only spatial generalized additive models (GAM) were used to evaluate and visualize spatial variation within the RdP community. We also examined how our estimates and conclusions about spatial variation might have been affected by lower-density sampling from random subsets street observations. Random subsets were selected to determine the robustness of study results in scenarios with sparser street sampling. Selected characteristics were estimated to be present for between 18% (unpleasant odor) to 59% (low-hanging or tangled wires) of the street segments in RdP; estimates remain similar (± 6%) when relying on a random subset created to simulate lower-density spatial sampling. Spatial patterns of variation based on predicted probabilities across RdP differed by indicator. Structural sinking and low-hanging or tangled wires demonstrated relatively consistent spatial distribution patterns across full and random subset sample sizes. Smartphone-based systematic observations represent an efficient and potentially feasible approach to systematically studying neighborhood environments within informal communities. Future deployment of such tools will benefit from incorporating data collection across multiple time points to explore reliability and quantify neighborhood change. These tools can prove useful means to assess street-level exposures that can be modifiable health determinants across a wide range of informal urban settings. Findings can contribute to improved urban planning and provide useful information for identifying potential locations for neighborhood-scaled interventions that can improve living conditions for residents in Rio das Pedras.

Electronic supplementary material

The online version of this article (10.1007/s11524-019-00351-7) contains supplementary material, which is available to authorized users.

Introduction

Neighborhood quality can influence the health and safety of urban residents [1–4]. In a seminal paper from Wilson and Kelling, observed and perceived forms of neighborhood degradation known as physical disorder (e.g., broken windows, abandoned cars, graffiti, litter, etc.) could attract criminal activity. This became known as the “broken windows theory” [5]. Though, the effects from physical disorder can extend beyond public safety concerns within neighborhoods. Historically, neighborhood health studies have incorporated indicators of physical disorder and deterioration used to describe its role in affecting community-level and individual-level health and risk [2, 6–8]. Neighborhood quality as a contextual determinant of health can be investigated through characterizing the distribution of physical and social disorder. As such, indicators of physical and social disorder are fundamental to the empirical understanding of urban neighborhoods [9].

Direct observation is a common approach used to collect community-level data that describe social and physical environmental conditions [10]. A method known as systematic social observation (SSO) collects prespecified observational characteristics of sampled locations such as streets [9]. Systematic social observation protocols, also known as neighborhood audits, are traditionally conducted by raters trained in logging, coding, and scoring information in the format of a survey or checklist [11]. A comprehensive review of SSOs in the context of neighborhoods found that there is a great deal of variability in the methods and instruments used to gather neighborhood observations [11]. As a result, comparability across studies is limited by methodological and analytical heterogeneity including by scale, SSO item definition, and measurement reliability. Data quality problems can lead to incorrect conclusions in determining the contextual role of neighborhood effects on residents. As new methods are developed, transparency in sharing assumptions and operational details can facilitate adoption and further improvements by the research community, or even use by community members themselves.

Assessing neighborhoods through indirect, systematic observational approaches by video recording [12] and web-based maps [13, 14] have emerged as efficient alternatives to field observation especially when covering large expanses. Virtual audits have garnered great popularity for using web-based proprietary mapping products such as Google Street View [15–17] and Google Earth [18] to study major cities and suburban regions.

However, coverage for marginalized areas, such as informal communities, is a potential “blind spot” that lack appropriate digital imagery and scale to conduct virtual audits. The collection of fine-scale intra-urban data to describe street-level conditions is needed to understand variation within and among informal communities, and the implications of such variation for health.

There is, therefore, an opportunity to apply emerging smartphone technology within informal communities to assess informal communities where large-scale digital mapping is not available. The utility of mobile-phone-based survey instruments has been recognized in both clinical and field settings [19–22]. An advantage to using smartphone technology for collecting field data is the simultaneous inclusion of geospatial metadata derived from built-in GPS for observations. Methodologically, in-person SSO data collection is likely a better option for informal communities where remotely sensed or precollected imagery are not adequately available for conducting virtual audits to collect observational data. In-person SSO data collection is better suited to capturing fine-scale and non-visual urban variation such as odor or noise as compared with virtual audits and can capture temporal variation and trends on an investigator-defined schedule. Such information is critical to understanding conditions and in identifying and evaluating changes that can elevate neighborhood-level quality of life. While handheld electronic devices have existed as an alternative tool for the digital collection of observational data [20, 23], studies using a smartphone to conduct SSO studies within an informal community are just beginning to emerge [24].

In this study, we sought to adapt smartphone-based SSO strategies to characterize the physical environment of a large informal community known as Rio das Pedras (RdP) in Rio de Janeiro, Brazil, also known as a favela. Our goal is to demonstrate the utility of using a smartphone-based SSO instrument and leverage the data to describe these limitations by assessing selected street-level characteristics. In this paper, we will describe our primary data collection application, and summary analysis and predictive probability maps based on selected SSO factors. We will focus on decision points and lessons that could shape future data collection efforts.

Methods

Study Area

Rio das Pedras is the third largest informal community in Rio de Janeiro with approximately 63,500 residents. The community has a vibrant local economy made up of approximately 4000 businesses but with limited access to municipal services. The Jacarepaguá lagoon and mangroves bound the geographical area to the southwest and steep hills to the north as seen in Fig. 1. As such, these natural barriers limit its urban growth and horizontal expansion potential. The topography is relatively flat compared to other informal communities in Rio de Janeiro.

Fig. 1.

Satellite image of Rio das Pedras (RdP) with overlaid boundary layer. Image was extracted through Google Maps

Overall, the built environment is composed of multistory buildings with a higher density of persons per square kilometer in comparison to more conventional urban spaces within Rio de Janeiro [25, 26]. Most of the land cover is composed of buildings in varying conditions. Storefronts generally comprise the ground level along larger roads, while small alleys are typical between exclusively residential buildings. Paved streets and sidewalks are in varying conditions with limited interior open spaces. Intermittent heavy tropical rainfall has contributed to seasonal flooding events within the neighborhoods of Areinha and Areal. The lower-lying region is characterized as marshland, and, consequently, predisposed to soil instability. This characteristic has resulted in the prolonged, slow sinking of multi-story buildings, also known as structural non-parallelism, where structural integrity becomes severely compromised.

Such climatic events, in combination with open sewerage release into a local canal, and damp and cool conditions along narrow alleyways, increases the vulnerability to waterborne and vector-borne diseases, and other flood-related injuries [27]. Prior to the neighborhood audit study, the community has undergone incremental and transitional changes such as formalized piped drinking water distribution systems, untreated domestic wastewater conveyance connection to a neighboring canal, and other physical infrastructure improvements.

Street-Level Characterization Approaches

Data for each street were systematically collected using Fulcrum, a mobile software application (app) used for field data collection [28]. Figure 2 illustrates an adapted workflow model that summarizes the utility of Fulcrum into three steps: build, collect, and extend. The “build” phase, as described in this section, relates to the development of the mobile data collection tool. The “collect” phase entails the deployment of the auditors for onsite data collection. Finally, the “extend” phase relates to the compilation and the integration of collected geocoded SSO data, annotated notes, and photos in a single comprehensive file structure that has been cleaned and prepared for data querying, analyses, and mapping. As such, we describe our efforts within three phases: build, collect, and extend the data collection system.

Fig. 2.

Workflow schematic for Fulcrum platform

Build

SSO Street Audit

As part of the Rio das Pedras Community Needs Assessment, a systematic observation protocol previously adapted from Zenk and colleagues [29] used for evaluating neighborhoods in Detroit, Michigan, USA, was modified for use in this study to gather street-level information in RdP. We selected neighborhood characteristics that represented physical disorder and physical deterioration within the community. Street-level items selected for conducting direct observations included, but not limited to the following: odor, litter, structural evidence of sinking, and wire conditions.

The aforementioned SSO items are specific to RdP and aid in characterizing the feasibility for a smartphone-based SSO methodology and in understanding neighborhood-level health implications. Foul odors, while potentially subjective, are a non-visual indicator for biological and chemical pollution sources to residents. The proximity to a raw wastewater-dominated canal and the practice trash burning warrant the inclusion of foul odors as a physical disorder item. Scattered litter and trash represent environmental quality, and potential sources for vectors such as mosquitoes and rats that could promote infectious diseases. Informal communities within Rio de Janeiro are vulnerable to poor urban drainage largely due to debris and trash accumulation and blockage. The presence of standing water and trash can pose as hospitable habitats for transmitting vector-borne diseases such as leptospirosis and dengue [30, 31]. Given its unique biophysical landscape and soil instability, the prevalence of structural sinking is more specific to Rio das Pedras. This physical disorder item is considered to be more particular to informal settlements since construction is not regulated and not typically subject to permitting from city development officials. Commercial and residential building integrity and safety have become severely compromised due soil instability and improvised construction. Thus, injuries and accidental deaths from irregular staircases (tripping/fall hazard) and lower-level flooding can disproportionately impact residential safety [32]. Lastly, illicit wire connections and dangling wires, while commonplace for household electricity, generate concern from possible electrocution and unintended electrical fires. This is a normalized practice within informal communities as demand for electricity within an informal community setting generally increases with population growth [33]. See the Supplementary information for detailed descriptions of each item and defined potential responses. A high-density convenience field sampling strategy was implemented by prioritizing commercial street segments.

Instrument Construction

Data for each street segment were systematically collected using Fulcrum [28]. We developed a customized data entry system through the app’s platform. In addition, a mapping interface was also created to help guide the field team and track the recorded geographic information in near real-time (15-min updates) through Fulcrum’s application programming interface (API). The interface’s base map layer and its labeled streets were generated with the collaboration of community partners.

Team Formation and Training

In addition to technical preparations, logistical plans were also established through on partnerships with colleagues at Oswaldo Cruz Foundation (Fiocruz) and community partner at the Center for Citizenship and Investigation (NUCLEO). Community involvement was critical for building trust within the neighborhood. Auditors were not residents of the area, and their fieldwork involved the use of their smartphones for surveying and for taking photographs. This type of presence warrants suspicious and mistrust among locals. The site is not subject to conventional state-sanctioned peacekeeping and can generate concern with respect to safety and receptiveness for outsiders.

Each auditor independently surveyed one street at a time with no rater overlap. Prior to the finalization of the SSO instrument, two of the auditors were involved in pilot testing an earlier version of the SSO instrument and provided feedback for modifications in March of 2015. Each auditor underwent mandatory training before data collection. As part of the training, auditors reviewed the neighborhood evaluation protocol and related materials describing the physical landscape in RdP. The training SSO handout can be found in Appendix (Appendix material 1). A team of three auditors covered RdP neighborhoods during the late fall between 5 May 2015 to 15 May 2015.

Auditors were typically a distance of one or two street segments apart during data collection, allowing for brief check-ins to work toward consistency and address other issues arising in the field. While in the field, auditors identified street segments as areas between intersections or other major street transitions (e.g., dead ends, sharp curves, etc.). Longer, street segments were broken into more than one segment for data collection.

Collect

Data Collection

Full coverage of commercial streets was prioritized to fully capture pedestrian-based travel, expected to be concentrated on commercial streets. The field team divided the community into two major neighborhoods (Areal and Areinha) based on input from NUCLEO study partners. Those neighborhoods are largely embedded with retail activity. Non-commercial streets were sampled with a pattern that was determined informally because of challenges from unanticipated field conditions. However, the goal was to spread out the non-commercial street observations for each area. For safety reasons, the team halted data collection before sunset and resumed coverage the following day.

An auditor completed the SSO audit and took photographs of potentially health-relevant features within the street segment for each sampled street segment. Each completed audit was recorded and geotagged as a single data point onto the underlying map layer which with location placement assisted by the global positioning systems (GPS) typically embedded in smartphones. A manual override option was available in the instrument for correcting positional errors common due to street canyons and multipath error [34]. The labeling of street names from our digitized RdP base map aided in verifying point positions after an individual street audit. Photographs for each recorded street segment were also geotagged. Field data collection spanned 10 days with three rained out days. Each auditor used their personal iPhone smartphones and downloaded the Fulcrum application to conduct audits.

Extend

Sampling Map Boundary

We created a map boundary specific to the study. The boundary is a hybrid between a geopolitical boundary of RdP and a convex hull of amassed data points representing observed street segments. This boundary map excludes the “empty space” of the northeast region where no data was collected. At the time of sampling, this steeply sloped region was mostly comprised of older buildings with ongoing re-development and construction. After primary data collection, the map boundary determined usable surveyed streets that laid within the RdP boundaries subject and were included for analysis as shown in Fig. 1.

Analytic Approach

Observed street segments are represented as discrete data points within RdP. Thus, relevant SSO data for each street segment was extracted and described in the analysis. In addition, we determined design weights to each street and applied heavier weights to non-commercial streets (1.00) since commercial streets (0.34) were emphasized for data collection. The design weight for non-commercial streets was set to 1, and design weight estimates for commercial street were set to the inverse of its sample probability within its respective sample size. Generalized additive models (GAM) [35] were fitted to selected SSO items to evaluate and visualize the spatial distribution of SSO items within the community. Selected SSO items were originally dichotomous or collapsed to dichotomous. See Appendix material 2 for the formal definitions of selected SSO items and for commercial streets used in the audit. For each SSO item, we fit an intercept-only logistic GAM model with a bivariate spline of the longitude and latitude coordinates to estimate and interpolate spatial trends. This statistical approach has been described and applied to similarly designed studies within Rio de Janeiro for collecting geocoded biological specimens [36, 37].

We generated predictive probability maps by interpolating geographical coordinate values within RdP. p values were based on spatial GAM fits intercept (log-odds) estimates were converted to prevalence estimates for each SSO item.

In the interest of informing the density of sampling for future fieldwork, we simulated reductions in sample size by subsetting the original SSO data set. Random subsets were selected at 75%, 50%, and 25% of the original sample size to determine overall sample size stability. This approach allows for an exploration of how lower density sampling would have affected our findings. Spatial GAM regression was used to generate probability maps for selected SSO items in the full sample and for the random subsets. We used R version 3.3.2 for statistical analyses and spatial visualization [38]. All statistical tests were two-tailed and based on an alpha of 0.05.

Results

Neighborhood auditors surveyed a total of 643 street segments (86% of all street segments in the community). Six hundred thirty data points fell within our study region after layering our RdP boundary map for analysis. As such, our sample sizes for 75%, 50%, and 25% subsets were 472, 315, and 157 points, respectively. Figure 3 depicts the original sampled data set and its subsets of geocoded data points of sampled streets within a defined RdP boundary.

Fig. 3.

Maps of Rio das Pedras and sampled points

Table 1 presents proportions for dichotomized SSO items across sampled street segments. Items are reported with respect to the presence of diminished qualities related to physical disorder or deterioration (e.g., foul odor, trash/litter, structural non-parallelism, and hazardous wires) to assure consistency and comparability. Across lower-density subsets, SSO item prevalence was estimated to be within 6% of the estimate from the full set of observations. Among SSO items for physical disorder and deterioration, hazardous wire conditions across all street segments had the highest proportion (0.59) followed by the presence of litter and trash (0.56) within RdP. The presence of unpleasant odor (0.18) and structural non-parallelism (0.21) had the lowest observed street segment proportions.

Table 1.

Prevalence for street characteristics across sample sizes

	Full sample	Random subsets			Commercial streets only
	All (N = 630)	75% (N = 472)	50% (N = 315)	25% (N = 157)	N = 470
Hazardous wire conditions	0.59	0.62	0.63	0.62	0.59
Trash/litter present	0.56	0.55	0.54	0.54	0.56
Structural non-parallelism present	0.22	0.21	0.21	0.22	0.20
Unpleasant odor present	0.18	0.18	0.20	0.24	0.17

Appendix Fig. 1 displays estimated unadjusted global probabilities, 95% confidence intervals, and significance for a spatial spline term for selected SSO items across the full sample and random subsets. Averaged probability estimates exhibited nearly consistent estimations across sample sizes. However, p values for the bivariate spline term “s(long, lat)” did vary. Significance tests at p < 0.05 determined whether the sampled surface is distinguishable from a randomized distribution of events.

Predictive maps are shown across selected items and sample sizes in Fig. 4 with black and red colors indicating highest and lowest probabilities, respectively. Spatial predictions for selected SSO items are based on weighted probabilities derived from estimated unadjusted odd values. We selected trash, odor, structural non-parallelism, and hanging wires, as items that represent physical disorder and deterioration specific to RdP. We observed that surface variation differed between SSO items. Within SSO item subsets, probability maps for unpleasant odor and trash presence exhibited differing spatial patterns. This suggests that decreased sample sizes may not adequately capture an accurate representation of its spatial variation for those SSO items. Structural non-parallelism and hazardous wires exhibited highly similar probability patterns across its sample sizes. A southwest-to-northeast gradient for structural non-parallelism spatial probabilities was consistently observed across each sample size. The highest probabilities of sinking buildings were concentrated in southwestern RdP, an area considered as lower-lying floodplains that were inhabited more recently and that has historically been subjected to seasonal flooding events. And, the increased probability of dangling, loose wires was mostly concentrated in northeastern RdP. This characteristic coincided with urban revitalization efforts was undertaken by a private-public electrical utility provider that services Rio de Janeiro known as Light [39]. Electrical infrastructure improvements included formal electrical connection, removal, and re-bundling dangling wires (commonly known as gatos) within informal communities to improve physical infrastructure quality and to prevent illicit electrical siphoning and electrocutions [33]. Sinking buildings and hazardous wires appear to maintain its respective spatial distribution signatures despite their varying sample sizes.

Fig. 4.

Probability maps of featured physical disorder/deterioration items based on all and subset samples. From top to bottom: presence of unpleasant odor, structural non-parallelism, trash, and hazardous wires. Probability values range between 0 to 1 for each plot. Panels with asterisks denote statistically significant spatial dependency (p < 0.05) with respect to longitude and latitude spline term “s(long, lat)”

Discussion

In this feasibility scale study, we demonstrated the utility of a smartphone-based SSO instrument used for direct observations of neighborhoods at the street level within an informal community. Overall, the experience showed that it is possible to gather spatial data by exploiting embedded GPS technology from smartphones. A benefit from using Fulcrum or other smartphone-based platforms (e.g., Open Data Kit [40]) is the incorporation of geocoded metadata to neighborhood audit information (e.g., street names as received from our community partner organization). Given the increasingly ubiquitous presence of smartphone technology, conducting neighborhood audits and other field-based urban studies shows promise. Future directions may incorporate short-term or long-term longitudinal-based data collection designs.

We were able to detect spatial variation for physical disorder and deterioration. We also noted that lower density sampling might have been sufficient to estimate proportions within the original SSO sample set and subsets for only selected items. Whereas for detecting spatial patterns and dependencies, we observed no significant spatial gradients and simpler patterns at lower simulated sampling densities. The decrease in sample size can likewise result in imprecise regression estimates as seen with hazardous wire presence at 25% subset where estimations demonstrated evidence of statistical instability. Thus, needed sampling density will depend on the goals of a study, specific measured SSO items, and the complexity of their spatial patterns. In addition, clustering of specific features can influence the necessity for increased sample sizes when compared to sampled streets with smooth trends or significant spatial dependence. Clustering attributed to odor (e.g., canal, sites of burning trash) is likely attributed to proximity to localized sources such as untreated sewage or burning trash from the canal, whereas the spatial clustering of structural sinking is likely explained by soil instability in the flood-prone southwest region of RdP. As shown in Fig. 4, a reduced sample of 115 streets would not have captured such an important spatial feature.

This study is novel in adopting an SSO to an informal community to explore its urban landscape features. The work also extended the traditional scope of a neighborhood audit by presenting estimated probabilities of SSO characteristics specific to an informal community across space. Predictive maps aided in visualizing spatial variation and gradients across RdP. Our initial findings suggest that reduced sample sizes could be considered for assessing prevalence or simple spatial patterns. When prevalence is of primary interest for future work in drawing spatial-based inferences, we note that street-level assessments are advantageous in representing the spatial variation of environmental hazards and resources for neighborhoods when compared to formalized, aggregated units of observations such as census blocks or administrative districts [25, 37, 41].

In addition to understanding spatial variation of environmental exposures, there can be gained knowledge of health-related risks specific to RdP based on the spatial distributions of selected physical disorder items. For example, the presence of dangling, unbundled wires, a potential proxy for injury risks from electrocution, exhibited a higher probability within the northeastern region of RdP when compared to other regions. This spatial characteristic may imply that the non-commercialized region of RdP might be more burdened with potential risks from dangling wires. The distribution maps generated in Fig. 3 highlighted regions that might be vulnerable to increased health-related risks from physical disorder presence. In future studies, data collection on health outcomes among residents residing in audited streets should be considered to estimate spatially distributed associations between street-level physical disorder items and selected health outcomes within RdP. There are ongoing efforts to harmonize health data to city, sub-city, and neighborhood characteristics using physical and social environment data within Latin America [42]. However, to detect important disorder-health correlations, one would likely need to cover a broader spatial area to have sufficient variation and statistical power or have outcomes that are fairly proximal (such as perceived safety). This work serves as a prelude to estimating street-level spatial-based exposure-health association analysis using SSO audit data.

Limitations

A limitation of this study is that there was no overlap in the street segments rated by auditors. Future work is needed to provide estimates of inter-rater and test-retest reliability to examine sources of observer bias attributed to the neighborhood audit. Inter-observer and intra-observer biases are common measurement bias most seen in systematic social observation methods [29, 43]. Another future direction is to use the collected photographs from street segments to verify the sighting of specific items. This approach could be conducted as a post-test to discern potential observer biases as discussed in the literature [15, 43] and to help further classify environmental features potentially. Future work can also focus on developing image recognition algorithms that can automate the detection of items through the use of photographs and map out its presence, thus simplifying the work in the field. Also, there is potential for incorporating photos into data quality monitoring and on-going trainings for future neighborhood audits.

Another limitation relates to the representation of characteristics as attributes of observed street segments. Future work could gain precision by geocoding individually observed SSO factor within each street segment. This could be particularly advantageous where street segments vary in length.

Smartphone battery life posed a logistical inconvenience during data. The length of time while in the field was dependent on the auditor’s battery life of their personal smartphones. At the time, the auditors did not carry back-up batteries and had to retreat to NUCLEO’s community office to recharge. We recommend in any future work to bring back-up batteries in addition to making prior arrangements with study partners to rest and to recharge their smartphone devices.

A technical shortcoming with smartphone-based apps was incorrectly positioned and geotagged sampled points due to street canyons and multi-path errors. As a group, we had to adapt our SSO protocol and verify each recorded geotagged points and make any positional corrections after each audited point. In future work, researchers should anticipate potential geotagging errors when characterizing streetscapes. As a general rule, researchers should pre-test their smartphone devices using their survey instruments in various urban settings (e.g., open air, low-rise building, mid-rise buildings, high-rise buildings, etc.) before deployment into the field to gain an understanding of positional accuracy and precision. Also, researchers should be familiarized in approaches to correct mispositioned points during data collection and use that knowledge to train data collectors.

Strengths

A benefit from using an app-based instrument for neighborhood audit is that the tool streamlined systematically collected field data. Also, the tool enabled the research team to forgo the two-step process of written recording (e.g., pen and paper) and later data entry. The incorporation of photos illustrated measured and unmeasured street characteristics for each street segment and was useful in verifying potential differences in interpretation of items across reviewers.

Another strength is the relatively high-density convenience sampling design we adopted for in-person measurements. This level of depth in capturing street-level neighborhood conditions can better characterize intra-urban variation for describing community health risks and resources. We were able to estimate spatial patterns by depicting the distribution of hot and cold spots of selected SSO items through probability maps. In future work, the collection of additional SSO surveys across RdP can potentially serve as tools for public health surveillance or identifying the prevalence of certain risk factors across space and over time.

Conclusion

Smartphone with GPS technology is an attractive tool for conducting field-based urban health studies. This can be applied even within informal communities that may otherwise be “blind spots” when using administrative data or when conducting virtual audits (which did not have coverage beyond the RdP’s main bus route) to characterize an informal community. With these tools, we were able to visualize variability in the built environment, and additionally capture geotagged photos to understand the context further. This approach has potential in identifying street-level exposures that vary within and between informal communities that can affect the health of urban residents. However, it is important to note that the distribution of particular street-level characteristics could drastically vary with respect to sample sizes and sampling locations. This approach has potential in identifying street-level exposures that vary within and between informal communities that can affect the health of urban residents. Future directions may incorporate short-term or long-term longitudinal-based data collection designs.