Technology and Data Collection in Chronic Disease Epidemiology

In this issue of Preventing Chronic Disease, Moodley et al (1) present the results of a spatial analysis of the locations of advertisements for sugar-sweetened beverages (SSBs) and vendors who sell SSBs in relation to the location of schools in 5 neighborhoods in South Africa. In their article, “Obesogenic Environments: Access to and Advertising of Sugar-Sweetened Beverages in Soweto, South Africa,” the authors used a global positioning system (GPS) and a digital camera to gather data on the locations of SSB advertisements and vendors. Their innovative and low-cost approach could be replicated in any setting, including the United States, where time-sensitive point-location data on environmental exposure are needed but are unavailable through more traditional data-collection sources. In this sense, their approach to gathering data is situated within the broader technological developments of volunteered geographic information, crowdsourced data, and GPS-enabled mobile technology for public health (2–6). 
 
Although the main objective of Moodley et al was to provide a descriptive analysis of the intensity of SSB advertising, their approach to using technology deserves to be highlighted because it may be of great value to public health practitioners. To this end, Preventing Chronic Disease readers may find valuable some additional examples of the use of handheld GPS devices or smartphones for data collection for chronic disease epidemiology. Smartphones are GPS-enabled, and photographs taken with smartphone cameras are encoded with a GPS location. Software applications for smartphones that allow photographs to be exported and their location information to be stored on a convenient database include commercial applications such as Collector for ArcGIS (Esri, http://doc.arcgis.com/en/collector/) and open-source free applications such as Ushahidi (www.ushahidi.com/product/ushahidi/). 
 
Many recently published studies illustrate how this technological approach has been used in the field. Braun et al (7) provided a comprehensive review of the use of mobile technology for field data collection among community health workers. Aanensen et al (8) described the development of a system to link smartphones to Web applications for the collection of field data, which can include GPS locations, photographs, videos, and audio. Chunara and colleagues (9) cited examples of mobile technology use for rapid reporting of outbreak information, such as malaria in Cambodia. Patel et al (10) described the development and implementation of a smartphone application to measure the presence of smoking in vehicles, in addition to the presence of adult passengers, child passengers, or both; they also stress the advantages of efficiency and standardization and the ability to transmit data from many remote locations to a centralized website for further analysis. Kanter and colleagues (11) developed, field tested, and evaluated a mobile telephone–based nutrition environment survey in Guatemalan supermarkets, and they noted that the mobile application had equivalent reliability and validity to a paper version of the survey and was also faster to use. King et al (12) reviewed advances in and issues related to using mobile technologies to assess the built environment for the purpose of improving active living and healthy eating. Eyler et al (13), in a presentation of case studies for the assessment of physical activity and the built environment, described the development of an iPad application (named iSOPARC) that enables users to collect and manage data elements for the System for Observing Play and Recreation in Communities (SOPARC), which has been validated and in use since 2004. The goal of the mobile application was to increase the use of SOPARC by making it more accessible to a broader range of end users. 
 
Bethlehem et al (14) discussed a different approach to using digital technology to assess neighborhood obesogenic characteristics; instead of collecting data in the field, they remotely analyzed digital photographs from Google Earth and Google Street View. In both cases, they found that assessments were valid and reliable and could be completed in roughly one-half the time as field-based data collection. One drawback of this approach is that not all potential areas of interest have been imaged for Google Earth and Google Street View, particularly areas where cars are prohibited. Furthermore, the date of image collection is not controlled by the health researcher, so a large study may contain data obtained at different times. Nonetheless, this approach is an interesting area for further development and highlights the innovative ways in which digital data are being used. 
 
Mobile technology is changing rapidly, and so are the innovative applications for using it. Curtis et al (15) collected street-level spatial video data in a Haitian community through the analysis of 4 automobile-mounted digital video cameras. The spatial video was viewed in Google Earth, and environmental attributes of interest (eg, standing water, trash, structural integrity of homes) were manually coded; the resulting data were exported to ArcGIS (Esri) for further spatial analysis. As a sign of applications to come, Igoe et al (16) discussed the feasibility of using smartphones for real-time measurement of ultraviolet A (UVA) radiation and aerosol optical depth, both of which are measures of the physical environment that can affect health. It is not too great a stretch to imagine the near future when our smartphones or wearable technology may be able to measure UVA radiation and provide real-time recommendations for limiting sun exposure. 
 
Moodley et al provide a case study of how a group of researchers with a defined research question for a well-documented public health concern used readily available low-cost technology to create a unique spatial database of environmental exposures. This approach has relevance to many different geographic settings and exposures, including data that may be available from commercial vendors but prohibitively expensive (eg, business and marketing data), as well as exposures that are more ephemeral, such as advertising billboards, where existing data sets may not be current for the period studied. Their data collection methods can be adopted by researchers and communities interested in various chronic disease–related exposures or assets (either harmful or protective), such as alcohol and tobacco advertising, fast-food outlets, and farmers markets. Public health practitioners could either adopt their approach of using a handheld GPS in combination with a digital camera or use similar approaches that are available through ready-made commercial or open-source smartphone applications, as this brief sampling of literature suggests.


Introduction
Rates of obesity and overweight among South Africans are increasing. Food marketing has a profound impact on children and affects their lifelong eating patterns; in urban areas of South Africa, disposable incomes are growing and ultra-processed food is increasingly available at low cost. The combination of these factors will strain an already fragile health system. Our aim was to investigate the density of outdoor sugar sweetened beverage (SSB) advertising and the number of formal and informal vendors selling SSBs in a transforming, historically disadvantaged urban setting of South Africa.

Methods
A digital camera and global positioning system navigation system were used to record the location of SSB advertisements and food vendors in a demarcated area in Soweto. Data were collected by walking or driving through each street; a food inventory was completed for every food vendor. Spatial analyses were conducted using a geographic information system.

Introduction
Non-communicable diseases (NCDs) will be the leading cause of death on the African continent by 2030 (1). Between 1992 and 2005, obesity prevalence increased by 35% in sub-Saharan Africa (2). In 2012, South Africa had an obesity prevalence of 39.2% among females and 10.6% among males (3). The level of obesity among adolescent girls has increased significantly; 25% of female adolescents are overweight or obese (4,5).
Most South Africans have poor dietary habits, many of which start during childhood; black South Africans have the lowest dietary diversity of all South Africans and a higher-than-average sugar intake (3). Analysis of sugar sweetened beverage (SSB) consumption in Soweto indicated that adolescents consume between 1.1 and 1.4 servings of SSBs daily (6). This amount translates into 10 to 12 teaspoons per day, which exceeds the proposed World Health Organization daily recommendation of 6 teaspoons of sugar per day (7). This high sugar intake from a single source significantly increases the risk of developing obesity-related NCDs, especially type 2 diabetes (6). Furthermore, sugar has been implicated as a contributor to obesity (8).
A combination of rising incomes and discretionary spending, coupled with marketing, advertising, and availability of high-energy, processed food and beverages, the biggest source of added sugar, play a role in fostering this trend (3). The increasing consumption of processed products is linked to commercial advertising and their greater availability. Marketing by the food and beverage industry strongly influences long-term food and beverage preferences, and its success relies on children's brand recognition and subsequent preference for familiar brand foods (9)(10)(11).
Neighborhoods of various socioeconomic statuses have different levels of exposure and intensity to advertisements of ultra-processed food and beverage products (10,11). The effect of advertising on creating and promoting an obesogenic environment has been demonstrated in the United States, where a 30% increase in food advertising resulted in increased obesity levels, and every 10% increase in the number of fast food advertisements was accompanied by a 6% increase in the consumption of SSBs (11). The higher density of advertisements of unhealthy foods in low-income areas was accompanied by an absence of exposure to goods and activities that promote healthier lifestyles (11,12).
In addition to selling SSBs, street vendors in South Africa sell other high-sugar-content items, such as candy and deep-fried doughnuts known as "vetkoek" (13). Other easily available street foods include burgers, deep-fried potato chips, and "kotas" (a quarter loaf of bread with a combination of deep fried chips, cheese, and meat fillings) (13). Advertising and access to obesity-promoting beverages and street foods contribute to obesity in South Africa (10).
Limited data are available on the density of outdoor advertising and vendors in South Africa. This study examined 2 aspects of the obesogenic environment in an urban setting in South Africa by exploring the frequency and location of outdoor advertising for SSBs and the proportion of food vendors selling SSBs. The goal was to understand how to best craft advocacy activities that limit the promotion of unhealthy products, particularly in settings in close proximity to schools.

Methods
Soweto is a historically disadvantaged area of Johannesburg that covers more than 200 square kilometers and has a population of 1.3 million. There are 1,776 households and 6,357 inhabitants per square kilometer (14). During the past decade, Soweto transformed economically; by 2013, four large shopping malls and several fast-food chains entered the market. This study covered 5 areas in Soweto: Klipspruit West, Mofolo South, Dube, Meadowlands, and Orlando East. During July and August 2013, data on all outdoor SSB advertising and SSB branding in a 38.3 km 2 area were collected. Ethics approval was granted by the University of the Witwatersrand Ethics Committee. The study did not include human participants, and both the advertisements of SSBs and the food and beverages being sold by vendors were in the public domain.
Data were collected by 3 trained research assistants who either walked or drove through each street in the study area. SSB advertisements and food vendors were identified, and data were collected using 2 separate data coding sheets. In addition to collecting information on the location (global positioning systems [GPS] coordinates were determined using a Garmin Nuvi 30, Garmin Ltd.), type, and size of the advertisement, a digital photograph was taken using a Sony Cybershot DSC-W270 12.1 megapixel camera (Sony Corporation). The study team completed an inventory of all food and drink items for sale from vendors. The type of vendor was noted as informal (ie, temporary building structure), formal (ie, permanent building structure with limited resources, for example, no access to electricity or refrigeration), or a shop (ie, permanent structure with access to electricity and refrigeration), and an inventory of beverages sold at these vendors was recorded.
For the purpose of this study, outdoor advertising was defined as billboards, bus stop advertisements, signs placed along the sidewalk, urban art on streets or buildings, large posters, and signage for restaurants or food vendors. Items excluded were branded clothing, packaging, and taxis and buses (ie, moving targets). Advertisements of SSBs and a combination of SSBs and fast foods were included in the analysis.
Researchers were unable to measure the size of the advertisements, so an estimation of each advertisement size was made. Advertisements were classified as small if their dimensions were less than approximately 70 by 40 centimeters or, in terms of paper and cardboard sizes, between A4 (8.3 × 11.7 in) and A3 (11.7 × 16.5 in). Medium advertisements had dimensions of A0 (33.1 ×46.6 in) and large advertisements had dimensions that were measured in meters (eg, billboards).
The GPS coordinates of SSB advertisements and vendors in the study area were used to create distinct spatial point pattern objects in the R library Spatstat (www.spatstat.org). A point pattern that consisted of all the advertisements and vendors in the study area was also created. Using Spatstat, the intensity (number of points per km) of each point pattern was computed. The intensity of each point pattern formed the outcome variables in the spatial point pattern analysis. To assess the association between the point patterns and the distribution of schools in the study, a covariate that measured distance from any given point in the study region to the PREVENTING CHRONIC DISEASE nearest school was created using the distmap function in Spatstat. We assessed the association between the intensity of each point pattern and the covariates of interest using the Kolmogorov-Smirnov test of goodness-of-fit (15,16).
Homogeneous and inhomogeneous Poisson models were fitted using the R library Spatstat (15)(16)(17)(18)(19), and the Akaike Information Criterion was used as a basis for selecting the best fitting model (20,21). In each model, the distance to the nearest school (z) was the primary covariate considered. Models 1, 2, and 3 were best-fitting inhomogeneous Poisson models that were fitted to the SSB, vendor, and both SSB and vendor point patterns, respectively. A fourth model (Model 4) was an inhomogeneous Poisson model with marks labeled as SSB and vendors.

Results
In total, 145 advertisements for SSBs were identified (Table 1). More than half (53%) of SSB advertisements were found outside houses. Many informal vendors operated stalls from their homes. Nearly two-thirds (62%) of branded SSB advertisements were part of a display sign for a shop name, including branded signs for tuckshops (ie, small shops located in or near a school that sell snacks, candies, beverages, and food items that target children) found outside houses. Half of the primary and high schools (14 of 28) in the sample area displayed advertisements of SSBs on school premises, and 13 of these were branded school signs.
A total of 180 vendors were included in the study area; 27% were informal fast-food outlets, 12% were formal outlets, and 61% were shops. More than 85% of shops stocked SSBs (Table 2). Formal and informal vendors both supplied fast food, although few of these vendors stocked SSBs because of lack of refrigeration.
The findings from the spatial analysis described the intensity of SSB advertisements and vendors in relation to schools. The intensity of the SSB point patterns in the study area of 38.3 km 2 was 3.58 points per square kilometer (Table 3). Figure 1 depicts the density of SSB advertisements and their distances to schools and vendors. The figures indicate 2 school clusters in the northwestern and southeastern parts of the study area. The vendor and SSB advertisements identified were distributed around school "hotspots" (Figure 2).

Discussion
Findings from this study indicate that vendors selling both SSBs and advertisements for SSBs are located in close proximity to primary and high schools in Soweto and that this placement is not random. Approximately each square kilometer contained 1 primary or high school, 4 SSB advertisements, and 5 vendors, 3 of which sold SSBs. The most frequent advertisements were for 1 beverage company. This is the first study of its kind in one of the most densely populated urban areas in South Africa in the process of economic transition. Findings provide an understanding of the obesogenic environment by determining the geospatial intensity and distribution of SSB advertising, as well as the availability of ultra-processed foods and beverages. By strategically positioning advertisements in schoolyards or in close proximity to schools, children are being targeted. In another South African-based study, conducted in Western Cape schools, more than 60% of schools had a branded food or beverage advertisement board displaying the school name (13). Principals of these schools indicated that they had not received any monetary or program support from the sponsoring food and beverage companies, but those advertisements send an implicit message to students and the community (13).
A similar study in Sydney, Australia, found that the most frequent outdoor advertisements in close proximity to schools were for SSBs and alcoholic beverages; 24% of the total number of food advertisements located around these primary schools promoted SSBs (22). In New Zealand, 22% of the outdoor advertisements in close proximity to secondary schools were for SSBs (23). The frequency of these marketing messages influences social norms and promotes the perception that calorie-dense, nutrient-poor beverages and food products are normal (24)(25)(26).
To ensure the development of healthy dietary practices, especially in transforming low-and middle-income households in urban areas of South Africa, resources and efforts should be directed toward preventing obesity in these communities and understanding the causes of social determinants of obesity at a population level (27). Policy makers should consider developing mandatory regula-tions that target advertising in and around schools (28). This type of regulation has a precedent in the example of legislation to restrict tobacco advertising (23). However, an unintended consequence of restrictions placed on outdoor advertising may encourage a switch by industry to other modalities of advertising, such as television or social media (27).
Preventing obesity cannot be solved with a single solution, and managing the obesity epidemic requires efforts at both the population and individual levels (1). The South African Department of Health's National Strategic Plan for Non-Communicable Diseases 2013-2017 calls for intersectoral and multidisciplinary action (1). Unhealthy product promotion should be limited and substituted with the sponsorship of healthy choices (28). A package of interventions directed toward making the environment healthier and making healthy eating a norm is needed (28). One of these interventions includes a ban on advertising ultra-processed products during television viewing hours for children. Others might involve regulating food and beverages in school vending machines. Finally, tuckshops could ensure that healthy, balanced school lunches are provided (29).
People feel most empowered when they make their own decisions. Children in particular are disempowered and are unable to negotiate the advertising content to which they are exposed (25). Ideally, prohibiting advertising of SSBs to children should be voluntary. If this does not occur in a short period, government should consider setting mandatory standards for the marketing of beverages and food to children and adolescents, as is already occurring in many world settings. The South African government released a set of draft guidelines on the labeling and advertising of food and beverages to children in 2014. Our research supplies data that will provide policy makers with evidence as they move forward (29). Given the growing burden of obesity in South Africa and the challenges of losing weight after adolescence, establishing these standards is now a matter of urgency (30).
A limitation of this study was that it focused on exploring the density and nature of outdoor advertising and did not account for other forms of advertising exposure, such as television, radio, telephone messaging, and print media. Further research is needed to develop a comprehensive picture of the exposure of children to advertising in other formal and informal settings and in rural and urban settings. In addition, the study determined the geospatial intensity and distribution of advertising and availability of SSBs but did not determine a causal relationship between these factors and the prevalence of obesity in Soweto. However, other studies have identified the effect of food availability and advertising on consumption patterns. Our findings have implications for policies that regulate SSB advertising, especially in the proximity of schools.