Food Vendors and the Obesogenic Food Environment of an Informal Settlement in Nairobi, Kenya: a Descriptive and Spatial Analysis

Abstract

Food environments of urban informal settlements are likely drivers of dietary intake among residents of such settlements. Yet, few attempts have been made to describe them. The objective of this study was to characterize the food environment of a densely-populated informal settlement in Nairobi, Kenya according to the obesogenic properties and spatial distribution of its food vendors. In July–August 2019, we identified food vendors in the settlement and classified them into obesogenic risk categories based on the types of food that they sold. We calculated descriptive statistics and assessed clustering according to obesogenic risk using Ripley’s K function. Foods most commonly sold among the 456 vendors in the analytic sample were sweets/confectionary (29% of vendors), raw vegetables (28%), fried starches (23%), and fruits (21%). Forty-four percent of vendors were classified as low-risk, protective; 34% as high-risk, non-protective; 16% as low-risk, non-protective; and 6% as high-risk, protective. The mean distance (95% confidence interval) to the nearest vendor of the same obesogenic risk category was 26 m (21, 31) for vendors in the low-risk, protective group; 29 m (25, 33) in the high-risk, non-protective group; 114 m (88, 139) in the high-risk, protective group; and 43 m (30, 56) in the low-risk, non-protective group. Clustering was significant for all obesogenic risk groups except for the high-risk, protective. Our findings indicate a duality of obesogenic and anti-obesogenic foods in this environment. Clustering of obesogenic foods highlights the need for local officials to take action to increase access to health-promoting foods throughout informal settlements.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11524-022-00687-7.

Introduction

Since the late twentieth century, the prevalence of overweight/obesity and non-communicable diseases, including cardiovascular disease, cancer, and diabetes mellitus, has risen steadily in low- and middle-income countries (LMICs), while morbidity and mortality from undernutrition have declined [1]. This epidemiologic transition is due, in part, to a concurrent nutrition transition, during which the prominence of traditional dishes composed of whole, minimally-processed foods and ingredients has waned, replaced by highly-processed foods with large amounts of saturated fat and added sugars [2], such as fast foods, packaged snacks, and sugar-sweetened beverages [3]. This shift in diet composition is a major contributor to the growing prevalence of overweight and obesity in LMICs [2].

Changes in the availability, accessibility, affordability, convenience, and desirability of foods—dimensions of the food environment—may contribute to shifting diets among populations in LMICs [4, 5]. Although studies of food environments in LMICs remain few in number, several have reported relationships between one or more dimensions of the food environment and dietary intake [6–12] and others have found an association between the food environment and overweight/obesity [13–15].

Only a handful of studies have examined food environments in Sub-Saharan Africa (SSA) [16]. In particular, little is known about food environments of urban, informal settlements in SSA, despite their role as a home to more than half of the urban population in the region [17]. In Nairobi, Kenya, it is estimated that 60% of the population—roughly 2.5 million people—live in informal settlements [18], faced with limited access to clean water, inadequate sanitation and housing, high unemployment, and food insecurity [17, 18].

Food environments of Nairobi’s informal settlements are likely to be a key determinant of diet quality [19] and, consequently, of the prevalence of overweight/obesity among slum residents, which has been estimated to be approximately 25% [20–22]. Slum dwellers frequently have limited income for travel outside of the informal settlement to purchase food [19, 23, 24] and few resources (e.g., land, tools, water) with which to produce their own food [25], which may increase their dependence on foods sold within the informal settlement.

The development of strategies for improving the healthfulness of foods sold in Nairobi’s informal settlements and reducing the burden of overweight/obesity among slum residents will require knowledge of the foods that are present and their obesogenic properties. Additionally, it will require an understanding of the types of vendors whose products pose the greatest obesogenic risk to slum residents and the spatial patterns in which these vendors are arranged, as the replacement of the obesogenic foods purveyed by these vendors with anti-obesogenic foods may yield the greatest improvements in diet quality. To date, studies of food environments of Nairobi’s informal settlements have been limited to descriptions of the availability of broad food groups in select villages within one or two informal settlements [26–28]. Given the small scale and high mobility of many vendors in these settings [27, 29], there is great potential for hyperlocal variation in food environments, which previous studies are unlikely to have captured. Furthermore, these studies did not consider the propensity of the available foods to promote overweight/obesity, the most prevalent form of malnutrition in Nairobi’s slums [20–22]. To address these gaps, we sought to characterize the food environment of a densely-populated, informal settlement in Nairobi, Kenya according to the obesogenic properties and spatial distribution of its food vendors.

Methods

Study Site

Kibera is an informal settlement composed of 15 autonomous villages in Nairobi County, Kenya, located 5 km southwest of Nairobi city center (Supplemental Fig. S1-S3) [29]. With nearly 1000 inhabitants per hectare and a population of more than 280,000 residents, Kibera is one of the most densely-populated informal settlements in SSA [30]. Similar to informal settlements elsewhere in the region, Kibera is characterized by extreme poverty, poor health, and a lack of many formal basic services, like running water [18]. We selected Kibera as the study site because of the longstanding presence of CFK Africa (https://cfkafrica.org/) in this community. CFK Africa (formerly, Carolina for Kibera) is a non-governmental organization with which several members of our study team are affiliated. Having implemented health, education, and empowerment programs in Kibera since 2001, CFK Africa is well-respected among Kibera residents, including food vendors. Many vendors in informal settlements are unable to obtain business licenses and may be unwilling to engage with researchers out of fear of jeopardizing their livelihood [27, 31]. Our study team’s reputation in Kibera helped to ensure that vendors were sampled as comprehensively as possible.

Data Collection

From July 15 to August 9, 2019, we identified 524 vendors. Data were collected daily between 8:00 and 16:00 local time by two research assistants using the mobile data-gathering platform, EpiCollect5 (https://five.epicollect.net/). Information collected included global positioning system (GPS) location, type of vendor, and types of foods and beverages sold by the vendor (hereafter, simplified to “foods” or “food products”). Vendor type and types of food sold were collected through direct observation, with permission from vendors; no vendors refused. While our goal was to identify vendors across the entire informal settlement, we prioritized areas that were known to be highly trafficked, as this strategy was expected to yield the largest sample. Data collection occurred during peak business hours, and few vendors were expected to be missed due to operating outside of the 8:00–16:00 time frame.

Classification of Vendors, Types of Food Sold, and Obesogenic Properties of Foods

We classified vendors into five categories: hawker, kiosk, stand/table, market stall, small restaurant. A description of each type of vendor is provided in Supplemental Table SI. This classification system was similar to those used in previous studies of urban food environments in SSA [32–34].

Next, we classified foods into 27 discrete categories, which were comparable to those used in similar studies [32, 33]. These categories and examples of foods in each are provided in Supplemental Table SII.

Finally, we classified foods according to their obesogenic properties, or their propensity for elevating risk of obesity. To do so, we used an approach developed by Kroll et al. [33], leveraging the NOVA classification system, which places foods into four classes based on degree of industrial processing [35, 36]. Class 1 foods are unprocessed or minimally processed and include fresh, frozen, or dried fruits, vegetables, grains, legumes, fish, poultry, meat, and dairy. Class 2 foods are processed culinary ingredients and include raw sugar, salt, oils, and other ingredients derived from plants or other natural sources. Class 3 foods are processed foods, including simple breads, cheeses, canned foods, and other foods manufactured with processed culinary ingredients. Finally, class 4 foods are ultra-processed foods composed of multiple ingredients, including additives, such as artificial colors, sweeteners, and emulsifiers, that are designed to imitate naturally-derived foods or ingredients [36, 37].

A growing body of evidence indicates that consumption of ultra-processed foods and certain processed foods is positively associated with the risk of obesity [38, 39]. By this rationale and in alignment with the method developed by Kroll et al. [33], we categorized vendors according to the obesogenic (high-risk vs. low-risk) and anti-obesogenic (protective vs. non-protective) properties of the foods that they sold. Foods for which the evidence was ambiguous, such as unprocessed dairy and fresh or cooked poultry and red meat, were deemed neither protective nor high-risk, following the example set by Kroll et al. [33]. Class 4 foods, class 3 foods high in salt, sugar, or saturated fats, and raw sugar, a class 2 food, were considered high-risk; all other foods were considered low-risk. Class 1 foods other than staples, class 2 foods other than raw sugar, and class 3 foods low in salt, sugar, and saturated fats were considered protective; all other foods were considered non-protective. Vendors for which high-risk foods comprised ≥ 1/3 of the total number of food types sold were classified as high-risk; the remaining vendors were classified as low-risk. Vendors for which protective foods comprised ≥ 1/3 of the total number of food types sold were classified as protective; the remaining vendors were classified as non-protective. From these two binary variables, we created a single, four-category variable with the following categories (Supplemental Table SIII): (1) low-risk, protective; (2) low-risk, non-protective; (3) high-risk, protective; and (4) high-risk, non-protective. Notably, our use of a proportion for the threshold differed from the method described by Kroll et al., who used an absolute number for the threshold [33]. We have observed that vendors in Kibera frequently sell a small number of distinct food types, with many selling only one or two. Consequently, use of an absolute threshold would have introduced non-positivity, as vendors selling a total number of food types below that threshold always would have been classified as low-risk and non-protective.

Missing Data

Several vendors had missing information because of unintended omissions during data entry or data loss on the EpiCollect5 storage platform. We do not expect that missingness was differential according to vendor type or foods sold. Vendors with missing information for foods sold (n = 68) were excluded because we were unable to classify them according to obesogenic risk. For vendors with missing GPS data (n = 5), longitude and latitude were calculated as the average of the longitude and latitude of the vendors sampled immediately before and after, as vendors were frequently positioned close to one another and sampled consecutively. Two vendors with missing GPS data were not sampled at the same time as other vendors; for these, longitude and latitude were assigned as the approximate geographic middle of the village from which they were sampled. For vendors with missing information about vendor type (n = 2), two authors (KRB and RL) reviewed the foods sold by the vendor and assigned a vendor type, as there were distinct patterns in the types of food sold across the vendor types.

Statistical and Geospatial Analysis

We calculated descriptive statistics to summarize the number and types of food sold, overall and by vendor type and obesogenic risk category. To identify spatial patterns, we plotted vendors on a map of Kibera, according to vendor type and obesogenic risk category. We calculated the means and 95% confidence intervals (CI) of the shortest distance between any two vendors, between vendors of the same type, and between vendors of the same obesogenic risk category. To determine whether vendors clustered with those of the same obesogenic risk category, we used Ripley’s univariate K function, which indicates whether the spatial arrangement of a set of points differs from the arrangement expected under the assumption of complete spatial randomness (CSR) [40]. Vendors in a given obesogenic risk category were considered to be clustered if the K(r) function exceeded the 95th percentile of the Monte-Carlo confidence envelope corresponding to the Poisson process representing CSR [40]. We performed 100 Monte Carlo simulations to calculate the CSR confidence envelope and corrected for potential edge effects (i.e., bias introduced when vendors just outside of the study area are unobserved) [40]. We obtained geographical boundaries of Kibera from Map Kibera (https://mapkibera.org/). Maps and distance estimates were generated in QGIS 3.16. Descriptive statistics and univariate K analyses were calculated in SAS 9.4.

Results

Vendor Characteristics

After excluding vendors with missing information for foods sold, the analytic sample included 456 vendors. A majority of vendors were stands/tables (29.8%) or kiosks (28.9%), while a smaller number were market stalls (17.1%), hawkers (12.5%), or small restaurants (11.6%). The median number of foods sold was highest among small restaurants (7; interquartile range, IQR: 5, 9) and kiosks (5; IQR: 3, 7), while hawkers (median = 1; IQR: 1, 2) and market stalls (median = 1; IQR: 1, 4) sold the fewest number of foods (Table 1).

Table 1.

Types of food sold and obesogenic risk category of vendors, overall and according to vendor type; Kibera informal settlement (Nairobi, Kenya), July–August 2019

				Vendor type
	All n (%)	Hawker n (%)a	Kiosk n (%)a	Stand/table n (%)a	Market stall n (%)a	Restaurant n (%)a
Total	456	57 (12.5)	132 (28.9)	136 (29.8)	78 (17.1)	53 (11.6)
No. of foods sold, median (IQR)	4 (1, 6)	1 (1, 2)	5 (3, 7)	4.5 (2, 6)	1 (1, 4)	7 (5, 9)
Food type
Vegetables, raw	129 (28.3)	1 (1.8)	27 (20.5)	92 (67.6)	4 (5.1)	5 (9.4)
Vegetables, cooked	36 (7.9)	5 (8.8)	1 (0.8)	0 (0.0)	4 (5.1)	26 (49.1)
Fruits	98 (21.5)	2 (3.5)	22 (16.7)	71 (52.2)	3 (3.8)	0 (0.0)
Meat/poultry, fresh	48 (10.5)	0 (0.0)	2 (1.5)	6 (4.4)	40 (51.3)	0 (0.0)
Meat/poultry, cooked	35 (7.7)	0 (0.0)	2 (1.5)	0 (0.0)	4 (5.1)	29 (54.7)
Fish, fresh	12 (2.6)	0 (0.0)	1 (0.8)	11 (8.1)	0 (0.0)	0 (0.0)
Fish, cooked	26 (5.7)	0 (0.0)	0 (0.0)	1 (0.7)	2 (2.6)	23 (43.4)
Meat/poultry/fish, fried or processed	29 (6.4)	9 (15.8)	9 (6.8)	6 (4.4)	3 (3.8)	2 (3.8)
Eggs	19 (4.2)	1 (1.8)	11 (8.3)	5 (3.7)	0 (0.0)	2 (3.8)
Staples, raw	62 (13.6)	1 (1.8)	10 (7.6)	32 (23.5)	18 (23.1)	1 (1.9)
Staples, cooked	27 (5.9)	1 (1.8)	2 (1.5)	0 (0.0)	2 (2.56)	22 (41.5)
Legumes/pulses, raw	26 (5.7)	0 (0.0)	6 (4.5)	7 (5.1)	13 (16.7)	0 (0.0)
Legumes/pulses, cooked	41 (9.0)	4 (7.0)	2 (1.5)	0 (0.0)	4 (5.1)	31 (58.5)
Nuts/seeds	5 (1.1)	0 (0.0)	0 (0.0)	3 (2.2)	2 (2.6)	0 (0.0)
Dairy, unprocessed	35 (7.7)	0 (0.0)	28 (21.2)	2 (1.5)	5 (6.4)	0 (0.0)
Dairy, processed	6 (1.3)	0 (0.0)	5 (3.8)	0 (0.0)	1 (1.3)	0 (0.0)
Beverages, sweetened	75 (16.4)	2 (3.5)	42 (31.8)	2 (1.5)	5 (6.4)	24 (45.3)
Beverages, unsweetened	19 (4.2)	0 (0.0)	2 (1.5)	1 (0.7)	3 (3.8)	13 (24.5)
Starches, fried	107 (23.5)	34 (59.6)	11 (8.3)	9 (6.6)	9 (11.5)	44 (83.0)
Sweets/confectionary	134 (29.4)	7 (12.3)	83 (62.9)	13 (9.6)	6 (7.7)	25 (47.2)
Traditional mixed dishes	61 (13.4)	4 (7.0)	1 (0.8)	1 (0.7)	9 (11.5)	46 (86.8)
Fats/oils	26 (5.7)	0 (0.0)	22 (16.7)	2 (1.5)	2 (2.6)	0 (0.0)
Bread	56 (12.3)	0 (0.0)	51 (38.6)	2 (1.5)	2 (2.6)	1 (1.9)
Herbs/spices	30 (6.6)	0 (0.0)	11 (8.3)	16 (11.8)	2 (2.6)	1 (1.9)
Condiments/sauces	1 (0.2)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (1.9)
Baking ingredients, raw	40 (8.8)	1 (1.8)	31 (23.5)	2 (1.5)	6 (7.7)	0 (0.0)
Other mass-produced foods	33 (7.2)	0 (0.0)	28 (21.2)	2 (1.5)	2 (2.6)	1 (1.9)
Obesogenic risk category
Low-risk, protective	199 (43.6)	11 (19.3)	29 (22.0)	106 (77.9)	20 (25.6)	33 (62.3)
Low-risk, non-protective	75 (16.4)	2 (3.5)	11 (8.3)	14 (10.3)	47 (60.3)	1 (1.9)
High-risk, protective	26 (5.7)	2 (3.5)	6 (4.5)	2 (1.5)	1 (1.3)	15 (28.3)
High-risk, non-protective	156 (34.2)	42 (73.7)	86 (65.2)	14 (10.3)	10 (12.8)	4 (7.5)

IQR interquartile range

^aPercentages are column percentages, except for those in the first row, which are row percentages; percentages may not sum to 100% because of rounding

The most common types of food sold among vendors were sweets/confectionary (29.4%), raw vegetables (28.3%), fried starches (23.5%), and fruits (21.5%), while the least common were unsweetened beverages (4.2%), eggs (4.2%), fresh fish (2.6%), processed dairy (1.3%), nuts/seeds (1.1%), and condiments/sauces (0.2%).

We observed differences in the most common types of food sold across the five types of vendors. Among hawkers, the most common type of food sold was fried starches (59.6%), and among kiosks, the most common type of food sold was sweets/confectionary (62.9%). Among stands/tables, market stalls, and restaurants, the types of food most commonly sold were raw vegetables (67.6%), fresh meat/poultry (51.3%), and traditional mixed dishes (86.8%), respectively.

With respect to obesogenic risk, most vendors were classified as low-risk, protective (43.6%), or high-risk, non-protective (34.2%). Few vendors were classified as high-risk, protective (5.7%). The proportion of vendors classified as high-risk, non-protective was highest among hawkers (73.7%), and kiosks (65.2%), while the proportion of vendors classified as low-risk, protective was highest among stands/tables (77.9%) and restaurants (62.3%). Among market stalls, the obesogenic risk category containing the largest proportion of vendors was low-risk, non-protective (60.3%).

The types of food most commonly sold among vendors in each obesogenic risk category are shown in Table 2. Among vendors classified as low-risk, protective, the median number of foods sold was 6 (IQR: 4, 8), and the most common types of food sold were raw vegetables (62.8%) and fruits (48.7%). Among vendors classified as low-risk, non-protective, the median number of foods sold was 1 (IQR: 1, 1), and the most common type of food sold was fresh meat/poultry (64.0%). Among vendors classified as high-risk, protective, the median number of foods sold was 6 (IQR: 3, 8), and the most common types of food sold were fried starches (73.1%) and sweets/confectionary (65.4%). Finally, among vendors classified as high-risk, non-protective, the median number of foods sold was 4 (IQR: 2, 5), and the most common type of food sold was sweets/confectionary (62.8%).

Table 2.

Types of food sold according to obesogenic risk category of vendor; Kibera informal settlement (Nairobi, Kenya), July–August 2019 (n = 456)

	Obesogenic risk category
	Low-risk, protective n (%)a	Low-risk, non-protective n (%)a	High-risk, protective n (%)a	High-risk, non-protective n (%)a
No. of foods sold, median (IQR)	6 (4, 8)	1 (1, 1)	6 (3, 8)	4 (2, 5)
Food type
Vegetables, raw	125 (62.8)	0 (0.0)	4 (15.4)	0 (0.0)
Vegetables, cooked	29 (14.6)	0 (0.0)	6 (23.1)	1 (0.6)
Fruits	97 (48.7)	0 (0.0)	0 (0.0)	1 (0.6)
Meat/poultry, fresh	0 (0.0)	48 (64.0)	0 (0.0)	0 (0.0)
Meat/poultry, cooked	23 (11.6)	1 (1.3)	5 (19.2)	6 (3.8)
Fish, fresh	12 (6.0)	0 (0.0)	0 (0.0)	0 (0.0)
Fish, cooked	19 (9.5)	0 (0.0)	7 (26.9)	0 (0.0)
Meat/poultry/fish, fried or processed	2 (1.0)	0 (0.0)	3 (11.5)	24 (15.4)
Eggs	12 (6.0)	0 (0.0)	2 (7.7)	5 (3.2)
Staples, raw	44 (22.1)	10 (13.3)	3 (11.5)	5 (3.2)
Staples, cooked	17 (8.5)	0 (0.0)	7 (26.9)	3 (1.9)
Legumes/pulses, raw	23 (11.6)	0 (0.0)	2 (7.7)	1 (0.6)
Legumes/pulses, cooked	29 (14.6)	0 (0.0)	11 (42.3)	1 (0.6)
Nuts/seeds	5 (2.5)	0 (0.0)	0 (0.0)	0 (0.0)
Dairy, unprocessed	2 (1.0)	13 (17.3)	1 (3.8)	19 (12.2)
Dairy, processed	0 (0.0)	0 (0.0)	0 (0.0)	6 (3.8)
Beverages, sweetened	12 (6.0)	0 (0.0)	13 (50.0)	50 (32.1)
Beverages, unsweetened	11 (5.5)	0 (0.0)	6 (23.1)	2 (1.3)
Starches, fried	30 (15.1)	0 (0.0)	19 (73.1)	58 (37.2)
Sweets/confectionary	19 (9.5)	0 (0.0)	17 (65.4)	98 (62.8)
Traditional mixed dishes	41 (20.6)	2 (2.7)	12 (46.2)	6 (3.8)
Fats/oils	2 (1.0)	0 (0.0)	2 (7.7)	22 (14.1)
Bread	2 (1.0)	0 (0.0)	3 (11.5)	51 (32.7)
Herbs/spices	29 (14.6)	1 (1.3)	0 (0.0)	0 (0.0)
Condiments/sauces	0 (0.0)	0 (0.0)	1 (3.8)	0 (0.0)
Baking ingredients, raw	5 (2.5)	3 (4.0)	1 (3.8)	31 (19.9)
Other mass-produced foods	1 (0.5)	0 (0.0)	2 (7.7)	30 (19.2)

IQR interquartile range

^aPercentages are column percentages; percentages may not sum to 100% because of rounding

Spatial Distribution of Vendors

Vendors were predominantly located along major roads of the informal settlement, with intersections appearing to have the highest density of vendors (Fig. 1). On average, the shortest distance between any two vendors was 12.3 m (95% CI: 11.4, 13.3). Shortest distances between vendors, overall, and according to vendor type and obesogenic risk category, are shown in Table 3.

Fig. 1.

Map of vendors according to vendor type; Kibera informal settlement (Nairobi, Kenya), July–August 2019 (n = 456)

Table 3.

Distance between vendors, overall, and according to vendor type and obesogenic risk category; Kibera informal settlement (Nairobi, Kenya), July–August 2019 (n = 456)

Category	Distance (m) to nearest vendor of same categorya
	Mean	95% CI
Any vendor	12.3	11.4, 13.3
Vendor type
Hawker	59.8	44.1, 75.5
Kiosk	29.9	24.7, 35.2
Stand/table	27.3	20.1, 34.4
Market stall	42.5	32.4, 52.6
Restaurant	66.5	48.3, 84.6
Obesogenic risk category
Low-risk, protective	26.0	21.1, 30.8
Low-risk, non-protective	43.2	30.4, 55.9
High-risk, protective	113.8	88.3, 139.3
High-risk, non-protective	28.6	24.5, 32.6

m meters; CI confidence interval

^aCategory refers to vendor type or obesogenic risk category

^bDistance to nearest vendor, regardless of vendor type or obesogenic risk category

Vendors were, in general, evenly dispersed throughout the informal settlement according to obesogenic risk category (Fig. 2), though there did appear to be a higher density of low-risk, protective and low-risk, non-protective vendors in the three easternmost villages of Kibera. In contrast, the villages in the southwestern part of the informal settlement were more dominated by vendors in the high-risk, protective and high-risk, non-protective categories. Finally, the villages in the central and northern parts of Kibera contained a relatively equal mix of vendors according to obesogenic risk category.

Fig. 2.

Map of vendors according to obesogenic risk category; Kibera informal settlement (Nairobi, Kenya), July–August 2019 (n = 456)

The mean shortest distances between vendors in the low-risk, protective (26.0 m; 95% CI: 21.1, 30.8) and the high-risk, non-protective (28.6 m; 95% CI: 24.5, 32.6) categories were lower than the mean shortest distance between vendors in the high-risk, protective category (113.8 m; 95% CI: 88.3, 139.3) and marginally lower than the mean shortest distance between vendors in the low-risk, non-protective category (43.2 m; 95% CI: 30.4, 55.9). Based on the results from the univariate K analyses, there was significant clustering of vendors of the same obesogenic risk category for all categories except high-risk, non-protective (Supplemental Fig. S4-S7).

Kiosks, stands/tables, and market stalls appeared to co-locate with vendors of the same type. On average, the shortest distance (95% CI) between any two kiosks, any two stands/tables, and any two market stalls was 29.9 m (24.7, 35.2), 27.3 m (20.1, 34.4), and 42.5 m (32.4, 52.6), respectively. Hawkers were relatively evenly dispersed along major roadways and at intersections, and the mean shortest distance (95% CI) between any two hawkers was 59.8 m (44.1, 75.5). In contrast to the other types of vendors, restaurants were more dispersed and often located away from major roads and junctions. On average, the shortest distance (95% CI) between any two restaurants was 66.5 m (48.3, 84.6).

Discussion

In this informal settlement in Nairobi, Kenya, we encountered a food environment with a high prevalence of both obesogenic and anti-obesogenic foods and with marked clustering of vendors selling foods with similar obesogenic properties. To our knowledge, our study is the first to have examined the food environment of an informal settlement in Nairobi according to the obesogenic properties of foods sold.

The food environment of this informal settlement was characterized by a duality of obesogenic and anti-obesogenic foods. This pattern of a large number of nutrient-poor, highly-processed foods alongside an abundance of nutrient-rich, minimally-processed foods is consistent with the nutrition transition that has been observed in LMICs throughout the world since the late twentieth century [3]. Although a number of studies have considered urban food environments in SSA [16], we are aware of only one other study that has described the food environment of an informal settlement in SSA according to obesogenic risk. Kroll et al. [33] examined the food environment of Khayelitsha, a township in Cape Town, South Africa with a large shantytown population, reporting that the most commonly available foods were fresh fruit, commercial breads, sugar, and sweetened beverages, and that fewer than 10% of vendors were low-risk, protective, a much smaller proportion than that which was observed in our study (44%). This difference in distribution of obesogenic risk across vendors may indicate greater availability of minimally-processed foods in Kibera, relative to Khayelitsha. It may also be explained by the low response rate (~ 50%) of vendors in the study by Kroll et al., if vendors selling fruits, vegetables, or other unprocessed foods were less likely to participate.

The finding that food vendors of this informal settlement comprised nearly as many high-risk, non-protective vendors as low-risk, protective vendors is novel; earlier investigations of food environments of Nairobi’s informal settlements described the availability of very broad food categories and did not consider obesogenic risk [26–28]. Although comparing the findings from our study to those of earlier studies is inadvisable because of differences in scope and methodology, we anticipate that the Kibera food environment is similar to those of Nairobi’s other informal settlements, given their geographic proximity, cultural similarities, and shared constraints on physical infrastructure and living conditions [18]. Nonetheless, additional studies of food environments of informal settlements are needed to better understand how Kibera’s food environment compares to those of Nairobi’s other informal settlements and of informal settlements elsewhere SSA.

In our spatial analysis, we observed that most vendors were situated along major roads and at intersections. This observation is consistent with the results of two previous studies of vendors in Nairobi’s informal settlements, which found that vendors tended to operate along major roads and footpaths [27, 28]. Major roads and intersections typically have a high volume of pedestrian and vehicle traffic, attracting vendors in search of customers [27, 28].

The results of our investigation of clustering of vendors according to obesogenic risk suggested that vendors selling low-risk, protective foods, like vegetables, fruits, or legumes, tended to be located near other vendors selling low-risk, protective foods. Similarly, vendors selling high-risk, non-protective foods, like confectionary, sweetened beverages, or fried starches, tended to cluster with one another. Earlier studies of food environments of urban informal settlements in SSA have not formally examined clustering of vendors according to obesogenic risk. However, a similar clustering pattern was reported in a study of the food environment of a large Brazilian city, in which the authors reported that retailers selling mainly unprocessed and minimally-processed foods were located near one another and retailers selling mainly ultra-processed foods were similarly clustered [41]. The implications of this clustering for the diet quality of slum residents are unknown, and future research should consider whether residents living closer to clusters of high-risk, non-protective vendors consume a greater number of ultra-processed and other obesogenic foods, compared to those living further from such clusters, and whether proximity to clusters is associated with risk of obesity.

The results of our study at once reveal the concerning state of Kibera’s food environment and highlight areas in which interventions may improve the availability of health-promoting foods. Of greatest concern is that more than one-third of vendors in this sample were selling primarily obesogenic foods; notably, a majority of these were kiosks and hawkers, with more than half of the sampled kiosks and hawkers classified as high-risk, non-protective. It stands to reason that substitution of obesogenic foods with anti-obesogenic foods among kiosks and hawkers may considerably reduce the number of high-risk, non-protective vendors in this informal settlement and reduce the obesogenic risk posed to slum residents. The creation of food hubs in areas that are furthest from existing markets has been suggested as a potential strategy for improving access to nutrient-rich, anti-obesogenic foods [19]. Food hubs could strengthen supply chains for unprocessed foods, like fruits and vegetables, and increase access to such foods among households that are currently furthest away from them. Using the findings from our spatial analysis, local government officials and non-governmental organizations, in consultation with community members and food vendors, should identify optimal locations for such hubs and support their operation through increasing vendors’ access to business licenses and infrastructure for basic food preparation and storage, which may have the added effect of improving food safety, a major concern among slum residents [19, 24]. Such an intervention would address the broader issue of food insecurity, an intractable public health issue in Nairobi’s slums [24], and may have the double-duty effect of alleviating both under- and overnutrition.

Ultimately, improvements in the Kibera food environment through food hubs or any other initiative must not come at the expense of existing vendors and their livelihoods; rather, the objective must be to help vendors replace obesogenic foods with anti-obesogenic foods. Concurrent with any intervention to improve the healthfulness of foods sold must be an educational campaign to encourage slum residents to choose anti-obesogenic foods, in recognition that food environments and dietary preferences have a bidirectional relationship [5] and that the success of vendors will likely depend on shifts in preferences among residents toward anti-obesogenic foods.

Our study has several strengths. First, the vendor food audits were collected via direct observation, without reliance on self-report. This prevented omission of foods that may be perceived as less healthy, and it ensured consistency in the names used for different foods or dishes, improving the accuracy of the obesogenic risk classification. Second, despite the well-known challenges associated with collecting food environment data in informal settlements [4], the sample size in our study was relatively large, allowing for a detailed characterization of the food environment. Vendors in informal settlements may not possess a license to operate and, consequently, may be unwilling to share information with researchers out of a concern that their operation may be shut down by local authorities [27, 31]. Members of our research team have spent decades working in Kibera and have gained residents’ trust, which facilitated the collection of these data. Finally, latitude and longitude were captured using a validated, mobile application at the exact location of the vendor at the moment of data collection. Vendors in Kibera may not have an official business address, preventing the use of methods that rely on government or third-party databases of vendor locations. Using the approach described here, we were able to capture precise spatial data for even the most mobile vendors.

Notwithstanding these strengths, this study had limitations. As noted, collecting food environment data in informal settlements is challenging, and there were vendors for whom food audit data were missing, leading to their exclusion from the analysis. Additionally, we likely failed to identify a number of vendors who were operating outside of the time period in which we collected data or in different areas each day. Missing data is a challenge in studies of food environments with many informal vendors [4], like Kibera. Although we mitigated the amount of missing information to the greatest extent achievable, it is possible that missing data biased the results if the excluded or unobserved vendors were systematically different from those in the analytic sample, with respect to types of food sold or location of operation.

Conclusions

The food environment of this informal settlement in Nairobi, Kenya was characterized by a duality of obesogenic and anti-obesogenic foods. This pattern of a large number of nutrient-poor, energy-dense foods alongside an abundance of nutrient-rich foods is consistent with the ongoing nutrition transition that has been observed in LMICs throughout the world. Clustering of vendors selling obesogenic foods in this food environment and in similar settings may have implications for the diet quality of people living near such clusters, highlighting the need for local officials and non-governmental organizations to implement strategies to ensure consistent access to health-promoting foods throughout informal settlements.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution

KRB contributed to the design of the study, performed the analyses, and wrote the manuscript; RL contributed to the design of the study and data collection and provided critical feedback on the manuscript; MO and HO contributed to the design of the study, provided logistical support for data collection, and provided critical feedback on the manuscript; EN provided critical feedback on the manuscript; ASA and SLM obtained funding to support data collection, contributed to the design of the study, and provided critical feedback on the manuscript.

Funding

This study was supported by a National Heart, Lung, and Blood Institute National Research Service Award predoctoral traineeship (grant #5T32HL129969), funding from the Duke University Robertson Scholars Leadership Program, and a Departmental Pilot Award from the University of North Carolina at Chapel Hill Department of Nutrition.