The global dairy industry, comprised of ~265 million cows, has continued to grow over the past decade, with milk production increasing from 590 million tons in 2009 to 683 million tons in 2018 (1). Of these, Kenya produced ~3.8 million tons, which represented about 10.8 and 0.6% of the African and global shares, respectively. While comparatively small in terms of global production, the dairy sector is significant to Kenyan livelihoods, both nutritionally and economically (2). The sector is one of the largest agricultural segments of the country, contributing about 4% of the national gross domestic product (GDP) and 14% of the agricultural GDP (3). Exponential growth of the dairy sector is expected with the predicted rise in demand for milk and other milk products influenced by growth in population, rapid urbanization, and desire for intake of livestock source foods (4). For example, Herrero et al. (4) predicted a triple increase in milk demand in Sub-Saharan Africa by 2050. Similarly, the Food and Agriculture Organization (FAO) of the United Nations forecasts a 175% rise in milk demand for Kenyans between 2010 and 2050 (5).

Increased demand for milk consumption coupled with its predicted low supply will put pressure to existing value chains, and this will trigger the evolution of more milk supply chains, which will complicate already complex food systems. Food systems present some of the most complicated networks, especially in urban areas where production and distribution are through simple to complex value chains [(6–8); Hueston and MacLeod, 2012]. Such system complexities are excellent avenues for the introduction and transmission of pathogens, including food hazards among other food safety risks (9, 10).

According to the national livestock production report of 2012, Nairobi, one of the fastest-growing urban cities in Africa (11) produced ~39 million L of milk (12). This was less than the required 388 million L based on the estimated 125 L per capita milk consumption in urban areas (13) for the 3.1 million city residents then (14). Nairobi, the largest city in Kenya and representing ~9.2% of the national population, has grown by 45% from 3.1 million people in 2009 (15) to 4.4 million in 2019 (15). This rapid population growth is expected to create a lot of pressure to produce and supply more food for city dwellers. This may trigger the evolution of more complex food chains which, in turn, may present challenges in food safety standards.

Based on the annual per capita consumption and projected growth rate of the city, approximately at 4% (11), it means that by 2050, there will be ~10.3 million residents in Nairobi and that they will require about 1.8 billion L of milk. It follows that more than 97% of milk consumed in the city will be sourced from production systems that are based outside the city. Currently, the proportion of milk that comes from outside Nairobi accounts for about 75–90% (6). The characteristics of milk chains supplying Nairobi have been described as highly complex and made up of multiple small-scale value chain actors who are highly interconnected and interdependent (16). The study by Kiambi et al. (16) described seven chain profiles (segments of the milk value chain) that form the Nairobi dairy food system. These include chain profiles for urban and peri-urban farming systems, profiles of traders affiliated to Dairy Traders Association (DTA) and those not belonging to this association (non-DTA), profiles associated with medium and large dairy cooperatives, and profiles of large processing companies. These chain profiles are tightly interlinked to form the vast Nairobi's dairy system within the free milk trading market (liberalized) (17).

Achieving food safety in such complex food systems is a challenge, particularly because milk is produced primarily by small-scale farmers, and marketing channels are dominated by informal systems (18–20). The rising milk demand (4, 5) coupled with unmatched production (4), complex interactions between value chain actors (16, 21), and compromised governance of the Nairobi dairy system (21) will put massive pressure on existing milk value chains, with a possible evolution of new ones to satisfy the rising demand. Complexities associated with urban food chains could provide excellent platforms to expand the range of food-borne pathogens as well as to amplify health and economic impacts of a single contamination incident (7). On the other hand, the degree of mixing and contact between human and livestock in urban environments has been shown to create ecological niches with opportunities for pathogen transmission, and some studies have linked urbanization to risk of emerging infectious diseases (22, 23).

There are many factors that may contribute to unsafe milk (24), and challenges in food safety in Africa are precipitated by poor food safety systems, lack of systematic surveillance, underdeveloped human resource, and insufficient capacity to determine the magnitude of the problem (25). Considering the uneven distribution of hazards/risks (26), competing priorities and inadequate resources in these countries, designing and implementation of interventions to promote food safety require a targeted risk-based approach that focuses on value chain analysis (27). This involves a thorough understanding of the “what” (e.g., contamination practices, quality deficiencies, and poor accessibility), “when” (risk seasonality), “where” (in which chains, chain nodes, areas it occurs), “who” (who creates it and who is exposed), “how” (practices and behaviors), and “how much/many” (e.g., how much contamination or how many people are exposed) (26). Several studies have mapped various value chains within food systems in Nairobi, namely, beef, sheep, and goat value chains (28), camel milk value chain (29), livestock keeping in the city (28), Nairobi dairy value chain (16), poultry value chain (8), pork value chain (30), and governance issues (21). These studies provide the critical frameworks needed for full analysis of food systems and to guide the development of necessary interventions along the value chains. In addition, such detailed scrutiny of the systems helps to understand the dynamics therein including assessment of any structural vulnerabilities (26, 28, 31).

Bacteriological characteristics of milk help to determine the quality of milk flowing through various nodes of the value chains. Bacteriological quality depends on various factors, among them is the health status of animals (32–36) as well as practices in milk handling and storage (37–39). Some studies on raw milk have demonstrated the presence of a wide range of microorganisms in marketed milk associated with poor hygiene standards (33, 40–42), or possible infection of cows in farms as shown by the presence of zoonotic pathogens like Mycobacterium species (33), Brucella abortus and E. coli O157:H7 (43, 44). Other hazards identified in raw milk at market level include the presence of aflatoxins (45, 46), antibiotic residues (47), and antibiotic-resistant bacteria (48). On the other hand, although pasteurized milk is expected to have minimal hazards, some studies have found excessive levels of bacteria in processed milk (19, 49), which may indicate system failure in terms of ensuring high quality of milk that arrives in processing or faulty pasteurization processes or failure in post pasteurization hygiene (50). In addition, compromised regulatory standards and procedures at pre- and post-marketing may also influence the quality of processed milk as shown in studies conducted in China on an infant formula that was adulterated with melamine (51, 52).

Investigation of bacterial load, specifically analysis of TBC and TCC, has been widely conducted as an indicator to determine the hygienic quality of milk (53). TBC in milk reflects the total number of bacteria that can grow to form countable colonies on standard methods agar when incubated aerobically at 32°C for 48 h (54). Milk stored at ambient temperatures with poor hygienic standards would favor bacterial growth and multiplication leading to its deterioration (55, 56). On the other hand, coliform bacteria are present in the environment and in the feces of all warm-blooded animals and humans (57–59). Therefore, while it is possible that infected cows could shed the bacteria in milk (60–62), detection of coliform in milk may indicate possible contamination with bacteria from the cow environment, including udder, milking utensils, water, or the handler (63).

This study overlays a bacteriological analysis onto the mapping framework developed by Kiambi et al. (16) to provide an indication of the quality and safety of milk flowing along the value chain in terms of TBC and TCC. The authors appreciate that TBC and TCC are not major public health hazards that would be present in the milk supplied to Nairobi, but their presence in higher than acceptable limits would indicate possible areas of system vulnerability and, therefore, point to critical areas of intervention. The research questions investigated in this study were: (i) how do TBC and TCC compare along the various nodes of the Nairobi milk value chain and (ii) what are the milk production and handling practices that would influence food safety along the value chain?

This study was part of the Urban Zoo project, which was based at the International Livestock Research Institute (ILRI), Kenya. The overall objective of the Urban Zoo project was to understand mechanisms leading to the introduction and transmission of pathogens to urban populations through livestock commodity value chains in 33 sub-locations of Nairobi. Most of the qualitative data used in this study were collected during the mapping of the dairy value chain (16) implemented in Nairobi County, the capital city of Kenya, between January 2014 and January 2015. The County, which is divided into nine administrative sub-counties (Figure 1), lies 696.1 km² of land located at 1.2921° S, 36.8219° E and at an altitude of 1,798 m above sea level.

To comprehensively map the bacteriological (TBC and TCC) landscape of milk in Nairobi, milk sampling was conducted on nodes that were identified during the initial mapping by Kiambi et al. (16). These included farmers (within the city and from peri-urban areas), traders whether fixed (milk bars, shops/kiosks, and restaurants) or mobile (hawkers or roadside vendors), milk vending machines (MVMs), milk collection centers, and large processing companies (mainly branded processed products).

By 2050, the demand for milk consumption will triple in Sub-Sharan Africa (4), with consumption projected to rise by 175% from 2010 to 2050 in Kenya (5). The current population in Nairobi (15), coupled with high per capita milk consumption (13) that is unmatched with production has resulted in most of milk being sourced from production systems that are based outside the city (6, 16). The situation is expected to worsen with the projected population growth (11) and increased demand for milk by city dwellers (5). Food chains associated with urban food systems are complex and are, thus, likely to provide excellent platforms to expand the range of food-borne pathogens (7, 22, 23).

This study utilized the previously developed framework of Nairobi's dairy value chain (16) to investigate milk quality and safety in complex food systems to identify food safety challenges. The mapped framework demonstrated the vastness and complexity of Nairobi's dairy value chain with multiple interactions between various actors and nodes of the value chain. Based on these complex interactions, the study concluded that a holistic approach would be required to address any intervention and policy decision. This approach has been supported by other studies (8, 26, 28, 29).

This study found that TBC levels in production nodes (farms and collections centers) were generally good and within the acceptable limits of the EAS for TBC, but that TCC limits were higher (than EAS) in farms. This agrees with several other studies that have demonstrated good bacterial quality of milk at farm levels (35, 55, 66, 67). However, this could be compromised by other factors like health status of the animals (32–36) and practices related to milk handling and storage practices (37–39). Apart from TBC and TCC, which only serve as an indicator of the robustness of the system in this study, some of the practices reported (and observed by the researchers) on farms warrant further investigation to inform broader interventions. For example, the availability of animal feeds was reported to be a challenge, and sometimes farmers fed cows with feeds sourced from dumping sites, leaking sewer lines, market leftovers, and obtained by the roadside. Such unconventional feed sources may present opportunities for the introduction of heavy metals and subsequent public health concerns (68–70). In addition, wastes swept from poultry houses were reported to be mixed with dairy commercial concentrate feeds, as this was perceived to increase milk production. This practice may result in the introduction and transmission of antimicrobial resistance (AMR), as several studies have demonstrated high levels of AMR in poultry and poultry environments (71–73).

We noted that the bacterial quality of milk deteriorated at retail level. For example, the logistic regression analysis for TBC showed that the odds that milk samples from retailers were acceptable were less than the odds of samples collected from farm as follows: 0.02 (milk bars), 0.02 (restaurants), 0.07 (shops/kiosks), 0.03 (roadside vendors), and 0.11 (supermarkets). For TCC, the model results indicated that the odds of milk samples from milk collection centers were acceptable were less than the odds of samples collected from farms (0.18), and that the odds of milk from milk bars and restaurants were less than the odds of farm samples (0.03 and 0.06, respectively). Similarly, the odds of samples from roadside vendors and shops/kiosks being acceptable were less than those of samples collected from farms (0.02 and 0.12, respectively). Our results agree with findings from other studies that show such bacterial deterioration of milk as it flows from farms through the retailing system (33, 35, 42, 74). Such deterioration may be influenced by, among other factors, heavy bacterial load in source (66, 67), adulteration (75), flow of milk over longer distances without cold chain, and poor handling practices during transportation and storage. In this study, there are some reported practices that may contribute to increased bacterial count in milk at retail level. For example, milk was stored in plastic (non-food grade) containers, and cold chain was deficient as reported in various KIIs and FGDs (see Table 7). Traders and retailers in Nairobi reported to depend on local agents based at the farms to source for milk on their behalf, which was then bulked into 20-L plastic containers and transported while stack under passengers' seats in public vehicles (FGDs with non-DTA traders and retailers, KII with DTA traders). Despite the law requiring milk transportation to be done in cold chain and by licensed vehicles (transport permit issued by KDB), the traders and retailers indicated that it was expensive to travel to farms daily to source for milk as well as to own or hire a vehicle to transport their milk. Probably, that is the reason some traders mentioned to use unorthodox methods for preservation of milk like the addition of hydrogen peroxide and formalin. Malicious addition of formalin and hydrogen peroxide has been mentioned as a way of increasing the shelf life of milk in the absence of cold chain (75). Apart from ethical aspect, researchers have reported that addition of artificial preservatives, such as chemicals, could result in serious health problems, including cancer (76, 77).

If the raw milk spoilt, it was reported as “accidentally curdled,” and it was reported to be converted into fermented milk (allowed to ferment for several more days) or yogurt (addition of flavors and colors to the fermented milk). This probably explains the high TBC and TCC in homemade fermented milk and homemade yogurt. On the regression analysis, our results indicated that the odds that the homemade (yogurt and fermented) and processed (yogurt and fermented) types of milk were acceptable for TBC was 97 and 90% lower, respectively, than raw milk, while for TCC, the homemade (yogurt and fermented) had lower odds (94%) than raw milk. While most Kenyans normally boil milk before consumption (78, 79), it is important to note that these products (home-made fermented and yogurt) were made from raw milk that had gone bad. Consequently, these could pose significant public health threats if the raw milk was sourced from cows infected by any zoonotic disease. For example, the consumption of raw milk has been associated with brucellosis in humans (80).

The logistic regression analysis of the processed milk products showed that they were generally within the acceptable limits for TBC and TCC. This agrees with other studies that have shown the importance of pasteurization and ultra-heat treatment in significantly reducing bacterial load in milk (81–83). However, other studies have shown significant levels of bacteria in pasteurized milk (19, 49, 84–86). This happens with poor quality of raw milk (35, 55, 66, 67) when there is a breakdown in the pasteurization process(50), or factors related to post-processing handling. In this study, the assessment of maximum values for the colony-forming units for TBC and TCC showed that there were some processed milk products that were highly contaminated (refer to Tables 2, 3 in the results section). These included pasteurized milk (4.1 × 10²), processed fermented milk (2.7 × 10¹), and processed yogurt (5.6 × 10⁴). Some practices elucidated during FGDs and KIIs with the dairy cooperatives and large processing companies that may contribute to high bacterial load in processed milk included lack of cold chain in collection centers and acceptance of milk that should be rejected so as not to benefit competitors who care less about quality. Unfair competition among value chain actors has been identified as one major factor that hinders achieving optimal food safety in the Nairobi dairy value chain (21, 84). Other factors that were mentioned in this study included lack of training on hygiene across the value chain, selling of milk by hawking from place to place or by the roadside, and selling of milk in unlicensed premises (hence such milk is not monitored by regulatory authorities). Therefore, addressing food safety requires concerted efforts of every actor in the value chain. Some strategies that have been deployed to improve milk safety elsewhere include improvement of infrastructures in farms and at collection, enhanced information on production, and improved frequency of milk collection to reduce build-up of bacteria in milk (87, 88). Undoubtedly, improved training programs and provision of extension services would enhance good practices in milk production and handling.

This study provides a detailed analysis of food quality and safety challenges in Nairobi's complex dairy value chain. The descriptive analysis of the mean values for TBC and TCC shows that raw milk collected from farms and processed products were relatively good, but these parameters deteriorated at the retail level. The logistic regression analysis on all sample types showed that the odds that milk samples from retail (milk bars, restaurants, shops/kiosks, roadside vendors, and traders) were acceptable were less than the odds of samples collected from farm being acceptable for TBC and TCC. Likewise, the odds that raw milk samples from retail (milk bars, restaurants, shops/kiosks, roadside vendors, and traders) were acceptable was less than the odds of raw samples collected from farm being acceptable for TBC and TCC. With raw milk as the reference, the analysis by milk type showed few differences in the odds of TBC and TCC acceptability across milk types.

Several practices with possible influence on food safety were mentioned. In production, these were related to keeping cows in very muddy cowsheds, obtaining animal feeds from dumpsites and market leftovers, obtaining feeds by the roadside, treatment of cows by unqualified personnel coupled with compromise on withdrawal periods following treatment, resale of milk that has been rejected from dairy cooperatives or allowing it to ferment further (and taken as fermented milk), and addition of water to increase the volume of milk. In bulking centers, the practices were related to accepting milk that should be rejected, lack of cold chain and basic screening tests, and lack of procedures for the management of milk that has been rejected. In retail, there was milk adulteration by addition of water and other chemicals, lack of cooling facilities during transportation and at sale points, sale to consumers of raw milk that got spoilt, and conversion into yogurt or fermented milk.

The analytical methodology presented in this study demonstrates a practical approach for strategic policy decisions. To achieve milk quality and safety, the authors suggest the implementation of more robust training of people involved in the milk system. However, this needs to be guided by a critical analysis of prevailing challenges in every segment of the value chain. Every node of the value chain should be considered prior to designing and implementing any intervention, as further underscored in Kiambi et al. (16), FAO (26), and Alarcon et al. (28). The framework and findings obtained can help future research and policymakers to reach an informed decision on possible points of intervention in improving the quality and safety of milk flowing through complex value chains.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: Data repository of the University of Liverpool available at https://doi.org/10.17638/datacat.liverpool.ac.uk/1639.

SK, EF, PA, JR, and EK designed the study and data collection tools. SK and JM collected data. SK, JO, PA, and EF drafted the manuscript. SK, GA, and NG developed the culture and isolation standard procedures, facilitated culture and isolation and interpreted TBD and TCC results. All authors read, commented on, and approved the final manuscript for publication.

This study was supported by the United Kingdom Medical Research Council, Biotechnology and Biological Science Research Council (United Kingdom), the Economic and Social Research Council (United Kingdom), the Natural Environment Research Council (United Kingdom), through the Environmental & Social Ecology of Human Infectious Diseases Initiative (ESEI), grant reference: G1100783/1. This study also received support from the CGIAR Research Program on Agriculture for Nutrition and Health (A4NH) led by the International Food Policy Research Institute (IFPRI).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.