Doctoral Thesis No. 2026:10 Faculty of Veterinary Medicine And Animal Science D octoral T h esis N o. 2026:10 • Food safety in inform al m arkets • M aureen K uboka Food safety in informal markets Maureen Kuboka Hazard assessment for food in East African Community countries Acta Universitatis Agriculturae Sueciae Doctoral Thesis No. 2026:10 Consumption of unsafe food leads to foodborne diseases (FBD), which cause economic and productivity losses in many countries. The East African Community bears a high burden of FBD, yet there is limited evidence on the food safety status in the region. This thesis uses an integrated approach to investigate food safety in informal markets, the main sources of food in the region. The findings are important for informing risk characterisation and intervention strategies to reduce the prevalence of FBD. Maureen Kuboka received her doctoral education at the Department of Animal Biosciences, Swedish University of Agricultural Sciences. She obtained her B.Sc. and M.Sc. degrees at the University of Nairobi, Kenya. Acta Universitatis Agriculturae Sueciae presents doctoral theses from the Swedish University of Agricultural Sciences (SLU). SLU generates knowledge for the sustainable use of biological natural resources. Research, education, extension, as well as environmental monitoring and assessment are used to achieve this goal. ISSN 1652-6880 ISBN (print version) 978-91-8124-207-2 ISBN (electronic version) 978-91-8124-227-0 Food safety in informal markets Hazard assessment for food in East African Community countries Maureen Kuboka Faculty of Veterinary Medicine and Animal Science Department of Animal Biosciences Uppsala DOCTORAL THESIS Uppsala 2026 Acta Universitatis Agriculturae Sueciae 2026:10 Cover: Open-air vegetable market in Bujumbura, Burundi (Photo credit: Maureen Kuboka, 2023) ISSN 1652-6880 ISBN (print version) 978-91-8124-207-2 ISBN (electronic version) 978-91-8124-227-0 https://doi.org/10.54612/a.647e7s3ve7 © 2026 Maureen Kuboka, https://orcid.org/0000-0003-0388-2398 Swedish University of Agricultural Sciences, Department of Animal Biosciences, Uppsala, Sweden The summary chapter is licensed under CC BY 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. Other licences or copyright may apply to illustrations and attached articles. Print: SLU Grafisk service, Uppsala 2026 Food safety in informal markets: Hazard assessment for food in East African Community countries Abstract Food-borne diseases result from consuming unsafe food. The burden of food-borne diseases is disproportionately high in Africa. Despite this, there has been a lack of empirical data on food safety and food-borne diseases. One challenge in food safety research in Africa has been the dominance of informal food markets. This thesis reports on the food safety status in informal markets in East African Community (EAC) countries, focusing on hazard levels, risk factors for contamination, stakeholder perceptions and effectiveness of a food safety intervention. A mixed methods approach was used. Firstly, a systematic review and meta-analysis were conducted across seven EAC countries. Of 4,134 initial records, 53 studies were included for review and analysis. Pooled prevalence estimates were 41% for Escherichia coli (E. coli) (95% CI: 34–52%), 12% for Salmonella spp. (95% CI: 12–27%), and 9% for Campylobacter spp. (95% CI: 7–32%). Secondly, 16 focus group discussions and 33 key informant interviews involving traders, consumers, and stakeholders in the food system were conducted in Burundi and Kenya. Consumers and traders were found to confuse food quality with food safety. Consequently, they relied on heuristic cues and physical assessment to guarantee food safety. Animal source foods were perceived as most likely to cause disease. Risky practices such as food adulteration, improper traditional processing, slaughter, and consumption of diseased or dead animals were reported. Thirdly, knowledge, attitude, and practice (KAP), alongside a prevalence survey, were conducted among 170 meat sellers in Kenya, categorised into treatment (n=59) and control (n=111). Additionally, 103 meat samples were collected in Burundi. Overall, meat contamination with E. coli, coliforms and Salmonella exceeded EAC standard limits. KAP survey revealed that higher practice scores were associated with reduced levels of E. coli and coliform in meat. Higher cleanliness scores and access to running water were associated with reduced odds of Salmonella contamination. The KAP and hazard prevalence results showed no significant difference between treatment and control groups. Our study reveals high contamination levels in foods, and that consumers are at risk of pathogenic infections through risky behaviours and unhygienic handling. These findings can inform effective risk communication and strategic interventions aimed at reducing the risk of foodborne diseases. Keywords: (Risk, foodborne diseases, food markets, food security, Africa) Livsmedelssäkerhet på informella marknader: Riskbedömning för livsmedel i länder i Östafrikanska gemenskapen Abstract Konsumtion av kontaminerad mat kan leda till livsmedelsburna sjukdomar, och bördan av livsmedelsburna sjukdomar är oproportionerligt hög i Afrika. Trots detta har det saknats empiriska data om livsmedelssäkerhet och livsmedelsburna infektioner. En bidragande faktor till både hög förekomst av livsmedelsburna sjukdomar och till brist på data kan vara dominansen av informella livsmedelsmarknader. Denna avhandling rapporterar om livsmedelssäkerheten på informella marknader i länderna i Östafrikanska gemenskapen (EAC), med fokus på prevalens av smittämnen och riskfaktorer för kontaminering, samt effektiviteten av interventioner och kunskap och åsikter hos olika aktörer. Flera olika metoder användes i studien. Först genomfördes en systematisk översikt och metaanalys i sju EAC-länder. Av 4134 identifierade artiklar inkluderades 53 studier för granskning och analys. Sammanslagen prevalensuppskattning var 41 % för E. coli (95 % konfidensintervall (KI): 34–52 %), 12 % för Salmonella-arter (95 % KI: 12– 27 %) och 9 % för Campylobakter-arter (95 % KI: 7–32 %). Därefter genomfördes 16 fokusgruppsdiskussioner och 33 intervjuer med viktiga intressenter i livsmedelssystemet i Burundi och Kenya. Konsumenter och försäljare visade sig blanda ihop livsmedelskvalitet och livsmedelssäkerhet, och därför förlitade de sig på observationer och annan fysisk bedömning av livsmedlen för att avgöra om livsmedlen var säkra att konsumera. Animaliska livsmedel ansågs mest sannolika att orsaka sjukdom. Riskfyllda metoder som felaktig traditionell bearbetning, slakt och konsumtion av sjuka eller döda djur rapporterades. Sedan utvärderades kunskap, attityd och praktik (KAP), tillsammans med en prevalensundersökning bland 170 köttförsäljare i Kenya, kategoriserade i grupperna intervention (n=59) och kontroll (n=111). Praktiska rutiner, tillsammans med rengöring och sanitet samt tillgång till rinnande vatten, kopplades till köttkontaminering med E. coli, totalantal koliforma bakterier och Salmonella. Resultaten från KAP och prevalensstudierna visade ingen signifikant skillnad mellan intervention och kontrollgrupperna. Vår studie visar höga föroreningsnivåer i kött och att konsumenter löper risk för patogena infektioner genom ohygienisk hantering och riskbeteenden. Dessa fynd ger grundläggande information för att kunna utforma effektiv riskkommunikation och strategiska insatser i syfte att minska riskerna för livsmedelsburna infektioner och förbättra folkhälsan. Nyckelord: (Risk, livsmedelsburna infektioner, livsmedelsmarknader, livsmedelssäkerhet, Afrika) La sécurité alimentaire dans les marchés informels : Évaluation des risques alimentaires dans les pays de la Communauté d'Afrique de l'Est Résumé Les maladies d'origine alimentaire résultent de la consommation d'aliments non sur. Le fardeau des maladies d'origine alimentaire est disproportionnellement élevé en Afrique. Malgré cela, il a existé un manque des données empiriques sur la sécurité alimentaire et les maladies d'origine alimentaire. Un des facteurs contributifs a été la domination des marchés alimentaires informels. Cette thèse rend compte de l'état de la sécurité alimentaire sur les marchés informels des pays de la Communauté d'Afrique de l'Est (EAC), en se focalisant sur les niveaux de dangerosité, les facteurs de risque de contamination, les perceptions des parties prenantes et l'efficacité d'une intervention en matière de sécurité alimentaire. Une approche mixte a été utilisée. Premièrement, une revue systématique et une méta-analyse ont été menées dans sept pays de la EAC. Sur 4 134 dossiers initiaux, 53 études ont été incluses pour examen et analyse. Les estimations de prévalence combinées étaient de 41 % pour E. coli (IC 95 %: 34–52 %), 12 % pour Salmonella spp. (IC 95 %: 12–27 %) et 9 % pour Campylobacter spp. (IC 95 %: 7–32 %). Deuxièmement, 16 discussions de focus group et 33 entretiens avec des informateurs clés impliquant des parties prenantes du système alimentaire ont été menés au Burundi et au Kenya. Consommateurs et commerçants confondaient qualité alimentaire et sécurité alimentaire. Par conséquent, ils s’appuyaient sur des indices heuristiques et des évaluations physiques pour garantir la sécurité alimentaire. Les aliments d'or igine animale étaient perçus comme les plus susceptibles de provoquer des maladies. Des pratiques à risque telles que l'adultération alimentaire, un traitement traditionnel inapproprié, l'abattage et la consommation d'animaux malades ou morts ont été signalées. Troisièmement, connaissances, attitudes et pratiques (KAP), ainsi qu'une enquête sur la prévalence, ont été menées auprès de 170 vendeurs de viande au Kenya, classés en traitement (n=59) et témoins (n=111). De plus, 103 échantillons de viande ont été prélevés au Burundi. Globalement, la contamination de la viande par E. coli, coliformes et Salmonella a dépassé les limites standard de la EAC. L'enquête KAP a révélé que des scores de pratique plus élevés étaient associés à des niveaux réduits E. coli et coliformes dans la viande. Des scores de propreté plus élevés et un accès à l'eau potable étaient associés à une réduction des risques de contamination par Salmonella. Les résultats de la KAP et de la prévalence des risques n'ont mis en évidence aucune différence significative entre les groupes traités et témoins. Notre étude révèle des niveaux élevés de contamination dans les aliments, et que les consommateurs sont exposés à des infections pathogènes en raison de comportements à risque et d'une manipulation insalubre. Ces résultats peuvent éclairer une communication efficace des risques et des interventions stratégiques visant à réduire le risque de maladies d'origine alimentaire. Mots-clés:( Risque, maladies d'origine alimentaire, marchés alimentaires, sécurité alimentaire, Afrique) Dedication To the memory of my late father, Mr Benjamin Kuboka, whose sacrifice and support were invaluable throughout my education. I come as one, but I stand as ten thousand − Maya Angelou. 6 Contents Dedication............................................................................................6 List of publications .............................................................................11 List of tables ......................................................................................13 List of Figures ....................................................................................15 Abbreviations .....................................................................................17 1. Introduction ..............................................................................19 1.1 Food safety globally ....................................................................... 19 1.1.1 Global food safety policy and regulation .......................... 19 1.1.2 Regional food safety control and legislation .................... 20 1.2 Burden of FBD in Africa ................................................................. 21 1.2.1 Food safety regulation in Africa ........................................ 22 1.3 Food safety and regulation in EAC ............................................... 22 1.3.1 Food supply and food safety regulations in Burundi ....... 23 1.3.2 Food supply and food safety regulations in Kenya ......... 24 1.4 Food safety in informal markets .................................................... 28 1.4.1 Actors.................................................................................. 28 1.4.2 Governance, policies and food safety control ................. 29 1.4.3 Infrastructure ...................................................................... 31 1.5 Application of risk assessment in informal markets ..................... 32 1.5.1 Risk assessment studies................................................... 32 2. Aims of the thesis ....................................................................39 7 3. Materials and methods ............................................................41 3.1 Study design ................................................................................... 41 3.2 Ethical Considerations (Papers II-IV) ............................................ 41 3.2.1 Informed consent (Papers II-IV) ....................................... 41 3.2.2 Anonymity and pseudonymisation (Papers II-IV) ............ 41 3.2.3 Compensation (Papers II-IV) ............................................ 42 3.2.4 Use of large language models and artificial intelligence. 42 3.3 Study sites (Papers I-IV) ................................................................ 42 3.4 Study populations and selection of participants (Papers I-IV) .... 44 3.5 Study populations and selection of participants (Papers I-IV) .... 45 3.6 Questionnaire development and interviews (Papers II and IV)... 47 3.6.1 Focus group discussion and key informant interview guides (Paper II and IV) ................................................................. 47 3.6.2 Focus group discussions and key informant interviews (Paper II and IV) ............................................................................. 47 3.7 Evaluation of the effectiveness of a food safety and hygiene intervention programme (Paper III) ........................................................... 48 3.7.1 Knowledge, attitude and practice questionnaire ............. 49 3.7.2 Questionnaire Interviews................................................... 49 3.8 Microbiological analyses (Papers III and IV) ................................ 50 3.9 Data analyses ................................................................................. 51 3.9.1 Meta-analyses (Paper I) .................................................... 51 3.9.2 Qualitative studies (Paper II and IV) ................................ 51 3.9.3 Quantitative analyses (Paper III) ...................................... 52 4. Results and discussions ..........................................................55 8 4.1 Prevalence of pathogens in foods (Paper I, III and IV)................ 55 4.2 Risk factors for contamination (Papers I and III) .......................... 57 4.3 Knowledge, attitudes and practice of meat sellers (Paper III) .... 59 4.3.1 Knowledge.......................................................................... 59 4.3.2 Attitudes.............................................................................. 60 4.3.3 Practices ............................................................................. 61 4.3.4 Observations ...................................................................... 65 4.3.5 Evaluation of food safety intervention .............................. 68 4.4 Food safety and foodborne illnesses from the perspective of stakeholders (Papers II and IV) ................................................................ 70 5. Conclusions .............................................................................75 6. Recommendations and policy implications .............................77 6.1 Training and capacity building ....................................................... 77 6.2 Surveillance..................................................................................... 77 6.3 Provision of necessary infrastructure ............................................ 78 6.4 Promote food safety standards and regulations .......................... 78 7. Future perspectives .................................................................81 References ........................................................................................83 Popular summary ........................................................................... 103 Populärvetenskaplig sammanfattning ............................................ 105 Acknowledgements ........................................................................ 107 9 11 List of publications This thesis is based on the work contained in the following papers, referred to by Roman numerals in the text: I. Maureen Kuboka, Ianetta Mutie, Karin Artursson, Johanna F. Lindahl, Gunnar Carlsson, Florence Mutua & Delia Grace (2026). Prevalence of Escherichia coli, Campylobacter and Salmonella spp. in the East African Community: A Systematic Literature Review and Meta-Analysis. (Food Microbiology doi:org:10.1016/j.fm.2025.105004) II. Maureen Kuboka, Delia Grace, Karin Artursson, Johanna Lindahl, Gunnar Carlsson & Florence Mutua (2024). Food safety in informal public markets in Kenya: perceptions of stakeholders in the food chain. (Frontiers in Sustainable Food Systems doi:10.3389/fsufs.2024.1411318) III. Maureen Kuboka, Karin Artursson, Linnet Ochieng, Florence Mutua, Delia Grace, and Johanna F. Lindahl. Knowledge, Attitudes and Practices of Meat Sellers in Nairobi and Risk Factors Associated with Microbial Hazards (Manuscript submitted for publication). IV. Maureen Kuboka, Florence Mutua, Delia Grace, Johanna F. Lindahl and Karin Artursson (other co-authors to be confirmed). Food safety in Burundi: a mixed methods study. (Draft manuscript) Paper I and II are reproduced with the permission of the publisher and published open access. 12 The contribution of Maureen Kuboka to the papers included in this thesis was as follows: I. MK was involved in the conceptualisation and coordination of study activities. Developed the study protocol with input from the supervisors. Involved in the screening of articles as one of the two independent reviewers. Oversaw data extraction exercise and performed statistical analysis with support from supervisors. MK is the main writer of the manuscript and revised it after feedback from co-authors and journal reviewers. II. MK was involved in the conceptualisation, planning and coordination of study activities. Developed data collection tools, with support from the supervisors. Led the data collection team. Performed qualitative analysis of the texts with support from a qualitative data specialist. MK is the main writer of the manuscript revised it after feedback from co-authors and journal reviewers. III. MK was involved in the conceptualisation, planning and coordination of study activities. Revised data collection tools (which were adapted from a previous study), with support from the supervisors. Led the data collection team. Performed lab analysis with the support of a research assistant. Performed statistical analysis with support from the supervisors. MK is the main writer of the manuscript and revised it after feedback from co-authors and journal reviewers. IV. MK was involved in conceptualisation, planning and coordination of study activities, with support from supervisors and partners in Burundi. Led the data collection team. Performed lab analysis with the support of a research assistant. Performed qualitative and statistical analysis with support from the supervisors. MK is the main writer of the manuscript and revised it after feedback from co-author. 13 List of tables Table 1: Technical regulations on food safety control of animal source foods in Burundi ................................................................................... 24 Table 2: Percentage market share of retail outlets by income quantiles for fresh produce in Nairobi, Kenya ............................................................ 25 Table 3: Technical regulations on food safety control of animal source foods in Kenya ..................................................................................... 26 Table 4: Average level of E. coli, Salmonella and Campylobacter spp. in pooled food samples in EAC, January 2000 to June 2022....................... 62 Table 5 : Observations on meat handlers' personal hygiene and use of personal protective equipment in Nairobi, Kenya (n=170) ....................... 66 Table 6: Observations on environmental hygiene and sanitation factors around butcher shops in Nairobi, Kenya (n=170) .................................... 67 Table 7: Comparison of knowledge, attitude, and practice scores, as well as point prevalence of hazards, between meat sellers in the treatment and control groups in Nairobi, Kenya ........................................................... 69 15 List of Figures Figure 1: Map of East African Community countries, nested within Africa. Created with QGIS (https://qgis.org/) ..................................................... 43 Figure 2: Map of the study sites in Kenya (left) and Burundi (right). Created using MapChart (https://www.mapchart.net/) .......................................... 45 Figure 3: Poster on .handwashing guidelines for meat sellers in Nairobi, Kenya. Photo credit: ILRI ...................................................................... 48 Figure 4: Typical layout of a butcher shop in Nairobi, Kenya. Photo credit: author ................................................................................................. 49 Figure 5: Causal diagram illustrating potential variables associated with meat contamination in butcher shops in Nairobi, Kenya. Created using DAGitty (https://www.dagitty.net/ ) ......................................................... 54 Figure 6: A mix of raw and roasted meat products on sale in a butcher outlet in Nairobi, Kenya. Photo credit: author.......................................... 63 Figure 7: Mix of raw and roasted meat on sale at an outdoor kiosk in Bururi, Burundi. Photo credit: author ................................................................ 64 Figure 8: Young men slaughtering a pig in an open field, adjacent to sanitation facilities in Bururi, Burundi. Photo credit: author ...................... 73 Figure 9: Proportional piling method for ranking risky foods during an FGD in Burundi. Photo credit: author ............................................................. 73 17 Abbreviations AIC AfCFTA ASF ASTM CAC CTT DALY DRC EAC EAEC EC EHEC EIEC EPEC EU FBD FAO FDA GAIN ICC ILRI IRT KAP LMIC Akaike information criterion Africa continental free trade area Animal source foods American Society for Testing and Materials Codex Alimentarius Commission Classical test theory Disability Adjusted Life Years Democratic Republic of Congo East African Community Enteroaggressive E. coli European Commission Enterohaemorragic E. coli Enteroinvasive E. coli Enteropathogenic E. coli European Union Foodborne diseases Food and Agriculture Authority of the United Nation Food and Drug Authority of the United States Global Alliance for Improved Nutrition Item characteristic curve International Livestock Research Institute Item response theory Knowledge, attitude and practice Low- and middle-income country 18 ODK OR PP RTE SSA SLR STEC TBC WHO WOAH UNEP Open data kit Odds Ratio Percentage point Ready-to-Eat Sub-Saharan Africa Systematic Literature Review Shiga-toxin-producing E. coli Total bacteria count World Health Organisation World Organisation of Animal Health United Nations Environmental Programme 19 1. Introduction 1.1 Food safety globally Food safety is the assurance that food will not cause harm to consumers when it is prepared and eaten as intended (FAO & WHO, 2023). Foodborne hazards are harmful agents that are either naturally present in food or contaminate it at any stage in the value chain, making the food unsafe. These hazards can be microbiological, chemical, or physical (Barlow et al., 2015). Consuming unsafe food can result in serious consequences, including illness and death (Bisht et al., 2021; ECDC and EFSA, 2023). Ensuring food safety is crucial for public health, food security, and the social and economic well- being of individuals and nations. In 2015, the World Health Organisation (WHO) identified 31 hazards associated with foodborne diseases worldwide. Diarrheal disease agents were the leading causes of these illnesses, with Norovirus, Campylobacter spp., Salmonella spp., and pathogenic E. coli topping the list. Additionally, Taenia solium, Vibrio cholerae, Hepatitis A, and aflatoxin have significantly contributed to the overall FBD burden (WHO, 2015). Globally, unsafe food causes an estimated 600 million illnesses and 420,000 deaths each year. Additionally, over one million illnesses annually result from chemicals, primarily heavy metals (Gibb et al., 2019). The burden of foodborne diseases (FBD) is particularly severe in low-and middle- income countries (WHO, 2015). 1.1.1 Global food safety policy and regulation Food safety policies and regulations are established to ensure consumer protection and promote fair trade through harmonised standards (CAC, n.d.). The United Nations (UN), through the joint efforts of the World Health Organisation (WHO) and the Food and Agriculture Organisation (FAO), holds the mandate for global food safety policy and regulation (FAO, 2023; WHO, 2022). 20 The FAO and WHO work with member countries by providing scientific evidence and strengthening their capacity to deliver safe food for all. A key responsibility is to support the integration of food safety into national policies and assist in establishing infrastructure that minimises food safety risks. By using the FAO/WHO Food Control System Assessment Tool, countries can monitor and evaluate the performance of their food control systems, thereby promoting continuous improvement (FAO, 2023; WHO, 2022). Additionally, the FAO and WHO ensure the implementation of active surveillance and response systems within countries to monitor and address foodborne disease events. Through the International Food Safety Authorities Network (INFOSAN), FAO and WHO facilitate information sharing among countries to prevent, manage, and respond to food safety incidents (Savelli et al., 2019). The Codex Alimentarius Commission (CAC), jointly established by the FAO and WHO as the standard-setting body, serves as the international reference for food safety, aiming to protect consumer health and promote fair trade (CAC, n.d.). The FAO and WHO participate in the CAC by providing recommendations based on independent scientific evidence regarding food safety risks, ensuring that standards align with public health considerations. The UN quadripartite, comprising the FAO, WHO, the World Organisation for Animal Health (WOAH, formerly OIE), and the UN Environment Programme (UNEP), formed a collaboration in 2022 under a joint One Health plan of action. For food safety interests, this initiative aims to ensure the protection of public health and sustainability throughout the agri-food supply system (FAO et al., 2022). 1.1.2 Regional food safety control and legislation In developed countries, food safety governance is structured with distinct laws and oversight bodies responsible for surveillance, control, and management of food safety priorities. For example, the general food law created by the European Commission (EC) sets out principles for food safety, consumer protection, and trade within the EU. The Commission also established the European Food Safety Authority (EFSA), which provides 21 oversight and independent scientific advice on matters related to food safety (FAO & WHO, 2002). In the United States, the primary law governing food is the Federal Food, Drug, and Cosmetic Act (FDCA). The main agencies responsible for food safety regulation, inspection, and surveillance are the Food and Drug Administration (FDA) and the USDA Food Safety and Inspection Service (FSIS) (Lam & Patel, 2023). In Southeast Asia, the ASEAN-Food Safety Regulatory Framework was developed to guide food safety authorities on priority activities for ensuring food safety and consumer protection (Lin, 2018). The African Food Safety Authority was recently established by the African Union in 2025, to provide oversight for the regional harmonisation of food safety legislation and promote intra-regional trade (African Union, 2025). 1.2 Burden of FBD in Africa The burden of disease is measured in Disability Adjusted Life Years (DALYs), which represent years of life lost due to illness or disability and premature death. One DALY equates to one year of life lost (Devleesschauwer et al., 2014). The African population bears the greatest burden of FBD, estimated at 1,200 DALYs per 100,000 population, with children disproportionately affected. For instance, of the 59,000 global deaths due to non-typhoidal Salmonella, 32,000 were in Africa. The overall FBD burden in Africa translates to 91 million cases and 142,000 deaths annually (WHO, 2015). The loss in productivity is estimated at USD 20 billion, with an additional USD 3.5 billion spent on treatment costs (Jaffee et al., 2019). In Kenya, the FBD burden, based on estimates from five of 47 counties, is extrapolated to be 1,140 DALYs per 100,000 population, mainly due to diarrhoeal diseases. This represents 1% of the annual GDP loss (Hoffman & Baral, 2019). In Rwanda, FBD burden from the consumption of milk and dairy products contaminated with five bacterial agents was estimated at 3,870 DALYs and resulted in 55 deaths in 2010 (Sapp et al., 2023). 22 The economic burden due to FBD is estimated at USD 391 million in Burkina Faso and USD 793 million in Ethiopia, equivalent to 3.0% and 0.9% of gross national income, respectively (van Wagenberg & Havelaar, 2023). Despite the FBD burden being highest in Africa (WHO, 2015), poor surveillance systems and under-reporting have resulted in limited reports on the occurrence of FBD events (Grace, 2017; Sosah & Donkor, 2025). Additionally, in most FBD cases, it has been challenging to establish the source of infection. In the absence of surveillance reports and empirical evidence, expert elicitation has been relied upon, which has associated most FBD events with animal-derived sources and products consumed raw, although processed products have also been linked to FBD (Sapp et al., 2022). 1.2.1 Food safety regulation in Africa The role of the African Union The African Union recently launched the African Food Safety Agency (AFSA) to harmonise food safety policies across the continent, aiming to promote trade under the African Continental Free Trade Area (AfCFTA). As this is yet to be operationalised, traditional oversight by national authorities is in place. These government agencies formulate regulations based on the Codex Alimentarius, often resulting in variations across countries that can hinder trade. Additionally, countries rely on regional regulations, such as those of the East African Community (EAC) (FAO, 2023; Kang’ethe et al., 2020). 1.3 Food safety and regulation in EAC The EAC is a regional bloc in Africa, made up of eight partner states: namely, Burundi, Democratic Republic of the Congo (DRC), Kenya, Rwanda, Somalia, South Sudan, Tanzania (United Republic of Tanzania), and Uganda. The regional bloc was established in 1999, with the core objectives of strengthening political, social and economic ties, and the establishment of a common market for trade of goods and services (East African Community, 2002). 23 Agricultural products are the top-most traded goods across EAC, with animal source foods (ASF) ranking third, including dairy products, honey and eggs. Additionally, the trade of live animals and meat has been rapidly increasing at a rate of 13% from 2017 to 2022 (Chatham House Resource Trade Database, 2022). Despite this, harmonised policies across countries are lacking; instead, each partner state applies its own regulations established by the national food control authority. In the EAC Secretariat, food safety is overseen by the Technical Working Group (TWG) on medicines and food safety, which is nested within the health program, with a focus on disease surveillance, epidemic prevention and control. The TWG is currently working on a draft food safety and quality policy (EAC, 2025). 1.3.1 Food supply and food safety regulations in Burundi Burundi is a country in EAC which has a total population of 14.4 million, with over 85% living in rural areas and depending on agriculture for food and livelihoods (FAO, 2015). Poverty levels are high, with about 70% of the population living on less than USD 3 per day (World Bank, 2025). The hunger index is provisionally estimated at > 40 due to a lack of data, compared to a global average of 18, thus considered alarming (GHI, 2025). Poverty and food insecurity have been identified as drivers of poor diets, which can further exacerbate foodborne infections (Stehl et al., 2025). Several policies are in place regarding agriculture, food security and nutrition in the country; however, these do not directly address food safety. Various technical regulations, including laws, decrees, and ministerial ordinances, govern the safety of animal-derived foods, as listed in Table 1. Responsibility for food safety is divided among different ministries: agriculture, health, and trade, which can create loopholes in the execution of mandates (Niragira et al., 2020). Research on food safety has been limited; consequently, there is a lack of understanding of the extent of food safety challenges and the burden of foodborne disease. Only a few studies are available on milk quality and aflatoxins (Iribagiza et al., 2024a; Iribagiza et al., 2024b; Udomkun et al., 2018). In addition, recent studies have reported the endemicity of foodborne parasites (Minani et al., 2022; Minani et al., 2024). The levels of pathogens 24 in foods and exposure pathways that can lead to public health risks are understudied. Table 1: Technical regulations on food safety control of animal source foods in Burundi Technical regulation Ministry in charge Purpose Law No. 1/28 of 24 December 2009 Ministry of Environment, Agriculture and Livestock Ensure the health of domestic, wild and aquaculture animals, and bees Ministerial Ordinance No. 710/652 of 3 April 2019 Ministry of Environment, Agriculture and Livestock Sets the conditions and procedures for the health inspection of animals and food products of animal origin Ministerial Ordinance No. 710/653 of 3 April 2019 Ministry of Environment, Agriculture and Livestock Examination of slaughter animals and veterinary health inspection of meat and meat products Ministerial Ordinance No. 710/654 of 3 April 2019 Ministry of Environment, Agriculture and Livestock Govern slaughterhouses and slaughter areas Law No. 1/17 of 30 November 2016 Ministry of Environment, Agriculture and Livestock Organisation of fisheries and aquaculture Decree-Law No. 1/16 of 17 May 1982 Ministry of Public Health and Acquired Immunodeficiency Syndrome (AIDS) Control Establish the public health code Modified from: Niragira et al., (2020). 1.3.2 Food supply and food safety regulations in Kenya Kenya has a population of approximately 47 million, with about 30% of people living below the poverty line (Kenya National Bureau of Satistics, 2023). Most of the low-income earners rely on informal markets for food and nutrition (Muunda et al., 2023; Tschirley et al., 2004; van der Lans et al., 2012). Informal markets are therefore common in Kenya and include open- air markets, small shops, kiosks, and eateries. 25 Table 2 displays the market share of various retail outlets for fresh produce. Fresh vegetables, fruits and fish are obtained from open-air markets and kiosks, while meat and milk are mainly sold in shops and kiosks (Koech et al., 2024; Muunda et al., 2023; van der Lans et al., 2012). In addition, supermarkets are emerging, supplying about 20% of fresh foods, and expanding with rising incomes and the growing middle class (Neven & Reardon, 2004; Tschirley et al., 2004). Table 2: Percentage market share of retail outlets by income quantiles for fresh produce in Nairobi, Kenya Source:(van der Lans et al., 2012) The main laws on food safety are the Food and Drug Act of 1978, revised in 2023, and the Public Health Act of 1921, revised in 2002. Additionally, various other laws and regulations govern food safety. Table 3 summarises the regulations enacted to promote the safety of ASF. Despite a well- established legal framework, there are gaps in enforcement, resulting in a weak control system. Furthermore, most of these regulations are outdated and have not addressed emerging issues such as the expanding informal food sector. The national food policy of 2021 acknowledges the scale and importance of the informal food sector; however, specific guidelines relevant to the sector are missing from the policy (The National Food Safety Policy, 2021). Similar to Burundi, food safety regulation is delegated to various departments within the ministries of agriculture, health, and trade. The multiplicity of agencies makes regulation and control complex, as it leads to duplication of roles and responsibilities, which can result in some areas being overlooked. Income quantile Super- market chains Mini markets Green- grocer shops Open- air markets Kiosks Hawkers 1-lowest 0.0 0.1 0.0 64.7 41.5 3.6 2 0.0 0.0 0.0 60.6 34.4 3.7 3 0.4 0.7 0.0 60.7 36.4 1.8 4 1.7 0.1 0.0 59.5 38.0 0.4 5-highest 13.7 0.4 1.7 47.8 32.6 2.4 26 Efforts to establish an overarching body to coordinate food safety activities have not yet been successful (Kang’ethe et al., 2020; Oloo, 2019). An effective national food safety control system can be divided into four pillars, including legal frameworks, surveillance and control, evidence and communication (Kwoba et al., 2023). Kenya seems to have a functional legal framework, but lacks in surveillance, evidence and communication (Kang’ethe et al., 2020 ; Oloo, 2019). Risk-based studies that generate evidence to drive investment in food safety surveillance and control systems are warranted. Table 3: Technical regulations on food safety control of animal source foods in Kenya Technical regulation Ministry in charge Purpose Public Health Act, Cap 242 Ministry of Health • To make provision for securing and maintaining health • Handle matters related to food (anything other than water and drugs intended for food preparation) Food, Drugs and Chemical Substance Act, Cap 254 Ministry of Health • To make provision for the prevention of adulteration of food, drugs and chemical substances • Prohibit the sale of food that is unfit for human consumption (unwholesome, adulterated, poisonous, harmful) or food not of the nature and quality demanded Animal Diseases Act, Cap 364 Ministry of Agriculture, Livestock and Fisheries • Isolate infected animals intended for slaughter • Bury or destroy carcasses • Slaughter and dispose of carcasses from animals declared infected, with compensation • Examine animals, carcasses, meat and animal products for evidence of notifiable diseases Standards Act, Cap 496 Ministry of Investment, Trade and Industry • Promote standardisation and provide for testing of products • Examine, test and sample of commodities • Inspect processes and manufacturing premises 27 Fisheries and Management and Development Act of 2016 Ministry of Investment, Trade and Industry • Conservation, management and development of fisheries and aquatic resources. Meat Control Act, Cap 356 Meat Regulations No 10 of 2010 Ministry of Investment, Trade and Industry • License, inspect and set standards for slaughterhouses and meat for local consumption and export • License and inspect slaughterhouses and premises where meat is processed for human consumption • License meat inspectors • Make regulations on import of meat and meat products, local slaughterhouse licensing and export slaughterhouse regulations • Ensure quality through setting of standards for processing meat products and additives; storage and transport of meat and meat products • Surveillance (sample meat products and food additives) • Control import and export of meat • Ensure meat is wholesome and fit for human consumption Dairy Industry Act, Cap 336 Ministry of Agriculture, Livestock and Fisheries • Improve efficiency of production, quality and marketing of dairy produce (milk, cream, butter, ghee, cheese and any other byproducts of milk) • Prescribe handling, transport and storage of dairy produce intended for consumption and the way produce shall be sold • Inspect dairy products to ensure they meet the quality standards • Surveillance (collect and test dairy) Kenya Meat Commission Act, Cap 363 Ministry of Agriculture, Livestock and Fisheries • Regulates purchase, slaughter stock (cattle, small stock, poultry), slaughter and market products, and ensure cold chain for slaughtered products • Establish and maintain abattoirs and cold storage for meat, meat products and eggs 28 Trade Description Act, Cap 505 Ministry of Investment, Trade and Industry • Prohibit false trade descriptions • Prohibit false indications of price • Prohibit import of goods with false descriptions of origin • Inspect premises and seize and detain goods and documents Cooperative Act, Cap 490 Ministry of Micro, Small and Medium Enterprises Development • Run societies by keeping and auditing accounts (dairy cooperatives are governed under this Act) Modified from: Kang’ethe et al., (2020). 1.4 Food safety in informal markets Informal food markets are defined as non-modern, unregistered outlets where food is sold to consumers as raw or processed products. The informal food sector is heterogeneous and includes open-air markets in urban and peri- urban areas, small shops and kiosks, eateries and food service centres and micro and small processing cottages (Roesel & Grace, 2014). Informal markets supply over 80% of food consumed in LMICs (Blackmore et al., 2020; Hannah et al., 2022; Tschirley et al., 2004) and act as a ready market for produce for smallholder farmers (Ferris et al., 2014). Due to the informality in structure, the sector often evades government control, taxation, social protection guidelines, labour laws, and employment benefits (Vanek et al., 2014). 1.4.1 Actors Employment in the agricultural sector accounts for 57% of total employment in the population of Sub-Saharan Africa (SSA), while informal employment in non-agricultural work represents 66% of total employment (Vanek et al., 2014). Despite this significance, there is limited information on actual employment rates in the informal food production and distribution sector; therefore, it is unclear where employment in the informal food supply chain is reflected in the data above. However, existing literature indicates that informal value chains provide numerous employment opportunities for individuals involved in production, processing, transportation, and market distribution as wholesalers and 29 retailers (Kiambi et al., 2018; Murungi et al., 2021; Nordhagen et al., 2023). Henson et al. (2023) argue that the number of retail vendors alone in mid- sized cities could be in the tens of thousands. Food supplied to markets is predominantly produced by smallholder farmers (Dixon et al., 2001; Lowder et al., 2021), who own about two hectares of land in rural areas (Lowder et al., 2016) and often engage in crop cultivation alongside livestock keeping, fishery, or poultry farming. These farmers are considered the primary actors in the informal system. Within the supply chain, multiple levels of traders operate between producers and consumers, making the food chain longer and traceability systems challenging (Carron et al., 2017; Kiambi et al., 2020; Murungi et al., 2021; Nordhagen et al., 2023). In addition, informal value chains are characterised by middlemen, who act as intermediaries between different levels of traders (van der Lans et al., 2012). Women and youth constitute a critical labour force in the informal food value chains, using them as a means of income generation and livelihood (Hall, 2017; Mugo & Kinyua, 2023; Raney et al., 2011). While underage children are sometimes involved in trading activities, this issue often goes undocumented for ethical reasons (Nordhagen et al., 2023). Some value chains, such as large animal meat, are more male-dominated (Murungi et al., 2021), while value chains such as poultry and milk sale are more female- dominated (Tangka et al., 2000; Tung, 2005). Low-income earners and poor households are the main consumers in informal markets, as they use this channel to meet their nutritional needs at low expense (Tschirley et al., 2004; van der Lans et al., 2012). Other reasons why informal markets thrive are that they offer variety of fresh foods, are conveniently located, and provide friendly pricing as well as social ties (GAIN, 2022; Isanovic et al., 2023). 1.4.2 Governance, policies and food safety control Informal value chains are characterised by numerous small players, which poses a challenge in food governance and control. In Kenya and other LMICs, food safety in domestic markets is managed at the local government level (The National Food Safety Policy, 2021). This decentralised approach is considered more suitable for governance as it enables closer administrative 30 support and service provision for local communities (Jafee et al., 2021). Local governments are mainly responsible for architectural planning of markets, providing sanitation services, conducting surveillance, and collecting levies. These markets are also overseen by market heads or market committees, who manage daily operations and maintain order (Davies et al., 2022; Henson et al., 2023). However, a significant challenge is that municipal governments often lack the capacity to effectively implement food safety measures as set out in laws and policies. Limited public resources, competing priorities, and lack of data to support the urgency of food safety matters further hinder progress in food safety governance in local markets. This inadequacy exposes the public to health risks (Jafee et al., 2021). One policy option presented by local governments is the formalisation of the sector through enforcement of regulations, registration, licensing or resettlement in new business areas (Henson et al., 2023). However, when this is done without the participation of the actors, it creates resistance and a disconnect between food control agencies and informal market actors (Grace et al., 2019; Resnick, 2019). Young & Crush (2020) note that governance of informal value chains must occur in an enabling environment that supports the core players. They further observe that the success of policy and regulation in the informal sector depends on the inclusion and cooperation of informal actors, with a participatory approach adopted in decision-making and design of interventions. Capacity building among the actors through the provision of training and technologies is also recommended as a more promising approach, which involves informal actors and improves knowledge and hygiene practices (Grace, 2023; Young & Crush, 2020). Additionally, reorganising actors into cooperatives as an initiative driven by the actors themselves and supported by the government and private sector, can enhance local government's ability to conduct surveillance, provide support more effectively, facilitate access to premium markets, and ensure product safety and quality (Kumar et al., 2013; Nyokabi et al., 2023). 31 1.4.3 Infrastructure Infrastructure at informal markets is essential for preserving quality, reducing food loss and waste, and preventing health risks. This includes hygiene and sanitation infrastructure, slaughter facilities, and storage and transport facilities. However, these have often been found to be inadequate, missing, or poorly maintained (Henson et al., 2023). Such infrastructural challenges hinder the provision of safe food in informal markets. For example, (Nyokabi et al., 2021) observed that milk in the informal chain was transported without cold-chain storage, which could accelerate its deterioration. In contrast, in more formal chains, milk cooling facilities were strategically located in rural areas where milk was aggregated to preserve its quality and shelf-life before being transported to processing plants. Additionally, lactometer and alcohol tests, along with other essential tests, were also conducted at milk reception centres to verify milk quality, which was not the case in informal chains. In Kenya, the Meat Control Act provides guidelines for the control of slaughter activities and the sale of meat (The Meat Control Act, 1972). Ante- mortem and post-mortem examinations are conducted at the slaughterhouse to ensure that the meat is safe for consumption. However, some animals bypass these inspections due to a shortage of veterinary inspectors, who may be required to work across multiple facilities, leaving certain areas unattended (Cook et al., 2017; Otoigo et al., 2024). Poor transport facilities, road and transport networks, and a lack of specialised tools can also undermine surveillance efforts. In the absence of testing and verification, informal actors tend to rely on sensory cues to assess food quality and safety (Bukachi et al., 2021), yet physical checks alone are insufficient to detect pathogenic contamination (Minani et al., 2024). Trained personnel, modern facilities, and rapid tests are essential to enhance surveillance activities, enabling the quick detection of foodborne risks and the implementation of appropriate prevention measures. Additionally, access to a stable supply of clean water and electricity supports hygiene measures and proper food storage. Economic support through credit facilities can provide capital to invest in simple technologies 32 and equipment necessary for food safety, as well as enable actors to comply with other regulatory requirements (Young & Crush, 2020). 1.5 Application of risk assessment in informal markets 1.5.1 Risk assessment studies Risk assessment is the gold standard in food safety research and has been used to identify, evaluate, and prioritise potential hazards to determine the likelihood and severity of harm. Risk assessment is a scientific process comprising four steps: hazard identification, hazard characterisation, exposure assessment, and risk characterisation. A hazard is not necessarily a risk; therefore, risk-based approaches, rather than hazard-based approaches, are applied in decision-making, establishing standards and guidelines to protect public health and facilitating trade (Barlow et al., 2015). Despite its widespread application in the formal food supply chain, risk assessment has been little applied in informal food systems. This may be due to the complexity of these systems, which often involve non-linear flow of goods and people, with limited traceability (Grace et al., 2010). However, adaptive risk-based approaches can be applied in informal markets. One such approach is hazard analysis critical control point (HACCP) (Oloo et al., 2017; Zakhia & Program, 2002). Another is an integrated approach in research involving quantitative and qualitative methodologies to generate empirical evidence for decision-making and risk mitigation in informal market systems (Häsler et al., 2019). In an integrated approach, focus group discussions (FGDs) with ranking techniques, ethnographic studies, prevalence surveys, literature reviews, and demographic surveys can be used for hazard identification and characterisation. Ethnography, observational studies, and modelling can be applied in exposure assessment. Risk characterisation can be calculated by integrating the results using probabilistic methods such as Monte Carlo simulation (Grace et al., 2008). Based on the results, proper interventions and mitigation strategies to reduce risk can be applied. In our case, an integrated risk-based approach was used for hazard assessment. This involved conducting systematic literature reviews to quantify hazard levels and identify priority foods. Participatory approaches 33 were used to determine priority foods from the community's perspective and to map exposure pathways. Additionally, a survey assessing knowledge, attitudes, and practices, along with a prevalence study, was conducted to understand the underlying cognitive factors among informal actors, the point prevalence of hazards, and their associated risk factors. Systematic literature reviews Systematic literature reviews (SLRs) are widely regarded as the most reliable source of evidence for evidence-based practice. SLRs aim to synthesise findings from multiple studies using a rigorous and transparent protocol, ensuring consistency and minimising bias. When combined with meta- analysis, SLRs provide quantitative summaries that strengthen the overall evidence base to answer specific research questions and identify the gaps (Dekkers et al., 2019; Paul & Leeflang, 2021). Having originated in the medical field, they have primarily been used to test different medical interventions (Higgins & Thomas, 2024). The value of SLRs lies in their ability to present comprehensive, unbiased evidence. However, their validity can be compromised by inadequate appraisal or the inclusion of low-quality studies (Dekkers et al., 2019; Higgins & Thomas, 2024). Few SLRs have been conducted to analyse levels of priority pathogens in foods on the African continent, where evidence on the burden of foodborne disease has been scarce. Even so, only a handful have been conducted in EAC. Our study was conducted to understand the levels of priority hazards in foods and their risk factors. This evidence can contribute to the estimation of disease burden in Africa. Participatory approaches Participatory epidemiology was first used in pastoral areas of East Africa to investigate livestock disease patterns (Catley, 2000; Chambers, 1994). It has since evolved and been applied in public health epidemiology, food security, nutrition, and socio-economic studies (Catley et al., 2018; Häsler et al., 2019; Roesel et al., 2019; Wong et al., 2018). Participatory tools include ethnography, FGDs, key informant interviews (KIIs), observations, ranking and scoring, and pairwise matrices. These tools promote adult learning and help bridge the gap between literate and illiterate 34 populations. Scores obtained can be analysed quantitatively, providing more insight into the question under study. Triangulation is also used to confirm findings and improve the reliability of results (Alders et al., 2020). Participatory approaches can be time-consuming, require proper logistical planning and support, and demand appropriate training and skill in the use of methods and tools. Additionally, due to their socio-cultural approach, some scientists have a negative attitude towards participatory approaches, which may limit their application (Catley, 2000). Despite their limitations, participatory methods are said to be more cost- effective and useful in low-resource settings, complementing conventional risk-based methods and integrating indigenous knowledge into epidemiology and risk prevention. They are also applied to understanding exposure pathways, social perceptions and drivers of risk in community settings, which may not be captured through quantitative methods or clinical diagnostics (Catley & Mariner, 2002). Participatory approaches have been recommended as a risk-based method for understanding exposure to risks and describing dynamic processes in informal market systems. They also involve communities and various levels of actors and are therefore considered a more ethical approach to governance, planning, and decision-making regarding food safety in informal systems (Grace et al., 2008). However, few studies have attempted to use these methods in food safety. Existing studies focus on animal health and disease patterns (Dione et al., 2014; Nthiwa et al., 2019). We employed a participatory approach through FGDs, KIIs and a stakeholder meeting to investigate perceptions on food and food safety and assess priority hazards and risk factors among communities in Burundi and Kenya. Questionnaire surveys Questionnaire surveys are used to collect data from a representative subset of the population, and the results can be generalised to the entire population. Questionnaire surveys should be probabilistic, with each participant having a non-zero and equal chance of participating, thereby reducing selection bias and providing accurate insights into the population. Questionnaire surveys involve three phases: question design, sampling, and data collection. 35 Participation involves providing information by answering questions (Fowler, 2009). An example of a questionnaire survey is a knowledge, attitude and practice (KAP) survey. This is a behaviour change learning model that posits knowledge and attitude as unobservable constructs. The model supports that knowledge and attitude are correlated and that knowledge forms and shapes attitudes, which in turn influence behaviour (Valente et al., 1998). KAP surveys rely on structured and standardised questionnaires to assess what people know, believe, and do within a given population. Knowledge refers to the understanding of a subject (which can be scientific, social, indigenous, implicit, or self-reflective) (Hulme, 2018). Attitude refers to personal feeling or positioning regarding a subject, while practice refers to regular activities that constitute a behaviour. Other factors that can influence practices include beliefs, socio-cultural norms, demographic factors and environmental factors (Macías & Glasauer, 2014). These types of questionnaire surveys can be limited because they are more extractive; people are unable to express their views or influence the process. Conventional food surveys also provide point prevalence of food contamination with limited information on seasonal variation. They are also time- and resource-consuming, and require logistical support, especially if face-to-face interviews are conducted. Other studies have questioned the validity of KAP, based on the reliability of measurement scales. However, this limitation can be addressed using psychometric scales (Dohoo & Emanuelson, 2021) Nonetheless, KAP surveys play an important role in public health research in situational analysis of knowledge gaps, cultural beliefs, and behavioural patterns that influence risk. They can also highlight priority needs and barriers to the effective delivery of intervention programmes. KAP surveys provide evidence for designing targeted interventions, communication strategies, and public health programmes to address risky practices. They are also valuable for establishing baseline levels and evaluating changes following interventions. (WHO, 2008). Fewer studies have been done based on the KAP learning model, especially in studying the safety of ASF in informal markets. Some studies have adopted a participatory approach to understand KAP among meat 36 consumers (Roesel et al., 2019). Others have used structured questionnaires (Kimindu et al., 2024), while some have focused on practices and their association with demographic characteristics among slaughterhouse workers (Wambui et al., 2017). We employed KAP and prevalence surveys to understand factors associated with meat contamination in the retail node in Kenya. Intervention studies Intervention studies are recommended to improve food safety knowledge and promote behaviour change among stakeholders. Common strategies employed in food safety interventions include; capacity building, communication campaigns, and the provision of technologies and infrastructure (Madjdian et al., 2024; Malavi et al., 2021; Takeuchi et al., 2017). While most interventions emphasise training, evidence suggests that education alone is not sufficient to influence behaviour (Singh et al., 2016). The most effective outcomes are achieved when multiple approaches are combined. These typically involve education-based sessions conducted by trained personnel, dissemination of media messages such as radio broadcasts and audio-visual content, provision of tools, and the creation of an enabling environment to reinforce key messages and sustain change (GAIN, 2021; Kwoba et al., 2023; Takeuchi et al., 2017). Several factors consistently contribute to the success of interventions. Programmes delivered regularly over extended periods, often using a longitudinal and participatory approach, tend to be more impactful (Lindahl et al., 2018). Stakeholder consultations and iterative implementation with monitoring and evaluation strengthen engagement (Samaan et al., 2012). Considering contextual factors, such as tailoring training to specific groups based on cultural norms, educational levels, and individual needs, is also important (Singh et al., 2016). The design of interventions is also significant. For example, randomised controlled trials are considered more reliable, while observational studies are more prone to bias (Grace et al., 2018). Despite achievements in food safety interventions, challenges remain. High turnover and the itinerant nature of market actors may disrupt continuity, whereas limited time and commitment to programme activities 37 can hinder long-term impact (Singh et al., 2016). Additionally, a poor policy environment and lack of support from regulatory authorities, key infrastructural challenges, and discontinuity in access to tools and technologies previously provided by research teams raise concerns about the scalability and sustainability of interventions (Bekele et al., 2025; Grace et al., 2019). 39 2. Aims of the thesis The overall aim was to understand the food safety situation in the domestic markets of East African Community countries, the hazards that occur in food, contamination levels, associated risk factors and effectiveness of a food safety intervention. The specific objectives were; 1. To synthesise literature on priority foodborne disease hazards and risk factors in seven EAC countries. 2. To understand food safety and foodborne hazards in Kenya from the perspective of key stakeholders in the food chain. 3. To assess the effectiveness of a non-randomised intervention programme in Kenya, based on KAP, handwashing practices, and bacterial prevalence between treatment and control groups. 4. To understand food safety and foodborne hazards in Burundi from the perspective of key stakeholders in the food chain. 5. To assess the prevalence of food safety hazards in meat in Burundi 41 3. Materials and methods 3.1 Study design The overall thesis followed a mixed methods design, encompassing a secondary review of literature, qualitative methods (FGDs, KIIs and stakeholder meetings), and quantitative methods (KAP and prevalence survey). 3.2 Ethical Considerations (Papers II-IV) All studies involving human participation (Papers II-IV) were conducted adhering to ethical standards and considerations. Institutional ethical approvals were obtained from the International Livestock Research Institute (ILRI), followed by national approvals according to established guidelines in the respective countries. 3.2.1 Informed consent (Papers II-IV) Participants were informed about the study and its objectives before data collection began. Those who agreed to participate provided written informed consent, which included consent to take pictures. Participation was voluntary, with no victimisation or consequences regarding future activities. Participants were free to withdraw at any time during the interviews. 3.2.2 Anonymity and pseudonymisation (Papers II-IV) Participants were assured of anonymity before taking part in the interviews. The identities and geographical locations (global positioning system, GPS coordinates) of the butcher shops and markets where meat sampling was done (Papers III and IV), were recorded during the visit for potential follow- up, such as future intervention programmes. These details were included in the dataset but were not presented in the results during manuscript preparation or journal publication. Names and phone numbers of participants were collected after the interviews solely for compensation purposes and were not included in the 42 datasets nor made public. Instead of their real identities, pseudonyms were used to identify butcher shops/meat stalls visited and individual participants during data analysis and manuscript preparation. 3.2.3 Compensation (Papers II-IV) Participants in FGDs and KIIs were compensated for the time spent during the interviews. Compensation was agreed upon with government and ILRI officials, ensuring it was clear this was not payment for information, but compensation for time spent during the interviews. A rate equivalent to USD 10 was used in both countries. Meat samples (100 g) used for microbiological analysis were purchased at prices set by the butcher shop or meat seller. 3.2.4 Use of large language models and artificial intelligence AI-powered Google Translate (https://translate.google.com/) and MS Word built-in document translator were used in language translation (Paper IV). Language proofing for the thesis and papers was done using Instatext (https://instatext.io/) and Microsoft Copilot (https://copilot.microsoft.com/). 3.3 Study sites (Papers I-IV) The studies were conducted in East Africa, with primary data collected in Burundi and Kenya, and secondary data gathered from five of the seven EAC countries. Due to limited food safety research in Burundi, as established by a previous scoping review (Mutua et al., 2021), the country was identified as a suitable site for generating evidence to inform policy and decision-making. Kenya was purposively selected as a potential representative of a vibrant informal economy within the EAC and other LMICs. Figure 1 is a map of Africa, with the EAC highlighted. Paper I: The systematic literature review was conducted across EAC countries to cover all seven member states at the time (as of 2022): Burundi, Democratic Republic of Congo, Kenya, Uganda, South Sudan, Tanzania, and Rwanda. Paper II: The setting was the informal markets in Kenya. Two markets were selected: Kangemi Market in Nairobi and Masii Market in Machakos, representing urban and rural settings, respectively. 43 Figure 2 (left) shows the locations of the study sites. Markets were selected to reflect urban and rural settings and for their broad coverage considering, adequate representation of retail and wholesale actors, and multiple value chains, including ASF, fruits and vegetables, and ready-to-eat foods. Figure 1: Map of East African Community countries, nested within Africa. Created with QGIS (https://qgis.org/) Paper III: The setting was Nairobi, Kenya. Four administrative units were chosen as a follow-up to a previous study (Koech et al., 2024) and a food safety intervention carried out in the area. Paper IV: Qualitative component The setting was informal markets in Burundi. Two markets were selected: Cotebu Market in Bujumbura and Matana Market in Bururi, representing urban and rural settings, respectively. Figure 2 (right) shows the location of the study sites. These markets were chosen for their strategic location, accessibility, daily sales, inflow of consumers and goods, the presence of 44 both retail and wholesale actors, and the opportunity to include actors representing various value chains, such as ASF, fruit and vegetables, and ready-to-eat food products. Quantitative component The setting was open-air markets selling meat in the capital, Bujumbura. Bujumbura Mairie, the urban province consisting of three communes (northern, southern and central), was selected due to ease of access, proximity to the laboratory, and the likelihood of finding a representative number of meat-selling stalls. All three communes were included in the sampling. 3.4 Study populations and selection of participants (Papers I-IV) Paper I: The SLR included foods sampled from domestic markets in EAC countries. Paper II: Traders and consumers were selected as participants in FGDs. In addition, stakeholders within the value chain, including animal health practitioners, public health officials, market and community leaders, and researchers, were involved in KIIs. Selection was done with the assistance of market heads, community leaders and officials from the Department of Public Health. Paper III: Sampling included the area where the qualitative study in Paper II was conducted to allow for a comprehensive understanding of the food safety situation. Meat sellers in butcher shops in Nairobi formed the sampling population. A mapping exercise had been conducted in the study area, which established that there were 430 butcher shops, which formed the 45 3.5 Study populations and selection of participants (Papers I-IV) Paper I: The SLR included foods sampled from domestic markets in EAC countries. Paper II: Traders and consumers were selected as participants in FGDs. In addition, stakeholders within the value chain, including animal health practitioners, public health officials, market and community leaders, and researchers, were involved in KIIs. Selection was done with the assistance of market heads, community leaders and officials from the Department of Public Health. Paper III: Sampling included the area where the qualitative study in Paper II was conducted to allow for a comprehensive understanding of the food safety situation. Meat sellers in butcher shops in Nairobi formed the sampling population. A mapping exercise had been conducted in the study area, which established that there were 430 butcher shops, which formed the sampling frame. In addition, a food safety intervention had been conducted Figure 2: Map of the study sites in Kenya (left) and Burundi (right). Created using MapChart (https://www.mapchart.net/) 46 among meat sellers in the area. Participants were therefore selected at random from the list and categorised into two; treatment (had participated in the intervention) and control groups (had not participated in the intervention). The minimum allowable statistical power of 0.8 and an expected effect size of at least 0.4 were used in sample calculation (Cohen et al., 2003). The total calculated sample size was 210, comprising 70 participants who had taken part in the intervention and 140 who had not. These sample sizes were considered sufficient to detect differences between participants in the intervention and control groups, minimise Type II errors and for practical feasibility. However, during data collection, only 59 of the 70 who participated in the intervention were available and willing to take part in this study (84% of the 70). Additionally, only 111 (79% of 140) butcher shops could participate as controls. This reduced the actual sample size to 170 butcher shops, compared to the calculated sample size of 210. This difference in sample size is a limitation of the study, as it may have affected statistical power and the ability to detect small effect sizes. Paper IV: Qualitative study Similar to Paper II, traders and consumers were involved in FGDs, while other stakeholders were involved in KIIs. Quantitative study A scoping visit was conducted before the main data collection, which identified a total population of 200 meat-selling stalls in nine markets across three communes. All nine markets were included in the sampling. Assuming a 50% prevalence of microbial pathogens, based on the literature (Dang- Xuan et al., 2019; Dinede et al., 2023; Gazu et al., 2023), with a 10% margin of error and a 95% confidence interval, a sample size of 96 was calculated. After applying the finite population correction factor, adjusting for clustering, and accounting for possible non-participation, a final sample size of 103 was used. Meat-selling stalls were selected randomly using random 47 numbers from the obtained list. In markets where there were too few stalls (<10), all were sampled. 3.6 Questionnaire development and interviews (Papers II and IV) 3.6.1 Focus group discussion and key informant interview guides (Paper II and IV) FGD and KII guides were developed in consultation with food safety experts. The guides included probing questions to ensure clarity, and the facilitator followed up with additional questions during the interviews as needed. The guides included questions on food choices, how and where food purchases are made, practices that compromise food safety along the chain, the occurrence of foodborne illness in the community, food safety challenges, and proposed solutions. These provided data about priority foods for food safety interventions, consideration of food safety in purchase decisions, risky practices that can compromise food safety, foodborne disease occurrence patterns and how the community manages food safety and foodborne illnesses. 3.6.2 Focus group discussions and key informant interviews (Paper II and IV) The tools were piloted before use in both countries. For use in Burundi, the tools were translated into French and Kirundi by local speakers. Interviews were therefore conducted in English and Swahili (in Kenya) and in French and Kirundi (in Burundi). For triangulation, questions asked of traders were also posed to consumers and key stakeholders. In addition, key informants were asked technical questions relevant to their fields of expertise, including veterinary public health, human health, community and market leaders, consumer organisation representatives, and researchers. In total, 16 FGDs and 33 KIIs were conducted in both countries. Two stakeholder meetings involving consumers, traders and key informants were held in Kenya to confirm results from the FGDs and KIIs. This was not possible in Burundi due to time and cost constraints. 48 3.7 Evaluation of the effectiveness of a food safety and hygiene intervention programme (Paper III) A non-randomised intervention programme was conducted in December 2023 by ILRI and other partners. The intervention included a training workshop for meat sellers, held at the four locations where a survey had previously been conducted. Meat sellers were trained on hygiene, handling, and proper handwashing to reduce microbial contamination in meat. Tools were provided to participants, including a poster with a message on proper handwashing (Figure 3), which was displayed on the walls of the meat shop, as well as personal protective equipment (PPE) and plastic and bamboo cutting boards. Based on the intervention, the participants were divided into treatment and control groups, as described in section 3.4. Our team utilised the KAP questionnaire and sample analysis to assess whether there were changes in KAP and meat contamination between the two groups, following the intervention (Paper III). In addition, hand rinse water was collected to assess the effectiveness of handwashing practices among meat sellers following the intervention programme. Figure 3: Poster on handwashing guidelines for meat sellers in Nairobi, Kenya. Photo credit: ILRI 49 3.7.1 Knowledge, attitude and practice questionnaire The questionnaire used was adapted from a previous study (Koech et al., 2024). Questions were reviewed for clarity in consultation with experts, and a pilot was done before actual use. Questions included demographic and business characteristics, source of meat and transport means, cleaning routine, KAP questions, and possible incentives for improving food safety. These provided data on demographic, business and behavioural factors that could be associated with meat contamination. 3.7.2 Questionnaire Interviews Face-to-face interviews with participants in both treatment and control groups were conducted using the open data kit (ODK) tool (https://getodk.org/ ). One-on-one interview approach minimised non- response errors, as the main questions were set as mandatory in the tool and therefore could not be skipped. Where the respondent was unfamiliar with Multipurpose cloth used for wiping hands and surfaces Tree stamp used as cutting board Offal for sale Bucket used for storage and transport of meat Carcass hang on hooks for sale Figure 4: Typical layout of a butcher shop in Nairobi, Kenya. Photo credit: author 50 English, the questions were translated into Swahili during the interviews, further allowing better comprehension and evading non-response. Alongside the questionnaire interviews, observation was conducted using a checklist that covered cleaning and sanitation, handwashing practices, use of PPE, and hygiene and handling of meat. 3.8 Microbiological analyses (Papers III and IV) The meat samples were collected at the end of the interviews and processed to evaluate the levels of E. coli and coliform bacteria using MC-Media Pads (MC-Media Pad®; Millipore, Merck Life Science, Germany). This method was selected as a convenient and rapid approach for detecting E. coli and coliforms. Based on colour coding, selected E. coli colonies were picked and sub-cultured on MacConkey agar for confirmation. Detection of Salmonella spp. was carried out according to the ISO 6579- 1:2017 horizontal method for the enumeration and detection of Salmonella spp. This culture-based method was selected because it includes enrichment steps that increase the likelihood of detecting Salmonella spp. Other rapid methods, such as PCR, are sensitive and provide more information on the genetic characteristics of the microbes. However, due to cost, time constraints, and limited access to the technology in some study areas (Burundi), culture-based methods were used. In addition to microbiological analyses of meat, hand rinse water was collected to examine the total bacterial count (TBC) of the hands (Paper III). The method used was adapted from that applied in healthcare (ASTM International, 2021). This method was considered appropriate as it allowed for serial dilution and enumeration of TBC, where contamination levels were too high. 51 3.9 Data analyses 3.9.1 Meta-analyses (Paper I) The effect estimate was calculated as the proportion of samples testing positive for each of the three hazards studied. To account for the binomial distribution of prevalence data and high between-study heterogeneity, we used metapreg in STATA® version 19.5 (https://www.stata.com/ ) to fit a logistic-normal random-effects model and estimate pooled prevalence with Wilson 95% confidence intervals. This model accounts for within-study binomial error and between-study heterogeneity without requiring continuity corrections or transformation. Meta set command uses logit-transformed inverse-variance meta-analysis. The former offers more robust modelling for sparse data; the latter enables subgroup comparisons. Estimates were similar across methods. Heterogeneity was further explored using subgroup analyses and metaregression. 3.9.2 Qualitative studies (Paper II and IV) Qualitative data were analysed using NVIVO®14 software (https://techcenter.qsrinternational.com/) and in accordance with COREQ (Consolidated Criteria for Reporting Qualitative Research) guidelines (Tong et al., 2007). This enables comprehensive reporting of the design of tools, data collection, and analysis processes. Grounded theory was chosen for data analysis, as it allows themes to emerge naturally from the data. Rather than starting with a hypothesis, grounded theory enables systematic, cyclical analysis, allowing patterns and themes to develop organically (Corbin & Strauss, 2008). This approach facilitates the exploration of concepts and provides insights without prior assumptions or hypotheses. In addition, content analysis was applied to derive quantitative statistics, which further provide insight into the data (Erlingsson & Brysiewicz, 2017; Hsieh & Shannon, 2005). 52 3.9.3 Quantitative analyses (Paper III) Item response theory and classical test theory Quantitative KAP data were analysed in R, version 4.4.0 (https://cran.r- project.org/) to compare differences between the treatment and control groups. Two-parameter Item response theory (IRT) was used to measure knowledge, while traditional classical test theory was applied to attitude and practice. The IRT (Item Response Theory) approach was chosen for the knowledge questions as a psychometric method to model the relationship between an underlying unobservable trait (latent trait, in this case, knowledge, represented as theta (θ)) and the probability of an individual responding to specific items (questionnaire questions). Theta denotes a 50 % probability of getting a particular question correct. The unidimensionality of the items, which refers to the consistency of the items in assessing the intended construct, was evaluated using principal component analysis (Cattell, 1966). In addition, cohesiveness of the items was evaluated using Cronbach's alpha test and was used as a first step to identifying misfit items. A value of 0.7 or higher is considered satisfactory (Muloi et al., 2023; Sharma et al., 2024). Theta scores are therefore calculated based on responses to a standardised scale, providing insights into an item's difficulty (how easy the item is) and discrimination ability (how effectively the item distinguishes between individuals with varying levels of ability, θ) (Yang & Kao, 2014). On the other hand, classical test theory (CTT) was used for the computation of scores for attitude and practice questions. Additionally, personnel hygiene and butcher shop cleanliness were evaluated using this approach. In this method, the underlying trait is measured as the average score of all the given items. CTT is based on the equation (X=T +e), where X is the observed or measured score, T is the true score (the individual's actual score), and e is the random error of measurement, which is assumed to be independent and has a mean of zero. Scores are given as a summation of scores to scale and represent the person’s average score (Hambleton et al., 1991). 53 CTT is more test and respondent-dependent; using CTT, the respondent’s ability will appear low if the test questions are difficult, while that same respondent will appear to have a high ability if the questions are easier. IRT is more robust than CTT, due to its ability to distinguish individuals with different abilities by separating the characteristics of the individuals and the questionnaire items (Yang & Kao, 2014). However, the application of IRT in our study was limited only to knowledge items because of the low Cronbach’s alpha, presented by attitude and practice scores, which fell below 0.7. Regression analyses Regression models were used to associate contamination levels with predictor variables, including demographic, infrastructural, and KAP variables. Predictor variables were selected using the Boruta machine learning algorithm. In this computation, for all features in the original dataset, random copies (shadow features) are created, and classifiers are trained on this extended dataset. The original features are then compared against the shadow features, and only those that are statistically more important than the shadow features are retained as important for modelling (Kursa et al., 2010). The selected features were then interlinked in a causal diagram or directed acyclic graph (DAG). A DAG displays assumptions about the relationships between variables (often called nodes). The relationships are indicated with a line and an arrow, pointing in the direction of the effect. A DAG is also acyclic, meaning there are no feedback loops. Figure 5 displays the causal diagram used to model the relationship between predictor variables and outcome variable (meat contamination), thereby identifying confounders (Textor et al., 2016). Model and regression analyses significance was considered at p-value=0.05. Additionally, differences in frequencies between treatment and control groups were evaluated based on percentage points (pp), which is the arithmetic difference of two percentages. 54 Blue=confounders Blue with a mark = outcome variable Green = indirect predictors Green with an arrowhead = exposure variable Pink = Predictor variables to adjust for The arrow points to the direction of association/effect Figure 5: Causal diagram illustrating potential variables associated with meat contamination in butcher shops in Nairobi, Kenya. Created using DAGitty (https://www.dagitty.net/ ) 55 4. Results and discussions 4.1 Prevalence of pathogens in foods (Paper I, III and IV) Paper I analysed levels of E. coli, Salmonella and Campylobacter spp. in foods, as these pathogens are most associated with diarrhoeal diseases (WHO, 2015). Many studies in our review (34 out of 53) examined E. coli in foods, while fewer studies (13 out of 53) focused on Campylobacter spp. in foods. Pooled proportion showed that the prevalence level of E. coli across all food types was 41% (95% CI: 34 – 52%). E. coli prevalence was particularly high in ready-to-drink beverages (66%; 95% CI: 31 – 89%), meat (65%; 95% CI: 30 – 87%), and chicken meat (63%; 95% CI: 36 – 82%). Cooked foods and flours had the lowest prevalence at 12% (95% CI: 2– 43%). These levels are comparable to findings from other African countries (Dinede et al., 2023; Paudyal et al., 2017). Additionally, pathogenic E. coli, including enterotoxigenic E. coli, enteroaggregative E. coli and enteropathogenic E. coli were found in chicken meat, milk and infant foods (Odwar et al., 2014; Tsai et al., 2022). Shiga- toxin producing E. coli prevalence was 2% (95% CI: 1– 10%), with the highest level in chicken meat and other meat types at 4% (Paper I). Similarly, high levels were observed in meat in prevalence studies in Burundi and Kenya (Papers III and IV), with 100% and 98% contamination at an average count of 5.17 ± 0.92 log cfu/g and 3.46 ± 1.21 cfu/g and in the two countries respectively. E. coli is used in the food industry as an indicator of hygiene and faecal contamination (Feng et al., 2020). With over 50% of some foods contaminated with E. coli as shown in our study, there is significant concern about hygiene and sanitation in food establishments in EAC. The presence of pathogenic E. coli in foods is more alarming as it poses a risk to human health and contributes to the high burden of diarrhoeal diseases in children under the age of five, which compromises their growth and development (Kyu et al., 2025). Paper I reported a pooled prevalence of 12% (95% CI: 12–27%) for Salmonella spp., and 9% (95% CI: 7–30%) for Campylobacter spp. across all food categories. The prevalence of Salmonella in our review (Paper I) was 56 similar to findings from the survey in Kenya (Paper III), which reported a 12% prevalence in meat. However, the prevalence survey in Burundi revealed a higher presumptive prevalence of 40% (Paper IV). Findings on Salmonella from Papers I and III are similar to results from Burkina Faso (Dinede et al., 2023), while findings in Paper IV are comparable to levels observed in South-East Asia (Dang-Xuan et al., 2019; Rortana et al., 2021). When the prevalence of Salmonella spp. is high (>40%), the risk of salmonellosis is estimated to be between 4% and 14.5% per person per year (Rortana et al., 2022). Our findings then indicate a potentially higher risk of salmonellosis from meat in Burundi compared to Kenya. This makes Salmonella spp. highly significant for public health, and elevated levels of the pathogen in foods are a cause for concern, necessitating increased efforts to reduce the risk of human infection. The pooled Campylobacter prevalence in our review (Paper I) was lower compared to findings from other African countries (Dinede et al., 2023; Gazu et al., 2021). This result can be linked to the samples analysed under our review, mostly dressed beef carcasses, which have a lower risk of Campylobacter contamination compared to other meat types, such as poultry (Havelaar et al., 2022; Inglis et al., 2020). Salmonella spp., particularly the non-typhoidal type, is a common cause of FBD illnesses and outbreaks in the European Union (EU) and the United States (US), whereas Campylobacter spp. is named the leading cause of gastrointestinal infection in the EU (EFSA and ECDC, 2021; Hoffmann et al., 2025). In our review (Paper I), fewer studies analysed Campylobacter spp., likely due to the specific conditions required for incubation and culturing, or a lack of necessary technical capacity, which may result in underestimating its burden in EAC. However, data from other countries show a high health and economic burden associated with Campylobacter and Salmonella spp. In Burkina Faso, the disease burden associated with poultry meat consumption due to Campylobacter and non-typhoidal Salmonella (NTS) was estimated at 42,600 DALYs in 2017. Economic costs estimated as productivity losses and willingness to pay for reduced risk of death and suffering, related to Campylobacter and NTS in Burkina Faso was USD 43 million and USD 280 million, respectively. In the same year, the disease burden from these two hazards linked to beef consumption in Ethiopia was 57 approximately 15,000 DALYs, while the burden associated with dairy products was slightly higher than that from beef consumption, at 17,000 DALYs (Havelaar et al., 2022). Altogether, the economic costs in Ethiopia were USD 288 million and USD 181 million for Campylobacter and NTS, respectively. The estimated economic burden of human salmonellosis, including loss of wages, reduced productivity, and medical costs, exceeds €3 billion in the EU and USD 17 billion in the US, while the economic costs of campylobacteriosis are estimated at €2.4 billion in the EU and USD 11 billion in the US (EFSA and ECDC, 2021; Hoffmann et al., 2025). Table 4 provides a summary of the levels of E. coli, Salmonella and Campylobacter spp. in the five countries where data were available (Paper I). The prevalence of E. coli was significantly higher in wild meat from the Democratic Republic of Congo (DRC) (86%) than in milk from Rwanda (13%). This difference is likely because wild meat analysed in the DRC may have been exposed to various environmental factors, as well as issues related to poor hygiene and inadequate sanitation infrastructure during harvesting and slaughter, which can compromise safety. In contrast, milk sampled from collection centres in Rwanda had lower contamination levels, possibly because higher quality standards were enforced by farmer organisations (Kumar et al., 2013; Nyokabi et al., 2023). Tanzania and Rwanda had higher prevalences of Salmonella, above the regional average. This may be associated with the food types analysed, mainly raw meat, fish and milk, which carry a high risk of Salmonella infection (EFSA and ECDC, 2021; Hoffmann et al., 2017). No studies were found for Campylobacter in Rwanda and Uganda, indicating a data gap. 4.2 Risk factors for contamination (Papers I and III) Of 53 studies reviewed, three analysed factors associated with food contamination (Paper I). Storage temperature, cleaning and disinfection, use of appropriate PPE and training were significantly associated with contamination. Additionally, lack of access to water and sanitation facilities, and sale of mixed meat products from different species, increased the odds of pathogenic contamination (Carron et al., 2018; Kariuki, 2018; Niyonzima et al., 2017). These findings were comparable to those revealed by the KAP 58 and prevalence survey (Paper III). Multivariable analyses showed that reduced likelihood of Salmonella contamination in meat was linked to access to running water (log OR -1.990; 95% CI: -4.059 to -0.206; p-value = 0.038), and better cleaning practices (log OR = -0.118; 95% CI: -0.225 to -0.030; p-value = 0.017). On the other hand, higher chances of meat contamination with E. coli and coliforms were associated with the location of butcher shops (β = 0.876; CI: 0.198 to 1.555; p-value = 0.012), and high bacterial counts on food handlers’ hands (β = 0.222; CI: 0.059 to 0.386, p- value = 0.008). Our findings are consistent with previous reports. For example, in Kenya, contamination of carcasses in slaughterhouses was linked to the personnel’s hand hygiene (Wambui et al., 2018). Similarly, in Mozambique, caregivers' practices of washing hands with soap and refrigerating food were associated with a lower likelihood of infant food contamination (Bick et al., 2020). These observations support the assertion that proper handwashing practice and cold chain maintenance remain paramount in food safety. Mixing and sale of meat from different species, as well as high temperature storage, were associated with an increased risk of Salmonella contamination in Vietnam (Ngo et al., 2021), which is consistent with findings from our review (Paper I). Although these practices were common among meat sellers in Kenya (Figure 6), no association with meat contamination was established in Paper III. However, univariable analyses showed that meat shops obtaining meat from a combination of slaughterhouses and storing end-of-day stock in the refrigerator only from the second day (the first day meat is left on hooks at room temperature, see Figure 4), had higher odds of contamination. While there was no direct observation of how meat was transported in our study, it is common for carcasses from different species to be mixed during transport (Gichure et al., 2014; Kunyanga et al., 2021), which can increase the risk of contamination. Additionally, temperature fluctuations between ambient and chilling conditions provide opportunities for microbial proliferation (Fahrion et al., 2013). Moreover, improving access to essential infrastructure, especially clean water (Paper III), and training of personnel (Paper II) were associated with safer food. These findings are consistent with a study in Nigeria, which 59 established that access to infrastructure and training promotes awareness and adoption of hygienic and safe handling practices in informal food establishments, thereby contributing to safer food (Grace et al., 2012). Consistent with our results (Paper III), findings from Uganda revealed that NTS infection was associated with particular locations (Kivali et al., 2024). This may require further investigation to understand environmental and other factors that may contribute to the endemicity of Salmonella in specific regions. 4.3 Knowledge, attitudes and practice of meat sellers (Paper III) 4.3.1 Knowledge Knowledge was evaluated using IRT. Of the 16 items and 5 pictorial statements, ten were selected to form the knowledge questions for analysis, based on a Cronbach's alpha of 0.70, which meets the minimum threshold for reliability (Nunnally & Bernstein, 1994). Item difficulty parameters ranged from -0.5 to -2.8, indicating a mix of easy and difficult knowledge questions. Item discrimination parameters ranged from 0.8 to 2.5, implying that most items had a moderate to high ability to differentiate between respondents with low and high food safety knowledge. Additionally, eigenvalue analysis of the item correlation matrix revealed a dominant first eigenvalue (2.8), followed by a marked decline to the second eigenvalue (1.2). The overall pattern and scree plot structure supported essential unidime