Risk factors for food contamination among children discharged from community management of acute malnutrition programmes in South Sudan: A cross-sectional study and hazard analysis critical control point approach

Abstract

Children under-5 years of age are particularly vulnerable to severe acute malnutrition (SAM), and the risk factors associated with relapse to SAM are poorly understood. Possible causes are asymptomatic or symptomatic infection with enteric pathogens, with contaminated food as a critical transmission route. This cross-sectional study comprised a household survey with samples of child food (n = 382) and structured observations of food preparation (n = 197) among children aged 6–59 months that were discharged from treatment in community management of acute malnutrition (CMAM) programmes in South Sudan. We quantified Escherichia coli and total coliforms (TCs), measured in colony forming units per g of food (CFU/g), as indicators of microbial contamination of child food. A modified hazard analysis critical control point (HACCP) approach was utilised to determine critical control points (CCPs) followed by multivariate logistic regression analysis to understand the risk factors associated with contamination. Over 40% (n = 164) of samples were contaminated with E. coli (43% >0 E. coli CFU/g, 95% CI 38%–48%), and 90% (n = 343) had >10 TCs (CFU/g) (>10 TC CFU/g, 95% CI 87%–93%). Risk factors associated (p < 0.05) with child food contamination included if the child fed themselves (9.05 RR, 95% CI [3.18, 31.16]) and exposure to animals (2.63 RR, 95% CI [1.33, 5.34]). This study highlights the risk factors and potential control strategies that can support interventions that reduce food contamination exposure in young children and help further protect those that are highly vulnerable to recurrent exposure to enteric pathogens.

INTRODUCTION

In 2022, estimates suggest 13.6 million children under the age of five suffered from severe acute malnutrition (SAM) (WHO et al., 2023), also referred to as severe wasting (WHO, 2023). Over 300,000 of these estimated cases were from children under-five in South Sudan (UNICEF, 2022). SAM is a short-term acute condition with a high case-fatality rate (CFR) that increases vulnerability to infections and disease (Olofin et al., 2013). There is a significant body of research addressing the risk factors, short-term health implications and treatment for SAM (O'Sullivan et al., 2018). However, little is known about children's health and nutrition following discharge from SAM/moderate acute malnutrition (MAM) treatment programmes or why some children are more at risk of relapse than others (Schaefer et al., 2021; Stobaugh et al., 2019).

Globally, relapse to SAM within the first 6 months post-treatment ranges from 0% to 37%, while the mortality rate ranges from 0.1% to 10.4% within 12 months post-discharge (Stobaugh et al., 2019). Evidence has suggested that severe and persistent diarrhoeal disease among children post-SAM is a key factor for relapse (Adegoke et al., 2021). In Bangladesh, (Ashraf et al., 2012) 20% of children reported diarrhoea in the 2-week post-discharge period, which decreased to 6% at 3 months post-discharge. Gastrointestinal disease is also likely to be underreported as enteric infections are often asymptomatic (Cichon et al., 2016). This highlights that there may be an association between enteric infections, whether symptomatic or asymptomatic and an increased risk of relapse.

Food is an important route of transmission for enteric diseases in children under-5 years of age (Pires et al., 2021). Infections often present as gastrointestinal syndromes—with symptoms such as diarrhoea, nausea and vomiting (Sharif et al., 2018). Enteric pathogens can weaken a child's immune system; impact their cognitive, motor and social-emotional development (Kotloff et al., 2013; Ngure et al., 2013); and, lead to subsequent or exacerbated malnutrition (Petri et al., 2008; Schaible & Kaufmann, 2007). Subclinical (symptoms are mild or absent, escaping diagnosis) and asymptomatic (patient infected and can infect others, but is not clinically recognisable) enteric infections are thought to be common in low and middle- income countries (LMICs) (Collard et al., 2022; Quilliam et al., 2013), and have been associated with acute malnutrition (AM) (Hasso-Agopsowicz et al., 2021). Crucially, these infections are likely to be a risk factor for relapse in children who have recovered from SAM.

Globally, there are an estimated 600 million cases of foodborne illnesses each year, with children under-five bearing 40% of this burden (Havelaar et al., 2015). Multiple studies in low- and middle-income countries (LMICs) have found high levels of microbial contamination in child foods (Bick et al., 2020; Ercumen et al., 2017; Gizaw et al., 2022; Islam et al., 2013; Touré et al., 2011; Tsai et al., 2019). For example, Escherichia coli was detected in 68% of child food samples in rural Ethiopia (Gizaw et al., 2022) and 48% of samples in rural Malawi (Taulo et al., 2008).

Modified hazard analysis and critical control point (HACCP) methods are recognised as an effective tool for understanding and preventing food contamination in LMIC households (Bick et al., 2020; Gautam & Curtis, 2021; Islam et al., 2013; Manjang et al., 2018; Touré et al., 2011). HACCP is a management system that addresses food safety through the analysis of food production processes. Critical control points (CCPs) are assigned to steps in the process at which controls could be applied to reduce, prevent or eliminate hazards to an acceptable level (FAO & WHO Codex Alimentarius Commission, 1997).

Studies employing HACCP and other quantitative approaches to explore child food contamination and hygiene in LMICs have identified common CCPs. Poor handwashing with soap (HWWS) practices (Bick et al., 2020; Davis et al., 2018; Islam et al., 2013), household animal ownership and the presence of animal waste (Barnes et al., 2018; Ercumen et al., 2017; Parvez et al., 2017) and poor storage practices (Ercumen et al., 2017; Touré et al., 2011) have all been identified as risk factors for child food contamination in both urban and rural contexts. However, the risk of contamination in food fed to children that have suffered and recovered from SAM in humanitarian settings has yet to be studied. To design and implement effective water, sanitation and hygiene (WASH) strategies to improve food hygiene, it is critical to understand the complex pathways through which food may become contaminated within households.

This cross-sectional study aimed to assess the prevalence and risk of E. coli and total coliform (TC) contamination, as indicators for faecal contamination, in child foods for children that are at risk of relapse to SAM. We used food sampling and analysis to determine prevalence of faecal contamination in food. Furthermore, we applied a modified HACCP approach, paired with a risk factor analysis, to explore potential practices and risk factors that would contribute to food contamination in the low-income setting of South Sudan.

METHODS Study design, setting and population

This was a cross-sectional study, nested within a multi-country prospective cohort study of children following anthropometric recovery from SAM aged 6–59 months in South Sudan, Somalia and Mali (King et al., 2022). The study was a collaboration between the London School of Hygiene and Tropical Medicine (LSHTM), Action Against Hunger (ACF) and the Ministry of Health in South Sudan.

Aweil East, South Sudan, is in Northern Bahr el Ghazal State which contains seven subdivisions, with a total population of 538,765 of which 102,365 are under-5 years of age. Aweil East has a total of 13 government health facilities and outreach clinics in the area, all of which are supported by ACF. Each health facility is approximately 20–30 km apart. Six out of the 13 health facilities were included in the study, with sites in urban, peri-urban and rural settings. Study sites were selected based on high SAM caseloads, established CMAM programmes, and accessibility by data collectors.

The study enroled children aged 6–59 months who have recovered from uncomplicated SAM at the point of discharge from treatment within a CMAM programme (discharge was defined as recovered according to the global WHO definition of recovery from SAM (World Health Organization, 2023): with a mid-upper arm circumference (MUAC) ≥ 125 mm, weight-for-height Z-score (WHZ) ≥ −2 and/or no oedema for two consecutive weeks). The inclusion criteria can be found in King et al. (2022).

Sample size calculation

The sample size for the cohort was calculated as part of a wider study to assess WASH-related risk factors for relapse to SAM and AM (King et al., 2022). This study used a power of 0.8, a level of significance of 0.05, and assuming the true proportion relapsing within 12 months of discharge to be 12% (Stobaugh et al., 2019), the minimum detectable differences (MDD) between exposed and unexposed groups under different exposure scenarios of WASH exposures (i.e., proportion exposed from 10%–80%) were calculated using the R package “powerMediation,” function “SSizeLogisticBin.” A total of 614 children were required for the study. The study enroled 618 children discharged from CMAM programmes in Aweil East, South Sudan. Six children did not meet the enrolment criteria and the study remained with 612 children. From these 612 households enroled, every third child was randomly selected (n = 200) to participate in the structured food observations.

Data collection: Household survey and structured observations

We conducted a household survey in 612 households to characterise sociodemographic factors, asset ownership, WASH conditions (including animal contact and exposure, hygiene and handwashing, sanitation, faeces disposal, water access, water treatment and storage), child intake and food preparation and storage practices (Figure 1). Trained data collectors administered the survey via a home visit within 2 weeks of discharge from CMAM programmes and as soon as possible after enrolment into the study.

In 197 of the 612 households, we conducted additional structured observations of food preparation, storage and hygiene practices (Figure 1). Data collectors arranged a time with the caregiver to observe the preparation and serving of the child's food. Observations lasted between 1 and 3 h, depending on the duration of preparation, feeding and food storage activities.

The survey and structured observations were recorded on tablets using Open Data Kit (ODK) submitted to and checked by the Field Research Coordinator, who uploaded the data to secure ODK servers at LSHTM. All data collection was conducted between June 2021 and October 2022.

Data collection: Food sample collection and analysis

Cooked food samples that had been prepared for the child were collected from each household. For the participants participating in the structured observations, food samples were collected at the end of the observations. For the remaining households, a food sample was collected during the household survey, if available and upon reasonable request. Samples were collected according to standardised and pre-tested protocols. An estimated 10 g of the child's food was placed in sterile 100 mL WhirlPak® bags and transported on ice to a field laboratory to be processed within 6 h of collection. All samples were homogenised by shaking the bags. Following standard Association of Official Analytical Chemists (AOAC) methods (#991.14) (Da Silva et al., 2018), a known amount of each food sample (11 g) was blended with 89 g of saline solution (0.9% Sodium chloride [NaCl]). The pH was adjusted to 6.6–7.2, with either hydrochloric acid (HCl) or sodium hydroxide (NaOH), and 1 mL of the diluted sample was placed onto a 3M^TM PetriFilm^TM for the simultaneous detection of the primary outcome, E. coli and secondary outcome, total coliforms (TCs), as indicators of faecal contamination (Paruch & Mæhlum, 2012). Bacteria were enumerated as colony-forming units (CFU) per gram (CFU/g) of the food sample. The limit of detection for 3MTM PetriFilm^TM is 1 CFU/g. Negative controls and duplicates were analysed in each batch of samples.

Modified HACCP methodology

HACCP is a management system in which food safety is addressed through analysing food product information and assessing physical, chemical and biological hazards throughout the product's lifetime. Critical limits are specified, and CCPs are identified (FAO & WHO Codex Alimentarius Commission, 1997) along the food preparation process.

A modified HACCP was conducted on the observation data. All foods were categorised based on ingredients and preparation processes (FAO & WHO, 2017). A food-flow diagram was constructed for each food category based on the information collected from the structured observation data. A general food-flow diagram was also constructed for all foods. Descriptive statistics were generated to describe common preparation, feeding and storage practices but also compare these practices between food groups. Evaluation of the food preparation process was used to identify a range of points that may impact the prevalence or magnitude of contamination. CCPs were then identified out of the range of control points using a decision tree and based on whether (a) preventive control measures can be applied, (b) the step specifically eliminates or reduces the occurrence of hazard to an acceptable level, (c) loss of control could increase contamination to unacceptable levels and (d) subsequent step eliminates identified hazard(s) or reduces occurrence to an acceptable level. Control measures, if any, were then identified for each CCP and then reported as part of the HACCP approach (FAO & WHO Codex Alimentarius Commission, 1997). Food-flow diagrams were annotated with CCPs, and feasible control measures were suggested to reduce E. coli and TC contamination in child foods.

Statistical analysis

All statistical analyses were conducted in RStudio 2022.02.2 Build 485 ‘Prairie Trillium’ (R Foundation for Statistical Computing).

To estimate relative wealth, sociodemographic, asset and wealth variables, commonly used in the Demographic Health Survey (DHS), were used to create a wealth index based on principal component analysis (FactoMineR package: ‘PCA’ function). Thirty-one variables were used to construct the index: asset ownership (12 binary variables), household size (continuous), number of rooms (continuous), housing characteristics (eight binary variables), animal ownership (binary) and ownership of different animal species (eight continuous variables).

Housing characteristics including roof, wall and floor materials were dichotomised into low- and high-quality construction materials (Khudri & Chowdhury, 2013). High-quality construction materials included concrete, laminated wood or tiles. Low-quality construction materials were associated with poorer structural stability and less easy to clean, such as mud, thatch or wooden planks. The exclusion of households based on missing socioeconomic data can significantly impact sample size and statistical power, so missing values were statistically imputed (missMDA package: ‘MIFAMD’ function) (Vyas & Kumaranayake, 2006). The first principal component analysis was used to group the households by socioeconomic status (SES) into relative wealth quintiles.

Arithmetic means were used to report colony counts (Haas, 1996). For E. coli, an outcome of >0 CFU/g was defined as ‘contaminated,’ while 0 CFU/g was defined as ‘non-contaminated.’ Guideline values for acceptable E. coli in food vary depending on the product and we decided conservatively to classify no detected E. coli as safe (Hasell & Salter, 2003). For TC, an outcome of >10 CFU/g was defined as ‘contaminated,’ while ≤10 CFU/g was defined as ‘non-contaminated’ as other studies have reported (Touré et al., 2013). The prevalence of E. coli and TC contamination in child foods was calculated alongside 95% confidence intervals (CIs). The proportion of samples contaminated was calculated with logic transformed 95% CIs.

All other variables were converted to binary or categorical variables, and food types were grouped based on common ingredients of samples and food preparation processes. Households were categorised into the WHO/UNICEF Joint Monitoring Programme (JMP) ladder definitions for drinking water, sanitation and hygiene (WHO & UNICEF, 2023).

Univariate and multivariable analysis was conducted to test for associations between variables and the primary and secondary outcomes. Individual variables were tested for their relative risk of association with contamination using univariate binomial logistical regression (stats package: ‘glm’ function). Variables with a p value below 0.1 in the binomial logistical regression and those variables that perfectly predicted the outcome were considered for inclusion in the binomial portion of the multivariable model. For each iteration of the multivariable model, variables were added individually in a stepwise procedure. Akaike Information Criterion (AIC) was used to compare the base model (Null model) with each new model. The variable was retained where the statistic was minimised by 1 AIC (ΔAIC = 1) but if the variable's p value scored >0.1, it was not retained. A likelihood ratio test (lmtest package: ‘lrtest’ function) was also used to understand whether each model was significantly improved, compared with the previous model. During each step, variables were checked to ensure they continued to have a p < 0.1, or otherwise, they were removed from the next step. The process was repeated until no variables improved the model fit and all scored a p < 0.1. Age, gender and SES were added to the model as confounders. The final multivariable model was checked for multi-collinearity, with variables exceeding a variance inflation factor of 10 removed (car package: ‘vif’ function).

Ethical statement

Ethical clearance was obtained both before the enrolment of study participants and before the analysis of child food contamination data from Solutions IRB, Yarnell, Arizona, US (March 2020, #20200310); London School of Hygiene and Tropical Medicine, Research Ethics Committee, London, UK (July 2020, #18059) and Ministry of Health, Juba, South Sudan (June 2020, # MOH/ERB 6/2020). Informed written consent was obtained from all study participants.

RESULTS Characteristics of study participants and exposures

A total of 612 households were included in the study, however, based on the availability of food in the households, we were only able to collect food samples from 382 of the 612 households enroled (Figure 1).

Of these 382 households, caregivers responsible for preparing and feeding the child were predominantly female (94%). The children in the study were 52% male and 48% female, with a mean age of 34 months (IQR: 11–47). The median household size was 7 (IQ range: 6–9), and the average number of rooms in the house was 2 (IQ range: 1–3). The floor (99%), walls (98%) and roof (98%) of houses were predominantly made of low-quality materials such as mud, thatch or wood planks. Households were split into the five wealth quintiles (1st = highest wealth quintile, 5th = lowest wealth quintile), with the highest proportion in the 1st quintile (23%) and the lowest proportion in the 5th quintile (16%).

Water, sanitation and hygiene

The majority of households had access to a basic drinking water service (79%), meaning that they were using an improved source with a round trip collection time of no more than 30 min including queueing. Few (24%) respondents reported water was treated before consumption, with the filtration of water through a cloth the most popular treatment practice (45%). Thirty-nine percent of household drinking water was stored with a cover or lid. All stored drinking water samples had a median free residual chlorine (FRC) concentration of <0.5 mg/L, indicating that water was not appropriately treated with chlorine, while 38% had a turbidity of more than five nephelometric turbidity units (NTU). Of the 379 drinking water samples, 80% had >0 E. coli (CFU/100 mL) and 98% were contaminated with >10 TC (CFU/100 mL). On average, households had 55 L (SD = 40) of water in the house of which 33 L (SD = 26) were reserved for drinking water. The majority (59%) of households had not experienced insufficient drinking water in the household within the past 4 weeks, but 32% had experienced water scarcity once or twice in this period. The main reason for experiencing water scarcity was that the water point needed to be fixed (77%). The majority of households (60%) had an alternative source of drinking water.

For sanitation, the majority (77%) of households reported open defecation (disposal of faeces in the environment e.g., fields, forests), with 14% reporting the use of unimproved facilities (pit latrine without a platform, hanging latrine or bucket latrine) and 9% limited facilities (improved facility, shared between two or more households). Child faeces were frequently disposed of outside the house in the open (80%), and in some cases, there were human faeces observed inside the compound (14%).

The majority (75%) had limited handwashing facilities (available handwashing facility lacking soap and/or water), 20% of households had a basic facility (available handwashing facility with soap and water present), and 5% had no facility whatsoever. Soap was present in 47% of households, with less than half (45%) of survey respondents using soap and water to wash their hands. Only 28% of households reported HWWS at the five key times (after defecation, after changing a child's nappy/diaper/faeces, before food preparation, before eating and after eating). Garbage in the compound was also observed in 69% of households.

Animal exposures

Households reported they were most commonly exposed to chickens (54%), goats/sheep (29%) and cattle (17%). In 36% of households, animal faeces were observed in the compound. During food preparation and feeding, chickens (54%), goats/sheep (15%) and dogs (10%) were most commonly observed in food preparation or feeding areas.

All exposures are reported in Tables 1 and 2.

Table 1 Characteristics of 382 households of children recovered from severe acute malnutrition (SAM), South Sudan.

	N	n
Household characteristics
Number of participants	382	382
Child age (intervals)	382
<12		7 (2%)
12–24		47 (12%)
24–35		176 (46%)
36+		152 (40%)
Child gender—Female	382	197 (52%)
Caregiver gender—Female	382	360 (94%)
Relative wealth quintile	382
1st		85 (23%)
2nd		82 (21%)
3rd		82 (22%)
4th		73 (19%)
5th		60 (16%)
Household size—Median (IQR)	382	7 (6–9)
Rooms—Median (IQR)	382	2 (1–3)
Household has exposure to any animals	382	245 (64%)
Chicken	382	207 (54%)
Goat/sheep	382	112 (29%)
Cattle	382	64 (17%)
Dog	382	51 (13%)
Other	382	5 (1%)
Presence of animal faeces in compound	382	136 (36%)
Water, sanitation and hygiene (WASH) conditions
Drinking water	382
Basic (improved source, <30 min round trip)		300 (79%)
Limited (improved source, >30 min round trip)		38 (10%)
Unimproved (unprotected well/spring)		29 (8%)
Surface (e.g., river, dam, lake)		15 (4%)
Water treatment practices—Yes	382	91 (24%)
Cloth filtration		41 (45%)
Filtration for example, uses a sand filter or ceramic filter		13 (14%)
Chlorine tablets for example, Aquatabs		23 (25%)
Other		14 (16%)
Volume of water in house (litres)—Mean (SD)	382	54.8 (40.3)
Volume of water in house available for drinking (litres)—Mean (SD)	382	33.0 (26.5)
Experience of water scarcity (in the past 4 weeks)	382
Never		227 (59%)
Once or twice		121 (32%)
Other		34 (9%)
Reason for water scarcity (in the past 4 weeks)	155
Water point is broken		119 (77%)
Water not available from source for example, dried up		17 (11%)
Other		19 (12%)
Drinking water contamination
Drinking water with >5 NTU	379	143 (38%)
Stored drinking water with median free residual chlorine concentration <0.5 mg/L FRC	379	379 (100%)
Drinking water with >0 Escherichia coli (CFU/100 mL)	379	302 (80%)
Drinking water with >10 TC (CFU/100 mL)	379	372 (98%)
Sanitation facility	382
Limited (improved facility, shared with two or more households)		35 (9%)
Unimproved (pit latrine without a platform, hanging latrine or bucket latrine)		52 (14%)
Open defecation (disposal of faeces in the environment (e.g., fields, forests)		295 (77%)
Site of child defecation	382
On the ground (inside the house)		61 (16%)
In the open (outside the house)		307 (80%)
Other		14 (4%)
Presence of human faeces in compound	382	53 (14%)
Handwashing facility	382
Basic (available handwashing facility with soap and water at home)		77 (20%)
Limited (available handwashing facility lacking soap and/or water)		285 (75%)
No handwashing facility		20 (5%)
Availability of soap	382	181 (47%)
Handwashing practices
Reports less than five key times for HWWS	382	275 (72%)
Does not use soap and water	360	212 (55%)
Presence of garbage or waste in compound	382	263 (69%)

Note: Full table can be found in Supporting Information S1: Appendix A: Table 1.

Abbreviation: HWWS, handwashing with soap.

Table 2 Food preparation, storage and feeding practices, including the prevalence of food contamination, in households of children recovered from severe acute malnutrition (SAM), South Sudan.

	N	n
Observed practices during food preparation
Fruit and vegetable preparation practices before feeding
Wash	382	362 (95%)
Peel	382	218 (57%)
Cook	382	280 (73%)
Nothing	382	43 (11%)
Food cooked at the following times
Early morning before sunrise	382	76 (20%)
Mid-morning	382	143 (37%)
Mid-afternoon	382	151 (40%)
Early evening before sunset	382	106 (28%)
Late evening after sunset	382	171 (45%)
Other	382	71 (19%)
Types of food provided to child	382
Cereals	382	177 (46%)
Cereals with meat, fish or milk	382	127 (33%)
Cereals with vegetables	382	67 (18%)
Fruit or vegetables	382	11 (3%)
Main ingredients in food
Cereals for example, sorghum, rice, wheat, millet, maize, porridge	197	193 (98%)
Fruit and/or vegetables	197	83 (17%)
Meats	197	33 (17%)
Fish	197	34 (17%)
Milk	197	54 (27%)
Other	197	62 (31%)
Additions to food before/during cooking
Water	197	164 (83%)
Milk	197	30 (15%)
Tea	197	10 (5%)
Nothing	197	28 (14%)
Food additions heated	197	77 (39%)
Cleaning with soap and water
Caregiver hands	197	42 (21%)
Utensils	197	30 (15%)
Surfaces	197	27 (14%)
Observed practices before feeding the child
Additional ingredients added to food after cooking
Water	197	101 (51%)
Milk	197	48 (24%)
Tea	197	12 (6%)
Nothing	197	79 (40%)
Washing of hands with soap and water
Caregiver	197	40 (20%)
Child	197	37 (19%)
Cleaning with soap and water
Utensils	197	183 (93%)
Surface	197	123 (62%)
Observed practices during feeding of the child
Who feeds the child	197
Caregiver		150 (76%)
Another adult		15 (8%)
Child		32 (16%)
Child seating location	197
Bare floor		99 (50%)
Mat on the floor		62 (31%)
Other		36 (19%)
Feeding vessel
Separate plate/bowl	197	100 (51%)
Child hand	197	71 (36%)
Caregiver hands	197	47 (24%)
Other	197	12 (6%)
Observed practices when storing and when eating remaining food
Leftover food retained and eaten again	382	131 (34%)
Leftovers stored in bowl or plate or pot (covered)	382	224 (59%)
Leftovers stored (duration)	382
Do not store prepared food		254 (66%)
Less than 1 day		101 (26%)
1–2 days		30 (8%)
Leftovers cooked (methods)	131
Nothing		64 (49%)
Reheat		25 (19%)
Boil		42 (32%)
Presence of animals during food preparation and feeding process
Presence of animals during child feeding area	197	89 (45%)
Chicken	197	80 (41%)
Goat/sheep	197	30 (15%)
Dog	197	19 (10%)
Other	197	12 (6%)
Food contamination
Food with >0 Escherichia coli (CFU/g)	382	164 (43%)
Food with >10 Total Coliforms (TC) (CFU/g)	382	343 (90%)

Note: Full table available in Supporting Information S1: Appendix A: Table 2.

Food preparation

The majority of households washed their fruit and vegetables before cooking or feeding their child. The most common time to cook food was in the late evening, after sunset (45%). Only 59% of households covered stored food. A minority of households (34%) reported eating leftovers, leftovers were commonly not re-cooked (49%) whereas some households reported boiling their leftovers (32%).

Of the 197 structured observations, the three most common ingredients of food fed to children included cereals (98%), milk (27%) and fish (17%). Water (83%) was added most frequently before or during cooking. When something was added to food, the additional ingredient was heated only 39% of the time. Food was most commonly prepared directly in a cooking pot (60%) compared to on the table (2%) or floor (12%).

Prevalence and quantity of food contamination

Food samples included: (a) cereals, including asida (a sorghum wheat-based pap), bread, rice and porridge (n = 177), (b) cereals with meat, fish and milk (n = 127), (c) cereals with vegetables (n = 67) and (d) fruit/vegetables (n = 11).

Of the 382 samples tested, 164 samples were contaminated with >0 E. coli CFU/g (43%, 95% CI [38%, 48%]) with a mean load of 41 CFU/g (SD: 126). A total of 343 samples were contaminated with >10 TC CFU/g (90%, 95% CI [87%, 93%]) with a mean load of 382 CFU/g (SD: 577). The proportion of samples contaminated varied across food types (Figure 2). Cereals with fruit and vegetables were contaminated the most (51%, 95% CI [39%, 63%]), followed by cereals with meat, fish or milk (50%, 95% CI [41%, 58%]), fruit and vegetables (45%, 95% CI [19%, 74%]), and lastly, cereals served without any accompaniments (35%, 95% CI [28%, 42%]).

HACCP approach

The frequencies of observed preparation, feeding and storing practices of children's food were reported for each of the four food categories (Table 3).

Table 3 Observed food preparation and feeding practices for food categories, among households of children recovered from severe acute malnutrition (SAM), South Sudan.

Observed risk	All samples (n = 197)	Cereal (n = 50)	Cereals with meat, fish or milk (n = 104)	Cereals with fruit or veg (n = 42)	Fruit/veg (n = 1)
Contamination
Escherichia coli (>0 CFU/g)	101 (51%)	25 (50%)	54 (52%)	22 (52%)	0 (0%)
Total coliforms (>10 CFU/g)	176 (89%)	43 (86%)	95 (91%)	37 (88%)	1 (100%)
Handwashing
No HWWS during preparation	155 (79%)	45 (90%)	70 (67%)	39 (93%)	1 (100%)
No HWWS before feeding	157 (80%)	44 (88%)	73 (70%)	39 (93%)	1 (100%)
No HWWS of child before eating	160 (81%)	46 (92%)	73 (70%)	40 (95%)	1 (100%)
Cleaning
Not washing utensils during preparation	167 (85%)	50 (100%)	76 (73%)	40 (95%)	1 (100%)
Not washing surfaces during preparation	170 (86%)	50 (100%)	78 (75%)	41 (98%)	1 (100%)
Feeding
Sitting directly on the ground	99 (50%)	35 (70%)	31 (30%)	32 (76%)	1 (100%)
Feeding by hand (adult)	47 (24%)	7 (14%)	36 (35%)	4 (10%)	0 (0%)
Feeding by hand (child)	71 (36%)	18 (36%)	37 (36%)	16 (38%)	0 (0%)
Reheating
Not reheating leftovers (n = 82)	45 (55%)	4 (100%)	27 (43%)	14 (93%)	NA
Not reheating leftovers to boiling (n = 37)	5 (14%)	NA	32 (89%)	0 (0%)	NA
Storage
Not covering leftovers	28 (14%)	13 (26%)	10 (10%)	4 (10%)	1 (100%)
Not separating leftovers from raw ingredients	69 (35%)	21 (42%)	34 (33%)	14 (33%)	0 (0%)
Animal contact and exposure
Contact with eating area	89 (45%)	19 (38%)	50 (48%)	19 (45%)	1 (100%)

Contamination with E. coli and TCs was similar for all food categories. HWWS was rare by the caregiver before food preparation (21%), before feeding (20%) and of the child's hands before eating (19%). Utensils (15%) and surfaces (14%) were not regularly washed before food preparation. Half (50%) of children were fed directly on the ground, with 24% of caregivers feeding the child with their hands, 3% feeding by spoon or 36% of children feeding themselves by hand. When stored, 14% of food across all categories was not covered, which increased to 26% for cereals with meat, fish or milk. Of the stored food, 55% of meals were not reheated, with 35% of reheated food not being done to boiling. Many households also had animals in the child-feeding area (45%).

Food flow diagrams were produced for each food category based on the structured observations in 197 households; cereals with meat, fish or milk (n = 104) (Supporting Information S1: Appendix B: Figure 1), cereal-based foods (n = 50) (Supporting Information S1: Appendix B: Figure 2), cereals with fruit or vegetables (n = 42) (Supporting Information S1: Appendix B: Figure 3), and fruit or vegetables (n = 1) (Supporting Information S1: Appendix B: Figure 4). Each flow diagram was annotated with potential areas for hand, animal, water and utensil contamination and CCPs. Specific food flow diagrams were combined to produce a general food flow diagram labelled with hypothesised areas of pathogen presence, growth and survival (Figure 3). The following control measures were suggested as part of the HACCP approach to counter the CCPs relevant for all food categories:

• 1.

  HWWS throughout preparation and feeding and using clean utensils.

• To limit CCP1: Handling (Figure 3).

• HWWS of caregivers was low during food preparation (21%), before feeding (20%) and of the child (19%). Utensils were rarely used to feed the child (3%).

• HWWS of both caregivers and children and using clean utensils are recommended to reduce food contamination.

2.

Cook all food to boiling.

• To limit CCP2: Cooking and Reheating (Figure 3).

• Although leftovers were often reheated (55%), they were rarely done so to boiling (14%).

• The reheating of prepared foods was often not observed, and when food was reheated, it was not always to boiling.

3.

Avoid animal contact with the child's eating area.

• To limit CCP3: Animal exposure (Figure 3).

• Various animal species were observed to come into contact with where the child is fed. Households often had at least one species come into contact with this area (45%).

• Avoiding animal contact with where the child is fed is suggested to reduce contamination of that area and food.

4.

Store food in covered containers.

• To limit CCP4: Storage (Figure 3).

• Although food was more commonly covered (86%), not covering food (14%) was still observed and considered a risk.

• Storing food in covered containers is justified to help reduce food contamination.

Univariate regression for risk factors for E. coli (>0 CFU/g) and TC contamination (>10 CFU/g) contamination of child foods

After the HACCP approach identified potential risk factors for food contamination, univariate regression models were run for potential risk factors and other covariates against the binary outcomes of E. coli (>0 CFU/g) and TC contamination (>10 CFU/g). A total of 18 variables were found to be significant (p < 0.1) when analysed against the binary outcome of E. coli contamination (>0 CFU/g) and eight variables for TC contamination (>10 CFU/g). Univariate results can be found in Supporting Information: Appendix C: Table 1.

Multivariable regression for risk factors for E. coli (>0 CFU/g) and TC contamination (>10 CFU/g) contamination of child foods

The multivariable regression model for food contamination >0 E. coli CFU/g included four variables (Table 4). Children feeding themselves led to a 9.1 times increased risk of E. coli contamination (9.05 aRR, 95% CI [3.18, 31.16]; p = <0.01). Animal presence in the child feeding area during mealtime led to a 2.6 times increased risk of E. coli contamination (2.63 aRR, 95% CI [1.33, 5.34]; p = <0.01).

Table 4 Multivariable binomial regression model comprising four risk factors for child food contamination with >0 Escherichia coli CFU/g in households of children recovered from severe acute malnutrition (SAM), South Sudan.

Risk factor	Univariate RR	aRR	2.5% CI	97.5% CI	p Value
Soap available—Yes (Ref: No)	0.66	0.52	0.26	1.02	0.06
Utensils washed before feeding—Yes (Ref: No)	3.76	3.58	0.88	18.87	0.09
Who is involved in feeding child (Ref: Primary caregiver)
Another adult	0.80	0.70	0.20	2.33	0.56
Child	6.51	9.05	3.18	31.16	<0.01
Any animal was present in child feeding area during mealtime—Yes (Ref: No)	2.18	2.63	1.33	5.34	<0.01

Note: Adjusted for confounders: age, gender and relative wealth index.

Abbreviations: aRR, adjusted risk ratio; CI, confidence interval; RR, risk ratio.

The multivariable regression model for food contamination >10 TCs CFU/g comprised two variables (Table 5). No variables were found to have a significant risk or protective effect on TC contamination. However, garbage in the compound was potentially associated with increased risk of contamination.

Table 5 Multivariable binomial regression model comprising two risk factors for child food contamination with >10 TC CFU/g in households of children recovered from severe acute malnutrition (SAM), South Sudan.

Risk factor	Univariate RR	aRR	2.5% CI	97.5% CI	p Value
Presences of garbage or waste in compound—Yes (Ref: No)	1.83	2.48	0.93	6.87	0.07
Feeding child preprepared food—Yes (Ref: No)	2.49	2.53	0.89	8.41	0.10

Note: Adjusted for confounders: age, gender and relative wealth index.

Abbreviations: aRR, adjusted risk ratio; CI, confidence interval; RR, risk ratio. DISCUSSION

This exploratory study assessed the contamination of food given to children aged 6–59 months that had recovered from SAM in South Sudan. To date, studies on food hygiene in similar settings or contexts are limited. The study combined a modified HACCP approach with qualitative structural observations and quantitative hazard data, collected through household surveys, to understand the factors contributing to the contamination of children's food. In our study, 43% of the samples were contaminated with E. coli (>0 CFU/g), and 90% were contaminated with TCs (>10 CFU/g). Four CCPs were identified during the preparation, feeding and storage of child food from the HACCP approach. These CCPs included the handling of raw and prepared food, cooking and reheating practices, animal contact during feeding, and storage of preprepared food. Child self-feeding and animal presence in the area during food preparation and feeding were risk factors for E. coli contamination.

At the time of the study, there had been no similar studies into food contamination with E. coli in South Sudan, so direct comparisons are difficult to make. However, the 43% E. coli contamination levels (>0 E. coli CFU/g) in our study population are similar to that recorded in Bangladesh, with studies recording E. coli contamination of child foods between 39% and 46% (Islam et al., 2012; Muller-Müller-Hauser et al., 2022). Other studies have shown higher contamination levels in urban Mozambique (53% >10 enterococci CFU/g) (Bick et al., 2020) and rural Ethiopia (68% E. coli >0 CFU/g) (Gizaw et al., 2022). This study also found that 90% of food was contaminated with TCs, slightly higher than the 84% of fresh and stored food contaminated with thermotolerant coliforms (TTC) (>10 TTC/g) found in peri-urban Mali (Touré et al., 2013). The contamination levels in South Sudan were in line with or higher than those found in other studies, and is especially concerning as expected levels of E. coli contamination in child food should be low or none (Hasell & Salter, 2003; Kennedy et al., 2011).

Other studies that carried out a modified HACCP approach of child foods observed similar food preparation, feeding and hygiene practices in their study populations, and drew similar conclusions to the CCPs in this study. Cooking, whether that is of fresh or preprepared food, food handling practices and storage were also identified in other studies (Bick et al., 2020; Gautam & Curtis, 2021; Islam et al., 2013; Manjang et al., 2018; Touré et al., 2011). Boiling of water or milk before being given to the child was also suggested as a CCP by other studies (Gautam & Curtis, 2021; Manjang et al., 2018). As this study did not assess the contamination of water or milk specifically, this was not found among our results but could be an area of further research. Our study also found animal contact with the food preparation and feeding area as a CCP whereas this was not identified in other studies (Bick et al., 2020).

The CCPs and risk factors identified in this study substantiate and build upon the existing WHO ‘Five Keys to Safer Food’ best practices for safer food, including (1) keeping hands, surfaces and equipment clean; (2) separating raw and cooked food; (3) cooking food thoroughly; (4) keeping food at safe temperatures and (5) using safe water (WHO, 2006). Feeding the child preprepared food was found to be associated with increased TC contamination. Food storage practices, including temperature and duration, have been associated with contamination (Ercumen et al., 2017; Touré et al., 2011). If these foods are not cooked sufficiently to the recommended 70°C (WHO, 1996), they will remain contaminated. Access to cooking fuel can be a barrier to cooking and heating food sufficiently (Che et al., 2021), so appropriate solutions would have to be assessed. Furthermore, various studies have shown the effectiveness of covering food to reduce contamination (Ercumen et al., 2018; Parvez et al., 2017; Touré et al., 2011). Providing households with lidded containers could be an appropriate intervention to reduce contamination. In particular, air-vented containers mitigate concerns of spoiling, as suggested by a study in Bangladesh (Rahman et al., 2016).

This study found a 9.1 times increased risk of E. coli contamination if child self-feeding was practiced. A study in Bangladesh also identified the same risk factor (Islam et al., 2012) with several studies showing that children's hands were typically contaminated with a high level of E. coli (Ercumen et al., 2017; Pickering et al., 2018). Children are likely to use their hands instead of utensils to eat, as well as touch surfaces including soil and the surrounding area (Davis et al., 2018). This creates a clear pathway for the contamination of children's hands. If children's hands are not cleaned before eating, as found in our study, there is a route for faecal contamination from their hands to food or directly to their mouths. HWWS of the child was rarely observed (19%), so improving the child's hand hygiene may be an appropriate method to reduce food contamination. A trial in Nepal found that improving the handwashing of children and adults through activities including games, storytelling and local rallies could reduce food contamination risks (Gautam et al., 2017).

The presence of animals within the child feeding area during mealtime was found to be associated with a 2.6 increased risk of E. coli contamination. Other studies in LMICs also found animals and the presence of animal faeces in the compound to be associated with food contamination (Barnes et al., 2018; Ercumen et al., 2017; Parvez et al., 2017). In this study, animals and their faeces were often not separated from the household, allowing for the contamination of soil and other surfaces (Penakalapati et al., 2017). Animals can host many foodborne enteric pathogens that can infect humans, including Campylobacter, E. coli and Salmonella (Heredia & García, 2018). Studies have also observed that animal ownership increases E. coli contamination of soil and stored water (Ercumen et al., 2017; Navab-Daneshmand et al., 2018). This creates a clear pathway in which animal faeces can contaminate soil and water, which can then contaminate children's hands and ultimately contaminate food. Improving animal husbandry practices such as keeping ruminant and avian species outside of the kitchen or eating areas may be an effective intervention to reducing child food contamination. This may include reducing the presence of animals within the compound or providing tools to dispose of faeces appropriately. However, local practices would need to be considered.

The availability of soap within the household was found to have some protective effect (p = 0.06) against E. coli food contamination, which aligns with similar studies where soap had protective effects against food contamination (Bick et al., 2020; Davis et al., 2018; Islam et al., 2013). A meta-analysis found promotion of handwashing with soap reduced diarrhoea risk by 30% in LMICs (Wolf et al., 2022). However, the presence of soap does not equate to adequate use, especially in households that do not have sufficient quantity and quality of water to aid in handwashing (Abdullahi et al., 2020). This issue is also of concern for the effective washing of utensils, such as spoons. The households included in this study had low volumes of water stored in the house at the time of the survey which may imply that water for handwashing is a low priority. Providing soap to households may help to reduce the risk factors of child self-feeding and animal exposure, but alone may likely be ineffective without adequate handwashing facilities including access to and use of improved water sources.

LIMITATIONS

As this is a cross-sectional study, the causality between risk factors and food contamination cannot be determined. Also, the results of this study are only generalisable for populations that live within similar settings (King et al., 2022). The observations also likely excluded variables that contributed to food contamination such as cooking and storage temperatures and the presence of flies in the household. Also, some variables do not reflect exact practice within the household. Questions such as soap availability did not capture proper usage, and washing of utensils did not capture whether utensils were free of pathogens.

As some of the data for this study were collected through structured observations, this may have led to reactivity through the Hawthorne effect, where participants modify an aspect of their behaviour in response to their awareness of being observed (Sedgwick & Greenwood, 2015). Furthermore, the observations by the reporter will likely have suffered from bias in their reporting, the enumerators may not have been able to observe all characteristics or there is the potential of ‘suppression’ of data as they build a relationship with the observed participants.

Only taking one sample before feeding also limited our understanding of how contamination changes throughout the food preparation and feeding process. It would be useful to take repeated samples to pinpoint where contamination is introduced or reduced. A HACCP approach could be applied in more depth and detail if we had available contamination data and CCPs developed to a greater degree. The temperature and duration of cooking and storage could also have been recorded to help understand how this affects food contamination.

Using E. coli and TCs as indicators for child food contamination provided only a limited understanding of the prevalence and diversity of other foodborne diseases in the child's food. Understanding the individual pathogens present can help to determine their origins and possible routes of food contamination, and further in-depth analysis of food samples with multi-pathogen PCR analysis or other techniques is recommended.

CONCLUSIONS

Developing our knowledge of the complex pathways in which food can become contaminated highlights that food hygiene interventions may complement more traditional WASH and infant and young child feeding (IYCF) interventions for children recovered from SAM. Future work should aim to understand how simple control measures such as reducing animal contact in the household and during feeding; improving knowledge and practices around cooking; and, hygiene specific to the child may reduce food contamination. Further work should also assess whether food contamination is a risk factor for SAM relapse.

Our study found that over 40% of child food samples were contaminated with E. coli, and that child self-feeding and animal exposure were key risk factors for increased contamination. Our findings are consistent with other studies exploring child food contamination, particularly in low-income areas. This study helps to build on the understanding of child food contamination and highlights one of the many risks within the household that this vulnerable population face once they have recovered from SAM.

AUTHOR CONTRIBUTIONS

Oliver Cumming and Heather Stobaugh developed the study concept. Oliver Cumming, Lauren D'Mello-Guyett and Karin Gallandat designed the research study and data collection processes. Lauren D'Mello-Guyett developed the data collection tools and databases. John Angong conducted data collection and environmental sample analysis. Nancy Grace Lamwaka, Sarah King, Lauren D'Mello-Guyett, Jackson Lwate Hassan, Lino Deng and Lino Deng oversaw data collection. Joseph Wells and Lauren D'Mello-Guyett analysed the data. Joseph Wells wrote the first draft of the manuscript, and all authors received and contributed to subsequent drafts. All authors read and approved the final manuscript.

ACKNOWLEDGEMENTS

We acknowledge the important contributions of all study participants and their families; the Ministries of Health in South Sudan; and the teams of dedicated staff who collected data for this study and provide treatment to malnourished children across all the study clinics. This study was made possible by the generous support of the American People through the United States Agency for International Development (USAID). The contents are the responsibility of the London School of Hygiene and Tropical Medicine (LSHTM) and Action Against Hunger (ACF) and do not necessarily reflect the views or official position of USAID or the United States Government. USAID played no role in the study's design, the collection, analysis and interpretation of data, or the writing of this manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.