Risk of adenovirus and Cryptosporidium ingestion to sanitation workers in a municipal scale non-sewered sanitation process: a case study from Kigali, Rwanda

Sanitation workers provide essential services that protect public health, often at the cost of their own health and safety. In this study, we evaluate occupational exposure to fecal pathogens at each stage in a non-sewered sanitation process. Bulk fecal waste samples were collected during waste collection and waste processing tasks and analyzed for Cryptosporidium , adenovirus, E. coli , and total coliforms using quantitative polymerase chain reaction and culture methods. Structured observations of worker hand-to-mouth behavior were conducted, and worker hand- and glove-rinse samples were collected and analyzed for E. coli and total coliforms. A Monte Carlo simulation was used to model the dose of pathogen ingested and the risk of disease across two waste collection and processing tasks. The model results show that the probability of disease was highest from exposure to adenovirus during collection. Our analysis highlights that pathogen-to-indicator ratios are useful for predicting the risk to adenovirus which has a high detection rate. On the other hand, the use of pathogen-to-indicator ratios to predict Cryptosporidium concentration is fraught due to variable detection rates and concentration. study outlines methods for conducting quantitative microbial risk assessments in low-resource settings. Pathogen-to-indicator ratios ﬁ c to this study are constructed from primary data in order to minimize model uncertainty.


INTRODUCTION
In the absence of a sewer network, non-sewered sanitation systems rely on workers to physically handle fecal sludge, a practice that can increase the risk of contracting enteric diseases (Stenström ). Despite this, risk assessments on occupational exposures to workers within non-sewered sani- The use of such ratios introduces uncertainty into models when ratios derived from contamination sources in one geographic and population context are used to estimate pathogen concentrations in another. However, this widespread use of pathogen-to-indicator ratios in contexts different from their origin has been called into question.
Relative concentrations of different organisms are known to be driven by context-dependent factors such as the source of the fecal input, pathogen endemicity in the population, concentrations of other naturally occurring organisms, and exposure of the medium to environmental factors which drive differential die-off rates (Cutolo et al. ; Silverman et al. ; Keuckelaere et al. ).
In this study, we develop a sampling and analysis framework for measuring pathogen-to-indicator ratios and quantifying pathogen specific risk of ingestion to workers in a fecal sludge collection and processing operation. In addition to the framework developed, the findings from this study can help governments, process designers, and occupational hygienists target resources toward controlling high-risk occupational exposures during waste collection and processing.

Study site
This study is a case study of a fecal waste collection and waste transformation process operating in Kigali, Rwanda, a city with no central sewer network (NIOSR ). When pit latrines and septic tanks fill, they are emptied by service providers who pump out the waste and transport it by truck to a waste-to-fuel facility which converts the fecal sludge into solid industrial-grade fuel.

Recruitment and ethics
Twelve workers were sampled while performing one of two types of work tasks in the waste-to-fuel process: (1) waste collection tasks and (2) waste processing tasks at the waste-to-fuel facility. See Supplementary Material for further description.
All workers consented to participating in the study, and all study protocols were approved by the UC Berkeley Committee for Protection of Human Subjects prior to the conducting of this research (CPHS/OPHS Protocol ID: 2017-06-10016). Potential candidates for inclusion were all workers performing jobs in the waste collection and transformation process over the age of 18. All workers who participated in the study were provided with information about reducing workplace exposures. All information was provided in Kinyarwanda, and written informed consent was collected from each participant prior to data collection activities. Our methodology combined a participatory research approach with capacity-building strategies which were designed to overcome knowledge gaps in microbial sampling, exposure assessment, and occupational health (Cahill ). This approach enabled the local project team to continue carrying out occupational and environmental monitoring beyond the scope of this study.

Model framework
A quantitative microbial risk assessment (QMRA) model was developed to estimate pathogen ingestion by workers due to contaminated hand-and glove-to-mouth contact events over an 8-h work period (Haas et al. ). Three types of inputs were measured directly in this study: (1) concentration of indicator bacteria on worker hands and gloves, (2) ratio of pathogens and indicator organisms in bulk

Pathogen-to-total coliform ratio in bulk samples
Concentration ratios of pathogens-to-total coliforms in bulk samples, R path:tc , were constructed as lognormal probability distributions fit to the empirical data of pathogen and indicator concentrations, C pathogen jg and C indicator jg . Concentrations lower than the LOD were replaced by a value equal to half the detection limit prior to fitting the lognormal distributions.
Indicator bacteria in worker hand-and glove-rinse samples Hand-rinse samples of ungloved hands (prior to glove application) and gloved hands were obtained from workers in collection and facility tasks using the USEPA method 1604 and a field protocol developed by Pickering et al.
(USEPA ; Pickering et al. ). For each task that was observed, four rinse samples were collecteda glove-and hand-rinse sample before the task started, and a glove-and hand-rinse sample when the task was completed. Hand washing in the middle of a task, or work shift, was not observed.
All hand-rinse samples were analyzed using a membrane filtration system with MI media (BD Difco, Franklin Lakes, NJ) and incubated at 35 ± 0.5 C for 24 h according to USEPA method 1604 (USEPA ). If the measured concentrations were lower than the LOD, then half the detection limit of 1 CFU/filter was assumed as the concentration. If the concentrations were too numerous to count (>500 CFU/filter), then the concentration of bacteria in the sample was calculated assuming 500 CFU/volume filtered. Additional information on the LOD and the full dataset of indicator bacteria results from hand-and gloverinse samples is available in the Supplementary Material.

Worker hand-to-mouth activity
One worker from the waste-to-fuel facility was trained to observe fellow workers and enumerate the number of hand-to-face activities observed, f jg , while wearing gloves (g ¼ gloves) or using ungloved hands (g ¼ hand) during collection (j ¼ collection) and facility tasks (j ¼ facility) (Gorman Ng et al. ; Kwong et al. ) . Each observation was limited to a single observer recording the hand-to-face activities of a worker doing a single task. We sought to minimize observers by hiring a familiar staff member to conduct observations.

Dose estimation
The total dose of pathogens, d pathogen j , ingested from both worker hand-and glove-to-mouth activities in a given task j, was calculated as follows: (1) where d pathogen j is the lognormal distribution of the pathogen ingested over an 8-h period in a given task type. Cryptosporidium and adenovirus is represented by v. As qPCR does not distinguish between viable and nonviable organisms, we made the health-protective assumption that 100% of the organisms were quantified are viable.
We assume that the transfer efficiency of all hand-or glove-to-face activities is equivalent to the 33%, the transfer efficiency reported for viruses between fingertips to lips, given the dearth of the literature on more specific transfer efficiencies (Nicas & Best ).

Risk of infection and disease
A 10,000-trial Monte Carlo simulation was used to estimate a worker's probability of disease, highly credible gastrointes-

Concentration of indicators and pathogens in bulk fecal sludge samples
The concentration of total coliforms detected was above the LOD in the vast majority of sample types, unlike E. coli which was below the LOD in 33% of samples (Table 1).
Surprisingly, the average concentration of total coliforms in facility samples was an order of magnitude higher than the concentrations in both collection and fuel samples.
The average total solids in each type of sample are shown in Supplementary Table S2.
Adenovirus was detected in 88% of the bulk samples. A high adenovirus detection rate was found in the samples from the collection (89%) and facility (100%) samples compared with the fuel samples (67%) ( Table 1). Within collection samples, the geometric mean of adenovirus samples above the detection limit (7.6 × 10 4 ) was an order of magnitude higher than that in the facility (9.10 × 10 3 ) and fuel samples (3.4 × 10 3 ).
Cryptosporidium was detected in 16% of bulk samples from all stages of the process. Most samples of Cryptosporidium did not amplify within the quantifiable range and those that did exhibited a very large spread.

Concentration of indicator bacteria in worker handrinse samples
The concentration of E. coli and total coliform concentrations measured in worker hand-and glove-rinse samples are shown in Table 2. The concentration of total coliform and E. coli measured in rinse sample of hands and gloves taken prior to the start of the work task was noticeably high, indicating a baseline contamination level at the start of the work shift. The geometric mean concentration was higher in after-task samples across both indicator and task types, except for E. coli measured on hands of facility workers, indicating a general accumulation of bacteria on both hands and gloves over the course of a work task.

Pathogen dose and risk estimation
The dose of Cryptosporidium and adenovirus was estimated based on observed worker hand-to-mouth activity (see Supplementary Table S4) and the measured concentrations of indicators in hand-rinse samples. The geometric mean dose of Cryptosporidium ingested by workers in the facility task group was an order of magnitude higher than in the collection task group, while the geometric mean dose of adenovirus was higher in the collection tasks (Table 3).
The risk of contracting gastrointestinal disease resulting from hand-to-mouth exposure events during an 8-h period is shown for both pathogens in two task groups. Between the two pathogens in both task groups, adenovirus posed a higher risk of gastrointestinal disease. Between the two task types, facility tasks posed a higher risk of gastrointestinal disease from both pathogens (Table 3).

DISCUSSION
The results of this risk assessment suggest that interventions should focus on reducing exposure for workers performing fecal waste collection and waste processing tasks. Of the pathogen task combinations explored in this study, the probability of disease was highest from exposure to adenovirus during collection tasks.
A high degree of contamination in the glove-rinse samples taken before the task suggests poor storage or poor sterilization of gloves in between tasks or workdays.
Storage of gloves in areas that are protected from airborne contamination as well incorporating regular handwashing and hand-sanitizing practices into standard operating procedures may encourage better decontamination of hands and gloves before, during, and after work tasks. Proper cleaning and storage of all personal protective equipment (PPE) is necessary to prevent other exposure pathways that were not explored in this study. Contaminated uniforms, eyewear, and boots may also elicit exposure pathways if they are improperly cleaned or stored.
framework for long-term exposure monitoring in lowresource settings.

CONCLUSION
Several interventions may be useful in controlling worker exposure to fecal sludge. Elimination and substitution are the most effective ways to reduce workplace hazards but are not applicable in this case. Therefore, engineering controls should be prioritized such as enclosure and automation of processes that require workers to handle or come into close contact with sludge. Administrative controls including education on the fecal-oral route or implementation of workplace hygiene programs can also prevent workplace exposures. Finally, protecting the worker with PPE that is properly worn, stored, and replaced at regular intervals can prevent exposure and disease to sanitation workers.

DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.