Faecal Pathogen Flows and Their Public Health Risks in Urban Environments: A Proposed Approach to Inform Sanitation Planning

Public health benefits are often a key political driver of urban sanitation investment in developing countries, however, pathogen flows are rarely taken systematically into account in sanitation investment choices. While several tools and approaches on sanitation and health risks have recently been developed, this research identified gaps in their ability to predict faecal pathogen flows, to relate exposure risks to the existing sanitation services, and to compare expected impacts of improvements. This paper outlines a conceptual approach that links faecal waste discharge patterns with potential pathogen exposure pathways to quantitatively compare urban sanitation improvement options. An illustrative application of the approach is presented, using a spreadsheet-based model to compare the relative effect on disability-adjusted life years of six sanitation improvement options for a hypothetical urban situation. The approach includes consideration of the persistence or removal of different pathogen classes in different environments; recognition of multiple interconnected sludge and effluent pathways, and of multiple potential sites for exposure; and use of quantitative microbial risk assessment to support prediction of relative health risks for each option. This research provides a step forward in applying current knowledge to better consider public health, alongside environmental and other objectives, in urban sanitation decision making. Further empirical research in specific locations is now required to refine the approach and address data gaps.


Selection of Reference Pathogens
According to the reference pathogen principle, pathogens from each microbial group (bacteria, viruses, protozoa and helminths) are selected that are of local regional significance and are assumed to be a conservative representative of its group. Consideration is also given to the available data to quantify the occurrence, persistence, infectivity and disease burden. For the present study representing an urban setting in the developing context, the following reference pathogens were selected:  [1] reported that ETEC concentration in faeces to range from 10 8 -10 9 per gram, however no citations were given. A value of 10 8 was selected.

Viruses
De Silva [2] reviewed the data on the concentration of rotaviruses in faeces and reported that persons can shed concentrations of 10 10 to 10 12 of virus per gram. However as virus excretion varies over the course of an infection [3,4], the reported concentrations are most likely peak excretion and therefore a lower value of 10 8 was applied to be representative of the overall loading.

Protozoa
Medema [5] reported that Cryptosporidium concentration in infected individuals to range from 10 5 to 10 7 oocysts per gram citing Chappell [6], a value of 10 6 oocysts per gram was selected.

Helminths
Feachem [1] reported that individuals excrete up to 300,000 eggs per gram of faeces, however no citations were given. 10 5 eggs per gram was selected as a starting point for the model.

Prevalence
Platts-Mills [7] investigated the pathogen-specific burdens of community diarrheoa in children (< 2 years) in developing countries covering eight contexts in Asia, Africa and South America. A total of 31,628 stools were tested of which 7318 were from symptomatic children. Prevalence of positive detection of pathogens was taken as an estimate of the different levels pathogen prevalence in the community. Prevalence of pathogenic E. coli was based on prevalence of enterotoxigenic E. coli (St-ETEC); viral prevalence based on Norovirus GII; and protozoa on Cryptosporidium (noting that Giardia was much more prevalent than Cryptosporidium at around 30% in 12-24 month old children). Percentages in Table S1 are approximate as the numbers were read off a low-resolution graphic ( Figure S1), however for the purpose of this illustrative case this was considered suitable. In addition, the prevalence of pathogens is likely to be higher amongst children in comparison to adults, and therefore the values selected are an overestimate for a mixed age population. In reality, the prevalence of different pathogens will vary between communities and also within a given community over time. The purpose of the modelling tool would be to explore this variability, rather than simply include it as a fixed value. As a starting point for Helminths, Pham-Duc [8] reported an Ascaris Lumbicoides prevalence of 24%, amongst agricultural communities in northern Vietnam (n = 1425). No representative values more suitable to the urban context were identified.

Faecal Excretion Per Day
Rose [9] reviewed the amount of faecal material excreted per person per day. For low income countries (n = 17) the mean was 243 with a range of 75-520 g·cap −1 ·day −1 , noting that this is the variability in the mean, individual variability would indeed exceed this range. Diarrhoea has an impact on stool production, structure, form and composition, leading to much higher faecal mass generation and water content. A point value equal to the mean of 243 g·cap −1 ·day −1 was selected.

Pathogen log10 Reduction
Inputs to the model for each pathogen class to account for the reduction in pathogens from excreta to exposure, including formal treatment systems and the pathogen reduction during conveyance or discharge to the environment. The range of data shown in Table S2 was sourced from a range of literature, with greater data availability for traditional treatment systems (i.e., wastewater or sludge treatment plants) than for conveyance processes (flows in drains, sewers, groundwater). Key limitations of the data were reference to log10 reduction without differentiating between pathogens (i.e., die off after irrigation 0.5-2 log10), not distinguishing whether the removal referred to the liquid or sludge components (particularly for septic tank and sludge treatment) and often only providing data for some pathogen classes. Reduction in soil, fresh produce and groundwater varies significantly depending on the local conditions, disposal or irrigation practices, and selected log10 reduction should be based on local data where available.
This table was developed for the purpose of the preliminary model testing and many assumptions have been made. The literature analysis used for this table was not extensive in the knowledge that the Global Water Pathogen Project was synthesising existing research in detail, and although not finalised at the time of writing, is expected to greatly inform this table. The numbers in brackets have been adopted in the preliminary model presented in the paper, however it would be important for further development of the model to test the sensitivity to the range of reductions.

. QMRA Calculations
Based on the standard QMRA methodology [17], the following the steps detailed in Table S3 below were applied to the calculated pathogen dose at each point of exposure for adults and children separately. Dose-response models were used to estimate the infection based on exposure to each pathogen, applying a beta Poisson model for E.coli, cryptosporidium and rotavirus [18,19,20] and an exponential model [17] with r = 1 for Ascaris. The illness infection ratio was assumed to be worst case of 1 for Pathogenic E. Coli and Ascaris [21] and the values for rotavirus and cryptosporidium were based on WHO Drinking water Guidelines [22]. While the ratio of DALY per infection can be estimated based on local probability estimates for the severity of diarrheal diseases and life expectancy, for this model of a hypothetical situation values were used from literature [23,24]. Table S3. Steps in the QMRA Calculations.
Step Equation

Exposure Assumptions
For the hypothetical case study, exposure inputs were based on literature from recent sanitation health risk assessments in similar low-income neighbourhoods or developing countries. This includes the estimated dose or volume consumed at each exposure point (detailed in Table S4), the expected frequency of exposure events per year (Table S5) and the likely proportion of population exposed to this pathway (Table S5). The data used in these references comes from both in-field surveys and literature. Due to the variability of all aspects of exposure on local conditions and behaviour and varying within cities and across seasons, Robb [25] argues for the use of local data, which have been determined in previous studies by questionnaires and field surveys and focus groups [25,26].   6 12 35% 35% Adults exposed during flooding 1-6 times per year and children exposed daily to every 4 months due to playing in drains [27,29]

Data for Validation
For a preliminary validation of the concentration of pathogens at the exposure point, the calculated pathogen concentrations at various points in the model were compared with literature to confirm the results were within the range of previously reported values. The data in Table S6 was developed for the purpose of developing the model and was drawn from a brief literature review (as this was not the focus of the research) which does not consider the full range of possible values when all variables are considered. As for Table S2, it is expected that this table would be significantly improved when the Global Water Pathogen Project is finished which will include detailed summary of the pathogen concentrations for the four considered pathogens (not yet available at the time of writing). As highlighted below, there is limited data on all pathogen classes for all pathways, with most research reporting E. coli concentrations only. For this reason, the ratio to other pathogens was calculated to provide a potential range for comparing calculated values. However it is recognised that these ratios would be expected to vary depending on the pathway and treatment stage. Dewatered and digested sludge [12] Bacteria data only given for extended aeration and was higher than inlet. River or downstream waterway E. coli: 10 4 -10 6.6 CFU/100 mL, Rotavirus: 10 −0.1 to −0.01 /100 mL, Ascaris 10 0.48-0.6 /L E. coli-Uganda water channel, wetland, swimming lagoon [29,33] Virus and Helminth-Drain and stream water reused in agriculture [37] Household environment E. coli: 10 5.4-6 CFU/100 mL Samples from soil in open space; playground for children [28] and unintentional ingestion flood water [29] Groundwater Dug well: Fecal coliform levels between 10 5 -10 6 CFU/mL Dug wells in a neighbourhood with septic systems (Gondwe 1997, referred to in [34]. Downstream environment E. coli 10 2.3 /g·soil, 10 5.6 /100 mL irrigation water Contamination of irrigation water and soil, Accra (Ghana) [38] Fresh produce E. coli 0.64 to 3.84 log10/g·produce [38] Helminths-10 3 /L and 10 6 /L faecal coliform [39] Rotavirus 10 −0.7 /100 g·wet·weight, Ascaris: 10 3.8-5.7 /100 g·wet·weight [37] E. coli-consumption 52-102 g/salad week Helminth and Faecal coliform: Mean contamination on produce at farmgate fed by stream water in South Africa [39] Virus and Ascaris wastewater irrigated lettuce in Accra (Ghana), values at Farm [37]. Ratio of pathogen concentration to E.coli E. coli O157: H7 7.6 × 10 −4 to 10 −2 . Cryptosporidium 5.5 × 10 −7 , Rotavirus 5.5 × 10 −6 , Ascaris 10 −6 per E.coli Used to compare E. coli concentration with other pathogens when data unavailable. [11,28,40]. However it is recognised that these ratios would not remain constant across the various pathways and treatment stages.

Sanitation scenario inputs -base case and improvement options
The illustrative case was primarily based on data from the Dhaka Bangladesh SFD report [32] with the division of wastewater flows between sanitation services shown in Figure S1, which is a visual representation of the data shown in Table 2 in the main report. Alternative sanitation improvement options were then developed on the basis of the most significant exposure pathways determined from the base case assessment. The options were tested by revising the model set-up or inputs such as the flow division (i.e., reducing % sewer flows flooding), adjusting pathogen log10 reductions (i.e., improving treatment efficacy) or exposure to reflect actual system improvements.
The details of these changes are shown in Table S7, with the results of the options analysis presented  in the main report Table 3. Figure S1. Inputs to model for Dhaka Bangladesh Base Case (alternative to Table 2 in main text). Table S7. Option modifications as analysed in the model and referred to in Table 3 of the main text.

Improvement Option
Changes made to the Base Case Inputs to Model Improvement Options 1a. Reduce leakage from sewer and drain Leakage from sewer and drain: based case 2%, change to 0.1% and flows shifted to continue 1b. Reduce groundwater use by half Base case proportion of population exposed to groundwater was 25%, changed to 12.5% 2a. Reduce exposure to local drain (i.e., Cover drains) Change exposure from 35% to 5% population exposed to local drain.
3a. Toilet and ST to sewer (not drain) Reduced toilet discharge to drain (21% to 5%) and septic tank effluent to drain (49% to 20%) to instead discharge to sewer. 3b. Improve conveyance-stop flooding and leakage Reduced flooding from 25% to 1% and leakage from 2% to 0.1%, 99% flows continue in open drain or sewer. 3c. Improve downstream conveyance-flows to treatment Increase flows to treatment-Sewer discharge to wastewater treatment increased from 43% to 95% and drain from 1% to 50% 3d. Improve wastewater conveyance to treatment (combine three above) Toilet and septic tank effluent to sewer (as for 3a), improve local conveyance (3b) and discharge flows to treatment (3c).

4a. Increase sludge emptying
Increase emptying from 12% to 95%, and reduce overflow to 1%. Improved septic tank effluent log10 reduction due to regular emptying (see Table S2 above) 4b. Improve sludge emptying and conveyance (not treatment) Emptying increased and improved septic tank effluent treatment (as for 4a). Reduced sludge emptied to household from 72% to 1%, increased sludge discharged to treatment from 1% to 70%. 5. Improve wastewater and faecal sludge treatment Traditional solution-Improved both wastewater and faecal sludge to combined primary and secondary treatment (See Table S2) 6. Reduce exposure to drain and leakage, stop untreated reuse Non-traditional solution: Reduce exposure to drain (as per 2a) reduce groundwater use (as per 1b), stop reuse of wastewater and sludge before treatment (5% from sewer, 10% drain and 1% sludge shifted from produce to treatment).