Quantitative Microbial Risk Assessment and Infectious Disease Transmission Modeling of Waterborne Enteric Pathogens

Abstract

Purpose of review

Waterborne enteric pathogens remain a global health threat. Increasingly, quantitative microbial risk assessment (QMRA) and infectious disease transmission modeling (IDTM) are used to assess waterborne pathogen risks and evaluate mitigation. These modeling efforts, however, have largely been conducted independently for different purposes and in different settings. In this review, we examine the settings where each modeling strategy is employed.

Recent findings

QMRA research has focused on food contamination and recreational water in high-income countries (HICs) and drinking water and wastewater in low- and middle-income countries (LMICs). IDTM research has focused on large outbreaks (predominately LMICs) and vaccine-preventable diseases (LMICs and HICs).

Summary

Human ecology determines the niches that pathogens exploit, leading researchers to focus on different risk assessment research strategies in different settings. To enhance risk modeling, QMRA and IDTM approaches should be integrated to include dynamics of pathogens in the environment and pathogen transmission through populations.

Introduction

Waterborne pathogens that cause enteric and diarrheal disease remain a global public health threat. The Global Burden of Disease study estimates that 1.3 million people, including 0.5 million children under five, died from diarrheal disease in 2015 [1]. Much of this burden is in low- and middle-income countries (LMICs) [2], largely due to inadequate water and sanitation infrastructure. Although the waterborne disease burden in high-income countries (HIC) is lower, outbreaks caused by emerging pathogens and aging infrastructure have resulted in an increase in research and public health activity. In both LMICs and HICs, global climate change (and resulting changes in temperature, rainfall, and extreme weather) has increased the burden of waterborne disease [3, 4]. In the coming decades, climate change is expected to continue to increase this burden and complicate control efforts [5, 6]. To address these public health issues, models have been used for assessing risks and evaluating mitigation efforts. The specific choice of modeling strategy depends on policy question of interest, which will vary by setting. Here, we conduct a literature review to examine the extent to which modeling strategy, as well as the pathogen and transmission pathways of interest, is influenced by the income setting of the study site.

Waterborne enteric pathogens include bacteria, viruses, and protozoa and can exploit a variety of transmission pathways. The major pathogens of concern are pathovars of Escherichia coli (EIEC, ETEC, EPEC, EHEC, EAEC, etc.), Salmonella spp., Shigella spp., Campylobacter spp., Vibrio cholerae, Cryptosporidium spp., Giardia spp., rotavirus, norovirus, poliovirus, and hepatitis A. These pathogens are transmitted through multiple pathways, including water (drinking or recreational water), food, fomites, soil, and contact with animals, which complicates intervention strategies and assessments of disease risk. Moreover, these multiple pathways can be interdependent [7]; e.g., water can be the driver of contamination even when it is not the proximate route of exposure to humans, as in the case of produce contaminated by irrigation. This complex system of transmission pathways can hinder pinpointing the “cause” of the disease, but it can also create opportunities for interventions at different points along the transmission chain. Increasingly, models are being used to examine modes of transmission and optimal intervention strategies. The choice of model framework is often dependent on the specific transmission pathways and patterns of disease observed in a given context.

In their seminal paper, Levine and Levine [8] argued that the distinct human ecologies in developing and developed country contexts have led to distinct patterns of enteric disease epidemiology. In LMICs, for example, increasing urbanization has driven increased population densities, often with insufficient water and sanitation infrastructure; decreasing breast-feeding has increased children’s susceptibility to infection and disease; changing agriculture practices have increased fecal contamination and exposure to zoonotic pathogens; and increasingly disruptive and destructive periods of violence and political strife can devastate infrastructure and create high pathogen transmission conditions. In HICs, on the other hand, access to developed and advanced water and sanitation infrastructure has largely eliminated waterborne enteric bacterial disease. Waterborne viruses and certain protozoa (e.g., rotavirus, Cryptosporidium), however, are still of concern. Ecological niches for pathogen transmission in HIC settings include concentrated animal feeding operations (CAFOs) and large-scale food-service operations; agricultural and food-based operations remain a vehicle for outbreaks of (largely bacterial) enteric illness. Moreover, the increasing dependence on care institutions (e.g., child or geriatric care centers) has left certain populations vulnerable to both acute and chronic enteric infections. These different ecological niches across different national income strata have resulted in different research priorities and strategies. Specifically, different risk models are implemented depending on many factors, including disease burden and environmental contamination levels. This review seeks to clarify how these niches have resulted in divergent use of two modeling strategies. Quantitative microbial risk assessment (QMRA) and infectious disease transmission modeling (IDTM) have both been used to better understand the magnitude and source of disease risk, the contributions of different transmission pathways, and the opportunities for interventions.

QMRA is a modeling framework for characterizing pathogen exposure from the environment through an exposure assessment and for characterizing human risk from that exposure using a dose–response model [9]. QMRA models are comprised of a series of probabilistic steps informed by the literature or primary data collection; the whole model is run many times to generate a distribution of outcomes (Figure 1a). QMRA was initially applied to estimate risk of disease from waterborne pathogens in the 1980s [10–12], using dose–response functions fit to data from the literature [13, 14], and it has since been used in a variety of risk settings. Technological and methodological advances have continued to improve QMRA, and it has been adopted as an assessment framework in many regulatory settings. QMRA models can be limited by lack of information for the specific context; for example, 1) the dose-response function may be estimated based on an animal model or on a different exposure medium; 2) the treatment efficacy may be only known under low turbidity conditions; or 3) the human-environment contact network may not be well-characterized. Furthermore, the uncertainty at each step is compounded. Even with these limitations, QMRA models excel at representing how heterogeneities interact to produce rare, stochastic harm events.

Figure 1.

Schematics of typical a) quantitative microbial risk assessment (QMRA) and b) infectious disease transmission model (IDTM) research studies.

IDTM has long been used as a tool in mathematical epidemiology to connect underlying transmission mechanisms to observed outbreaks, to make predictions of incidence, and to investigate the likely impact of control strategies. The fundamental framework is the susceptible–infectious–recovered, or SIR, model developed in the early 1900s [15] that models person-to-person contact and infection transmission in a population of individuals (Figure 1b). Although preliminary work modeling waterborne pathogens in humans began to gain traction in the early 2000s, including retrospective analysis of the 1994 Cryptosporidium outbreak in Milwaukee, WI [16, 17], it was the 2010 cholera outbreak in Haiti that marked a turning point for the field. The 2010 Haiti outbreak prompted an increased interest in IDTM methods for waterborne risk characterization and methods development. For waterborne enteric pathogens, direct person-to-person transmission is typically only one of multiple exposure pathways; environmental exposures—through water, food, or fomites—are a significant, or even the primary, source of infection in many contexts. Modelers have only recently begun to develop and adopt SIR-type models that include an explicit description of an environmental transmission pathway to address this concern [18–21]. In particular, the field of cholera modeling (responding to the 2010 outbreak in Haiti), has been an important adopter of this environmental framework and has a significant presence in the literature (citations are included in supplementary information). While IDT model are typically limited to describing average behaviors, they have been powerful tools for analyzing population level-dynamics.

Despite similarities in conceptual framework and goals, there has been relatively little methodological crossover between QMRA and IDTM. For instance, QMRA rarely considers possible secondary transmission in estimates of risk, while IDTM is only beginning to incorporate pathogen measurement and environmental surveillance. Through a literature review, we explore how research in different settings has leveraged varied modeling strategies and how these strategies differ specifically according to income context (HIC or LMIC), pathogens of study, and environmental transmission pathway.

Literature search methods

We identified references for this review through a search of PubMed for articles published between January 1, 2012 and September 16, 2017 with search terms “E. coli” or “Escherichia coli” or “Salmonella” or “typhoid” or “Shigella” or “Campylobacter” or “Cholera” or “Vibrio cholerae” or “Norovirus” or “Cryptosporidium” or “Giardia” or “polio” or “hepatitis A” or “waterborne pathogens” or “enteric pathogens” or “enteropathogens” AND “risk assessment” or “quantitative microbial risk assessment” or “QMRA” or “transmission dynamics” or “mathematical model” or “infectious disease model” or “transmission model”. We chose this set of waterborne pathogens for their important role in enteric disease globally; although other pathogens may share some common transmission pathways (e.g., foodborne pathogens), we only included pathogens for which water was a significant transmission pathway. After reviewing the results of this initial search, we identified 220 publications that form the basis of this review. Although not every reviewed paper is discussed, a full bibliography is provided in the supplementary material. Because there were relatively few IDTM compared with QMRA studies, citations for IDTM papers prior to 2012 are also included in the supplementary material. This review does not include studies using IDTM models with environmental pathways if they were published in journals not indexed by PubMed or if they were not used in a waterborne disease context.

Since environmental pathways exploited by pathogens are multiple and interdependent, the pathway involved in contamination is not necessarily the pathway involved in proximate exposure. For example, contaminated agricultural runoff may be the source of river contamination, though exposure occurs through recreational activities in or on the river, such as swimming or boating. In this review, we classify the transmission pathway of an article by its main emphasis, which, in this example, might be wastewater if it focuses on agricultural contamination or might be recreational water if the focus is on recreational exposure.

Results

QMRA and IDTM Research Objectives

There were 149 QMRA articles that met our study criteria, 93 of which were conducted in HICs (62%). The number of QMRA papers in HICs has been slowly increasing over the past 5 years but the number of published articles in LMICs has remained largely constant (supplement Figure S1a). There were 72 IDTM articles that met our study criteria; most were grounded in LMIC contexts (47%), with the remainder split between HIC (24%) and generic/unspecified (29%) contexts (supplement Figure S1b). QMRA studies explicitly focus on environmental risk and are designed to assess risk and probability of rare events under heterogeneous conditions. They are, therefore, largely focused on evaluation of an existing, large-scale process or treatment or a novel, smaller-scale treatment. The research objectives for IDTM studies, on the other hand, are focused on population-level dynamics. IDTM studies, therefore, are concerned with understanding epidemiological patterns of outbreaks or evaluating the impacts of possible policies or interventions, primarily vaccination. The advancement of modeling methods was explored for both QMRA and IDTM. These papers are not discussed in this review, but are included in the supplementary information.

QMRA studies that evaluated treatment or intervention strategies disproportionately came from HICs. In the HIC context, most studies considered wastewater treatment or other water management strategies [22–32] or the treatment of food products before consumption [33–38]. A handful of papers considered hand or fomite disinfection [39–41], household-level water treatment [42], or vaccination of animals against waterborne pathogens [43]. In LMICs, there was more emphasis on household water treatment [44–47] and less on wastewater treatment or water management [48–52], the latter primarily coming from upper-middle-income countries. Studies also considered treatment of livestock waste [53] and disinfection of kitchen clothing [54]. Papers evaluating current practices or infrastructure, on the other hand, were dominated by assessment of wastewater irrigation in both HICs [55–64] and LMICs [30, 65–73]. Studies evaluating other wastewater management practices [74, 75], rainwater collection [76–78], and food preparation and supply management [79–81] were done in both HICs and LMICs.

The research objectives for IDTM studies were primarily outbreak assessment and vaccination or other intervention assessment. Outbreak assessment papers were dominated by cholera [82–93] and polio [94–96]. Cholera [97–102] and polio [103–108] were also well represented among vaccination and other policy assessments; vaccination (and vaccine cost-effectiveness in particular) was the primary focus of rotavirus IDTM analyses [109–113]. Other objectives included dynamics analysis or risk-of-outbreak assessment.

Pathways and pathogens: research priorities in LMIC and HIC settings

QMRA

When comparing research using QMRA in LMIC and HIC research settings, we find a greater focus on food (38% HIC vs. 16% LMIC) and recreational water (21% HIC vs. 12% LMIC) in HICs (N=93) relative to LMICs (Figure 2a). These exposure pathways are exploited in developed countries because of the high-intensity farming practices and remaining vulnerabilities with non-point and point source contamination. In LMICs (N=50), on the other hand, the focus shifts toward wastewater contamination (36% LMIC vs. 23% HIC) and drinking water (29% LMIC vs. 7% HIC), likely because of the prevalent use of wastewater in agriculture and the limited availability of effective water treatment. These priorities are consistent with the human–pathogen ecology described by Levine and Levine [8].

Figure 2.

Distribution of a) quantitative microbial risk assessment (QMRA) research studies by transmission pathway and b) infectious disease transmission model (IDTM) research studies by study objective. The distributions are stratified by income context (low- and middle-income (LMIC) and high-income (HIC) countries). IDTM articles without a specific country context are not included in this figure.

HICs have robust drinking water and sanitation infrastructures but also have industrialized food systems that are vulnerable to large-scale contamination and citizens with generally more leisure time to enjoy recreational water through swimming, paddling, or other activities on rivers, lakes, and beaches. The most commonly studied waterborne pathogens in the HIC context, therefore, were either associated with food (Campylobacter (N=28), norovirus (N=24), and non-typhi Salmonella (N=22)), recreational water (Cryptosporidium (N=28)), or both (E. coli (N=30)). Even if food is the means of transmission of the pathogen to the population, it is not necessarily the source of contamination. Contamination sources may be wastewater irrigation, food handlers, etc. Foodstuffs of concern were primarily related to animal products—pork [35, 81, 114–117], beef [118–121], chicken [79, 122–125], shellfish [37, 126], and eggs and dairy [127–133]—but contaminated vegetable crops, such as nuts [33, 34, 134] and greens [29, 58, 61, 63, 135–137] were also studied. The pathogens most often investigated in food-related transmission context were Salmonella (N=8), Campylobacter (N=7), and E. coli (N=6). There are many other pathogens (e.g., Listeria) and toxins that are of concern in food-based risk assessment, but they were not included in this review as they are not considered waterborne pathogens.

Risk from exposure to water, particularly recreational water but also drinking water, was another major focus of QMRA in HICs and includes exposure from rivers, streams, or beaches [62, 138–154], urban infrastructure [155–157], and swimming pools or water parks [158, 159]. Sources of contamination ranged from agricultural to urban, and the primary pathogens of concern were Cryptosporidium (N=14), Giardia (N=10), E. coli (N=11), norovirus (N=8), and Campylobacter (N=9). Drinking water risk is still a concern in HICs because agencies like the U.S. EPA regulate drinking water treatment plants at low risk thresholds, on the order of 1 case in 10,000. Most work in this area has focused on wastewater reuse [23–25, 64, 74, 160], drinking water sourced from surface water [27, 32, 146, 161], and agricultural [57] or urban run-off or discharge [28, 142]. The pathogens most often considered in this context are Cryptosporidium (N=9) and E. coli (N=5), with lesser representation of norovirus (N=4) and Giardia (N=3), among others.

In LMICs, on the other hand, water and sanitation infrastructure can vary quite widely within a country, ranging from unimproved to safely managed. Increased pathogen load and transmission caused by poor infrastructure, as well as an international focus on epidemics flourishing in the wake of natural disasters or war—such as cholera in Haiti and Yemen and polio in Syria [162]—has resulted in research efforts directed at drinking water (29%) and wastewater (36%) (Figure 1). The major pathogens studied in the LMIC context are E. coli (N=20), Giardia (N=15), Cryptosporidium (N=13), Salmonella (including S. typhi; N=13), and rotavirus (N=12).

Drinking water was the major exposure pathway of concern in LMICs. Recent risk assessments have also emphasized the source of the drinking water contamination pathway. For example, studies in LMICs have considered contaminated surface water [163–166], groundwater [167, 168], rainwater [78], treated water [169], and multisource periurban water [48, 170, 171]. Researchers have also considered the impacts of intermittent water supplies [172] and household water treatment [45–47, 173]. Major pathogens considered in the drinking water context were E. coli (N=10), Cryptosporidium (N=7), Giardia (N=6), rotavirus (N=4), and Salmonella (N=4).

Wastewater was another common focus of risk assessment in LMIC, covering wastewater irrigation [30, 51, 52, 66, 67, 71–73, 174], livestock waste [53, 68], non-potable reuse or greywater [44, 175], treatment failure [49], and effluent discharge [50]. The pathogens of interest were varied and included E. coli (N=5), Giardia (N=4), Salmonella (N=3), rotavirus (N=3), and Cryptosporidium (N=3). The majority of wastewater risk assessments considered crop irrigation as the final exposure route, but we classify these as focusing on wastewater rather than foodborne exposures because of the focus on the source of contamination and the consideration of other relevant pathways, such as accidental ingestion by farm workers and their children.

IDTM

In this review, we found IDTM analyses contributing a wide variety of insights, from the role of drug resistance in recent Typhoid fever outbreaks [111], to optimization of oral cholera vaccination campaigns [98], to the risk of spread of polio from Hajj pilgrims and other international travelers [106]. The IDTM papers in our literature search differed from the QMRA papers in their conceptualization of transmission pathways. Compared to QMRA, exposure pathways are typically abstracted in IDTM into direct, person-to-person transmission and, sometimes, indirect, environmentally mediated transmission. Even when the transmission pathways are contextualized for the pathogen and population, the modeling mechanics typically remain the same. Unlike QMRA, inclusion of any kind of environmental data was rare, even when an environmental compartment was explicitly included in the model.

Recent IDTM work differs from QMRA not only in terms of transmission pathways but also in the set of pathogens than that it focuses on: cholera [82–93, 97–102, 176–179], polio [94–96, 103–108, 180–184], rotavirus [109–113, 185], and hepatitis A [186–190] were the most common, though Salmonella typhi [191–193] and norovirus [194–196] were also represented. The pathogens represented in IDTM papers were primarily either associated with recent, highly politicized outbreaks (e.g., cholera) or vaccine-preventable diseases (e.g., polio, rotavirus), in stark contrast to the pathogen distribution among QMRA papers. IDTM studies were overwhelmingly set in LMICs, perhaps due to the fact that there are more epidemiological studies in LMICs collecting health data that are exploited by IDTM and less of the environmental surveillance data that is generally exploited by QMRA studies. Exceptions where we see IDTM studies in HIC settings are for those pathogens where a vaccine has been or is currently being developed, such as rotavirus and norovirus, and for those pathogens where this is risk of an outbreak after importation into a HIC (e.g., [181]). Indeed, while outbreak assessment was primarily seen in the LMIC context, policy-driven analyses made up a large fraction of papers in both HIC and LMIC context (Figure 2b). IDTM methods papers often did not specify an income context.

Conclusions

Human ecology shapes the ecological niches exploited by human pathogens [8], which has led to different risk assessment research objectives in LMICs and HICs. In HICs, food and recreational water are the major sources of exposure, and risk assessment is often closely tied to regulation. In LMICs, research has focused on drinking water and wastewater and is particularly concerned with sources of contamination and developing intervention strategies. While these specific research objectives are largely natural extensions of the human and pathogen ecology, it is important to recognize that risks of waterborne disease are not homogeneous, particularly in LMICs.

We found very little crossover between QMRA and IDTM approaches to waterborne risk characterization and assessment. Only one recent QMRA study explicitly assessed an outbreak (salmonellosis in beef burgers in France [120]). Secondary transmission is typically neglected in QMRA, especially in HICs, and, when it is considered, it is typically in the form of a secondary attack rate parameter (e.g., [63]) and not dynamic modeling. Although the benefits of including IDTM thinking and techniques into QMRA have been previously examined [19], the practice has not yet caught on. The lack of consideration of secondary transmission may stem from the perception that secondary transmission is negligible, particularly in HICs where conditions force outbreaks to die out. However, even if an outbreak is not transiently amplified, there maybe be a significant burden of secondary infections as the outbreak dies out (e.g., as much as 20% of E. coli cases in outbreaks [197]). Moreover, vulnerable or high-risk populations can sustain outbreaks in settings where an outbreak would not take-off among general population. For example, 2017–18 Hepatitis A outbreaks in California, Michigan, and Utah were predominantly concentrated among homeless people and drug users [198]. Increasing rates of vaccine refusal have also created pockets of higher transmission potential.

IDTM can also benefit by adopting some of the exposure assessment approaches used in of QMRA, such as the modeling of environmental pathways. Although a number of waterborne disease IDTMs now include environmental compartments and transmission pathways, these choices have largely been driven by the desire to improve fitting and parameter estimation (e.g., [91]) or to better model spatial spread (e.g., [199, 200]). Leveraging environmental sampling to inform and verify modeled exposures has thus far been rare (e.g., [201]) but has potential to significantly increase the impact of and confidence in IDTM results.

Real advances in risk assessment will soon be possible by incorporating faster and more precise environmental pathogen measurement into both QMRA and IDTM. Although many recent QMRAs do measure pathogens in the environment as a basis for risk assessment, many do not, and the use of proxy indicator pathogens is still prevalent despite being an unreliable surrogate. Assays, such as those using the TaqMan Array Card [202], are being developed to quickly quantify multiple pathogens of interest. Methodological work, such as hydrodynamic modeling or the refinement of the assumptions of exposure quantification, has also advanced and is an important part of improving modeling practices. Ultimately, by reducing the uncertainty in exposure, risk assessors will be able to more precisely characterize and quantify risks from waterborne pathogens.

Historically, QMRA and IDTM studies have been developed and conducted independently with differing research objectives. We argue that risk modeling activities will be enhanced if these modeling approaches are integrated to include dynamics of both the fate and transport of pathogens in the environment as well as the transmission of pathogen through populations.