Public Health Quantitative Microbial Risk Assessment
Patrick Gurian
  • LAST MODIFIED: 26 July 2017
  • DOI: 10.1093/obo/9780199756797-0166


Quantitative microbial risk assessment (QMRA) is the application of mathematical models of exposure and dose response to predict the likelihood of adverse outcomes due to exposure to pathogens. These adverse outcomes include infection (the microorganisms replicate in or on the host organism), morbidity (illness, the microorganisms induce disease in the host), and mortality (the host dies due to the effects of the microorganisms). QMRAs have addressed a variety of pathogens, including viruses, bacteria, protozoa, and prions, and produce probabilistic estimates of harm. In other words, QMRA generally does not indicate if an adverse outcome will occur or not, but will instead indicates the probability that it will occur. As pathogens are present in many environmental media, avoiding all potential exposures with pathogens is not a feasible goal. QMRA provides a way to assess the impacts of these many potential routes of exposures in order to inform decisions about which risks are significant enough to merit efforts to avoid or mitigate them. The term QMRA is generally applied only to the calculation of the probability of harm, while efforts to prioritize and make decisions about risk are referred to as “risk management.” QMRA is often used synonymously with the term “microbial risk assessment,” given that the microbial risk assessment framework (described below) includes inherently quantitative steps. However, non-quantitative approaches to risk are recognized as an important component of risk assessment and management (see Hazard Identification section below). Microbial risks may also be identified, and some cases quantified, using epidemiological methods that correlate exposures and risks without employing specific models of exposure and dose that are part of the QMRA framework (see below). QMRA approaches have been applied to inform standards for microbiological quality of food, water, air, and touched surfaces, such as counters, doorknobs, and so on. QMRA approaches are seen as valuable because they allow for hypothetical cases to be considered (i.e., scenarios for which there are no available data. QMRA also allows for an assessment of the exposures associated with very low risks, such as 1 in 1 million or 1 in 10,000, that may be desired targets for risk mitigation efforts but are too low to be realistically measured. QMRA approaches may be criticized as being frequently applied without validation to scenarios substantially different from the circumstances under which the models were developed. Consensus views see a role for QMRA in decision making, while also recognizing that in many cases QMRA estimates are subject to substantial uncertainty and should be interpreted cautiously.

Risk Assessment Framework

A common conceptualization of the framework for developing quantitative microbial risk assessments includes four steps: hazard identification, exposure assessment, dose response, and risk characterization (QMRA Wiki 2016). These steps are essentially the same as those commonly used for chemical risk assessment. Each of these steps is described in more detail below.

  • QMRA Wiki. 2016. Quantitative Microbial Risk Assessment (QMRA) Wiki. East Lansing, MI: Center for Advancing Microbial Risk Assessment.

    Save Citation »Export Citation »E-mail Citation »

    This website provides information on QMRA methods and serves as a repository of inputs and data for QMRA. Both raw data and fitted parameters are available for dose response models for a large number of pathogens and host species. An archive of case studies provides many examples of how QMRA methods can be applied in different contexts.

    Find this resource:

    Hazard Identification

    Hazard Identification consists of identifying the organisms, exposure pathways, and disease outcomes of concern for an individual or population. This is generally a qualitative process. The most important hazards identified during this step are then evaluated quantitatively in subsequent steps. A summary of pathogens and their effects is provided as part of the QMRA Wiki 2016.

    Exposure Assessment

    The number of organisms to which the individual or population of concern is exposed is quantified. “Exposure” refers to the time-weighted concentration in the environment. Exposure assessments take many different forms, depending on the environmental medium being considered, but they generally require a consideration of both dispersion (how organisms move in the environment) and attenuation (how organisms die off or multiply in the environment). More information on environmental fate and transport can be found in sources such as Walton 2008, which provides a relatively succinct overview, and Ramaswami, et al. 2005, which provides a more extended treatment of the subject that integrates environmental processes with risk assessment. Specific issues that may affect microbes are that they may be subject to accumulation at interfaces (such as straining in a porous media, see Bradford, et al. 2014 for an overview of microbial transport in the subsurface), may replicate in the environment, may attenuate in complex manners rather than following simple first order decay models, may cluster or clump, and measurements may not correspond to viable organisms (a PCR signal may not indicate a viable organism, culture may fail to detect all viable organisms). For an analytical example of how water and air transport models may be integrated into a microbial risk assessment see Teng, et al. 2013.

    • Bradford, S. A., Y. Wang, H. Kim, S. Torkzaban, and J. Simunek. 2014. Modeling microbial transport and survival in the subsurface. Journal of Environmental Quality 43:421–440.

      DOI: 10.2134/jeq2013.05.0212Save Citation »Export Citation »E-mail Citation »

      This is a detailed and mathematically sophisticated review of mathematical models of microbial fate in groundwater.

      Find this resource:

      • Ramaswami, A., J. B. Milford, and M. J. Small. 2005. Integrated environmental modeling: Pollutant fate transport and risk in the environment. Hoboken, NJ: Wiley.

        Save Citation »Export Citation »E-mail Citation »

        This is an extensive synthesis of environmental modeling knowledge that integrates fate and transport with risk assessment. It is not specific to microbes but covers general environmental processes.

        Find this resource:

        • Teng, J., A. Kumar, P. L. Gurian, and M. S. Olson. 2013. A spreadsheet-based site specific risk assessment tool for land-applied biosolids. Open Journal of Environmental Engineering 6:7–13.

          DOI: 10.2174/1874829501306010007Save Citation »Export Citation »E-mail Citation »

          This is a brief and readable case study of how environmental transport models may be applied to an exposure assessment for a QMRA. The case considered is land application of biosolids, which introduces the reader to a variety of exposure assessment techniques, including air dispersion modeling, groundwater modeling, transport to surface water, and contamination of food crops.

          Find this resource:

          • Walton, J. C. 2008. Fate and transport of contaminants in the environment. Glenn Allen, VA: College Publishing.

            Save Citation »Export Citation »E-mail Citation »

            This is a clear and concise introduction to the basic processes governing the fate of pollutants in the environment. It is not specific to microbes but rather provides a general overview of environmental processes.

            Find this resource:

            Dose Response

            The environmental concentration identified during the exposure assessment is integrated with the amount of the medium contacted by the host organism (that is, the volume of water ingested, volume of air inhaled, etc.) to develop a dose, or quantity of organisms reaching the receptor. In this step, a suitable mathematical relationship is identified between the dose (number of organisms to which the receptor is exposed) and probability of an adverse outcome. Unlike chemical risk assessment, doses are not normalized to body weight and interspecies scaling factors are not used. The most commonly used dose response models in QMRA are the exponential and the beta Poisson models (see QMRA Wiki 2017 for the mathematical forms of these models). Values for the model parameters are specific to the pathogen being considered and may vary based on factors such as the strain of the pathogen, the host species, and the status of the host (age, immune status, etc.). Specific values for parameters are generally identified by statistical analysis of controlled dosing trials, although, in some cases, epidemiological studies may be used instead. In the above models, the “risk” term may refer to different endpoints, including the risk of infection (the microorganisms is able to reproduce in the host), risk of illness (the infection produces disease in the host), and risk of death. The endpoint is specific to the dose response model used and corresponds to the endpoint of the database used in fitting the parameter values (that is, a databased on doses and infections produces a dose response model for the endpoint of infection). For further details on the derivation of these dose response models and for statistical approaches to estimates of parameters for the models, see Haas, et al. 2014.

            • Haas, C. N., J. B. Rose, and C. P. Gerba. 2014. Quantitative microbial risk assessment. 2d ed. Hoboken, NJ: Wiley.

              DOI: 10.1002/9781118910030Save Citation »Export Citation »E-mail Citation »

              This is a widely used reference on QMRA that provides guidance on conducting QMRAs and covers the mathematical basis of QMRA models. It also contains a chapter on using QMRA in risk-management decision making.

              Find this resource:

              • QMRA Wiki. 2017. Dose response assessment. East Lansing, MI: Center for Advancing Microbial Risk Assessment.

                Save Citation »Export Citation »E-mail Citation »

                This web page provides the mathematical formula for the commonly used dose response models, recommended models and parameters for a wide range of pathogens, and an overviews of the mathematics of fitting dose response models.

                Find this resource:

                Risk Characterization

                The dose identified from the exposure assessment is input into the dose-response function and a probability of the relevant endpoint (illness, disease, or death) is estimated. A full risk characterization should extend beyond a single computation to address issues such as the uncertainty in the risk estimate, the variability in risk among different individuals and subpopulations, and the factors that contribute the most to uncertainty in the estimate. QMRA Wiki 2016 contains example calculations of risk and a tutorial on a common uncertainty propagation technique, Monte Carlo analysis. QMRA analyses may be complicated by the need to account for risk due to secondary transmission of pathogens (i.e., an infected host transmits the infection to another host). Generally, the environmental transport and dose response processes described above continue to operate in these situations, but they may be modeled in more or less detail given the diversity of exposures that may occur in these situations. For a discussion of common modeling approaches, see Haas, et al. 2014, cited under Dose Response.

                Risk Management

                As noted in the Introduction, risk management is separate from QMRA, but the latter is generally performed in order to complement a risk management effort. Typically, QMRAs are performed for different scenarios involving varying stringency of risk reduction measures. The risks, as well as the economic costs of different control options, are estimated under different scenarios (see Haas, et al. 2014, cited under Dose Response). Tradeoffs between risks and economic costs are made either explicitly or implicitly to choose a preferred risk management option. Risk communication, including two-way communications with a broad range of stakeholders, is an important component of risk management. While the procedure is described as a step-by-step, linear process, in reality it is often best implemented in an iterative fashion in which preliminary estimates are first made for each component of the framework. Based on these preliminary estimates, the important assumptions and parameters are identified and subsequent iterations focus on improving estimates for these key parameters or establishing the validity of key assumptions. Similarly, an iterative approach to risk management allows risk communication efforts to identify risks of concern to affected parties. While risk communication is often placed after risk assessment, an iterative approach will allow risks identified through communication efforts to be evaluated in the risk assessment framework (Morgan, et al. 2002).

                • Morgan, M. G., B. Fishhoff, A. Bostrom, and C. J. Atman. 2002. Risk communication: A mental models approach. New York: Cambridge Univ. Press.

                  Save Citation »Export Citation »E-mail Citation »

                  This reference describes an iterative, empirically grounded approach to risk communication. The process beings with a review of the literature and consultation with experts to develop a conceptual diagram of the causal processes influencing the risk. Interviews and surveys are used to assess public understanding. Risk communications are designed to address consequential deficits in the public’s understanding, and the performance of these risk communications is evaluated.

                  Find this resource:

                  Application Areas

                  Pathogens that have been addressed by microbial risk assessment include bacteria, viruses, protozoa, and prions. Microbial risk assessments have been applied in a wide variety of domains, including exposure to pathogens in food (see Teng, et al. 2013, under Exposure Assessment), drinking water (Ryan, et al. 2013), air (Hong, et al. 2010), and through activities such as swimming (Betancourt, et al. 2014) and contact with surfaces (fomites) (Ryan, et al. 2014).

                  • Betancourt, W., D. C. Duarte, R. C Vasquez, and P. L. Gurian. 2014. Cryptosporidium and Giardia in tropical recreational marine waters contaminated with domestic sewage: Estimation of bathing-associated disease risks. Marine Pollution Bulletin 85.1: 268–273.

                    DOI: 10.1016/j.marpolbul.2014.05.059Save Citation »Export Citation »E-mail Citation »

                    This is an example application of using QMRA methods to interpret monitoring data. Concentrations of giardia and cryptosporidium are reported for a number of beaches in Venezuela and then QMRA is used to estimate risks of swimming at those beaches.

                    Find this resource:

                    • Hong, T., P. L. Gurian, and N. F. Dudley Ward. 2010. Setting risk-informed environmental standards for Bacillus anthracis spores. Risk Analysis 30.10: 1602–1622.

                      DOI: 10.1111/j.1539-6924.2010.01443.xSave Citation »Export Citation »E-mail Citation »

                      This example analysis considered risks due to inhalation exposure to B. anthracis spores. Re-aerosolization of spores from surfaces is modeled and a method is presented to link surface concentrations with inhalation risks.

                      Find this resource:

                      • Ryan, M. O., P. J. Duzinski, P. L. Gurian, C. N. Haas, and J. B. Rose. 2013. Acceptable microbial risk: Cost-benefit analysis of a boil water order for Cryptosporidium. Journal of the American Water Works Association 105.4: E189–E194.

                        DOI: 10.5942/jawwa.2013.105.0020Save Citation »Export Citation »E-mail Citation »

                        This example application conducts a QMRA for Cryptosporidium in drinking water. The risks from the QMRA are then input into a cost-benefit analysis of the impacts of a boil water order to identify the circumstances under which risks are high enough to justify the costs of the order.

                        Find this resource:

                        • Ryan, M. O., C. N. Haas, P. L. Gurian, C. P. Gerba, B. M. Panzl, and J. B. Rose. 2014. Application of quantitative microbial risk assessment for selection of microbial reduction targets for hard surface disinfectants. American Journal of Infection Control 42.11: 1165–1172.

                          DOI: 10.1016/j.ajic.2014.07.024Save Citation »Export Citation »E-mail Citation »

                          This example analysis conducts a QMRA for pathogens transmitted by touched surfaces and then evaluates what level of disinfection of these surfaces is required to reduce risks to below 1 in 1 million.

                          Find this resource:

                          Validation Challenges

                          QMRA is widely used in research, and especially in health, safety, and environmental standards. Nevertheless, there are potential issues with the application of these methods, which include: (1) Extrapolation to low doses: Dose response models are generally formulated based on high-dosage experimental trials (i.e., doses producing positive responses in >10% of the experimental animals). The use of doses that produce the outcome of concern in a sizeable fraction of subjects is necessary to observe statistically significant associations between dose and risk given the limited number of animals that may feasibly be included in dosing trials. The dose response models are then extrapolated to doses of regulatory concern that are much lower, generally lower than 1 in 100 and often as low as 1 in 10,000 or 1 in 1 million. The validity of this extrapolation cannot be rigorously established by feasible experiments. (2) Extrapolation to humans: Dose response functions are sometimes based on human dosing trials, but more frequently on nonhuman animal trials due to ethical concerns and because the validation of such procedures is problematic. The relative vulnerability of one species relative to another is a complex function of the interaction between the host and pathogen and cannot be readily predicted (see Bartrand, et al. 2008). (3) Heterogeneity of host response: Host susceptibility may vary greatly due to factors such as age, immune competence, and prior exposure to the pathogen. While the general microbial risk assessment approach allows for such heterogeneity (see Weir and Haas 2009 for an example analysis), in most cases specific information on how these factors affect dose response relationships is not available. (4) Heterogeneity of infectious agents: Microbes have short reproductive cycles and mutate rapidly. There is a great deal of variability among different strains (see Mitchell-Blackwood, et al. 2012). Thus, risks may vary substantially based on the exact nature of a particular strain. As with host variability, the microbial risk assessment approach is, in theory, capable of handling such variability, but in practice, data on specific strain characteristics are often not available. These challenges mean that quantitative microbial risk assessments can often not be rigorously validated and may be subject to substantial uncertainties. Despite these limitations, microbial risk assessments have been found to be a useful input into many decision-making processes.

                          • Bartrand, T. A., M. H. Weir, and C. N. Haas. 2008. Dose-response models for inhalation of Bacillus anthracis spores: Interspecies comparisons. Risk Analysis 28:1115–1124.

                            DOI: 10.1111/j.1539-6924.2008.01067.xSave Citation »Export Citation »E-mail Citation »

                            This study fits dose response models to data for guinea pigs and monkeys. In some cases, data can be pooled across species, but not in all cases. The study notes the need for the development of general interspecies correction factors for dose response models.

                            Find this resource:

                            • Mitchell-Blackwood, J., P. Gurian, R. Lee, and B. Thran. 2012. Variance in Bacillus anthracis virulence assessed through Bayesian hierarchical dose-response modelling. Journal of Applied Microbiology 113.2: 265–275.

                              DOI: 10.1111/j.1365-2672.2012.05311.xSave Citation »Export Citation »E-mail Citation »

                              This is an example analysis of how Bayesian statistical techniques may be used to describe variability in dose response parameters across strains of the same species of pathogen. Data on exposure and mortality for guinea pigs exposed to different strains of B. anthracis are analyzed and found to produce very different dose response model parameters, indicating a large variation in the risk presented by different strains of B. anthracis.

                              Find this resource:

                              • Weir, M. H., and C. N. Haas. 2009. Quantification of the effects of age on the dose response of Variola major in suckling mice. Human and Ecological Risk Assessment 15.6: 1245–1256.

                                DOI: 10.1080/10807030903304906Save Citation »Export Citation »E-mail Citation »

                                This paper extends common dose response models to consider the effects of age on dose response for smallpox exposure in mice.

                                Find this resource:

                                Available Guidance

                                Guidance on implementation of quantitative microbial risk assessment methods is available from a variety of sources. Regulatory agencies have developed guidance documents (US Department of Agriculture and US Environmental Protection Agency 2012, US Environmental Protection Agency 2014), a standard text is available (Haas, et al. 2014, cited under Dose Response) and an internet wiki site has been established to provide guidance on methods as well as summaries of inputs commonly required for microbial risk assessments (QMRA Wiki 2016, under Risk Assessment Framework).

                                back to top