Team Members: Leak, E., Abla, M., Meschke, J.S., Jackson, C., Pogreba-Brown, K., Eltholth, M.M.
Problem Statement
Schools, particularly schools in inner cities and rural areas do not have safe potable water for students. While the Safe Drinking Water Act and the Ground Water Rule covers some of the microorganisms that may contaminate the water, viruses are not commonly monitored. The question remains, is the health of students being properly protected if these viruses are not tested for? This case study consisted in the evaluation the risk posed by the contamination with Norovirus of a well that serves an elementary rural school in the state of Wisconsin. It was taken into account only the exposure of school population to drinking water. The others routes of exposure were not considered due to the period of time available for this work. The obtained results were used to suggest measures to mitigate this risk.
We consider two possible scenarios to evaluate the exposure:
- Scenario A: Joan Rose Elementary is a school of 300 students located in rural Wisconsin. The school’s water supply comes from a private well. On a random test by environmental health, the well was tested. Results from the lab showed Norovirus to be present at DOSE X. Knowing the benefactor of the school, the teacher called in some assistance to determine what risk the students were at for becoming ill from drinking water at the school and where the contamination came from.
- Scenario B: A child returns to Joan Rose Elementary only a day after suffering a severe bout of vomiting and diarrhea. The child has two episodes of diarrhea (in the toilet) before being sent home. Unfortunately, there was a septic tank leak into the school’s well. What is the risk to the school population for both infection and disease
The methodology used was a Quantitative Microbial Risk Assessment – QMRA (Hass, Rose and Gerba), comprised of four stages, based on the National Academy of Sciences framework for Quantitative Risk Analysis, but modified to account for the properties of living organisms like BAC:
1. Hazard Identification: Describe a microorganism and the disease it causes, including symptoms, severity, and death rates from the microbe. Identify sensitive populations that are particularly prone to infection.
2. Dose-Response: The relationship between the dose (number of microbes) received and the resulting health effects. Data sets from human and animal studies allow the construction of mathematical models to predict dose-response.
3. Exposure Assessment: Describe the pathways that allow a microbe to reach people and cause infection (through the air, through drinking water, by touch, etc.). Determine the size and duration of exposure by each pathway. Estimate the number of people exposed and the categories of people affected.
4. Risk Characterization: Integrate the information from steps 1, 2, and 3 into a single mathematical model to calculate risk -- the probability of an outcome like infection, illness or death. Since steps 1, 2 and 3 will not provide a single value, but a range of values for exposure, dose, and hazard, risk needs to be calculated for all values across those ranges. This is called Monte Carlo Analysis, and the result is a full range of possible risks, including average and worst-case scenarios. These are the risks decision makers look at when choosing policy and that scientists use to determine if additional experiments are needed to refine the model’s parameters.
Hazard Identification
Norovirus is a calicivirus that as a group consists of five genogroups, three of which contain human strains. In the environment, viral particles are very stable which accounts for a high rate of infection due to fomite contamination (Jones). One of the challenges with studying noroviruses is that they cannot currently be grown in culture. Historically, feline calicivirus has been used as surrogate but more recently RT-PCR is used to detect genetic material and is then used to extrapolate the total number of viral particles present in the sample (Teunis). Norovirus is estimated to cause 21 million cases of illness in the United States every year (Norovirus Technical Fact Sheet). These estimates include both outbreak and sporadic cases. From 1997-2003 there were 232 outbreaks, 50% were attributed to a food source and 3% to water sources. Norovirus was reported as the etiologic agent in 38% of cases related to water-borne disease outbreaks, and related to ground water, accounted for 43% of all cases reported from 2005-2006 (Yoder). Attack rates are often very high given the low infectious dose.
Table 1.
| % Attack Rate | ||||
|---|---|---|---|---|
| Group | N0Cases | Median | 25th | 75th |
| Oyster-associated outbreaks | 95 | 58.3 | 40 | 75 |
| Food handler-associated outbreaks | 195 | 47.2 | 33.3 | 67.7 |
| GII/4-associated outbreaks | 27 | 41 | 29.9 | 54.5 |
| Other outbreaks | 136 | 56.9 | 40.0 | 75 |
| GII/3-associated outbreaks | 20 | 64.8 | 40.2 | 81.7 |
| Other outbreaks | 143 | 53.2 | 37.6 | 73.8 |
Nationally, Norovirus accounted for 3,673 outbreaks (106,412 cases) from 1998 to 2009 and 150 outbreaks (4,153 cases) in Wisconsin. These high values are accounted for the prevalence of the virus in food and water sources as well the high infection rate of the organism. Feeding studies have found that as low as 18 virus particles are capable of causing infection (Teunis) and viral shedding from diarrhea or vomiting can exceed 1.0 x 1010 viral particles per gram of stool.
Exposure Assessment
The exposure route for this risk assessment was defined as the risk of exposure via drinking well water. The model input parameters (Table1) were collected from literature and for those where there were no data available assumptions were used.
Figure 1. Dose response
The exposure pathway starts with contaminated drinking water. Virus in the drinking water infects a person (or more than one), who sheds the virus in their stool. The virus then gets into the septic tank, leaks out into the soil, into the groundwater, and then back into the well.
Virus concentration in wells is an assumption parameter extrapolated from studies looking at norovirus and overall enterovirus concentrations in drinking water wells (Hunt et al and Vaughn) The daily risk is a forecast of our model, and annual risk is calculated from the daily risk. The risk with chlorine intervention is a combination of the daily risk and the amount of reduction that is possible.
Most of the values associated with the school population were assumed. The shedding rate came from a paper on a German outbreak of norovirus by Hohne and Schrier. We found several values for the shedding rate, but chose this particular set of data because it included both genotypes of norovirus and is from an outbreak, rather than a feeding study or stool survey. We used the data set and created a distribution for shedding rate. We also set up a distribution for the number of diarrheal stools at school and the mass of individual diarrheal stools (for both students and teachers and staff). The mass of individual stools for students was an average of two values from two different sources (one was for healthy children ages 6mo to 5yrs, and the other was for healthy adults 23-28yrs of age) (Akinbami et al., Ogunbiyi). This average mass value was set as the minimum since it is for healthy individuals, and we assumed a maximum and likeliest value. The loading value is calculated based on the shedding rate and the mass of shedding stool. The effluent flow per day is calculated based on a value of 10gal per person per day, and adjusted for the number of individuals at the school and the length of the school day. The groundwater parameters were all extrapolated from the literature. The time to groundwater and time to well, as well as the viral concentration at the tap were forecast cells in our model. See Appendix 1 for model input parameters.
Dose Response
Due to the inability to culture the virus, dose-response data is not abundant for norovirus. However a single published human feeding study exists (Teunis). The published parameters contained a calculation error so it was necessary to adjust the dose parameter. The data was analyzed using the modified CAMRA dose-response R code (file provided to CAMRA). In order to examine the uncertainty in the dataset a bootstrap technique was used using 10,000 observations. The best fit model as the approximate beta-Poisson.
Where d is the dose administrated to the population, N50 is the median effective dose and α is the slope parameter. Models were built for both infection and illness. Figure 1 shows the dose-response curve.
Results for Dose-Response
| Alpha | N50 | |
|---|---|---|
| Infection | 0.110891 | 16963 |
| Illness | .071391 | 1465320 |
Risk Characterization
We ran a 2D simulation for each of 8 distinct scenarios. Scenario 1 and Scenario 2, with and without chlorine for infection and illness. We ran 59 outer by 400 inner loops to achieve tolerance bounds around our median risk curve. Parameters were attributed to uncertainty or variability based on the quality of the data.
| Scenario 1-III | No Chlorine | Chlorine | Scenario 2-III | No Chlorine | Chlorine |
|---|---|---|---|---|---|
| 5% | 3.21E-04 | 3.41E-10 | 5% | 7.06E-07 | 8.46E-13 |
| 50% | 5.62E-03 | 6.27E-09 | 50% | 1.66E-03 | 8.49E-10 |
| 95% | 3.94E-02 | 6.05E-08 | 95% | 2.33E-01 | 2.60E-06 |
| Scenario 1-Infection | No Chlorine | Chlorine | Scenario 2-Infection | No Chlorine | Chlorine |
| 5% | 4.25E-03 | 4.30E-09 | 5% | 7.97E-07 | 4.30E-12 |
| 50% | 2.80E-02 | 3.40E-08 | 50% | 5.27E-03 | 8.32E-09 |
| 95% | 9.05E-02 | 1.61E-07 | 95% | 4.29E-01 | 1.69E-05 |
Appendix 1. Model input parameters. Green cells are assumption cells, and blue cells are forecast cells. Distributions: 1= triangle, 2= logistic, 3= discrete uniform, 4= uniform
Risk Management & Communication
Groundwater is a major source of drinking water. Especially in rural areas and in metropolitan areas where there may not be sufficient amounts of surface water. (e.g. Houston, TX) Among other states that withdrawal groundwater are Florida, Arkansas, Idaho, Nebraska, California, etc. Groundwater is also used for irrigation in addition to public water supply. The use of groundwater for drinking purposes is highly favorable because of its generally good microbial quality. Most bacterium are filtered through soil and are not transported into aquifers. Due to viral microorganisms being so small, most disinfection methods fail to fully inactivate some viral types. In which viruses tend to vertically transport through soils. Septic tank leachate is the most frequently reported cause of groundwater contamination. (Yates) According to a study done by the U.S. Environmental Protection Agency (USEPA) in 1977, an estimated total volume of waste disposed via septic tanks is approximately 800 billion gallons per year. (U.S. EPA) Naturally this number has exponentially increased according to population growth some 34 years later. Since most septic tanks are placed underground, they are a major contributor to wastewater being discharged directly to groundwater. Several features of septic systems attribute to potential groundwater contamination and they include: improper construction, siting, installation, and maintenance of the septic tank; depth to water; climate; and geology of the site. (Yates) There are many types of disinfection methods utilized for water remediation. For the purpose of our case study, we as a group decided to use chlorination over the use of UV irradiation.
Chlorination is the least costly method of disinfection. The issue with using chlorine as a disinfection method is the fact that hazardous oxidation by-products also known as disinfection by-products (DBPs) are created, which can have a tremendous effect on the biological stability of water. Those by-products are Haloacetic acids (HAAs), Trihalonethanes (THMs), Dichloracetic acids (DCAA), Trichloracetic acids (TCAA), and chloroforms. THMs, DCAA, TCAA, chloroforms, HAAs, along with bromodichloromethane are all suspected to be carcinogens in humans, but occur in drinking water at very low dosages. (IARC; WHO; Nissinen et al.) Our system is able to achieve a 6 log reduction.