Inferences drawn from a risk assessment compared directly with a randomized trial of a home drinking water intervention.Risk assessments and intervention trials have been used by the U.S. Environmental Protection Agency Environmental Protection Agency (EPA), independent agency of the U.S. government, with headquarters in Washington, D.C. It was established in 1970 to reduce and control air and water pollution, noise pollution, and radiation and to ensure the safe handling and to estimate drinking water drinking water supply of water available to animals for drinking supplied via nipples, in troughs, dams, ponds and larger natural water sources; an insufficient supply leads to dehydration; it can be the source of infection, e.g. leptospirosis, salmonellosis, or of poisoning, e.g. health risks. Seldom are both methods used concurrently. Between 2001 and 2003, illness data from a trial were collected simultaneously with exposure data, providing a unique opportunity to compare direct risk estimates of waterborne disease from the intervention trial with indirect estimates from a risk assessment. Comparing the group with water treatment (active) with that without water treatment (sham False; without substance. A sham Pleading is one that is good in form but is so clearly false in fact that it does not raise any genuine issue. ), the estimated annual attributable disease rate (cases per 10,000 persons per year) from the trial provided no evidence of a significantly elevated drinking water risk [attributable risk attributable risk Epidemiology Any factor which ↑ the risk of suffering a particular condition. See Relative risk, Risk factor. Cf Nonattributable risk Statistics The rate of a disorder in exposed subjects that is attributable to the exposure derived from = -365 cases/year, sham minus active; 95% confidence interval confidence interval, n a statistical device used to determine the range within which an acceptable datum would fall. Confidence intervals are usually expressed in percentages, typically 95% or 99%. (CI), -2,555 to 1,825]. The predicted mean rate of disease per 10,000 persons per person-year from the risk assessment was 13.9 (2.5, 97.5 percentiles: 1.6, 37.7) assuming 4 log removal due to viral disinfection disinfection, n the process of destroying pathogenic organisms or rendering them inert. disinfection, full oral cavity, n a procedure used to reduce active periodontal disease, usually completed within a certain short time frame. and 5.5 (2.5, 97.5 percentiles: 1.4, 19.2) assuming 6 log removal. Risk assessments are important under conditions of low risk when estimates are difficult to attain from trials. In particular, this assessment pointed toward the importance of attaining site-specific treatment data and the clear need for a better understanding of viral removal by disinfection. Trials provide direct risk estimates, and the upper confidence limit estimates, even if not statistically significant, are informative about possible upper estimates of likely risk. These differences suggest that conclusions about waterborne disease risk may be strengthened by the joint use of these two approaches. Key words: drinking water, gastrointestinal, intervention trial, microbial microbial pertaining to or emanating from a microbe. microbial digestion the breakdown of organic material, especially feedstuffs, by microbial organisms. risk assessment, waterborne pathogens. Environ Health Perspect 114:1199-1204 (2006). doi:10.1289/ehp.8682 available via http://dx.doi.org/ [Online 4 April 2006] ********** Continued reporting of outbreaks of disease from consumption of drinking water (Barwick et al. 2000; Lee et al. 2002; Levy et al. 1998; Yoder et al. 2004) in the United States United States, officially United States of America, republic (2005 est. pop. 295,734,000), 3,539,227 sq mi (9,166,598 sq km), North America. The United States is the world's third largest country in population and the fourth largest country in area. has fueled the need for regulatory action through risk assessments as mandated by the Safe Drinking Water Act The Safe Drinking Water Act (SDWA) is a United States federal law passed by the U.S. Congress on December 16, 1974. It is the main federal law that ensures safe drinking water for Americans. (SDWA SDWA Safe Drinking Water Act of 1974 SDWA System Diagnostic Work Area (IBM) SDWA Sun Data Warehouse Appliance 1996). Risk assessments historically have been used to evaluate the health risks of properly treated drinking water because of the general belief that drinking water risks were too low to be detected through epidemiology studies. Recent drinking water intervention trials, however, have begun to question the assumption that there is little or no risk of infectious gastrointestinal (GI) illness attributable to the consumption of drinking water when water treatment systems are functioning properly (Payment et al. 1991, 1997). In contrast, other trials have suggested that there is little or no risk (Colford et al. 2005; Hellard et al. 2001). Based on these findings and in response to the 1996 Congressional amendment to the SDWA that emphasizes the need for sound science and risk-based standard settings [U.S. Environmental Protection Agency (EPA EPA eicosapentaenoic acid. EPA abbr. eicosapentaenoic acid EPA, n.pr See acid, eicosapentaenoic. EPA, n. ) 1989], there has been increased interest in evaluating methodologies to help estimate the risk of GI illness attributable to drinking water in communities. In the present study we compare and contrast two approaches for the assessment of risk of diarrhea caused by drinking water--a microbial risk assessment and a randomized ran·dom·ize tr.v. ran·dom·ized, ran·dom·iz·ing, ran·dom·iz·es To make random in arrangement, especially in order to control the variables in an experiment. intervention trial design. Using data collected in Davenport, Iowa Davenport is a city in the American state of Iowa that borders the Mississippi River. As of the 2000 census, the city had a total population of 98,359. A 2006 estimate tells that the city had grown slightly to 99,514. (Colford et al. 2005), we compared the two techniques to estimate the risk from waterborne pathogens due to exposure to drinking water. For this study, risk assessment is based on the integration of several independent sources of exposure information to estimate dose (i.e., water quality, drinking water treatment plant efficiency, and tap water consumption patterns). We then used the dose information in a health effects model to predict the risk of illness due to drinking tap water. The randomized intervention trial directly measures the impact of drinking water on diarrhea and compares the incidence of GI illness between intervention and control subjects. Both approaches have wide appeal. The randomized trial is considered the "gold standard" for providing unconfounded causal risk estimates associated with a particular exposure. When lacking these direct estimates of risk, quantitative risk assessment is the preferred method for attaining risk estimates and is used by the U.S. EPA, U.S. Food and Drug Administration, World Health Organization, and other stakeholders Stakeholders All parties that have an interest, financial or otherwise, in a firm-stockholders, creditors, bondholders, employees, customers, management, the community, and the government. for regulatory and operational purposes. Although these approaches are widely accepted, they also have many limitations. Low sensitivity because of sample size constraints, and biases due to both exposure and outcome misclassification must be acknowledged when interpreting randomized trial results. Similarly, risk assessments are model-based estimates and rely on water quality data as input, and so must be interpreted in this context. Both approaches have their strengths and weaknesses. In the present study our goal was to compare and contrast the two approaches for obtaining estimates of drinking water risk when coincident co·in·ci·dent adj. 1. Occupying the same area in space or happening at the same time: a series of coincident events. See Synonyms at contemporary. 2. data are available. Several authors have proposed methods for estimating the risk of drinking water (Haas et al. 1993; Messner et al. 2001; Regli et al. 1999). Our study differs from these previous studies in that we incorporated additional detailed local information relevant to risk assessment, including measurements of pathogen Pathogen Any agent capable of causing disease. The term pathogen is usually restricted to living agents, which include viruses, rickettsia, bacteria, fungi, yeasts, protozoa, helminths, and certain insect larval stages. levels in the source water over a 1-year period, pathogen removal efficiency of the Davenport Davenport, city (1990 pop. 95,333), seat of Scott co., E central Iowa, on the Mississippi River; inc. 1836. Bridges connect it with the Illinois cities of Rock Island and Moline; the three communities and neighboring Bettendorf, Iowa, are known as the Quad Cities. drinking water treatment plant (which uses sedimentation sedimentation In geology, the process of deposition of a solid material from a state of suspension or solution in a fluid (usually air or water). Broadly defined it also includes deposits from glacial ice and materials collected under the effect of gravity alone, as in talus , filtration, and chlorine disinfection), and data on local tap water consumption. Materials and Methods Attributable risk from intervention trial (Davenport, Iowa). The study design of the intervention trial in Davenport is similar to those of previously published drinking water intervention trials (Colford et al. 2002; Hellard et al. 2001; Payment et al. 1991, 1997). Unlike prior randomized trials, however, a crossover Crossover The point on a stock chart when a security and an indicator intersect. Crossovers are used by technical analysts to aid in forecasting the future movements in the price of a stock. In most technical analysis models, a crossover is a signal to either buy or sell. design was used where, for each intervention period (~6 months), half the enrolled cohort had a water treatment device installed at their kitchen faucet and half had a sham device installed that resembled the real device but provided no water treatment. At the end of the first treatment period, the device in each subject's household was switched to the opposite type, and illness was monitored for another 6 months. Participants were blinded throughout the study to their specific device type, and they recorded their daily occurrence of GI symptoms (e.g., diarrhea, nausea, vomiting vomiting, ejection of food and other matter from the stomach through the mouth, often preceded by nausea. The process is initiated by stimulation of the vomiting center of the brain by nerve impulses from the gastrointestinal tract or other part of the body. , cramps) in a personal health diary. The study resulted in treatment assignment and illness data for 1,296 subjects in 456 households. For further details of the Davenport intervention trial, see Colford et al. (2005). As part of the Davenport intervention study, a separate random digit dial (RDD RDD Random Digit Dialing RDD RDF (Resource Description Framework) Declarative Description RDD Radiological Dispersal Device RDD Rights Data Dictionary RDD Radiological Dispersion Device RDD Respiratory Drug Delivery ) telephone survey was conducted in the Davenport area. The goal of the survey was to obtain population-based estimates of the use of various home water treatments, water consumption, and the monthly occurrence of GI illnesses (Wade et al. 2004). We define attributable risk (AR) for the trial subjects as the estimated risk difference in daily rates of highly credible GI illness (HCGI) (Colford et al. 2005) among the subjects with the treatment device versus those with the sham device. HCGI is defined as the presence of any one of the following syndromic manifestations of GI illness: vomiting, watery wa·ter·y adj. 1. Filled with, consisting of, or soaked with water; wet or soggy. 2. Secreting or discharging water or watery fluid, especially as a symptom of disease. diarrhea, soft diarrhea with abdominal cramps, and nausea with abdominal cramps. The AR was estimated using a linear model with binomial binomial (bī'nō`mēəl), polynomial expression (see polynomial) containing two terms, for example, x+y. The binomial theorem, or binomial formula, gives the expansion of the nth power of a binomial (x+ errors and accounting for correlation using a generalized estimating equation (Zeger et al. 1988). Risk assessment model. The risk assessment was conducted without knowledge of the results of the Davenport trial. Figure 1 is a schematic A graphical representation of a system. It often refers to electronic circuits on a printed circuit board or in an integrated circuit (chip). See logic gate and HDL. of the general model for generating GI illness cases due to drinking water. Methods used to derive the model parameters are discussed later in this article. The model uses a population of 10,000 and a risk period of 1 year (365 days). The model is a simple linear process and works as described below. A concentration of the specific source water distribution of pathogens (e.g., Giardia Giardia /Gi·ar·dia/ (je-ahr´de-ah) a genus of flagellate protozoa parasitic in the intestinal tract of humans and other animals, which may cause giardiasis; G. lam´blia (G. intestina´lis) is the species found in humans. , Cryptosporidium cryptosporidium (krĭp'tōspərĭd`ēəm), genus of protozoans having at least four species; they are waterborne parasites that cause the disease cryptosporidiosis. , and culturable viruses) is randomly sampled for the day. On the basis of previous studies and goodness-of-fit tests of the source water data collected in Davenport, we assumed that the average concentrations of source water for a day followed a lognormal distribution Lognormal distribution Pattern of frequency of occurrence in which the logarithm of the variable follows a normal distribution. Lognormal distributions are used to describe returns calculated over periods of a year or more. (LeChevallier et al. 2003b). This distribution was estimated using the constant recovery rates shown in Table 1. We assumed that treatment efficiency due to sedimentation and filtration remained constant during the day but itself was a random draw from a Weibull distribution In probability theory and statistics, the Weibull distribution[1] (named after Waloddi Weibull) is a continuous probability distribution with the probability density function The concentration of pathogens in the resulting drinking water, [D.sub.i], for day i, was [D.sub.i] = [S.sub.i][T.sub.i][C.sub.i], [1] where [S.sub.i], [T.sub.i], and [C.sub.i] are the (daily) randomly drawn source water concentration, treatment efficiency, and disinfection, respectively. For each day i, for each of 10,000 individuals j, we randomly drew a volume of water consumption, [V.sub.ij], from a lognormal distribution (Rosebury and Burmaster 1992) based on data from the RDD telephone survey in Davenport (Wade et al. 2004). A random number of pathogens, [P.sub.ij], ingested in·gest tr.v. in·gest·ed, in·gest·ing, in·gests 1. To take into the body by the mouth for digestion or absorption. See Synonyms at eat. 2. for each subject i, on each day j, was generated from a Poisson distribution A statistical method developed by the 18th century French mathematician S. D. Poisson, which is used for predicting the probable distribution of a series of events. For example, when the average transaction volume in a communications system can be estimated, Poisson distribution is used with mean, [V.sub.ij] x [D.sub.i]. We generated a random (yes/no) indicator of illness, [I.sub.ij], based on the number of pathogens and the probability of illness given [P.sub.ij]. This probability was derived from separate dose-response curves dose-response curve A graphic representation of the effects that varous doses of an agent–eg, ionizing radiation or a chemotherapeutic agent, have on a given parameter–eg, cell viability, mutation frequency, DNA damage, tumor growth or metastasis or (probability of infection for a given ingested pathogen dose) and morbidity ratios (the ratio of those who become ill to those who are infected in·fect tr.v. in·fect·ed, in·fect·ing, in·fects 1. To contaminate with a pathogenic microorganism or agent. 2. To communicate a pathogen or disease to. 3. To invade and produce infection in. ) for each pathogen, which were based on published dose-response data (DuPont et al. 1995; Rendtroff 1954; Teunis et al. 1986; Ward et al. 1986). The final step, after generating data for 10,000 subjects and 365 days, is to count the number of events and divide by the time at risk to derive an estimate of disease incidence due to exposure to the specific pathogen in drinking water for the year: [1/[10,000 x 365]] [365.summation summation n. the final argument of an attorney at the close of a trial in which he/she attempts to convince the judge and/or jury of the virtues of the client's case. (See: closing argument) over (i=1)] [10,000.summation over (j=1)] [I.sub.ij]. [2] Parameter estimates in risk assessment model. Each step of the above model relies on parameter estimates. We derived almost all of these estimates from site-specific (Davenport) data. When site-specific data were not available, we used data from the literature. Source water concentration. Water quality data from the source water serving the study area came from the Davenport intervention study. These included approximately weekly measurements of Cryptosporidium and Giardia concentrations, as well as monthly measurements of culturable viruses (LeChevallier et al. 2003b). Figure 2 shows the raw data for both Giardia and Cryptosporidium, [X.sub.k], collected at different days, k, and Figure 3 shows similar data for the culturable viruses. These represent counts of pathogens in a fixed volume, Q, of sampled source water with assumed recovery rate, R. We assume that the counts of pathogens, [X.sub.k], are derived from an underlying Poisson distribution with mean [S.sub.k] x Q x R, where [S.sub.k] is the average source water concentration for day k. We assumed that [S.sub.k] follows a lognormal distribution, suggesting that a marginal likelihood In Bayesian probability theory, a marginal likelihood function is a likelihood function integrated over some variables, typically model parameters. Integrated likelihood is a synonym for marginal likelihood. of X is [L.sub.[[mu].sub.s][[sigma].sub.s]] (x) = [[infinity].[integral].0][[[e.sup.-sQR](sQR)[.sup.x]]/[x!]] [phi] ([log(s) - [[mu].sub.s]]/[[sigma].sub.s])ds, [3] where [phi] represents the standard normal density, and [[mu].sub.s] and [[sigma].sub.s] are the underlying mean and standard deviation In statistics, the average amount a number varies from the average number in a series of numbers. (statistics) standard deviation - (SD) A measure of the range of values in a set of numbers. , respectively, of the log(S) distribution. Equation 3 represents the likelihood contribution of one observation of the raw counts, [X.sub.k]; the parameters of interest, [[mu].sub.s] and [[sigma].sub.s], are estimated by maximizing the likelihood. The estimates of all parameters in the model are presented in Table 1. Treatment efficiency. Direct estimates of treatment efficiency with respect to Cryptosporidium species, Giardia species, and culturable viruses were not possible from the Davenport treatment facility because levels in effluent effluent waste from an abattoir carried away in liquid form. Disposal is a major problem because of the need to avoid pollution of waterways. See aerobic effluent treatment, anaerobic effluent treatment. water samples were uniformly below detection across the study period. Estimates of the efficiency of Bacillus subtilis Noun 1. Bacillus subtilis - a species of bacillus found in soil and decomposing organic matter; some strains produce antibiotics Bacillus globigii, grass bacillus, hay bacillus treatment, obtained from weekly measurements of source water and plant effluent data (LeChevallier et al. 2003b), were used as a surrogate surrogate n. 1) a person acting on behalf of another or a substitute, including a woman who gives birth to a baby of a mother who is unable to carry the child. 2) a judge in some states (notably New York) responsible only for probates, estates, and adoptions. for Cryptosporidium and Giardia treatment efficiency. Similarly, removal of somatic somatic /so·mat·ic/ (so-mat´ik) 1. pertaining to or characteristic of the soma or body. 2. pertaining to the body wall in contrast to the viscera. so·mat·ic adj. coliphage coliphage /col·i·phage/ (kol´i-faj) any bacteriophage that infects Escherichia coli. co·li·phage n. A bacteriophage with an affinity for a strain of Escherichia coli. from waters passing through the Davenport facility was used to approximate a distribution of treatment efficiency with respect to culturable viruses. The log removal for Cryptosporidium from chlorine disinfection was assumed to be zero, whereas the log removal for Giardia and enteric viruses enteric virus n. See enterovirus. was estimated from chlorine concentration time (CT) values collected in Davenport. The CT values were estimated for Giardia and were therefore directly applied to estimate disinfection efficacy for Giardia. Because there were no equivalent data for viruses, virus disinfection was assumed to have the same distribution as Giardia but a modulated mod·u·late v. mod·u·lat·ed, mod·u·lat·ing, mod·u·lates v.tr. 1. To adjust or adapt to a certain proportion; regulate or temper. 2. mean value. One way to establish specific viral log removal values for the model is to rely on data presented in a U.S. EPA guidance document (U.S. EPA 1991). Table E-7 in this guidance document suggested that 4 log removal of viruses would be achieved at 20[degrees]C, a pH of 6-9, and a CT value of 3. This table was based on hepatitis A Hepatitis A Definition Hepatitis A is an inflammation of the liver caused by a virus, the hepatitis A virus (HAV). It varies in severity, running an acute course, generally starting within two to six weeks after contact with the virus, and lasting no data (Sobsey 1988) assuming both a 3-fold safety factor and a 2-fold decrease in CT for every 10[degrees]C increase in temperature. Using Table E-4 from this same guidance document, we can estimate that the CT value in Davenport was approximately 13. Assuming a linear relationship between viral log removal and CT would suggest that the log removal of viruses by disinfection was > 12. This result assumes that viruses are dispersed dis·perse v. dis·persed, dis·pers·ing, dis·pers·es v.tr. 1. a. To drive off or scatter in different directions: The police dispersed the crowd. b. in chlorine-demand-free water and is not valid for viruses that occur in nature aggregated and associated with organic particles. Given the uncertainties associated with all of these assumptions, we chose to examine a variety of viral log removal values ranging from 4, the minimum required by the U.S. EPA, to the 13 log removal treatment level estimated above. Water consumption. During the period of the intervention trial, home tap water consumption data were collected from the RDD telephone survey. The estimate of the distribution of regular tap water consumption was obtained from 4,756 interviews. The water consumption distribution was assumed to be lognormal log·nor·mal adj. Mathematics Of, relating to, or being a logarithmic function with a normal distribution. log (Rosebury and Burmaster 1992), and we estimated the mean and standard deviation of this distribution. The RDD survey respondents were asked how much water was consumed in discrete glasses: < 1, 1-2, 3-5, and > 5. We took the number of respondents in each of these categories, made the categories contiguous (i.e., < 1, 1-2.5, 2.5-5, > 5), and estimated the mean and standard deviation of log consumption using maximum likelihood. Probability of disease. The functions used to generate a probability of disease given a quantity of pathogen ingested (dose response) were derived from dosing trials where healthy volunteers were given known quantities of pathogens. Specifically, DuPont et al. (1995) published data for a sample of healthy volunteers infected by known numbers of Cryptosporidium oocysts, Rendtroff (1954) reported similar data for Giardia, and Ward et al. (1986) for rotavirus rotavirus /ro·ta·vi·rus/ (ro´tah-vi?rus) any member of the genus Rotavirus. ro´taviral Rotavirus /Ro·ta·vi·rus/ (ro´tah-vi?rus , which we used as a surrogate for culturable viruses (Regli et al. 1991). Exponential functions exponential function In mathematics, a function in which a constant base is raised to a variable power. Exponential functions are used to model changes in population size, in the spread of diseases, and in the growth of investments. were used for Giardia and Cryptosporidium, and a beta-Poisson function for culturable enteric viruses (Teunis et al. 1996). Results The estimated attributable (annual) rate of disease per 10,000 people from the Davenport trial (expressed as the rate in the sham group minus the rate in the active group) was -365 [95% confidence interval (CI), -2,555 to 1,825], which provided no evidence of a significant association of the use of drinking water with disease. The result was negative because there were more cases reported from the active than from the sham group. Based on the upper value of the 95% CI, the trial was statistically consistent with as many as 1,825 cases per 10,000 people per year attributable to drinking water. These estimates were calculated from a cohort of 1,296 persons that reported 394 episodes of HCGI while in the active group and 350 while in the sham group (Colford et al. 2005). Table 2 is a summary of the estimated cases of illness from our risk assessment models based on pathogens. Assuming a 4 log removal of viruses from disinfection (U.S. EPA regulatory limit), the predicted risk was 13.9 (2.5, 97.5 percentiles: 1.6, 37.7) cases per 10,000 persons per year (due to Cryptosporidium, Giardia, and culturable enteric viruses), whereas assuming an 6 log removal, the predicted risk dropped to 5.5 (2.5, 97.5 percentiles: 1.4, 19.2) cases. At 6 log removal there was less than 1 case associated with viral exposure. Results from higher viral log removal did not vary from the results using 6 log removal. The width of the CI values from the Davenport trial and risk assessment should not be compared, as the former incorporates sources of variation and uncertainty that are not relevant in the latter. We also examined the sensitivity of our risk assessment results to alternative parameterizations of the model by conducting the following sensitivity analyses: a) instead of assuming a Poisson distribution, we modeled pathogen density using a negative binomial distribution In probability and statistics the negative binomial distribution is a discrete probability distribution. The Pascal distribution and the Polya distribution are special cases of the negative binomial. with different levels of aggregation; b) rather than using Davenport-specific treatment efficiency values, we used published values (Rachwal et al. 1996); c) rather than using site-specific data from the RDD telephone survey, we based the mean and standard deviation of the estimated average daily water consumption on reported U.S. EPA values (U.S. EPA 2000); and d) we varied the two dose-response parameters by a factor of 10. Results based on the above variations increased predicted Cryptosporidium cases to be as high as 25 cases per 10,000 persons per year, Giardia cases to as high as 100, and culturable enteric viruses to as high as 15. This brings the predicted risk to as high as 140 cases per 10,000 persons per year. There was little effect from adding overdispersion to the pathogen distribution using the negative binomial model. The higher estimates were primarily because of the use of non-Davenport-specific treatment values. Discussion In this study, both risk assessments and intervention trials are used to obtain health risk estimates. The interpretation of the results obtained from these two approaches, however, can often differ. The data collected in Davenport provided a unique opportunity to compare and contrast these two approaches. Even though there was no evidence of a significant association in the Davenport analysis, the upper bound risk estimate from the intervention trial (based on the 95% CI) was higher than the drinking water standards provided by the U.S. EPA. Under these rigorous standards, the Davenport analysis provides a useful upper bound on the risk; however, a risk assessment is needed to estimate the risk within the limits set by regulatory agencies regulatory agency Independent government commission charged by the legislature with setting and enforcing standards for specific industries in the private sector. The concept was invented by the U.S. . Specifically for Davenport, the trial estimated an upper-end risk of 1,825 cases per 10,000 persons per year, whereas the risk assessment predicted 5-14 cases per 10,000 persons per year attributable to drinking water from Giardia, Cryptosporidium, and culturable viruses. An additional finding in our work was a difference in the estimation of illnesses provided by risk assessment when using site-specific water quality data rather than generally available estimates of treatment efficacies. Because of the different approaches used by risk assessments and intervention trials, the analytic results from each approach often have different interpretations. These differences in the two approaches are summarized in Table 3 and discussed below. Sensitivity. Historically, drinking water regulations have been based on a tolerable tol·er·a·ble adj. 1. Capable of being tolerated; endurable. 2. Fairly good; passable. See Synonyms at average. tol annual risk of 1 case per 10,000 persons, that is, a goal of fewer than one case of infection with a particular pathogen per 10,000 persons attributable to drinking water (Regli et al. 1999). Although this value is not explicitly used by the U.S. EPA, it is consistent with their regulatory guidelines guidelines, n.pl a set of standards, criteria, or specifications to be used or followed in the performance of certain tasks. (Regli et al. 1999). Epidemiologic studies epidemiologic study A study that compares 2 groups of people who are alike except for one factor, such as exposure to a chemical or the presence of a health effect; the investigators try to determine if any factor is associated with the health effect , such as the intervention trial conducted in Davenport, generally cannot measure such low-magnitude risks. The Davenport trial was powered to detect approximately 1,100 illnesses per 10,000 persons--a smaller risk difference than that observed in previous studies (Payment et al. 1991). To illustrate this lack of sensitivity, we estimated that to power the Davenport trial to detect an AR of 20 cases per 10,000 persons per year, a risk similar to that estimated by the risk assessment, would require a sample size of 8 million individuals; to detect an AR of 100 cases per 10,000 persons would require 416,000 individuals. The intervention trial, using traditional levels of statistical significance, lacks the sensitivity to detect the low number of cases predicted from the risk assessment. In addition to a limited sample size, the sensitivity of a trial may be decreased because of biases caused by, for example, misclassification of disease outcomes and exposure (i.e., people with disease are more or less likely to change their drinking water patterns). Because of randomization randomization (ranˈ·d  ·m , most misclassification in the trial
was likely to be nondifferential; that is, if subjects are
underreporting disease, they are likely to be doing so equally while in
the both the active treatment and the sham group. Because
nondifferential misclassification biases the estimate of AR toward the
null A character that is all 0 bits. Also written as "NUL," it is the first character in the ASCII and EBCDIC data codes. In hex, it displays and prints as 00; in decimal, it may appear as a single zero in a chart of codes, but displays and prints as a blank space. (Rothman and Greenland 1998), estimates that do not account for
this misclassification (e.g., the estimate in this study) would likely
underestimate the magnitude of these estimates. Adjusted risk estimates
would increase both point estimates and the upper end of the 95% CI.
Even if the sensitivity of the intervention trial precludes us from making any inference within the narrow range of regulatory limits, the trial data do provide a rigorous direct estimate of the upper bound to the risk from drinking water. For example, previous drinking water intervention trials in Canada showed that up to 35% of GI illnesses were transmitted through a public drinking water system (Payment et al. 1991). The Davenport intervention study was adequately powered to detect differences considerably smaller than this level and clearly demonstrates that the drinking water risks estimated in Davenport were below those observed by Payment (1991); that is, the upper end of the 95% CI for the percent AR reported by Colford et al. (2005) was 10%, whereas the point estimate reported by Payment et al. (1991) was 35%. For regulatory purposes the upper bound estimate from Davenport can be interpreted as the largest risk estimate that is still consistent with the intervention trial results. As mentioned above, this upper bound estimate is not only based on random error, due to sample size, but also on systematic error in a trial, such as the biases due to nondifferential misclassification. Causal evidence and pathogen-specific versus symptomatic outcomes. Risk assessment methods also have limitations. For example, risk estimates are model based and provide indirect evidence of risk based on water quality data, whereas the trials focus on direct estimates of illness. Additionally, the risk estimates include only a subset of the potential pathogens, compared with the intervention trial, which provides a risk estimate for diarrheal disease integrated over all pathogens as well as nonpathogenic causes of diarrhea Diarrhea (in American English) or diarrhoea (in British English) is a condition in which the sufferer has frequent watery, loose bowel movements. Many things can cause diarrhea, which can make diagnosis complex. . In this risk assessment we were able to provide estimates for two protozoan protozoan (prō'təzō`ən), informal term for the unicellular heterotrophs of the kingdom Protista. Protozoans comprise a large, diverse assortment of microscopic or near-microscopic organisms that live as single cells or in simple pathogens, Cryptosporidium and Giardia, as well as for culturable viruses. The microbiologic methods, however, identify all species of Cryptosporidium and Giardia, not all of which cause illness. Likewise, viruses that are culturable are only a subset of known viral pathogens; for example, noroviruses are a major cause of waterborne viral infections viral infection, n an infection by a pathogenic virus. A virus acts on the cell nucleus, taking over the genetic material within the nucleus and replicating itself. but are not culturable. These limitations can therefore lead to both under- and overestimates of risk. Model specification and the inclusion of other sources of risk. The specific model structure used in any risk assessment carries with it many assumptions. For example, the model in this study was based on risks associated with source water contamination and did not consider the potential contamination within the distribution system (LeChevallier et al. 2003a). Another assumption implicit in Adj. 1. implicit in - in the nature of something though not readily apparent; "shortcomings inherent in our approach"; "an underlying meaning" underlying, inherent the model structure is that secondary transmission is negligible (Eisenberg et al. 2002, 2003). Both of these assumptions can lead to biased results. By incorporating such processes, risk assessment models can be made more complicated and perhaps more accurate. Some environmental processes, such as risk incurred by the distribution system, can be also captured by the intervention trial and in fact was accounted for in the Davenport trial. Other processes, such as transmission processes, can be addressed only in a limited way by observing within-household transmission. Specifically, the standard intervention trial design is focused on individual-level risk, assuming that disease outcomes of different individuals are independent and therefore cannot capture population-level processes such as secondary transmission. Risk assessment models that incorporate disease transmission processes are the only models that account for these population-level conditions. Increasing the complexity of the model, however, leads to additional uncertainty with respect to model specification and potentially parameter specification. In general, it is impossible to account for all sources of variation. Models often must rely on estimates from small studies (e.g., dosing trials conducted on healthy individuals) and on very strong modeling assumptions. Sparse sparse - A sparse matrix (or vector, or array) is one in which most of the elements are zero. If storage space is more important than access speed, it may be preferable to store a sparse matrix as a list of (index, value) pairs or use some kind of hash scheme or associative memory. site-specific data, such as with the source water measurements of enteric viruses, increase the uncertainty of model-based estimates. Most of these limitations underestimate risk. There are additional uncertainties that result in overestimates of risk. For example, uncertainties in treatment--Bacillus spores are considered conservative indicators of Cryptosporidium removal, and disinfection CT values are based on half-lives, not the full integration of disinfection contact times. Site-specific versus general estimates in risk assessment. Our finding of a difference in estimates of illness from using site-specific rather than general U.S. EPA estimates for treatment efficiencies highlights the importance of a clear definition for a risk assessment. If the goal of a risk assessment is focused only on risk within a specified community (or similar communities), then site-specific data may be most appropriate and worth the additional effort to obtain. If, however, the goal is to generalize generalize /gen·er·al·ize/ (-iz) 1. to spread throughout the body, as when local disease becomes systemic. 2. to form a general principle; to reason inductively. about risk across multiple communities or large areas, then the general parameter estimates provided by the U.S. EPA are likely to be more appropriate. Examining alternative control strategies. One advantage of a model-based risk assessment is that alternative control strategies can be examined. For example, the pathogen-specific risk estimates from the risk assessment provided additional information for focused waterborne disease control strategies. Given that the viral log removal by disinfection is on the order of 6 or more, the predicted risk above 1 in 10,000 persons comes from exposure to protozoan species. Thus, if the risk levels presented in Table 2 are of concern, control efforts should be focused on treatment technologies that address protozoa rather than virus removal. If, on the other hand, viral log removal by disinfection is [less than or equal to] 4, control efforts should be focused on treatment technologies that address viral removal. Given the limited data to inform this assumption, resources could be focused on collecting more viral disinfection data. Cost and time. An additional limitation of intervention trials is that they are costly and time-consuming to conduct. In contrast, risk assessments are relatively inexpensive and quick to conduct. Conclusions Risk assessment and intervention trials provide complementary approaches to the estimation of a community's burden of disease attributable to drinking water. Risk assessments can provide estimates of low-risk situations; require data that are neither difficult nor expensive to collect; permit the evaluation of scenarios outside the conditions under which the data were collected and are therefore an attractive method for characterizing both existing and potential risk from contamination of drinking water; and can capture population-level processes such as secondary transmission. Intervention trials provide direct measures of AR within communities and provide risk estimates based on all causes of illness attributable to the drinking water. Even when point estimates of risk are not significant, these direct measures of risk can provide valuable upper bound estimates. Given their expense, intervention trials must be judiciously ju·di·cious adj. Having or exhibiting sound judgment; prudent. [From French judicieux, from Latin i applied. Risk assessments can be used to specify the conditions in which future trials are justified; that is, they can be used to identify high-risk conditions based on demographics The attributes of people in a particular geographic area. Used for marketing purposes, population, ethnic origins, religion, spoken language, income and age range are examples of demographic data. , magnitude and sources of environmental contamination, and types of treatment processes. Risk assessment can also provide information on where are the important data gaps. In particular, this assessment pointed toward the importance of attaining site-specific treatment data and the clear need for a better understanding of viral removal by disinfection. Ultimately, the choice of risk assessment, intervention trials, or both methods used jointly to evaluate waterborne disease risks depends upon specific research needs and available funding. REFERENCES Barwick RS, Levy DA, Craun GF, Beach MJ, Calderon RL. 2000. Surveillance for waterborne-disease outbreaks--United States, 1997-1998. MMWR MMWR Morbidity & Mortality Weekly Report Epidemiology A news bulletin published by the CDC, which provides epidemiologic data–eg, statistics on the incidence of AIDS, rabies, rubella, STDs and other communicable diseases, causes of mortality–eg, CDC See Control Data, century date change and Back Orifice. CDC - Control Data Corporation Surveill Summ 49:1-21. Colford JM Jr, Rees JR, Wade TJ, Khalakdina A, Hilton JF, Ergas IJ, et al. 2002. Participant blinding and gastrointestinal illness in a randomized, controlled trial controlled trial Clinical research A clinical study in which one group of participants receives an experimental drug while the other receives either a placebo or an approved–'gold standard' therapy. See Blinding, Double-blinded. of an in-home drinking water intervention. Emerg Infect infect /in·fect/ (in-fekt´) 1. to invade and produce infection in. 2. to transmit a pathogen or disease to. in·fect v. 1. Dis 8:29-36. Colford JM Jr, Wade TJ, Sandhu SK, Wright CC, Lee S, Shaw S, et al. 2005. A randomized controlled trial A randomized controlled trial (RCT) is a scientific procedure most commonly used in testing medicines or medical procedures. RCTs are considered the most reliable form of scientific evidence because it eliminates all forms of spurious causality. of in-home drinking water intervention to reduce gastrointestinal illness. Am J Epidemiol 161:472-482. DuPont HL, Chappell CL, Sterling CR, Okhuysen PC, Rose JB, Jakubowski W. 1995. The infectivity infectivity ability of an agent to infect. of Cryptosporidium parvum Cryptosporidium parvum is one of several species that cause cryptosporidiosis. Cryptosporidium parvum is a protozoal infection which causes an acute, watery, and non-bloody diarrhoea in immunocompromised patients. in healthy volunteers. N Engl J Med 332:855-859. Eisenberg JN, Brookhart MA, Rice G, Brown M, Colford JM Jr. 2002. Disease transmission models for public health decision making: analysis of epidemic and endemic endemic /en·dem·ic/ (en-dem´ik) present or usually prevalent in a population at all times. en·dem·ic adj. 1. conditions caused by waterborne pathogens. Environ Health Perspect 110:783-790. Eisenberg JN, Lewis BL, Porco TC, Hubbard AH, Colford JM Jr. 2003. Bias due to secondary transmission in estimation of attributable risk from intervention trials. Epidemiology 14:442-450. Haas CN, Rose JB, Gerba C, Regli S. 1993. Risk assessment of virus in drinking water. Risk Anal 13:545-552. Hellard ME, Sinclair MI, Forbes AB, Fairley CK. 2001. A randomized, blinded, controlled trial investigating the gastrointestinal health effects of drinking water quality. Environ Health Perspect 109:773-778. LeChevallier MW, Gullick RW, Karim MR, Friedman M, Junk JE. 2003a. The potential for health risks from intrusion of contaminants into the distribution system from pressure transients. J Water Health 1:3-14. LeChevallier MW, Karim M, Aboytes R, Gullick R, Weihe J, Earnhardt B, et al. 2003b. Profiling Water Quality Parameters: From Source Water to the Household Tap. Denver, CO:Research Foundation and American Water Works Association American Water Works Association (AWWA) is an international nonprofit professional organization dedicated to the improvement of drinking water quality and supply. It was founded in 1881 and, as of 2007, there are approximately 60,000 AWWA members world-wide. . Lee SH, Levy DA, Craun GF, Beach MJ, Calderon RL. 2002. Surveillance for waterborne-disease outbreaks--United States, 1999-2000. MMWR Surveill Summ 51:1-47. Levy DA, Bens MS, Craun GF, Calderon RL, Herwaldt BL. 1998. Surveillance for waterborne-disease outbreaks--United States, 1995-1996. MMWR CDC Surveill Summ 47:1-34. Messner MJ, Chappell CL, Okhuysen PC. 2001. Risk assessment for Cryptosporidium: a hierarchical Bayesian analysis Bayesian analysis A decision-making analysis that '…permits the calculation of the probability that one treatment is superior based on the observed data and prior beliefs…subjectivity of beliefs is not a liability, but rather explicitly allows of human dose response data. Water Res 35:3934-3940. Nash TE, Herrington DA, Losonsky GA, Levine MM. 1987. Experimental human infections with Giardia lamblia Giardia lamblia or G. intestinalis Single-celled protozoan parasite. Pear- or beet-shaped, the cells have two nuclei and eight flagella and attach with a sucking organ to human intestinal mucous membranes. They cause the disease giardiasis. . J Infect Dis 156:974-984. Payment P, Richardson L, Siemiatycki J, Dewar R, Edwardes M, Franco E. 1991. A randomized trial to evaluate the risk of gastrointestinal disease gastrointestinal disease, n an abnormal state or function of the GI system. due to consumption of drinking water meeting current microbiological standards. Am J Public Health 81:703-708. Payment P, Siemiatycki J, Richardson L, Renaud G, Franco E, Prevost M. 1997. A prospective epidemiological study An Epidemiological study is a statistical study on human populations, which attempts to link human health effects to a specified cause. of gastrointestinal health effects due to the consumption of drinking water. Int J Environ Health Res 7:5-31. Rachwal AJ, Bauer MJ, Chipps MJ, Colbourne JS, Foster DM. 1996. Comparisons between slow sand and high rate biofiltration. In: Advances in Slow Sand and Alternative Biological Filtration (Graham N, Collins R, eds). New York New York, state, United States New York, Middle Atlantic state of the United States. It is bordered by Vermont, Massachusetts, Connecticut, and the Atlantic Ocean (E), New Jersey and Pennsylvania (S), Lakes Erie and Ontario and the Canadian province of :John Wiley John Wiley may refer to:
Regli S, Odem R, Cromwell J, Lustic M, Blank V. 1999. Benefits and costs of the IESWTR IESWTR Interim Enhanced Surface Water Treatment Rule (US EPA) . J Am Water Works Assoc 91:148-158. Regli S, Rose JB, Haas CN, Gerba CP. 1991. Modeling the risk from Giardia and viruses in drinking water. J Am Water Works Assoc 83:76-84. Rendtroff RC. 1954. The experimental transmission of human intestinal protozoan parasites. II. Giardia lamblia cysts given in capsules. Am J Hyg 59:209-220. Rosebury A, Burmaster D. 1992. Log-normal distributions In probability and statistics, the log-normal distribution is the single-tailed probability distribution of any random variable whose logarithm is normally distributed. If Y is a random variable with a normal distribution, then X = exp(Y for water intake by children and adults. Risk Anal 12:99-104. Ross SM. 1985. Introduction to Probability Models. Orlando, FL:Academic Press. Rothman KJ, Greenland S. 1998. Modern Epidemiology. Philadelphia:Lippincott-Raven. SDWA. 1996. Safe Drinking Water Act. Public Law 104-182. Available: http://www.epa.gov/safewater/sdwa/text.html [accessed 20 June 2006]. Sobsey M. 1988. Detection and Chlorine Disinfection of Hepatitis A in Water. EPA Quarterly Report. Washington, DC:U.S. Environmental Protection Agency. Teunis PFM, van der Heijden OB, van der Giessen JWB JWB Jewish Welfare Board (now Jewish Community Centers Association) JWB John Wilkes Booth JWB Johnny Walker Black (Whiskey brand) JWB Jewelbox (C++ class library) , Havelaar AH. 1996. The Dose-Response Relation in Human Volunteers for Gastro-Intestinal Pathogens. Bilthoven, the Netherlands: National Institute of Public Health and the Environment. U.S. EPA. 1989. Risk Assessment Guidance for Superfund. Vol 1. Human Health Evaluation Manual (Pt A). Washington, DC:U.S. Environmental Protection Agency. U.S. EPA. 1991. Guidence Manual for Compliance with the Filtration and Disinfection Requirements for Public Water Systems using Surface Water Sources. Washington, DC:U.S. Environmental Protection Agency, 580. U.S. EPA. 2000. Estimated per Capita [Latin, By the heads or polls.] A term used in the Descent and Distribution of the estate of one who dies without a will. It means to share and share alike according to the number of individuals. Water Ingestion ingestion /in·ges·tion/ (-chun) the taking of food, drugs, etc., into the body by mouth. in·ges·tion n. 1. The act of taking food and drink into the body by the mouth. 2. in the United States: Based on Data Collected by the United States Department of Agriculture's 1994-96 Continuing Survey of Food Intakes by Individuals. Washington, DC:U.S. Environmental Protection Agency. U.S. EPA. 2001. Implementation and Results of the Information Collection Rule Supplemental Surveys. EPA 815-R-01-003. Washington, DC:Office of Water, U.S. Environmental Protection Agency,. Wade TJ, Sandhu SK, Levy D, Lee S, LeChevallier MW, Katz L, et al. 2004. Did a severe flood in Verb 1. flood in - arrive in great numbers arrive, come, get - reach a destination; arrive by movement or progress; "She arrived home at 7 o'clock"; "She didn't get to Chicago until after midnight" the Midwest cause an increase in the incidence of gastrointestinal symptoms? Am J Epidemiol 159:398-405. Ward RL, Bernstein DI, Young EC, Sherwood JR, Knowlton DR, Schiff GM. 1986. Human rotavirus studies in volunteers: determination of infectious dose and serological serological pertaining to or emanating from serology. serological test one involving examination of blood serum usually for antibody. response to infection. J Infect Dis 154:871-880. Yoder JS, Blackburn BG, Craun GF, Hill V, Levy DA, Chen N, et al. 2004. Surveillance for waterborne-disease outbreaks associated with recreational water--United States, 2001-2002. MMWR Surveill Summ 53:1-22. Zeger SL, Liang KY, Albert PS. 1988. Models for longitudinal data: a generalized estimating equation approach. Biometrics 44:1049-1060. Joseph N.S. Eisenberg, (1) Alan Hubbard, (2) Timothy J. Wade, (3) Matthew D. Sylvester, (2) Mark W. LeChevallier, (4) Deborah A. Levy, (5) and John M. Colford Jr. (2) (1) Department of Epidemiology, School of Public Health, University of Michigan (body, education) University of Michigan - A large cosmopolitan university in the Midwest USA. Over 50000 students are enrolled at the University of Michigan's three campuses. The students come from 50 states and over 100 foreign countries. , Ann Arbor, Michigan  “Ann Arbor” redirects here. For other uses, see Ann Arbor (disambiguation). Ann Arbor is a city in the U.S. state of Michigan and the county seat of Washtenaw County. , USA; (2) Center for Occupational and Environmental Health and Division of Epidemiology and Environmental Health Sciences, School of Public Health, University of California-Berkeley, Berkeley, California Berkeley is a city on the east shore of San Francisco Bay in Northern California, in the United States. Its neighbors to the south are the cities of Oakland and Emeryville. To the north is the city of Albany and the unincorporated community of Kensington. , USA; (3) National Health and Environmental Effects Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, Chapel Hill, North Carolina Chapel Hill is a town in North Carolina and the home of the University of North Carolina at Chapel Hill (UNC-CH), the oldest state-supported university in the United States. As of the 2000 census, it had a population of 48,715. As of 2004 its estimated population was 52,440. , USA; (4) American Water, Voorhees, New Jersey, USA; (5) Division of Healthcare Quality Promotion, National Center of Infectious Diseases infectious diseases: see communicable diseases. , Centers for Disease Control and Prevention Centers for Disease Control and Prevention (CDC), agency of the U.S. Public Health Service since 1973, with headquarters in Atlanta; it was established in 1946 as the Communicable Disease Center. , Atlanta, Georgia, USA Address correspondence to J.N.S. Eisenberg, Department of Epidemiology, School of Public Health, University of Michigan, 611 Church St., Ann Arbor Ann Arbor, city (1990 pop. 109,592), seat of Washtenaw co., S Mich., on the Huron River; inc. 1851. It is a research and educational center, with a large number of government and industrial research and development firms, many in high-technology fields such as , MI 48104 USA. Telephone: (734) 615-1625. Fax: (734) 998-6837. E-mail: jnse@umich.edu We acknowledge A. Phipps and M. Birkner for conducting the preliminary simulations and C. Wright for final editing and formatting. This work was partially supported by cooperative agreement U50/CCU916961 from the CDC and partially by grant RD-83172701 from the U.S. EPA. The authors declare they have no competing financial interests. Received 25 September 2005; accepted 4 April 2006.
Table 1. Values used for risk assessment models for the different
pathogens.
Model
Model component Cryptosporidium Giardia
Source water
Concentration (organisms/L, 1.06 [+ or -] 2.24 2.68 [+ or -] 24.20
mean [+ or -] SD) (a)
Recovery rate (b) 0.40 0.40
Treatment efficiency (log
removal)
Sedimentation and filtration 3.84 [+ or -] 0.59 3.84 [+ or -] 0.59
(mean [+ or -] SD) (a)
Chlorination (mean 0 3.5 [+ or -] 2.93
[+ or -] SD) (a)
Water consumption (L/day, mean 1.2 [+ or -] 1.2 1.2 [+ or -] 1.2
[+ or -] SD) (c)
Dose response (d) [lambda] = 0.004078 [lambda] = 0.01982
Morbidity ratio (e) 0.39 0.40
Model
Model component Viruses
Source water
Concentration (organisms/L, 0.93 [+ or -] 3.00
mean [+ or -] SD) (a)
Recovery rate (b) 0.48
Treatment efficiency (log
removal)
Sedimentation and filtration 1.99 [+ or -] 0.52
(mean [+ or -] SD) (a)
Chlorination (mean 4 [+ or -] 2.93
[+ or -] SD) (a)
Water consumption (L/day, mean 1.2 [+ or -] 1.2
[+ or -] SD) (c)
Dose response (d) [alpha], [beta], = 0.26, 0.42
Morbidity ratio (e) 0.57
Sample mean [+ or -] SD values are reported.
(a) Where [lambda] was estimated using data from DuPont et al. (1995)
and Rendtroff (1954), respectively. Estimates using data collected in
Davenport (LeChevallier et al. 2003b). All source water data were
modeled using a lognormal distribution. A Weibull distribution was used
for all treatment data. Disinfection for Cryptosporidium was assumed to
be zero. (b) Where [alpha] and [beta] were estimated using data from
Ward et al. (1986). Fraction of pathogens recovered. Data were from the
Information Collection Rule Supplemental Survey (U.S. EPA 2001) after
eliminating extreme observations (i.e., some samples reported a recovery
rate > 100% or < 0%). (c) Consumption of untreated water based on data
from an RDD survey conducted in parallel with the trial. All pathogen
models used the same lognormal distribution. (d) The Cryptosporidium and
Giardia dose-response models used an exponential function [Pr(D|X) = 1
- exp(-[lambda]X)] where [lambda] was identified using data from DuPont
et al. (1995) and Rendtroff et al. (1954), respectively. The rotavirus
dose-response model used a beta-Poisson function [Pr(D|X) = 1 - [1 + (X/
[beta])][.sup.-[alpha]]] where [alpha] and [beta] were identified using
data from Ward et al. (1986). D is disease, and X is dose. (e) The ratio
of those who become ill to those who are infected: Cryptosporidium
(DuPont et al. 1995), Giardia (Nash et al. 1987), and viruses (Ward et
al. 1986).
Table 2. Illness risk estimates associated with drinking water (cases
per 10,000 persons per year) predicted by the risk assessment model.
Cases of illness
Pathogen Mean 2.5-97.5 Percentile range
Cryptosporidium 2.1 0.8-3.5
Giardia 3.4 0.6-15.5
Enteric viruses (a) 8.4 0.2-18.7
Enteric viruses (b) 0 0-0.2
The percentile reflects the variability of the predicted mean estimate.
(a) Assumes that disinfection results in a 4 log removal.
(b) Assumes that disinfection results in a 6 log removal.
Table 3. Comparison of methodologic considerations between drinking
water risk assessment models and intervention trials.
Methodologic considerations Risk assessment Intervention trials
Sensitivity Not relevant Low
Causal evidence Indirect Direct
Pathogen inclusion Few Many
Model specification Adds uncertainty Not relevant
Transmission processes Can be included (a) Only in a limited
way
Distribution system effects Can be included (a) Was included
Examining alternative control Yes No
strategies
Expense Low High
Time Fast Slow
(a) Not included in this study.
|
|
||||||||||||||||||
Printer friendly
Cite/link
Email
Feedback
Reader Opinion