Week 1 Project Assignment Due July 20 at 11:59 PM Assignment 3—Waterborne Outbreak of Cryptosporidium InfectionsPublic health response is crucial in the event of an infectious disease outbreak to re

CHAPTER 14 The Massive Waterborne Outbreak of Cryptosporidium Infections, Milwaukee, Wisconsin, 1993

Jeffrey P. Davis, MD

Wisconsin, 1993

INTRODUCTION: APRIL 5, 1993

On Monday morning, April 5, Dr. Gerald Sedmak, a virologist with the City of Milwaukee Health Department (MHD), received calls from individual citizens complaining of gastrointestinal illnesses. Gerry told other MHD staff about these complaints, and in turn, these staff members began an effort to see whether there was an unusual number of individuals with similar gastrointestinal illness—indeed, there were. Kathy Blair, a registered nurse and epidemiologist with the MHD, received calls from the South Milwaukee Health Department, other local agencies, and private citizens regarding widespread occupational and school absenteeism related to diarrheal illness. Gerry, Kathy, and their colleagues could not have known at that time that these were the initial reports of what would unfold as a historically large and unprecedented waterborne outbreak in the United States. Many pharmacies on the south side of Milwaukee had or were nearly sold out of antimotility medications (medication to slow down or stop diarrhea). Additionally, many health care providers were ill and unable to work. Concomitantly, Dr. Steve Gradus, director of the MHD Bureau of Laboratories, and his colleague, Dr.Ajaib Singh, conducted telephone surveys of hospital emergency rooms and laboratories. They learned of an extreme number of weekend visits for diarrhea-related illnesses to emergency departments and increased numbers of requisitions for bacterial culture of stool specimens. The St. Luke’s Hospital microbiology laboratory exhausted its supply of media used to isolate enteric bacterial pathogens. Based on this information, MHD officials suspected that the outbreak was predominantly affecting the southern part of Milwaukee and included other municipalities in southern Milwaukee County. Initial newspaper and other media reports circulating on April 5 focused on the unusual number of diarrhea illnesses, the unknown etiology of the illness, and the shortage of antidiarrhea medications. Media interest intensified rapidly.

Kathy called Jim Kazmierczak, DVM, an epidemiologist and colleague in the Bureau of Public Health (BPH), Wisconsin Division of Health (DOH; now named the Wisconsin Division of Public Health), notified him of these events, and inquired whether DOH staff were aware of an unusual occurrence of similar illnesses elsewhere in the state. Up to then we had not been told of similar illnesses elsewhere.

During my initial conversations with Kathy and Jim that morning, we discussed the potential of any one of several etiologic agents to be associated with these events and the need for good laboratory data and illness characterization. Although the agent was not yet known, my initial impression was the illness must be considered to be water-borne until proven otherwise because of the magnitude and widespread occurrence of diarrhea illness. Concomitantly, Drs. Gradus and Singh discovered that most hospital microbiology laboratory staff members were not conducting virus culture of stool specimens and that tests for ova and parasites were infrequently ordered. Accordingly, infection control practitioners were selected to facilitate collection of stool specimens for virus culture at the MHD from the next 10 patients presenting at their facilities with diarrhea illnesses.

Although the agent was not yet known, my initial impression was the illness must be considered to be waterborne until proven otherwise because of the magnitude and widespread occurrence of diarrhea illness.

Later, we discussed the MHD plan to widely distribute stool kits for virus testing to clinics to test individuals with acute diarrhea illnesses and the results to date of testing for bacterial enteric pathogens at St. Luke’s Hospital. We sensed that this was a really big outbreak when we learned that from this one hospital on one weekend about 200 stool specimens were obtained for bacterial culture, and we were confronted with a really big clue when we learned that, despite this high volume of testing, all cultures were negative to date for bacterial enteric pathogens. I suggested testing stool specimens already known to be negative for bacterial enteric pathogens and still remaining at the St. Luke’s laboratory for parasitic and protozoan infections at the MHD laboratory. The MHD laboratory was one of two public health laboratories in Wisconsin. The emerging scope and breadth of this outbreak diminished the likelihood that this was a viral illness, and it was important to consider protozoan infections, as they could be associated with large community outbreaks.

Steve arranged for this testing and also requested that microbiology laboratory supervisors begin aggressively testing diarrhea stool specimens for protozoan infections, with specific emphasis on testing for Cryptosporidium. Steve also planned to review water treatment records that would be available from the City of Milwaukee Water Works (MWW), the municipal water utility.

Because of the apparent magnitude of the outbreak and implications for residents in multiple jurisdictions, I offered the onsite assistance of DOH staff to join the MHD in the investigation. We would need Tuesday, April 6 to structure a team and plan our activities in Milwaukee and would drive to Milwaukee from Madison early on Wednesday, April 7. The offer of assistance was accepted.

APRIL 6, 1993

On April 6, I spoke with Steve to discuss the findings of his review of the MWW water treatment records. He described some initial resistance to review the records and defense of the quality of Milwaukee’s water by the MWW administrators, but he was able to review some recent water treatment records. Generally, during recent weeks, there were increases in coliform counts that were consistent with substantial rainfall at water intake points; however, chlorine levels were high, and E. coli counts were within Department of Natural Resources (DNR) recommendations after treatment. Steve also noted spikes in treated water turbidity (one of many measures of water quality) occurred in late March primarily at the South plant, one of the two municipal water treatment facilities operated by the MWW. Turbidity is a technical measure of particles suspended in water. Turbidity can be measured in raw, untreated water and in finished, treated water. Steve was impressed with peaks in turbidity values of treated water on successive days.

Turbidity is a technical measure of particles suspended in water.

While we were discussing these unusual turbidity test results, Steve recalled the Carrollton, Georgia, outbreak of waterborne Cryptosporidium infections involving an estimated 13,000 diarrhea illnesses—a lot of illness by any measure.1 The Carrollton outbreak was associated with a filtered water supply; however, the peak filtered water turbidity was less than 1 nephelometric turbidity unit (NTU). That was surprisingly low. The NTU values in Milwaukee were considerably greater. Based on Steve’s preliminary information, the potential that Cryptosporidium was the etiologic agent in the Milwaukee outbreak seemed increasingly likely. I requested Steve to arrange a meeting early on April 7 with the MWW directors and an opportunity for our staff to review the water treatment records. This would be our first meeting in Milwaukee after arriving on site.

Typically, notifiable enteric diseases are reported when an etiologic agent is laboratory confirmed, although in Wisconsin outbreaks were to be reported upon suspicion regardless of whether the etiology is known. Also, relatively few individuals with true notifiable infections ever get tested, and diarrhea without laboratory data is not reported by physicians. Thus, several strategic surveillance activities were planned. I contacted Dr. Dennis Juranek, the chief of the Parasitic Diseases Division in the National Center for Infectious Diseases, the Centers for Disease Control and Prevention (CDC). Dennis had extensive experience with waterborne diseases and had directed the CDC’s investigation of the Carollton outbreak.1 We discussed the Carrollton outbreak, and Dennis suggested examining illness occurrence in nursing home populations, which were geographically fixed and not likely to obtain their drinking water from other sources or sites.

Later that day Steve set up acid fast smears from three stool specimens known to be bacterial culture negative. The test would determine whether the patients were infected with Cryptosporidium. The specimen staining procedure required an overnight interval before the results would be known.

APRIL 7, 1993

While driving to Milwaukee early on April 7 with Mary Proctor, PhD, MPH, and chief of the BPH Communicable Diseases Epidemiology Section, I discussed our prospective and retrospective surveillance needs. We decided it would be valuable to establish immediately a surveillance focus in two settings. Mary would establish and maintain surveillance for diarrhea illness in nursing homes and emergency departments throughout Milwaukee County and its four contiguous counties.

Bill Mac Kenzie, MD, was the CDC Epidemic Intelligence Service (EIS) officer assigned to the DOH who I supervised. Bill, Jim Kazmierczak, Mary and I from the DOH, Steve and Kathy of the MHD, and Wisconsin DNR staff met with MWW officials, and we were briefed on their water treatment methods and distribution system. The DNR had purview over regulation of drinking water utilities.

Drinking water for the City of Milwaukee (1993 population estimate of 630,000) and many of the other 18 municipalities in Milwaukee County was supplied by two MWW treatment plants, one in the northern part of the city (Linnwood Avenue Purification Plant, North Plant) and one in the southern part of the city (Howard Avenue Purification Plant, South Plant). Each plant had a submerged water intake grid in Lake Michigan about 1.25 (North) and 1.44 (South) miles offshore, respectively, where water entered an enormous tunnel and flowed by gravity through additional tunnels until it reached stations to pump water to the respective plants. The North Plant was located just offshore, but the South plant was located 3.5 miles inland from the Lake Michigan south shore (Figure 14-1).

The North Plant was a strikingly beautiful structure situated on a prominence projecting into Lake Michigan that was initially opened in the 1930s, and it could be viewed from the hills overlooking Lake Michigan. I grew up in a village along the north shore in Milwaukee County, passed by the treatment plant many times, and truly appreciated the majesty of this municipal water treatment facility. The same could not be said for the South Plant, built during the 1960s. Nonetheless, both plants housed modern, large treatment facilities.

Treatment capacities of each plant were sufficiently large to supply the entire water district fully. Treated water needs in Milwaukee are great because of its large population and industrial base, which included industries such as brewing that required large volumes of pure water. Should an outage occur in one plant, the distribution infrastructures from each plant were interconnected so that either plant could supply the water needs for the City of Milwaukee and its retail water customers elsewhere in Milwaukee County. With both plants in simultaneous operation, the South Plant predominantly supplies water to the southern portion of the district, and the North Plant predominantly supplies the northern portion. Central Milwaukee was typically supplied by both plants.

At the time of outbreak occurrence, water treatment in both plants followed the same sequence: the intake of raw water, the addition of chlorine as a disinfectant and polyaluminum chloride (PAC) as a coagulant to the raw water followed by rapid mixing, mechanical flocculation to remove solid and particulate material, sedimentation of the flocculent, and rapid filtration. The South Plant had 8 filters, and the North Plant had 16, each of which was enormous. After filtration, the water was pooled in a massive clear well at each plant (35 million gallons at the South Plant) from which it was distributed to customers (Figure 14-2). Parts of the water distribution infrastructure were very old, including large pipes, and MWW staff members were concerned that lead and copper could be leached if the pH of the water was too low. To address this concern, the MWW changed coagulants in late 1992 from the venerable and time tested alum to PAC.

Testing treated water for a variety of water quality indices was required by the DNR and the federal Environmental Protection Agency (EPA) and was done three times each day at each plant before water was released from the clear wells. Tests of water quality included bacteriologic (E. coli testing and coliform counts), chemical (residual chlorine, residual fluoride, alkalinity, and pH), and physical (color, threshold odor, raw water temperature, and turbidity) tests. The treated water was then distributed from the clear wells. Treated water leaving the plant was referred to as plant effluent. I do not believe I have ever asked anyone whether they would like a nice tall glass of cold plant effluent.

During March and April 1993, the turbidity of water treated at the South Plant and distributed to customers increased with spikes to historically high values. Early in March, there were no significant increases in finished water turbidity despite turbidity spikes in raw water; however, on March 23, the turbidity of South Plant treated water exceeded 0.4 NTU. This had not occurred in more than 10 years. Furthermore, the peak daily turbidity on March 28 and March 30 reached 1.7 NTU, even though the dosages of PAC were adjusted. When turbidity rises, the concern is that something dirty is getting into the system. The goal is to get it to precipitate out or be filtered out before it gets into someone’s nice tall glass of effluent. Chemists at the South Plant aggressively tried to control the turbidity by changing the dosage of coagulant, but PAC was not the coagulant they were used to using, particularly under such extenuating circumstances. Plant staff resumed use of alum instead of PAC on April 2, but a spike in finished water turbidity to 1.5 NTU occurred on April 5

There were substantial differences in daily comparisons of South plant and North Plant finished water turbidity (Figure 14-3); the North Plant treated water turbidity did not exceed 0.45 NTU. The MWW administrators, although mindful that treated water turbidity at the South Plant was unusually high, noted the turbidity results and other water quality measures were in compliance with state and federal regulatory standards. Turbidity related compliance was based on average results over one month. Bill Mac Kenzie and Steve Gradus thought MWW administrators viewed turbidity as less important than other measures of water quality, and more as a measure of clarity. Notably, continually during February through April 1993, samples of treated water for water quality testing obtained from both plants were negative for coliforms. Coliforms are a group of related bacteria whose presence in water may indicate contamination by disease causing microorganisms. Thus, it became strikingly apparent that the likely focus of the outbreak was the water treated in and distributed from the South Plant. Using water quality indices, there were clear differences in finished water quality between the North and South Plants during the same time interval. Nonetheless, this did not exonerate treated water from the North Plant.

Later, on April 7, raw and treated water quality records for an interval that exceeded 10 years were available from the MWW. I examined the turbidity data and plotted the monthly peaks in finished water turbidity from each plant for the past 10 years. Indeed, the initial spike in South Plant finished water turbidity of greater than 0.4 NTU represented the first time in more than 10 years that the finished water from this plant exceeded 0.4 NTU. Thus, the three major spikes occurring during March 28 through April 5 were truly historic peaks in turbidity.

After several internal planning meetings early in the afternoon, MHD Commissioner Paul Nannis and I met with the press regarding our initial impressions. I discussed the strong likelihood that this was a waterborne outbreak and that the prime focus of our investigation was the South Plant, but we would also pursue additional hypotheses. I noted that testing of stool specimens for bacterial enteric pathogens was all negative; however, results of testing these negative samples for a variety of other pathogens were pending.

While I met the press, Jim Kazmierczak and Bill Mac Kenzie were meeting with MHD staff when they received a call from Steve Gradus who reported MHD Bureau of Laboratories staff found Cryptosporidium oocysts in the three stool samples known to be bacterial enteric pathogen negative that he had set up the day before. In addition, staff from the St. Luke’s Hospital laboratory notified Steve that they had detected Cryptosporidium oocysts in stools from 4 patients. These stools were obtained from 7 healthy adults who resided in the Southern half of Milwaukee. Steve also received a report of a case of Cryptosporidium infection in an older resident of West Allis in southwest Milwaukee County. That laboratory diagnosis was made by an astute microbiologist at West Allis Memorial Hospital who tested the stool specimen for Cryptosporidium oocysts even though the test was not specifically ordered. Steve, Bill, and Jim recognized the significance of these laboratory findings. Given our findings and recalling those from the Carroll-ton outbreak that affected an estimated 13,000 people,1 these laboratory results from only eight adult individuals were very significant.

After the press session, Paul and I were to meet with Milwaukee Mayor John Norquist to discuss our findings and investigation plans. While walking to the mayor’s office, I was met by Bill and Jim and was informed of the eight laboratory-confirmed cases of Cryptosporidium infection. They believed Cryptosporidium was a highly plausible etiology for this outbreak, and I concurred. Bill raised the issue that a widespread boil-water advisory involving all users of City of Milwaukee municipal water would be needed to prevent additional cases.

The meeting with the mayor involved a substantial number of MDH, MWW, DNR, mayor’s administration officials, and our DOH team. The water treatment and quality data and state and federal water related regulations were discussed in detail with the mayor. Based on preliminary findings and illness characteristics, an outbreak of this nature that was so widespread would be considered as waterborne unless proven otherwise. We discussed the likelihood that this large waterborne outbreak was caused by Cryptosporidium infection based on laboratory data that we had just become aware of. We informed the mayor of our investigation plans to consider all possible sources of these infections.

The discussion ultimately focused on what could and should be done to control the outbreak. A variety of approaches were discussed, including disinfection and a boil-water advisory. Cryptosporidium was a highly chlorine-resistant protozoan, thus disinfection would not be effective. To be clear, chlorine will kill Cryptosporidium; however, the concentration needed would be great, and treating water with a concentration of chlorine that can quickly kill Cryptosporidium would make the water unsafe to drink or bathe in for too long of an interval and at great expense given the magnitude of the water supply. The pros and cons of a boil-water advisory were considered. Although Cryptosporidium oocysts were heat sensitive and inactivated with boiling, the downside involved the need to boil all water treated in MWW plants to be used for eating and drinking, the personal injury risks associated with boiling water and with consuming recently boiled water, and the bump in energy use that would be associated with a prolonged advisory. The educational needs to conduct this activity effectively would be enormous. Plus, all of these considerations were based on limited and newly available information. Ultimately, the mayor focused his attention on a soft drink can that I brought to the meeting and asked me, “Would you drink a glass of water here, right now?” I replied, “No.” Mayor Norquist then stated, “That’s it. We need to go public with what our suspicions are.”4 Clearly, if I would not drink the water, he would not let Milwaukee residents and visitors drink it unless it was safe to do so.

With no pun intended, this was a watershed moment. The mayor would invoke a boil-water order, and his staff notified the media of a press conference that evening. Precision was needed to provide clear instructions to all users of the water. The implications of this massive boil-water advisory were enormous. Municipal staff members would need to answer many questions. That evening during the press conference, the mayor told all city residents and all users of MWW water to boil their drinking water for 5 minutes and discard all ice.

Because of the need to obtain critical water samples to document any presence of Cryptosporidium in water from each of the implicated treatment plants, systematic sample collections were planned for April 8, and the South Plant would be closed for an undetermined interval beginning on April 9. From April 8 throughout this undetermined interval, all drinking water in Milwaukee would be supplied by the North Plant.

This was an enormous news story with national and international implications. The local media gave this virtually unprecedented coverage. The two major (and competing) Milwaukee newspapers collaborated in publishing a special issue in Spanish. Plans for daily media updates were announced. A half-page ad for an over-the-counter antimotility medication appeared, but now that a boil-water order was invoked, what would be needed to lift it? It was my responsibility to answer that question, but there were so many questions to answer and avenues to explore to describe fully the scope and all aspects of this outbreak, to generate and test hypotheses regarding how it occurred and why it was so large, and to determine whether measures were to be effective in controlling it.

THE OUTBREAK AND THE OUTBREAK INVESTIGATION

This is what was known about Cryptosporidium before April, 1993:

We needed to learn quickly what was known about Cryptosporidium, particularly regarding its associations with waterborne disease outbreaks. At the time of this outbreak, relatively few public health officials had much knowledge of this protozoan parasite and its associated illnesses, nor was it well known as a pathogen by those charged with keeping local supplies of drinking water safe.

Although initially detected in animals in 1907 and for years thought not to affect humans, Cryptosporidium was first reported as a human pathogen in separate case reports of enterocolitis in immune competent5–7 and compromised8,9 patients. Cryptosporidium became a prominent pathogen when it was recognized in 1981–1982 as an AIDS-defining illness and opportunistic infection. The initial report in 1984 of a waterborne disease outbreak of cryptosporidiosis was associated with an artesian well in San Antonio, Texas. During 1986 to 1992, waterborne Cryptosporidium infections were associated with surface water exposure in New Mexico (1986), post-treatment contamination of drinking water in Aryshire, UK (1988), and filtered water supplies in Carrollton, Georgia (surface, 1987), Swindon and Oxfordshire, UK (1989), the Isle of Thanet, UK (1991), and most recently in Jackson County, Oregon.10–14 The outbreak that was most similar to ours was the one in Carrollton.

Early Logistics

It was rapidly apparent this outbreak investigation would require work well beyond even the larger outbreaks I had focused on during my tenure as a state epidemiologist.

It was rapidly apparent this outbreak investigation would require work well beyond even the larger outbreaks I had focused on during my tenure as a state epidemiologist. Each of the outbreak’s many facets would need a person in charge. Information by necessity would flow in one direction, but nominally, and those providing data needed to understand this; however, information would be continually shared with those who needed to know. Thus, organization and appropriate use of human resources were critical elements of this investigation. As the state epidemiologist and the state’s leader of this multijurisdictional outbreak investigation, my time was focused on generating hypotheses, overseeing the generation of outbreak related data and information, communicating important findings, whether positive or negative, and assuring that all aspects of the investigation were proceeding in a timely, efficient way. My time would not be well spent working out the logistics of having enough staff available to conduct certain aspects of the investigation, although it was important for me to request and obtain the resources when needed. Ivan Imm, BPH Director, and John Chapin, DOH Deputy Administrator, were masterful in generating the personnel and supply resources needed and would remain on site for nearly 2 weeks to run interference. We had a large room in the Milwaukee Municipal Building for our investigation with dedicated phones, computer resources, and even a new fax machine needed on short notice for our surveillance activities that was purchased by the State Health Officer, Ann Haney. Several additional venues were established for activities requiring phone banks and dedicated teams could be rapidly deployed for survey purposes. Ultimately, nine teams, each of which focused on separate investigation components and all of which involved public health personnel, were created during the early phase of the outbreak investigation and were staffed by various combinations of City of Milwaukee and DOH personnel (Exhibit 14-1).

Early on we needed an experienced investigator to join our investigation and focus on special laboratory and surveillance projects. I had worked with David Addiss, MD, since 1985 when he began his 2-year term as an EIS officer assigned to our health department, and I was his field supervisor. We collaborated on many investigations since then. Dave was working as a medical epidemiologist in the Parasitic Diseases Division at the CDC and was extremely familiar with systems in Wisconsin. I invited Dave to join our investigation in Milwaukee, and fortunately, this was possible.

Review of MWW Data

Raw Water Source

EXHIBIT 14-1 Investigation of the Massive Waterborne Outbreak of Cryptosporidium Infections: Epidemiologic and Laboratory Studies Initiated and Health Education and Logistic Functions (Selected)

Studies to define the outbreak

• • Laboratory confirmed case surveillance: determine outbreak trends, characterize the illness in healthy populations and in immunocompromised and other special populations, and establish case definition, collection, and storage of serologic and other specimens for future studies

• • Emergency room log database and case surveillance

• • Nursing home database and surveillance

• • Random digit dialing (RDD) surveys:

  • • RDD 1: description of clinical disease, risk factors for acquiring Cryptosporidium infection in the community, early estimate of magnitude

  • • RDD 2: description of trends, age and zip code-specific attack rates, epidemic curve, estimates of outbreak magnitude, use of clinical services, hospitalization, economic impact, secondary transmission

  • • RDD 3 and 4: survey to assess morbidity before and after reopening the South Plant

• • Single day (short duration) of exposure database and surveillance: determine when problems with water treatment began, specific days that water was contaminated, incubation period calculations, quantify occurrence of secondary transmission

• • Immunocompromised, AIDS/HIV assess multiple risk factors, natural history, prevention measures—large ill and well cohorts

• • Surveillance in child care and daycare settings

• • Satellite outbreak investigations in communities outside of greater Milwaukee areas

Environmental Studies

• • MWW: plant protocols and engineering reviews, water quality data reviews

• • River and estuaries data: cooperative study to monitor the Milwaukee River watershed and subwatershed, the sewage treatment plant influent and effluent, beach sites and MSS treatment plant influents and effluents. Five samplings at 21 sites through Spring 1994

• • Efficacy of point of use filters in an outbreak setting

• • Meteorological data analysis

• • Laboratory testing of stored ice for Cryptosporidium and additional environmental testing

City of Milwaukee government activities (selected)

• • Press-related coordination and releases, risk communication, phone banks coordination, fact sheets development, and translations and other health education related functions

• • Quality assurance and inspections

• • All agencies availability (24/7) and interagency coordination and committees

• • Infection control and coordination with all health related facilities

The Milwaukee watershed includes three rivers that flow through Milwaukee County: the Milwaukee, Menomonee, and Kinnikinnick Rivers. Environmental sources of Cryptosporidium oocysts that could impact on these rivers and associated watersheds could have been agricultural, industrial (meat packing), and wildlife related. The three rivers converge a short distance before emptying into Lake Michigan within a harbor area protected by a large breakfront (see Figure 14-1). There are three major gaps in the breakfront, including a large central gap allowing boat access and egress, and smaller north and south fair weather gaps. The ambient current in the harbor and within the breakfront was southerly. The flow of water from within the breakfront through the south fair weather gap created a plume that was typically directly toward the raw water intake grid for the South Plant, which was located 7,600 feet offshore at the bottom of Lake Michigan under 50 feet of water.

The central sewage treatment plant operated by the Milwaukee Metropolitan Sewage District is located at the point where the river, created by the confluence of the three rivers, empties into Lake Michigan, and the effluent from the sewage treatment plant empties into the harbor within the confines of the breakfront.

Laboratory-Based Active Surveillance and Testing2,15

Steve transmitted information on Cryptosporidium detection to the directors of 14 clinical microbiology laboratories throughout the five county greater Milwaukee area and established laboratory-based surveillance with data to be reported to him. Compared with the usual parasitic testing of stool specimens, an additional flocculation step and special staining of the pellet with Kinyoun acid fast stain was needed to detect Cryptosporidium oocysts, and although it was the gold standard for testing at the time, the test was insensitive. This test was used in 13 of the laboratories. Because of the nature of the test, relatively few tests of stool for Cryptosporidium were requested before the occurrence of this outbreak, and these were primarily for testing patients with AIDS who had diarrhea, particularly chronic diarrhea.

Preoutbreak (prepublicity) baseline Cryptosporidium-related data were generated by retrospectively examining results of testing during March 1 through April 6 and were then compared with results of prospective testing during April 8 through April 16. During March 1–April 6, among 42 stool specimens submitted for Cryptosporidium testing, 12 (29%) were positive. This compares to 331 of 1009 specimens (33%) tested during April 8 to April 16. Of note, other pathogens accounted for only a very small proportion of the Cryptosporidium test-negative illnesses, and the percentage of specimens Cryptosporidium test positive (39%) during the Carrollton outbreak was similar. The similarity in the proportion of test positive principally among immunocompromised individuals with persistent diarrhea prior to the outbreak (29%) and among principally previously healthy persons of all ages in a wide range of demographic settings (33%) was striking and perplexing.

One question to be answered prospectively was the efficacy of Cryptosporidium-related laboratory tests that were or would be in development. To address this, Dave Addiss and a microbiologist from the CDC worked with Steve and other MHD laboratory staff members to obtain and archive stool and serum specimens for current and future testing.

Clinical Characteristics of Illness and Generation of a Case Definition2

The magnitude of this outbreak provided an opportunity to examine the clinical spectrum of illness among many individuals with laboratory-confirmed or clinically defined infections. Early in our investigation, we examined clinical signs and symptoms of illness and illness onset data among those with laboratory-confirmed Cryptosporidium infection to generate a reliable clinical case definition. We observed a virtual 100% occurrence of watery diarrhea among these individuals. Onset data suggested this outbreak emerged after March 1, 1993. We determined the illness onset interval of March 1 to April 9 would be inclusive of outbreak-related exposure until the South Plant was closed.

Although we recognized many thousands of cases likely occurred through May 15, the MHD received reports of 739 individuals with laboratory-confirmed Cryptosporidium infections; a more workable sample of 312 of these patients was selected, and extensive interviews were completed on 285, which was nearly a 40% sample and highly adequate for our purposes. This would be our study cohort with laboratory-confirmed infection. To understand the occurrence of illness in the community, we conducted the first of our four random digit dialed phoned surveys during April 9 to 12 to help describe clinical and other features of outbreak-related infections and generate a cohort of members with clinically defined illness to compare to the cohort with laboratory-confirmed illnesses. This first random digit dialed phoned survey (RDD 1) was administered by the DOH Sexually Transmitted Disease program staff who had substantial interviewing skills and by public health nurses, including an energetic group of retired public health nurses in southeastern Wisconsin who were a great resource and donated much time. The retired nurses, all of whom had long public health careers, literally stepped forward and asked us if they could be of service. Among 482 adult Milwaukee city residents, 42% had illness meeting the clinical case definition of cryptosporidiosis, and 6% of those individuals were hospitalized for their illnesses. We recognized the laboratory-confirmed illness cohort were ill enough to seek medical care and have their physicians order appropriate stool exams; indeed, 17% were immunocompromised, and 46% were hospitalized.

TABLE 14-1 Clinical Characteristics of Case Patients with Laboratory-Confirmed Cryptosporidium Infection and Survey Respondents with Clinical Infection

* The criterion for clinical infection was the reported presence of watery diarrhea.

†Unless otherwise noted, Yates’ correction has been applied to Pvalues. NA denotes not applicable.

‡Data are from 101 case patients interviewed during phase 1 of the study.

§ By Kruskal-Wallis test.

From Mac Kenzie WR, Hoxie NJ, Proctor ME, et al. A massive outbreak in Milwaukee of Cryptosporidium infection transmitted through the public water supply (Table 1). N Engl J Med 1994;331:161–167.

We found the cohorts to be similar in mean age, gender distribution, dates of illness onset, and in occurrence of abdominal cramps, fatigue, and muscle aches (Table 14-1). As anticipated, because the case group was comprised of patients ill enough to have consulted physicians and then be tested, those with laboratory-confirmed illness had diarrhea of longer duration (9 vs. 3 days), more frequent stools (median maximum of 12 vs. 5 stools daily), and higher rates of vomiting (48% vs. 18%) and fever (57% vs. 36%). Nonetheless, those with clinically defined illness were also quite ill.

Nursing Home Surveillance2,3,15

To determine whether and for how long gastrointestinal illness was associated with drinking water supplied by the South Plant, Mary Proctor established the aforementioned retrospective and prospective surveillance system for diarrhea illness among residents of representative samples of northern Milwaukee and southern Milwaukee nursing homes. Nursing home residents are relatively nonmobile (fixed populations) among whom information on diarrhea is collected routinely by nursing home staff. With their limited possibilities of exposures outside of their nursing homes, they represented an ideal human population to study. We defined diarrhea as three or more loose stools in a 24-hour period. Among residents of the nine nursing homes in northern Milwaukee, the prevalence of diarrhea remained less than 2% during March and April. In stark contrast, the prevalence of diarrhea among residents of six of the seven southern Milwaukee nursing homes increased to 16% during the first week of April and remained high until 2 weeks after the boil water advisory was issued (Figure 14-4c). Mary was particularly excited when she noted that the one nursing home in southern Milwaukee that did not observe an increase had a private well, and the prevalence of diarrhea among residents at that home remained less than 2%…serendipity at its finest. Furthermore, of stool samples prospectively collected from 69 southern Milwaukee nursing home residents with diarrhea (during April), 51% were positive for Cryptosporidium compared with none among samples from 12 northern Milwaukee nursing home residents with diarrhea. Clearly, if there were any doubts about the early association of illness with the South Plant, these data dispelled them.

Nursing home residents are relatively nonmobile (fixed populations) among whom information on diarrhea is collected routinely by nursing home staff. With their limited possibilities of exposures outside of their nursing homes, they represented an ideal human population to study.

Emergency Department Surveillance

While classically emergency department surveillance is useful during outbreak investigations, this was of greatest value to us only during the early phase of our investigation to facilitate case finding, clinical characterization, and case definition generation. To monitor the course and impact of the outbreak, other more focused surveillance activities proved to be superior.

Estimation of Magnitude of the Outbreak2

Because the outbreak magnitude was so large that it precluded any reasonable estimate through standard illness reporting means, we planned to generate this estimate by conducting the second of our RDDs (RDD 2). Bill Mac Kenzie, working closely with Neil Hoxie of the DOH, was the lead on this critical component of our outbreak investigation. With objectives of estimating the magnitude of the outbreak and its temporal and geographic impact, a questionnaire was created for use by telephone surveyors, and sample selection was conducted using computer-generated, random telephone numbers for the greater Milwaukee area. The survey was written by the University of Wisconsin Survey Research Institute (SRI) based on a set of key data needed by the DOH. DOH and SRI staff generated the final instrument. SRI staff members were finely honed to conduct these types of surveys and conducted all of the calls and completed all forms and data entry. Based on RDD 1 data, watery diarrhea with onset between March 1 and April 28 was used as a reliable definition of cryptosporidiosis. The survey included questions about watery diarrhea occurrence and frequency, other signs and symptoms, illness duration, health care visits, hospital stays, mortality, and demographic features. The survey sample included households in Milwaukee County and the four counties contiguous to Milwaukee County; the five-county greater Milwaukee area population was an estimated 1.61 million people. During April 28 through May 2, among the 840 households contacted, 613 (73%) participated representing 1662 household members from Milwaukee and the four surrounding counties. Household-related information was provided by interviewing the most knowledgeable adult member. The sample distribution was similar to the five county population distribution based on 1990 census data.

The clinical definition of cryptosporidiosis was watery diarrhea, but the time frame of onset was March 1 through April 28 to assess the impact of the boil-water order and other recommended control measures. Among the 1662 household member sample, 436 (26%) had illness meeting the definition of clinical cryptosporidiosis. These data were excellent for generating an epidemic curve to examine outbreak trends (Figure 14-5). The occurrence of watery diarrhea began with a small increase in morbidity beginning about March 18 followed by a larger increase in the five-county area beginning about March 24, a peak during April 1–April 7 with a distinct mode on April 4 followed by a rather sudden decrease in morbidity on April 13. This observed decrease in morbidity provided strong evidence that the boil-water order was effective.

Among participants residing within the MWW service area in Milwaukee County, the attack rate of watery diarrhea was greatest (52%) among residents in southern Milwaukee, lowest (26%) among residents of northern Milwaukee, and intermediate (35%) in the mid-zone region in which water could be supplied by either the South or North Plants. The attack rate among participants residing outside the MWW service area was 15%; however, among residents who lived outside the MWW service area but worked within southern Milwaukee, the attack rate was 39%.

To digress momentarily, it is logical to wonder why there was such a big difference in attack rates among residents in southern Milwaukee households (52%) and residents of the southern Milwaukee nursing homes (16%), who are typically thought of as being more vulnerable. We observed attack rates in this outbreak to be age dependent, and the older population had the lowest attack rates. This is likely associated with prior immunity. There were similar outbreaks during the 1930s, each involving tens of thousands of people prior to the construction of the north plant, and I suspect that individuals who lived in Milwaukee at that time were more likely to be immune.

By projecting the overall attack rate (26%) onto the five-county population (1.61 million), we estimated that during March 1 through April 28, the number of persons with watery diarrhea who resided in this area was 419,000. Subsequently, in another RDD survey (RDD 3), a background rate of watery diarrhea among residents in this area was estimated to be 0.5% per month. Application of these background data resulted in a final estimate of 403,000 residents of the five county area with watery diarrhea attributed to this truly massive outbreak. The RDD 2 data were used to estimate health care seeking and occupational impacts. Among the 436 household participants with watery diarrhea, 50 (11%) saw health care providers (estimated 44,000 persons seen as outpatients) for their illnesses, and 5 (1%) were hospitalized (estimated 4,400 persons hospitalized). Furthermore, for each case individual, the mean days of reported lost productivity (work or school) was 1.8 days, which projected to an estimated 725,000 days lost productivity attributed to this outbreak.2,16

These were astounding numbers. The outbreak was the largest waterborne outbreak documented in the United States, and perhaps in the developed world! Using these data as the estimates of morbidity, several years later after appropriate hospital-, workplace-, insurance-, and government-related records were reviewed, estimates of the fiscal impact of this outbreak were generated.16

Cases in Visitors to Milwaukee17

While most BPH communicable disease epidemiologists were involved in the onsite investigation, one, Wendy Schell, remained in Madison. In addition to memoranda circulated widely to physicians, public health officials, and infection control practitioners throughout Wisconsin, on the day after the announcement of the boil water advisory, I generated a memo. This memo included information regarding the occurrence of the large waterborne outbreak, descriptive epidemiologic data, and data pertaining to the Cryptosporidium etiology and the boil water advisory. Wendy distributed this memo nationwide to all state and territorial epidemiologists with instructions for the reporting of cases. These memoranda were used to establish regional and national surveillance for confirmed and suspected cases of Cryptosporidium infection among individuals residing outside of the five-county greater Milwaukee area. This surveillance activity was very fruitful. Numerous reports of cases were received from state and local departments, and after regional and national press coverage of the outbreak and related events, the DOH received numerous self-reports of illness from individuals who resided outside the five-county region but visited Milwaukee and subsequently became ill.

As questions and hypotheses mounted, we recognized that residents of the greater Milwaukee area had too many opportunities to become infected to be helpful in addressing some of them; however, interviewing individuals with short term visits to Milwaukee would be key to determining the incubation period, how long Cryptosporidium was present in the water supply, and the extent of secondary transmission in households.17

Interviewing individuals with short term visits to Milwaukee would be key to determining the incubation period, how long Cryptosporidium was present in the water supply, and the extent of secondary transmission in households.

During April 12 through May 20, out of region participants with suspected or confirmed Cryptosporidium infections were interviewed by phone using a standardized questionnaire focusing on demographic features, illness characteristics, length and site of stay, and water consumption; 130 of those interviewed had laboratory or clinically confirmed cryptosporidiosis, and 94 case individuals had brief (<48 hours) visits to the MWW service area. We had a unique opportunity to measure the interval during which Cryptosporidium was present in the water by noting the dates of the brief visits of these 94 case individuals. Among those with these brief visits, particularly the 63 with visits of 24 hours or less, we determined the incubation period by examining the intervals between dates of illness onset and dates of arrival to Milwaukee. The median incubation period was 7 days (range, 1–14 days). Among those with brief exposures, the earliest date of exposure was March 24, and the last date of exposure (with one exception) was April 5. The peak days of exposure were March 27 and 28, and based on arrival date information, Cryptosporidium was present in the municipal water supply continuously and daily during March 24 through April 5. Because of the brief durations of exposure, including several strikingly brief airport layovers, we could quantify consumption of tap water and beverages containing tap water while in the MWW service area. The median amount consumed was 16 ounces, and 23% of brief exposure participants drank 8 ounces or less.17

Important clinical observations among this cohort included the low (5%) risk of secondary transmission within households where the index case patient was an adult; however, cryptosporidiosis during this outbreak appeared more severe when compared with cases described in previous case series reports. The recurrence of watery diarrhea after apparent recovery from clinical illness was a frequent event among short-term visitors to Milwaukee with laboratory-confirmed infections (39%) as well as in Milwaukee County residents with clinical infections (21%) and contributed a significant proportion of cases of postoutbreak diarrhea in the community.17 The significance of this observation among the short-term visitors was that it excluded re-infection with Cryptosporidium as a cause of recurrence.

Response to Questions from Business

The boil-water order affected many businesses in Milwaukee County. Because of the importance of beer (as an industry and as a beverage to be depended on more often than usual in the days to come), one early concern involved the cold-filtered brewing process used by a major Milwaukee brewing company; however, this company pasteurized its water before it was used in the brewing process, and its filters were sufficiently small to remove Cryptosporidium oocysts effectively. Before and during the outbreak, this brewing company donated a very large volume of pasteurized, purified water for individuals who were immunocompromised and needed a reliable source of pure drinking water.

Many questions came from food processors and the health care industry. The only food-related industries with prevailing concerns were those that prepared food that was not to be heated to an adequate temperature to kill Cryptosporidium oocysts. Regional and nationwide recalls (all voluntary) of many of these types of products, such as prepared salads and dips, were needed and we depended on the Food and Drug Administration and U.S. Department of Agriculture to help with these actions.

Studies of Special Populations

Impact on Children and in Childcare Settings18,19

We were very cognizant of the importance of the observation of widespread school absenteeism in the recognition of this outbreak. Although much effort was focused on understanding the occurrence of illness in fixed populations, we needed to understand the impact of Cryptosporidium infection on children, particularly because of hygienic issues that may contribute to transmission in households, day care facilities and other settings, but also to determine the extent of asymptomatic infection and persistent shedding of Cryptosporidium in this population. To address these issues, two very specific Epidemic Aid investigations were generated with Dave Addiss serving as the CDC supervisor; Ralph Cordell focused on the impact of Cryptosporidium infections in child day care settings,18 and Helen Cicirello focused on broader impacts among children.19

Ralph coordinated the screening of 129 diapered attendees of 11 day care centers in metropolitan Milwaukee and found 35 (27%) with Cryptosporidium oocysts in their stool, 10 of whom did not have diarrhea during the outbreak. Thus, at least in this very young population, asymptomatic or minimally symptomatic Cryptosporidium infection appeared to be somewhat frequent.18 There was no systematic assessment of asymptomatic shedding of Cryptosporidium among those without diarrhea.

Helen coordinated a study of the clinical, laboratory, and epidemiologic features of outbreak-associated cryptosporidiosis among children who sought medical care at the Wisconsin Children’s Hospital during the outbreak. As with other diarrhea illnesses for which multiple stool examinations increases the opportunity to make a diagnosis, those children who had laboratory-confirmed Cryptosporidium infections had stools specimens submitted more frequently and later in their illnesses than those with stool exams that were negative.19

HIV Infection20

In outbreak settings, it was not known whether Cryptosporidium-related diarrhea illness attack rates were greater in HIV-infected persons than in the general population. There was great concern among the public health community and the HIV/AIDS advocacy community regarding the impact of this outbreak on those with HIV-infection and what could be done to prevent Cryptosporidium infections from occurring. Chronic diarrhea caused by Cryptosporidium is an AIDS-defining condition, and the massive exposure to Cryptosporidium among the HIV-infected community in Milwaukee could result in many new AIDS cases with substantial suffering because of the lack of an effective treatment. This outbreak occurrence provided an opportunity to survey a large community of HIV-infected individuals exposed to a point source of Cryptosporidium contaminated water to examine epidemiologic features and severity of clinical illness.

A group headed by Holly Frisby, DVM, an epidemiologist in the DOH AIDS/HIV Program, conducted a case control study incorporating a randomized sample of 263 among 703 HIV case-management clients in the five-county greater Milwaukee and one age- and gender-matched control selected from the general population using RDD methods. To facilitate participation and assure patient confidentiality, case patients were known only to their case managers. Case and control survey questions were similar. The survey participation among the sample of HIV-infected individuals was high (82%). During this outbreak, the attack rate of watery diarrhea among HIV-infected persons (32%) was less than among matched controls (51%); however, although HIV-infected individuals were not more likely to experience symptomatic Cryptosporidium infection than persons in the general population, once infected with Cryptosporidium, the duration and severity of illness were greater in HIV-infected individuals, particularly among those with CD4+ T-lymphocytes < 200 per mL.20

Mortality21

The media focused on severe and ultimately fatal Cryptosporidium infections in several individuals with previously diagnosed AIDS. During the first several weeks of our investigation, we were not aware of any deaths attributable to this outbreak that involved individuals who were not immunocompromise or had some other serious underlying illness. We realized measurement of mortality attributed to this outbreak must be deferred because of delays inherent in obtaining accurate mortality data. When final state mortality data were available for 1993–1994, we were able to measure this. I was continually concerned regarding the arbitrary and unfortunate number of 100 deaths attributed to the outbreak and circulated in the media, as there were no data to support this, and we would be the ones to provide an accurate assessment. Our approach was to assess death certificate data that specifically indicated cryptosporidiosis as a cause of death that could conclusively be attributed to the outbreak, and to measure the AIDS-related mortality during the first 6 months after the outbreak (the near term observed mortality that likely would be acutely associated with the outbreak) and compare that with expected AIDS-related mortality during this interval and to the AIDS-related expected mortality during the subsequent 6 months. We observed greater than expected AIDS-related mortality during the first 6-month interval and less than expected AIDS-related mortality during the subsequent 6-month interval. Thus, in addition to death certificate data (50 outbreak attributable deaths, 86% among persons with AIDS), we were able to measure premature mortality attributable to this outbreak among persons with AIDS with no death certificate mention of cryptosporidiosis (19 deaths).21 The “official” outbreak-related attributable mortality was 69 deaths, of which 93% occurred in persons with AIDS.

The “official” outbreak-related attributable mortality was 69 deaths, of which 93% occurred in persons with AIDS.

Detection of Cryptosporidium in the Water Supply

Another of our objectives was to demonstrate Cryptosporidium in water or other material sampled during the actual outbreak interval. Our first lead was a company that manufactured very fine absolute pore-size filters. For many days in a row, their test filters were discolored by an apparent water impurity. Fortunately, the filters were dated and saved by the company and available to us.

Perhaps the most fortuitous opportunity involved a call from an ice manufacturing company to Greg Carmichael, a quality assurance officer who worked for the City of Milwaukee. The company, located near the South Plant, manufactured large (50 gallon) blocks of ice for sculpture. During March 25 through April 9, impurities in the ice spoiled its color and clarity; however, representative blocks were saved! A tremendous opportunity existed, and we generated a plan to use these slabs. Under Dave’s supervision, each block of ice would be melted separately, and the resulting water would be divided into two aliquots, one of which would be filtered using a standard, nominal spun polypropylene cartridge filter, which was notoriously unwieldy to use and assay and crude in accuracy of results. The other aliquot would be filtered using a large (11.5 inch) Millipore membrane filter with a 0.45-micron fixed pore size that was suspected to be superior. Unfortunately, despite these elegant plans, there was confusion within the company, and all but the blocks of ice made on March 25 and April 9 were discarded. Nonetheless, the manufacturing dates of the remaining blocks were well timed for our purposes, and the study was completed according to plan.2 Dave reported that the concentration of Cryptosporidium oocysts in ice produced March 25 and April 9 and filtered using the membrane filter was 13.2 and 6.7 oocysts per 100 liters compared with 2.6 and 0.7 oocysts per 100 liters, respectively, using the spun cartridge filter. Regardless of the filter used, Cryptosporidium was confirmed to be present in each block at substantial concentrations even though freezing reduces the recovery of oocysts. The March 25 data were notable as the peak turbidity of South plant water that day was 0.5 NTU, which was substantially less than the NTU peak days later when the measurable concentration of oocysts would be projected to be considerably higher. As expected, the submicron membrane filter was functionally far superior to the standard spun cartridge filter.

Investigation of the Milwaukee Water Works South Plant

During my first conversation with Dennis Juranek of the CDC regarding the outbreak, he provided me with the names of raw water and drinking water experts: Walt Jakubowski (EPA, Cincinnati), Stig Regli (EPA, Washington, DC), Joan Rose (University of South Florida), Mark LeChevalier (American Water Works Association [AWWA]), and others. I called each of these individuals shortly after my initial review of the finished water quality and the invoking of the boil-water order. One aspect of the conversation with Walt was the need to conduct an in-plant inspection of both of the MWW plants by an EPA engineer who was expert in all aspects of the drinking water treatment process. Walt mentioned that Kim Fox would be particularly skilled to conduct this aspect of the investigation. Kim and an EPA colleague Darren Lytle arrived in Milwaukee on April 12 and began their investigation by reviewing detailed operational and water-quality data from both plants followed by in-depth onsite inspections of the South and North plants and extensive interviewing of plant and other MWW personnel. In the report of this investigation, Kim noted that the South Plant received a highly variable quality of raw water from Lake Michigan (the influent) for processing and noted that during March 18–April 9, 1993, the raw water turbidity levels ranged from 1.5 to 44 NTU (usually 3 to 4 NTU in prior months), and raw water coliform concentrations were also quite variable. Specific deficiencies that may have contributed either to a delay in recognition of a problem or the inability to bring finished water turbidity to plant baseline values included the lack of historical use records for the coagulant used (PAC was used only since September, 1992). Also, the residence time for the water in the plant was relatively short, and the time required to see a result in treated water quality after chemical adjustment was relatively long. These factors contributed to difficulty in optimizing the coagulant dose. In addition, the finished water turbidity was measured in the clear-well rather than as the water left each filter. The clearwell was the massive receptacle where water from all the filters was pooled before it was released as effluent from the treatment plant. Also, although the filters were frequently backwashed to remove impurities and maintain optimal filtering capability, the backwash water was recycled through the plant instead of being discarded.

Comparison of the Efficacy of Different Surveillance Methods Used in this Outbreak Investigation15

The availability of a rich array of surveillance data from seven categories of data source provided us with an opportunity to compare the efficacies of using each category of surveillance data during this waterborne outbreak investigation. We found that surveillance systems that could be easily linked with laboratory data were flexible in adding new variables, and those that demonstrated low baseline variability were most useful. Notably, geographically fixed nursing home residents served as an ideal population with nonconfounded exposures; however, the signals that were most timely (i.e., the shortest interval needed to learn about the peak) were consumer complaints to the South Plant utility (the best) and aberrant and peak finished water turbidity (Figure 14-4). Although not indicators of disease, these signals can be effectively used in stimulating heightened surveillance for human illness and generating timely messages to the public and persons at greater risk of water related illness and help reduce potential outbreak-related morbidity. This would be particularly helpful to implement in communities with populations greater than 100,000 with water supplies derived from surface water.

Unusual Clusters and Anecdotes

There were several unusual clusters. The Finals (Frozen Four) of the NCAA hockey tournament were held in Milwaukee April 1–3. We conducted a follow-up with cooperation from respective state epidemiologists to determine whether diarrhea illnesses occurred in any team or entourage members. Indeed, cases occurred among members or associates of three teams, but not among the team from the University of Maine that brought its own bottled water supply as a standard practice, and they won the NCAA championship. Bottled water generally should not be viewed as a panacea, but it sure was healthful in this instance. A large number of individuals with watery diarrhea illness were seen in the clinic facility at the speedway in North Wilkesboro, NC during a NASCAR event. All ill individuals had attended the funeral of the beloved NASCAR driver and Milwaukee resident Allen Kulwicki (killed in a crash of a private airplane) in Milwaukee 1 week previously. An outbreak of Cryptosporidium infections investigated in Michigan occurred among a Coast Guard crew on a boat that had its potable water supply tank filled in Milwaukee.

Professional sports were not spared. Several Milwaukee professional athletes became ill, and there were concerns raised among teams that had recently traveled or would be traveling to Milwaukee. The largest event of concern for which numerous special precautions were taken was the Milwaukee Brewers home season opening game that occurred before the boil-water advisory was lifted. Of course, the beer was safe to drink.

Invoking and Lifting the Boil-Water Advisory

When boil-water advisories are enacted, it is important to envision the terms and conditions needed to discontinue (lift) them. After becoming familiar with water testing processes and having numerous phone conversations with experts at the AWWA, EPA, CDC, the Wisconsin DNR, and in academia, I began the process of generating a Delphi survey that involved the repeated interviewing of these experts regarding what they felt the standard should be for lifting the boil-water order. The South Plant was closed, and thus, all of our attention was on the North Plant. We needed to demonstrate that the North Plant water was safe. This would involve repeated measurement of Cryptosporidium oocysts in samples of North Plant effluent water and representative samples of water obtained from distal sites in the municipal water distribution system. Also, spun cartridge filters, the standard at that time, would be used to filter samples of finished water and would then be tested for the presence of Cryptosporidium oocysts. The filters were sent to the AWWA laboratory in Belleville, Illinois. Mark LeChevallier and his colleagues in Belleville provided exceptional expertise and service during this outbreak.

After several rounds of calls, the Delphi process ultimately focused on the demonstration of less than one Cryptosporidium oocyst per 100 liters of filtered water at each of the sampling sites (one central and four distal) and in two consecutive samples at each of the central and distal sites. The samples would involve 12-hour collections at each site. With shipping, the cycle to process each filter took 2 to 3 days. I realized that it would probably take a minimum of 5 days to have the data needed to lift the advisory, and one sample from any site with results that exceeded the threshold could add another 3 days to the process. Sometimes my patience in waiting for test results wore rather thin.

Finally, the boil-water advisory was lifted on the evening of April 14 after receipt late on April 14 of the results of processing samples collected during the third round on April 13. The city was served exclusively by the North Plant until June, when the water filters replacement and refurbishing was completed at the South Plant.

Additional Investigations and Studies

Protection from Point-of-Use Filters22

During a press conference in the early phases of the outbreak, I requested that individuals who had point-of-use (filters that were installed in homes or workplace settings) water filters installed in their homes before the outbreak contact us because we were interested in evaluating whether they were effective in preventing diarrhea illness during this outbreak. The response to this request was overwhelming, and many individuals volunteered to actually donate their filter if needed. Bob Pond, an Atlanta-based EIS officer, joined Dave Addiss and me in Milwaukee to conduct this evaluation. We surveyed 155 filter owners and 99 completed the self-administered questionnaire. Among residents and users of water in the southern or central Milwaukee, we found users of submicron (pore size less than one micron) point-of-use filters during the outbreak were significantly less likely to experience watery diarrhea than those who consumed unfiltered tap water in a public building and those who had home water filters with pore sizes greater than 1 micron. Being conservative in interpreting our data, we concluded that (even in these extraordinarily adverse field conditions) submicron point-of-use water filters may reduce the risk of waterborne cryptosporidiosis.22

Postoutbreak Transmission of Cryptosporidium23

Surveillance for new infections was sustained for many months. Postoutbreak secondary transmission of Cryptosporidium was expected to occur; however, of great concern was the potential of sustained waterborne transmission of Cryptosporidium that could be occurring with clinical muting of symptomatic illness related to the widespread emerging immunity to Cryptosporidium. In a follow-up investigation of 33 individuals with onsets of laboratory-confirmed Cryptosporidium infection during May 1 to June 23 and neighborhood and household control subjects, we found waterborne transmission was not associated with these late illnesses. Risk factors for postoutbreak illness included immune compromise and living in a household with one or more children less than 5 years old suggesting person-to-person transmission. Despite biases introduced by the greater likelihood that immune-compromised individuals would be tested during the postoutbreak period, this study provided reassurance that new Cryptosporidium infections were not occurring as a result of drinking MWW water.23

Swimming Pool-Related Outbreaks

We were mindful that chains of transmission could be sustained as Cryptosporidium can be transmitted from person to person or an infected individual can contaminate a new, more limited water supply. Shortly after the Milwaukee outbreak and throughout the summer of 1993, outbreaks of Cryptosporidium infection among users of public or hotel pools were reported from several counties.24,25 The earliest such outbreak north of Milwaukee could be directly attributed to the Milwaukee outbreak; however, the phenomenon continued through the summer as additional pools in Wisconsin and elsewhere were seeded with the chlorine-resistant Cryptosporidium oocysts and pool users became infected and then contributors to multiple generations of illness. The optimal concentration of chlorine in swimming pools was not sufficiently high to kill the oocysts before an unsuspecting swimmer ingested a small volume of pool water.

Genotyping

Speciation and further characterization of the outbreak strain of Cryptosporidium were another objective of this investigation. To accomplish this, three volunteers who had been ill during the outbreak and had laboratory-confirmed infections were recruited to donate large volumes of stool. To our disappointment, after shipment to and maintenance of these specimens at a CDC laboratory, the Cryptosporidium in these specimens became nonviable. These specimens still proved useful because they were ultimately (more than a year later) used to test the Cryptosporidium parvum DNA to determine genotype of the strains. An important goal of the genotyping was to determine the source of the Cryptosporidium. Although limited, genotype information provides clues regarding whether the source strain was human or bovine. In Milwaukee, the source strain was suggested to be human based on the genotype testing26;however, I believe the extremely small number of specimens from the Milwaukee outbreak used in the genotype investigation and perhaps the sources of stool specimens tested make these observations more conjectural. The suggestion of a human source is supported by the limited data available, but I do not believe that it is proven. It is conceivable that multiple genotypes were circulating.

Serologic Studies

We were very interested in the prospect of testing banked sera from ill and well individuals during this outbreak to assess immune responses when a reliable assay became available at some unspecified time in the future. More than 5 years after the outbreak, plasma samples obtained from 553 Milwaukee children aged 6 months to 12 years old who were being tested for lead exposure during the time of the outbreak were tested at CDC laboratories for Cryptosporidium antibodies27 using an enzyme-linked immunosorbent assay to detect IgG to two immunodominant antigens of C. parvum. Samples were obtained during five distinct periods in a 5-week interval between March 1993 and May 1993. Each child was bled once, but the prevalence of antibody among children bled during each distinct period could be measured. Data were also available to compare antibody prevalence by region of residence in Milwaukee. The prevalence of antibodies increased from about 16% to 85% during the 5-week interval among children from southern Milwaukee ZIP codes and from about 21% to 45% during the 5-week interval among children from northern Milwaukee ZIP codes. Median antibody reactivity was also substantially greater among the children from southern ZIP codes compared with northern ZIP codes. In addition to corroborating our earlier epidemiologic observations, these data suggest that C. parvum infection was more widespread than initially measured based on our RDD studies.27 Because the samples were obtained without accompanying clinical information, it is not known what the proportions were of subclinical or asymptomatic infection.27

Theories on Why This Massive Outbreak Occurred: A Perfect Storm2,3

Outbreaks of great magnitude are generally not easily explained, nor was this one; however, a confluence of likely contributing factors and events during March and April 1993 was exceptionally compelling and provided unique opportunities to dramatically amplify case occurrence.

The South Plant Intake Grid Location and Unusual Weather Conditions

The placement of the South Plant intake grid was unfortunate at best. As previously noted, the three rivers flowing through Milwaukee County join together to flow into a bay and harbor along the Lake Michigan shore. This harbor is protected by a large breakfront. The breakfront has three large outer gaps through which water can flow in and out of the harbor. As previously noted, the ambient current in the harbor contained within the large breakfront was southerly. Water flowing out the south fair weather gap typically flowed directly toward the South Plant intake grid. During and after rainy and other high-flow periods when increases in runoff and storm sewer overflow in the rivers are noted, the discharges of dirt and particulate matter into the harbor can be striking (Figure 14-6). This was occurring during a prolonged, intense high flow period compounded by unusual weather conditions:

• • There was a high snow pack during the winter that melted rapidly while the frost line was still high. The runoff was excessive, and its impact was likely compounded by runoff manure spread onto the snow by farm workers intent on getting rid of it and concurrently enriching their fields.

• • There was record setting rainfall in March and early April that contributed to runoff within the river watersheds. Within the City of Milwaukee, this rainfall contributed to widespread storm sewer overflows resulting in vast volumes of sewage that could be disinfected but otherwise bypassed treatment in the Milwaukee Metropolitan Sewage District facility before it drained into the bay that was protected by the breakfront.

• • The wind conditions and direction were highly unusual. There was a prolonged period of northeasterly winds occurring in late March and early April. During this time, there was likely accentuation of the southerly flow of water within the breakfront with more flow of water in the bay out the south fair weather gap, which likely amplified plumes flowing directly toward the South Plant intake grid. Prevailing winds in other directions would have facilitated their dispersal.

• • These conditions would have enhanced the transport of Cryptosporidium oocysts toward the South Plant water intake.

Cross Connection Between a Sanitary Sewer and a Storm Sewer

In early March 1993, during the construction of soccer fields near the Menomonee River and close to downtown Milwaukee, a linkage of a storm sewer draining the fields with a central main sewer was being created when a large volume of impacted contents was noted in the main sewer. The contents included animal material, particularly animal intestines. In addition, there were many rubber rings that were used to prevent spillage of enteric contents when bovines were slaughtered and eviscerated. Further investigation by local officials resulted in the detection of a cross connection between an abattoir kill floor sanitary sewer and the storm sewer. After elimination of this cross connection and correction of sewage flow, the storm sewer was cleaned during a multiple week process. The cross connection was estimated to have existed for years. Material from the abattoir kill floor that impacted in the storm sewer or flowed directly to the river via the cross connection would now travel through the sanitary sewer and be treated at the sewage treatment plant or be disinfected and bypass treatment during periods of high flow (heavy rains). It is not known whether Cryptosporidium oocysts were released directly through the storm sewer into the Menomonee River during or preceding these events or whether a bolus of oocysts properly flowing through the sanitary sewer during cleanup procedures may have bypassed sewage treatment during a period of high flow.

Change in Coagulant

The change in coagulant routinely used in both the North and South Plants 6 months before the outbreak and the difficulty in coagulant dosing during a period of abnormal turbidity to bring the turbidity under control were factors. Furthermore, the Milwaukee Water Works interpretation of finished water turbidity as an aesthetic indicator may have contributed to late recognition of the difficulties (or later recognition of the importance of the difficulties) they were having with water treatment. The occurrence of this outbreak provided a focus on the importance of turbidity as a water quality indicator.

Human Amplification

The amplification of the burden of Cryptosporidium oocysts among residents of the greater Milwaukee area was a critical factor. In humans, the infectious dose of C. parvum that can result in illness is small. Among healthy adult volunteers with no serologic evidence of past infection with C. parvum, the median infectious dose has been determined to be 132 oocysts28;however, billions of oocysts are excreted each day in the stools of symptomatic individuals, and excretion of oocysts typically continues for several weeks after symptoms resolve. Shedding can be more prolonged in those who are immunocompromised. Additionally, the oocysts can remain infective in moist environments for 2 to 6 months. Thus, the opportunity for infection in this outbreak was extraordinarily, perhaps incomprehensibly, high. This was compounded by the aforementioned rainy, high-flow conditions when storm sewers typically overflowed and, other than disinfection, most sewage was not treated before discharge into the harbor. This all contributed to a sustained vicious cycle of oocyst and illness amplification.

Limited Testing Prior to Outbreak Recognition

There was some delay in recognition of this outbreak. Based on the sporadic occurrences of diarrhea illnesses weeks prior to the outbreak peak, the outbreak probably began in early March but was initially recognized on April 5. This delay can in part be explained by understanding that testing stool specimens for Cryptosporidium was not commonplace in medical practice at the time of this outbreak. Typically, testing for Cryptosporidium infection was requested in individuals with HIV infection who were experiencing diarrhea. The number of tests requested per month was relatively small, and most healthcare providers would not readily have suspected this diagnosis in healthy patients with brief or even prolonged diarrhea disease. If they requested obtaining a stool specimen to test for parasites, Cryptosporidium was not a pathogen routinely looked for. Tests for Cryptosporidium required an additional requisition. Indeed, a major factor in the limited number of test requisitions was the complexity caused by the additional pelleting and staining procedures and the attendant greater expense of testing. These factors may have contributed to this insufficient testing demand with consequent delay in ascertainment of an outbreak.

POLICY IMPACTS AND INFRASTRUCTURE IMPROVEMENTS

Standards

The massive Milwaukee cryptosporidiosis outbreak was historic. It is the largest known outbreak of waterborne disease ever documented in the United States and possibly in the developed world. The magnitude of this outbreak coupled with the direct association of illness with a municipal water treatment plant that was operating within existing state and federal regulatory standards at that time had an immediate powerful impact that focused widespread public concern on the quality of drinking water, underscored the need to have far more stringent regulatory standards, and raised awareness of cryptosporidiosis as a diarrheal illness. A specific need to create specific regulatory standards for Cryptosporidium in drinking water was apparent.

The Milwaukee Water Works expeditiously adopted standards and quality indices that were maximally stringent and far more stringent than state or federal standards. Wisconsin state standards were also strengthened and clarified. The antiquated and confusing use of monthly or other interval averages in turbidity as a threshold standard was eliminated.

The EPA was in the process of generating and releasing its Information Collection Rule, which provided water treatment facilities with a list of laboratory tests and measures that would be needed to ultimately generate a final set of standards. The occurrence of this massive outbreak in the midst of this process had bearing on the content of the Information Collection Rule.

Infrastructure Improvements

Virtually immediately after the boil-water order was invoked and with the assistance of many experts, officials began planning to make extensive changes to the Milwaukee water treatment infrastructure. Nearer term improvements included the revision of filters and filter beds (removal of all media, sandblasting all filter beds, structural changes, installation of new media) and installation of particle counters and automatic alarms for each filter (when very low threshold turbidity or particle counts are exceeded, there is automatic filter shutdown and diversion of affected water so that it did not enter the clear well). Longer term projects that took several years to complete included the construction of ozonation facilities at each treatment plant. Because Cryptosporidium oocysts are highly chlorine resistant, standard disinfection procedures are not sufficient to inactivate them. Ozonation as an adjunct to disinfection is effective because it inactivates oocysts by disrupting their cell walls and it is effective in disrupting other microorganisms found in raw water. The North and South Plants are now state of the art facilities and have been visited by utility related personnel from many states and countries.

CONCLUDING REFLECTIONS

The Milwaukee outbreak, the results of our collective investigations, and the local and state responses were of great interest to drinking water related regulators, utility (public and private) administrators, and municipalities throughout the United States, Canada, Europe, Japan, Australia, and elsewhere. Although fear of such a large outbreak is a great motivator, the significant and tangible improvements in water treatment authorized by the City of Milwaukee administration has had substantial impact on voluntary and required improvements in water processing and in water quality in communities in many states and countries. The transparency of these processes has improved as have communications between the water treatment and public health communities. These efforts were certainly apparent throughout Wisconsin when communities recognized that Lake Michigan turbidity events were not limited to Milwaukee (Wisconsin DPH, unpublished data). Clearly, we learned the importance of the need for improved surveillance for infections that may result from waterborne pathogens and for aberrations in water quality. This requires cooperation, coordination, and communication among public health agency and water utility staffs.

The high level of civic resolve to focus on solutions to rapidly control this massive outbreak and prevent even teeny ones from occurring in the future was extraordinary. As a native of Milwaukee County, I was deeply moved and very proud of this enormous yet efficient effort. Our team was very impressed with the exceptional cooperation received from local government, businesses large and small, and private citizens. We could not have accomplished as much as we did without the degree of cooperation and collaboration that we experienced. As expected, the greater Milwaukee community did this with a great sense of humor, at times magnificently sharp and self-deprecating, and civic pride. In one photograph featured in the newspaper at that time, an illuminated sign from a pharmacy-food mart says simply, “We have Immodium AD. Open Easter.” In another photo of a man walking to the front of a bar, the sign overhead says simply, “Mexican vacation? Why bother. Stay here.”

Without question, application of the lessons learned from this outbreak has had a sustained impact on preventing subsequent outbreaks in the United States and other developed countries.

REFERENCES

1.

Hayes EB, Matte TD, O’Brien TR, et al. Large community outbreak of cryptosporidiosis due to contamination of a filtered public water supply. N Engl J Med 1989;320:1372–1376.

2.

Mac Kenzie WR, Hoxie NJ, Proctor ME, et al. A massive outbreak in Milwaukee of Cryptosporidium infection transmitted through the public water supply. N Engl J Med 1994;331:161–167.

3.

Addiss DG, Mac Kenzie WR, Hoxie NJ, et al. Epidemiologic features and implications of the Milwaukee cryptosporidiosis outbreak. In Betts WB, Casemore D, Fricker C, Smith H, Watkins J, eds. Protozoan Parasites and Water. Cambridge, England: The Royal Society of Chemistry, 1995:19–25.

4.

Marchione M. Detective story: Skill, luck come to fore. Milwaukee J 1993;April 11:1, 12.

5.

Nime FA, Burek JD, Page DL, Holscher MA, Yardley JH. Acute enterocolitis in a human being infected with the protozoan Cryptosporidium. Gastroenterology 1976;70:592–598.

6.

Jokipii L, Jokipii AMM. Timing of symptoms and oocyst excretion in human cryptosporidiosis. N Engl J Med 1986;315:1643–1647.

7.

Wolfson JS, Richter JM, Waldron MA, Weber DJ, McCarthy DM, Hopkins CC, Cryptosporidiosis in immunocompetent patients. N Engl J Med 1985;312:1278–1282.

8.

Meisel JL, Perera DR, Meligro C, Rubin CE. Overwhelming watery diarrhea associated with cryptosporidium in an immunosuppressed patient. Gastroenterology 1976;70:1156–1160.

9.

Current WL, Reese NC, Ernst JV, Bailey WS, Heyman MB, Weinstein WM. Human cryptosporidiosis in immunocompetent and immunodeficient persons: studies of an outbreak and experimental transmission. N Engl J Med 1983; 308:1252–1257.

10.

D’Antonio RG, Winn RE, Taylor JP, et al. A waterborne outbreak of cryptosporidiosis in normal hosts. Ann Intern Med 1985;103:886–888.

11.

Gallagher MM, Herndon JL, Nims IJ, Sterling CR, Grabowski DJ, Hull HF. Cryptosporidiosis and surface water. Am J Public Health 1989;79:39–42.

12.

Richardson AJ, Frankenberg RA, Buck AC, et al. An outbreak of waterborne cryptosporidiosis in Swindon and Oxfordshire. Epidemiol Infect 1991;107:485–495.

13.

Joseph C, Hamilton G, O’Connor M, et al. Cryptosporidiosis in the Isle of Thanet: an outbreak associated with local drinking water. Epidemiol Infect 1991;107:509–519.

14.

Leland D, McAnulty J, Keene W, Sterens G. A cryptosporidiosis outbreak in a filtered-water supply. J Am Water Works Assoc 1993;85:34–42.

15.

Proctor ME, Blair KA, Davis JP. Surveillance data for water-borne illness: an assessment following a massive water-borne outbreak of Cryptosporidium infection. Epidemiol Infect 1998;20:43–54.

16.

Corso PS, Kramer MH, Blair KA, Addiss DG, Davis JP, Haddix AC. The cost of illness associated with the massive waterborne outbreak of Cryptosporidium infections in 1993 in Milwaukee, Wisconsin. Emerg Infect Dis 2003;9:426–431.

17.

Mac Kenzie WR, Schell WL, Blair KA, et al. Massive outbreak of waterborne Cryptosporidium infection in Milwaukee, Wisconsin: Recurrence of illness and risk of secondary transmission. Clin Infect Dis 1995;21:57–62.

18.

Cordell RL, Thor PM, Addiss DG, et al. Impact of a massive waterborne cryptosporidiosis outbreak on child care facilities in Metropolitan Milwaukee, Wisconsin. Pediatr Infect Dis J 1997;16:639–644.

19.

Cicirello HG, Kehl KS, Addiss DG, et al. Cryptosporidiosis in children during a massive waterborne outbreak, Milwaukee, Wisconsin: clinical, laboratory, and epidemiologic findings. Epidemiol Infect 1997; 119:53–60.

20.

Frisby HR, Addiss DG, Reiser WJ, et al. Clinical and epidemiologic features of a massive waterborne outbreak of cryptosporidiosis among persons with human immunodeficiency virus (HIV) infection. J Acq Immun Defic Syndr Hum Retrovirol 1997;16:367–373.

21.

Hoxie NJ, Davis JP, Vergeront JM, Nashold RD, Blair KA. Cryptosporidiosis-associated mortality following a massive waterborne outbreak in Milwaukee, Wisconsin. Am J Public Health 1997;87:2032–2038.

22.

Addiss DG, Pond RS, Remshak M, Juranek D, Stokes S, Davis JP. Reduction of risk of watery diarrhea with point-of-use water filters during a massive outbreak of water-borne Cryptosporidium infection in Milwaukee, 1993. Am J Trop Med Hyg 1996;54:549–553.

23.

Osewe P, Addiss DG, Blair KA, Hightower A, Kamb ML, Davis JP. Cryptosporidiosis in Wisconsin: a case-control study of post-outbreak transmission. Epidemiol Infect 1996;117:297–304.

24.

Mac Kenzie WR, Kazmierczak JJ, Davis JP. Cryptosporidiosis associated with a resort swimming pool. Epidemiol Infect 1995; 115:545–553.

25.

Cryptosporidium infections associated with swimming pools—Dane County, Wisconsin, 1993. MMWR 1994;43: 561–563.

26.

Sulaiman IM, Xiao L, Yang C, et al. Differentiating human from animal solates of Cryptosporidium parvum. Emerg Infect Dis 1998;4:681–685.

27.

McDonald AC, Mac Kenzie WR, Addiss DG, et al. Cryptosporidium parvum-specific antibody responses among children residing in Milwaukee during the 1993 waterborne outbreak. J Infect Dis 2001;183:1373–1379.

28.

Du Pont HL, Chappell, Sterling CR, Okhuysen PC, Rose JB, Jakubowski. The infectivity of Cryptosporidium parvum in healthy volunteers. N Engl J Med 1995;332:855–859.