We apportioned waterborne diseases to recreational, drinking, or NRND water categories as previously described in a structured expert judgment study (19) (Appendix Table 1). Recreational water is defined as water used for recreational activities, such as aquatic venues or natural bodies of water. Drinking water is used primarily for drinking but can include other domestic uses, such as washing or showering. Drinking water can come from a public water system, private well, or commercially bottled sources. NRND water is used for purposes other than recreation or consumption, such as agriculture, manufacturing or cooling equipment, or medical treatment; this category also includes backcountry streams and flood waters. Examples of diseases transmitted by NRND water include Legionnaires’ disease associated with a cooling tower, giardiasis transmitted by drinking untreated water from a stream, and Vibrio spp. wound infections after wading through flood waters.

We estimated the total number of cases for 17 diseases or syndromes in 2014 in the United States (Appendix): acute otitis externa, campylobacteriosis, cryptosporidiosis, giardiasis, Legionnaires’ disease, NTM infection, norovirus infection, Pseudomonas pneumonia, Pseudomonas septicemia, salmonellosis (nontyphoidal), Shiga toxin–producing Escherichia coli infection with an O157 or non-O157 serogroup, shigellosis, and vibriosis caused by Vibrio alginolyticus, V. parahaemolyticus, V. vulnificus, and other non–V. cholerae spp. (1) (Appendix Table 2). We divided total waterborne disease estimates previously reported (1) by structured expert judgment study estimates (19) to produce 2014 burden estimates for recreational, drinking, and NRND water exposure (Appendix Table 3). We quantified direct healthcare costs of treat-and-release emergency department (ED) visits and hospitalizations, measured by insurers and out-of-pocket payments, by using MarketScan data, as previously described (1).

We used anonymized administrative surveillance and literature data from 2000–2015 and based all estimates on the 2014 US population (318.6 million persons); at the time of analysis, 2014 was the most recent year for which all data were available. Data sources have been described previously (1). We estimated values by using probabilistic multiplier models (Appendix); each model input had an associated uncertainty estimate represented by a distribution of plausible values (Appendix Figure). Uncertainty in final estimates for each type of water exposure was a cumulative effect, incorporating the uncertainties of each model input (1,19). We obtained count distributions by using 100,000 iterations to generate point estimates of means and corresponding 95% credible intervals (CrIs; the 2.5th–97.5th percentiles of empirical distribution). We generated all-disease totals for each outcome by sampling from the distributions generated for each disease. We used SAS version 9.4 (SAS Institute, https://www.sas.com) and R version 3.5.1 (The R Project for Statistical Computing, https://www.r-project.org) for analyses.

Of the estimated 7.15 million infectious waterborne illnesses in 2014 in the United States, 78% of illnesses were attributed to recreational water, 16% to drinking water, and 6% to NRND water. Drinking water exposure caused 40% of hospitalizations and 50% of deaths, recreational water exposure 36% of hospitalizations and 20% of deaths, and NRND exposure 24% of hospitalizations and 30% of deaths. ED visits and hospitalizations from exposure to drinking or recreational water exposure cost >$2 billion annually.

Acute otitis externa, which can be acutely painful and cause itchiness, drainage, and swelling, was the most common recreational water–associated illness. Risks for acute otitis externa have been correlated with water quality measures, levels of Pseudomonas and other pathogens in water, and increased bather load in recreational water venues in some studies but not others (20–23). Water can displace the protective coating of cerumen (ear wax) in the ear canal, leaving the outer ear more vulnerable to infection by endogenous flora. Whether otitis externa is caused by endogenous flora or contamination introduced by water remains unclear. In the structured expert judgment study (19), experts were asked to consider this question while estimating the proportion of otitis externa transmitted through water versus other pathways. They estimated 81% (95% uncertainty interval 67%–95%) of Pseudomonas otitis externa was transmitted through water (19). In response to the absence of national standards for recreational water venues, the Centers for Disease Control and Prevention (CDC) participated in the development and updating of the Model Aquatic Health Code (MAHC) (24). For treated water venues, the MAHC includes guidance for health authorities and aquatics sector to minimize risks for acute otitis externa and other illnesses and injuries (24,25). For both treated and untreated venues, risk for acute otitis externa can be minimized by keeping ears as dry as possible when swimming and ensuring ears are dry after swimming. Recreational swimmers and parents of young swimmers can educate themselves about steps to minimize the risks for infection from enteric pathogens at trusted sites, such as the CDC’s healthy swimming webpage (26). Signage at treated and untreated recreational water venues (12) and adoption of CDC’s MAHC for treated public recreational water venues can further improve prevention. The number of illnesses and ED visits for acute otitis externa caused by drinking water exposure might seem counterintuitive, but drinking water was defined as water used for drinking, bathing, or showering.

Cryptosporidium and Legionella spp. infections in humans were identified in the late 1970s, and outbreaks associated with those pathogens in treated recreational water venues were identified soon after. In subsequent decades, reported incidences of treated recreational water–associated outbreaks increased substantially; Cryptosporidium, Legionella, and Pseudomonas spp. were the predominant etiologies (27). P. aeruginosa is a major cause of acute otitis externa and folliculitis or hot tub rash in addition to pneumonia and septicemia (28). Infectious Cryptosporidium sp. oocysts have tough outer shells that provide extreme chlorine tolerance, whereas Legionella and Pseudomonas spp. thrive in biofilms, where they are protected from chlorine inactivation.

Reported infectious disease outbreaks associated with public drinking water systems have decreased since the 1970s, likely because of federal management and treatment regulations that address enteric pathogens (3). However, outbreaks associated with individual water systems, such as private wells, and premise plumbing deficiencies have not decreased (3). Our analysis affirms that most hospitalizations and deaths associated with drinking water exposure were linked to biofilm-associated pathogens, not to pathogens causing enteric disease. Legionella bacteria are biofilm-associated pathogens and have become the most frequently reported cause of disease outbreaks associated with drinking water. Our analysis revealed that NTM infections were the leading cause of drinking water–associated illnesses, ED visits, hospitalizations, and deaths; NTM infections were the most common cause of death associated with all 3 water exposure routes. However, NTM infections are not nationally notifiable diseases; thus, cases are not consistently reported to public health authorities and outbreaks might remain undetected. Improving NTM illness reporting could lead to greater outbreak detection and inform disease prevention strategies for water systems. The Council of State and Territorial Epidemiologists has a standardized case definition for extrapulmonary NTM infections (opportunistic infections of wounds, soft tissue, or joints) to increase reportability, ensure consistency in reporting, and help identify outbreaks (29). Thus far, the state of Oregon has adopted this definition for their surveillance program (30), and other sites are piloting NTM surveillance programs (31).

NRND water includes water used for agriculture, such as for irrigation or livestock; industry, such as manufacturing or cooling equipment; medical procedures, such as medical devices, washing surgical tools and equipment, and hydrotherapy; and backcountry streams or floodwaters (19). NRND water made up 87% of the 322 billion gallons of water used each day in the United States in 2015; of that volume, 2 billion gallons were used for livestock, 118 billion gallons for irrigation, 7.55 billion gallons for aquaculture, and 133 billion gallons for thermoelectric power (6). NRND water is associated with a small portion of reported waterborne illnesses but is responsible for one quarter of waterborne disease–associated hospitalizations and one third of associated deaths.

Waterborne pathogens, such as NTM, Pseudomonas spp., and Legionella spp., occur naturally in freshwater and can colonize other environments, particularly large complex equipment that is not properly maintained. Those pathogens are aerosolized in water droplets produced during use, which can be inhaled. Water management programs are recommended to minimize growth and spread of pathogens in engineered systems (32–34). Resources can be found on the CDC Environmental Health Services safe water program webpage (35), guide for developing a water management program to reduce Legionella growth and spread in buildings (36), Legionella control toolkit (37), and guide for reducing risk from water (38). Water management programs in healthcare facilities are also critical for protecting vulnerable patient populations, staff, and visitors and might require specific considerations for a wide range of biofilm-associated pathogens, including NTM, Pseudomonas, and Legionella spp. (39–42). Interdisciplinary collaboration is needed for planning future management of complex water systems and investigating illnesses and outbreaks.

Reducing risks for illness from biofilms in water systems is difficult. Biofilms can persist despite standard water treatment processes. Biofilms are complex ecosystems; control is complicated by structural issues (e.g., pipe characteristics), system operational issues (e.g., water age, temperature, and residual disinfectants), and water end-user behaviors. Changes designed to reduce prevalence of 1 microbial constituent can sometimes produce unintended consequences, such as proliferation of NTM, if a comprehensive approach to biofilm control is not considered (43). NTM are persistent pathogens; despite standard treatment, NTM species have been recovered from surface water treatment plants, biofilms in sand filters, ozonated water, biofilms from granular activated carbon filters, and activated carbon-filtered water. Even municipal systems that incorporate full treatment chains, including 2 filtration and 2 disinfection stages, might not effectively prevent NTM growth (44). Because climate change continues to create extreme weather events and increase water temperatures, more stress will be placed on aging water infrastructure; biofilm-associated pathogens will continue to proliferate (5), highlighting the need for biofilm-focused control programs.

Prevention of biofilm-associated organisms is an emerging field, and more scientific evidence is needed to determine best practices for public water systems and premise plumbing and identify effective prevention strategies for homeowners and building managers. Persons can reduce their risk for illness from a biofilm-associated pathogen by flushing rarely used faucets and showerheads, flushing water heaters, and maintaining specific water heater temperatures. Resources for risk reduction can be found on CDC’s Preventing Waterborne Germs at Home webpage (45).

Most previous waterborne infectious disease estimates are not directly comparable with results from this study because of differing methods used to generate estimates. Many studies have attempted to estimate a burden for 1 disease or aspect of drinking or recreational water exposure. Two studies from a workshop convened by CDC and the Environmental Protection Agency used 2 different methods to estimate the number of cases of acute gastrointestinal illness in public drinking water systems only. Those studies estimated 4–16 million acute gastrointestinal illness cases per year (13,16). A 2008 study estimated all illnesses associated with public drinking water systems at 19 million cases per year (14). Previous publications have used cohort studies to estimate gastrointestinal illness from water recreational activities and percentages of economic burden from recreational water exposure (17,46). In Canada, rates for waterborne disease per 100,000 persons were estimated by using literature, clinical input, and administrative data (47). Our estimates differ from previous work because we focused on specific pathogens, including those causing nongastrointestinal diseases, and we provide estimates for recreational water, drinking water, and NRND water exposure routes.

The first limitation of our study is that the estimates relied on a series of multipliers. Although we attempted to account for uncertainty in the multiplier estimates, biased multipliers will yield biased estimates. We used attribution estimates for proportions of disease caused by each water exposure route, derived from a structured expert judgment study. Structured expert judgment is used when data are sparse and relies on expert opinion. Statistical techniques are used to combine expert opinions into a single estimate with an uncertainty interval. Various weighting techniques can be used to increase accuracy or informativeness (narrowness of the uncertainty interval), but tradeoffs between accuracy and informativeness occur. For some individual estimates, including norovirus infection, NTM infection, and acute otitis externa, uncertainty intervals were wide, reflecting uncertainty about the proportion of disease transmitted via water. Second, as with previous waterborne disease burden estimates (1), we included estimates of 17 infectious diseases that had adequate data available. Other waterborne infectious diseases were excluded, such as Pseudomonas folliculitis or viral gastroenteritis not caused by norovirus. We also excluded noninfectious diseases caused by exposure to contaminated water, such as from lead or harmful algae-derived toxins, and did not evaluate long-term chronic effects of infectious or noninfectious diseases. In addition, we excluded foodborne illnesses, such as salmonellosis from lettuce or norovirus disease from shellfish, where contaminated water might have been responsible. Third, this study used data from hospital and ED billing databases. Billing records are not medical records and might reflect diagnoses that result in reimbursement from insurance companies rather than true assessments by clinicians. Fourth, those billing databases use International Classification of Diseases, 9th Revision, Clinical Modification, coding to classify diagnoses; some illnesses might not perfectly match existing codes. Fifth, we provide estimates for 2014, and illness burdens, healthcare costs, and clinical practices have likely changed since then. Sixth, we assumed that severity of illness was independent of exposure route. For example, if salmonellosis was more severe when transmitted by drinking water than by recreational water, the multipliers used in calculating estimates did not account for this. Seventh, many diseases in this analysis might be more severe in persons with compromised immunity, but we did not attempt to apportion the burden of illness between waterborne illnesses and immune status. Finally, reliably distinguishing between illnesses linked to premise plumbing water exposure and distribution system exposure was difficult, which was reflected in 95% CrI widths for drinking water and NRND estimates for some pathogens, including NTM. More research is needed in this area because federal regulations for distribution systems differ from building codes and state and local laws that regulate premise plumbing.

In conclusion, quantifying the burden of illness linked to drinking, recreational, and NRND water is an essential factor for addressing biofilm-associated pathogens, because prevention measures and vital partners differ for each exposure route. Estimating disease burden by water exposure route will help direct prevention activities and prioritize limited resources. Measures for preventing illnesses from recreational water use, such as those in the MAHC for treated recreational water venues, could reduce disease burden associated with recreational water exposure. Incorporating water management programs and control measures for systems that are common sources of biofilm pathogen exposure into building and public health codes is crucial for reducing risks for biofilm-associated pathogen exposure in drinking water and NRND water systems (34). In addition, improved surveillance and reporting of biofilm-associated illnesses could lead to the timely detection and investigation of outbreaks and inform disease prevention activities. Interdisciplinary partnerships among public health agencies and industries that use NRND water are pivotal to reduce incidences of waterborne diseases. For drinking water and NRND water systems, addressing factors that promote biofilm growth (stagnation, temperature, or lack of disinfection) could reduce growth of biofilm-associated pathogens. Facilities that serve persons at increased risk for infections, such as healthcare facilities, should review water system design and operations to decrease risks from biofilm-associated pathogens. Additional research on the roles of pipe material, water chemistry, and optimal water distribution system management could provide insights into improving prevention actions. As climate change continues to place increased stress on water systems, improved prevention and maintenance strategies should be developed to prevent illness from all types of water exposure.