Mortality from selected diseases that can be transmitted by water – United States, 2003–2009

In the United States, mortality from classic enteric waterborne infections such as cholera and typhoid fever declined greatly during the 20th century following the advent of municipal drinking water disinfection and filtration and concurrent improvements in wastewater treatment (Cutler & Miller 2005). However, deaths from waterborne disease, as documented through waterborne disease and outbreak surveillance, continue to occur. In the Centers for Disease Control and Prevention's (CDC) Waterborne Disease and Outbreak Surveillance System (WBDOSS), waterborne transmission (for both individual cases of disease and outbreaks) includes ingestion, inhalation, or contact, and the type of water can include drinking water, recreational water, or environmental water (a broad term used to describe a variety of other exposures including consuming water directly from backcountry or wilderness streams, or inhaling aerosols from cooling towers and ornamental fountains) (Beer et al. 2015b). The scope of pathogens associated with waterborne disease and outbreaks is broad. In recent years, deaths have been reported as part of outbreaks caused by Legionella (Legionnaires' disease), hepatitis A, Salmonella serotype Typhimurium, and Vibrio spp. (Beer et al. 2015a, 2015b; CDC 2015b; Hlavsa et al. 2015). Fatal cases of other waterborne infections have also been documented by WBDOSS, including individual cases of Naegleria fowleri infection, healthcare-associated outbreaks of nontuberculous mycobacteria (NTM), or Pseudomonas infections associated with water (Williams et al. 2013; Capewell et al. 2015; CDC 2016).

Waterborne disease outbreaks and individual, sporadic cases (i.e., those that are not outbreak associated) of waterborne disease are reported by states to CDC through a variety of active or passive surveillance systems, such as WBDOSS (outbreaks), the Foodborne Diseases Active Surveillance Network (FoodNet) (sporadic and outbreak-associated cases), and the National Notifiable Diseases Surveillance System (NNDSS) (sporadic and outbreak-associated cases) (Table 1) (Yoder et al. 2010; CDC 2011, 2012, 2014; Adams et al. 2014; Crim et al. 2015). While outbreak and case surveillance reporting can reflect high-level trends in the epidemiology of waterborne disease in the United States, surveillance data do not adequately capture all illnesses or deaths from the large scope of waterborne diseases. Waterborne disease and outbreaks are investigated by local and state public health officials and voluntarily reported to the CDC. Prioritization and resources for waterborne disease and outbreak surveillance activities vary by state and by pathogen, thus these data are subject to under-reporting and can be inconsistent across jurisdictions. Additionally, these surveillance systems provide limited information on deaths, the information gathered on the case-patients can differ by pathogen, and national case surveillance systems do not exist for all waterborne pathogens of interest.

Administrative data are an alternative to surveillance data that can be used to summarize the national numbers of deaths for diseases, even those that occur rarely or are not captured in outbreak or case surveillance systems (Jhung & Banerjee 2009; Pine et al. 2012). To our knowledge, national numbers of deaths from pathogens that can be transmitted by water have not been previously summarized; additionally, there has been no comparison of national mortality from infectious waterborne pathogens spread by classical fecal–oral routes to environmental pathogens, such as Legionella and NTM, that are transmitted by other routes, including contact with and inhalation of aerosols. To better understand the recent epidemiology of waterborne disease in the United States, we used administrative data to determine which pathogens that can be transmitted by water were most associated with mortality on a national scale, to describe the characteristics of persons that died, and to tabulate the numbers of annual deaths (transmitted through all routes) documented nationally from these pathogens.

For this analysis, 13 pathogens or syndromes frequently implicated in waterborne disease and outbreak surveillance, reports, or case studies in the United States were selected for inclusion (Table 1). This included infections from pathogens that can be transmitted by any type of water and any exposure route (e.g., consumption of drinking water; or inhalation of aerosolized water from a cooling tower, fountain, or sink; direct contact with recreational water). Most of the pathogens are also transmitted in other ways in addition to water exposures. The 13 infections include seven with fecal–oral transmission and six transmitted via other routes (Table 1). The goal was to investigate mortality using a subset of infections with the best available data, thus this is not an exhaustive list of infections from pathogens that can be transmitted by water. Additionally, all deaths for each infection were included. No attempt was made to separate deaths from waterborne transmission of these pathogens from deaths due to transmission via other routes.

We analyzed data for 2003–2009 from the Agency for Healthcare Research and Quality's Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS), the National Center for Health Statistics' National Vital Statistics System (NVSS) death certificate records (Murphy et al. 2013), and selected disease surveillance systems (Yoder et al. 2010; CDC 2012, 2016) (Table 2).

The NIS provides de-identified, record-level data on hospitalizations, in-hospital deaths, and selected patient demographic characteristics. It is the largest all-payer database of US inpatient stays, and can be weighted to provide an estimate of all visits from US community hospitals. The NIS uses International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) to code clinical diagnoses. For the years 2003–2008, NIS included up to 15 diagnoses for each hospital stay; for 2009, NIS listed up to 25 diagnoses. We evaluated two outcome categories based on the NIS data: the principal diagnosis, i.e., the diagnosis deemed chiefly responsible for hospital admittance, and all-listed diagnosis, i.e., all diagnoses on the administrative record during the hospital stay.

NVSS death certificate data (Multiple Cause of Death files) include information on all deaths recorded in the United States, including place of death (e.g., home or hospital), age of individual at death, and up to 20 conditions that contributed to the death, coded using 4-digit International Classification of Diseases, 10th Revision (ICD-10) codes (CDC 2015a). We evaluated two outcomes derived from the death certificate data: the underlying cause of death, defined as the single cause that initiated a series of events that led directly to death; and any cause, assigned when one of the infections of interest was coded as an underlying or contributing cause.

Additional data from national disease-specific case surveillance systems were used to supplement NVSS and NIS data, where they were available. Data on deaths caused by Vibrio spp. were obtained from Cholera and Other Vibrio Illness Surveillance (COVIS) for the years 2003–2009 (CDC 2012). Data on infections with the free-living amebae Naegleria fowleri and Acanthamoeba spp. were obtained from CDC's Free-living Ameba Surveillance System (Yoder et al. 2010; CDC 2016). Balamuthia mandrillaris deaths, while collected in this system, were not included because this organism is believed to be primarily a soil pathogen rather than a waterborne pathogen (Visvesvara et al. 2007).

All analyses were performed using SAS 9.3 (SAS Institute Inc., Cary, NC). We calculated the average annual numbers of deaths associated with the selected waterborne infections using death certificate data. We then compared aggregate frequencies of in-hospital deaths for the selected waterborne infections identified in the two data sources over the seven-year study period. Weighted estimates of the average annual number of deaths, case fatality proportions, and patient characteristics including age and immune compromise status were obtained using the NIS all-listed diagnosis. Presence of an immune-compromising condition was identified using a standard list of ICD-9-CM codes (US AHRQ 2007).

In the death certificate data, using underlying and any cause of death, the infections associated with the most annual deaths were Pseudomonas pneumonia, NTM, Legionella (Legionnaires' disease), hepatitis A, and Salmonella; all other infections were associated with fewer than 15 deaths per year, on average (Table 2). Including any cause of death from the multiple cause of death data identified more deaths for all infections except free-living ameba infections.

Table 3 facilitates comparison between hospital discharge data and death certificate data for enumerating in-hospital deaths using the first-listed cause only (principal diagnosis for hospital discharge data, underlying cause for death certificates) and any diagnosis or cause of death listed in the respective records. Using a single diagnosis or cause of death only, the hospital data identified more deaths than the death certificates for all infections except infection with hepatitis A and NTM. Using the all-listed diagnosis or any causes of death, hospital data identified more deaths for all infections.

In the hospital discharge data, the annual number of deaths among patients ranged from 11 or fewer (i.e., below the limit of reporting for NIS data) for Campylobacter, Shigella, free-living ameba, and Vibrio spp., to 3,160 (95% CI 2,961–3,360) for Pseudomonas septicemia (Table 4). Among patients who died, median ages were lowest for Shigella (35 years), Cryptosporidium (39 years), free-living ameba (43 years), disseminated mycobacterium avium–intracellulare complex (MAC) (47 years), and Vibrio (54 years). For all other infections in this analysis, the median age of patients who died was 60 years and over. The highest proportions of patients with documented immune-compromising conditions were among patients who died with Cryptosporidium (86.2%), MAC (68.8%), Giardia (56.2%), and NTM (49.5%). Other conditions with more than a quarter of decedents with documented immune-compromising conditions were Campylobacter, Shigella, Legionnaires' disease, and Pseudomonas septicemia. As a percentage of all hospitalized cases (not shown), case fatality proportions were highest for free-living ameba infections (46.3%), Pseudomonas septicemia (17.3%), and Legionnaires' disease (7.9%).

Using data from disease-specific surveillance, during 2003–2009, an average of four deaths per year from infections with the free-living amebae Naegleria fowleri and Acanthamoeba spp. were reported to CDC (Table 5). During 2003–2009, an average of 42 deaths per year from Vibrio infection were reported to COVIS; the majority of deaths were due to V. vulnificus.

Table 5 lists the average annual number of deaths for each infection, as determined by the two analytic approaches used to combine the data sources. Using analytic approach 1 based primarily on death certificate data, we documented that 2,444 deaths occurred annually due to the 13 infections. Of these, 180 (7%) were from the seven infections associated with fecal–oral transmission, and 2,264 (93%) were linked to the six infections with other transmission routes. Infections from three pathogens, Legionella, Pseudomonas, and NTM, were responsible for 2,204 (90%) of the deaths. Using analytic approach 2 that relied on hospital discharge data for in-hospital deaths, we documented 6,939 annual deaths from the 13 infections. Of these, 493 (7%) were associated with fecal–oral transmission and 6,446 (93%) were associated with other routes, including 6,301 (91%) of all deaths occurring with Legionella, Pseudomonas, and NTM.

By both analytic methods used to enumerate annual deaths, most deaths were from infections that result from environmental pathogens: Pseudomonas, Legionella, and NTM (90% when relying primarily on death certificate data and 91% by incorporating hospital discharge data). The administrative data used in this analysis do not contain information on transmission route, thus deaths transmitted by water and by other sources were counted. However, this analysis echoes recent findings from surveillance for waterborne disease outbreaks in the United States, in which Legionella has emerged as the leading cause of death in reported waterborne disease outbreaks associated with drinking water and environmental water sources (Beer et al. 2015a, 2015b).

The widespread use of water for industrial, healthcare, cooling, decoration, and other purposes in high-income, developed settings has created ecological niches for Legionella spp., as well as other pathogens including Pseudomonas spp. and NTM. These environmental pathogens do not rely on human hosts for their life cycle, and, as such, do not have a fecal source. They have emerged as important causes of non-gastrointestinal illness associated with water systems in the built environment because they commonly grow in biofilms and multiply in plumbing systems and other built environment water systems such as cooling towers (Cangelosi et al. 2005; Falkinham et al. 2015). However, few literature estimates exist for the proportion of transmission due to water for these three pathogens.

US Environmental Protection Agency (EPA) regulations that have reduced drinking water-associated outbreaks of gastroenteritis do not apply to the plumbing pipes within homes or other buildings (US EPA 2012). Within buildings, plumbing systems have more stagnant water, a lower disinfectant residual (i.e., lower or non-detectable levels of chlorine or other disinfectant), and often warmer temperatures, all of which contribute to biofilm development and growth (Thomas & Ashbolt 2011; Williams et al. 2013). Biofilms have a complex ecology, and can occur in various settings and throughout water systems (Thomas & Ashbolt 2011; Chowdhury 2012; Lu et al. 2015). Biofilm-associated pathogens can reach a susceptible host when they are sloughed or disrupted from these biofilms and released into piped water. Furthermore, biofilm-associated pathogens can grow in recreational and other environmental water sources such as poorly maintained hot tubs, cooling towers, and decorative fountains, and can give rise to cases of illness in a variety of settings when aerosols are produced (Beer et al. 2015b; Hlavsa et al. 2015). The emerging exposures to biofilms present in plumbing systems, cooling towers, fountains, and hot tubs may have contributed to the predominance of environmental pathogens as causes of death.

Environmental pathogens differ from most fecal–oral pathogens in that they result in longer hospitalizations and more severe outcomes for persons who are older or have immune-compromising conditions, (i.e., they are, to some extent, opportunistic pathogens) (Collier et al. 2012; Falkinham et al. 2015). The previously documented association of these infections with older age was borne out of our analysis of hospital discharge data; however, the proportion documented as immune-compromised was not consistently higher in these pathogens than the others in our analysis. For example, in our analysis, Giardia, which is not typically considered opportunistic, was associated with higher proportions of decedents having documented immune-compromising conditions when compared to decedents with Legionnaires' disease, which is considered opportunistic.

Our findings suggest that mortality from the selected pathogens in the United States is not primarily attributable to fecal–oral pathogens. This observation may reflect, in part, access to healthcare and treatments that make it relatively unlikely that gastroenteritis results in death (CDC 1999). In addition, low mortality from gastrointestinal infections may stem from successes of drinking water treatment regulations from the US EPA, which have reduced microbial contamination and, in consequence, waterborne disease transmission of gastrointestinal diseases resulting from the ingestion of fecally contaminated water (Craun et al. 2010; Beer et al. 2015a; CDC 2015b). EPA regulations do not apply to individuals served by private wells and other individual water systems, and a large number of waterborne disease outbreaks associated with drinking water occur in these systems; however, these serve a minority of households in the United States (Craun et al. 2010). Chlorine-sensitive gastrointestinal pathogens resulted in only eight drinking water-associated outbreaks during the 2011–2012 outbreak reporting cycle, although these accounted for over half of the total outbreak-cases (Beer et al. 2015a), demonstrating that these pathogens still have potential to drive high-impact waterborne disease outbreaks when failures in improved water treatment and management occur. Cryptosporidium, a highly chlorine-tolerant parasite, arose as an important cause of diarrhea and severe illness or death in the immuno-compromised during the 1980s and 1990s. Cryptosporidium infections have become especially linked with swimming in chlorinated pools, an activity that gained popularity with increased leisure time. Since 1988, when the first reported cryptosporidiosis outbreak from treated recreational water occurred, the annual number of these outbreaks has increased significantly, although recreational water-associated outbreaks rarely have fatal cases. In 2011–2012, 90 recreational water-associated outbreaks from 32 states and Puerto Rico were linked to 1,800 cases of illness and one death (Hlavsa et al. 2015).

This analysis did not directly document deaths due to waterborne transmission. All deaths for infections of each pathogen were included, and no attempt was made to separate deaths from waterborne transmission from deaths due to transmission via other routes. In part, this is because few or no estimates of the proportion of transmission attributable to water exist for several pathogens of interest, including Pseudomonas spp. and NTMs. However, for some infections studied, the attributable waterborne fraction is likely to be high (e.g., Legionella, Naegleria), while for others it is probably much lower (e.g., non-Typhoidal Salmonella, hepatitis A). Published attribution estimates for several of the fecal–oral pathogens included in our study show that the proportion of waterborne transmission varies by pathogen as follows: 1–6% for Hepatitis A virus (0–27% CI range) (Vally et al. 2014; Butler et al. 2015; WHO 2015), 2–8% for Salmonella (0–35% CI range) (Vally et al. 2014; Butler et al. 2015; WHO 2015), 6–11% for Campylobacter (0–32% CI range) (Vally et al. 2014; Butler et al. 2015; WHO 2015), 37% for Cryptosporidium (8–72% CI range) (Butler et al. 2015; WHO 2015), and 42–48% for Giardia (5–75% CI range) (Butler et al. 2015; WHO 2015). Thus, it is important to consider that the deaths reported here represent the documented totals for the pathogens and syndromes, and therefore include resulting infections from all possible modes of transmission. Additionally, the datasets used in this analysis are expected to underestimate the number of deaths attributed to specific pathogens. Thus, the numbers of deaths we document here are likely to be conservative estimates of total deaths due to underdiagnosis, limitations to the sensitivity of death certificates and hospital discharge data for identifying deaths from infections (Barber et al. 2013; Govindan et al. 2014), and limited information on deaths that occur outside of a hospital setting. Not surprisingly, our analytic approach based on hospital discharge data identified nearly three times as many deaths as the approach based on death certificates. Individuals often die having serious chronic health conditions in addition to infections; death certificates tend to give more weight to the chronic conditions, and less to infections that lead to death more proximally (Govindan et al. 2014). Furthermore, for patients who die out of hospital, and those who die in hospital with both infections and terminal chronic diseases, the precise etiology of infections might not be investigated thoroughly. Although hospital discharge data also have low sensitivity for identifying infections and do not supply information on deaths occurring outside of the hospital, they may be a better source of pathogen-specific data than death certificates. A review of studies on the validity of administrative data to identify cases of severe sepsis concluded that ICD-9 codes for these conditions have high specificity (median of five studies, 98.5%) and lower sensitivity (median of seven studies, 42.4%) (Jolley et al. 2015). A Swedish registry-based study found that sensitivity was limited for identifying primary pneumonia in administrative data (38.2%) but specificity was high (98.6%) (Gedeborg et al. 2007). Finally, a study of administrative data from the United States Department of Veterans Affairs calculated a 70% positive predictive value of both pneumonia (not specific to pathogen) and NTM infection codes (Schneeweiss et al. 2007).

In spite of limitations to the sensitivity and specificity of hospital discharge data and death certificate data already noted, the stability and availability of these data makes them useful for monitoring disease outcomes across a wide range of conditions. Outbreak surveillance is suitable for documenting disease transmission through water; however, it is not a reliable source of information on deaths from pathogens with potential for waterborne transmission. For the United States, the combination of administrative data and death certificate data used here may be the best means of obtaining systematic data on deaths from pathogens that can be transmitted by water; importantly, these methods can be repeated annually as new data are made available, allowing for the tracking of waterborne disease-associated mortality over time. These data allow for a look at waterborne diseases beyond the enteric illnesses transmitted through drinking water and provide insight into the epidemiology of pathogens that can be transmitted by water.

In summary, we documented that infections from the 13 selected pathogens and syndromes were associated with nearly 7,000 deaths each year in the United States. Environmental and biofilm-associated pathogens, to which individuals are exposed through inhalation, contact, and other routes, are associated with over 90% of the documented deaths. Current regulations for drinking water systems and recreational water venues have largely had a goal of limiting the transmission of pathogens by the fecal–oral route; efforts to reduce illness and death from environmental pathogens will require a variety of behavioral and engineering strategies depending on their environmental niches and human interactions with water. For example, engineering standards have been developed to control Legionella growth in building water systems (ASHRAE 2015), and limiting the amount of water entering the nose while swimming in warm freshwater venues could reduce the risk of acquiring Naegleria infections (Yoder et al. 2010). The prominence of environmentally associated pathogens suggests that biofilms, commonly found in water, could be an important focal point for prevention efforts. Additionally, work is needed to better understand attributable proportions of waterborne transmission for these infections, so that targeted prevention efforts can reduce the mortality associated with these infections in the United States.

The findings and conclusions in this article are those of the authors, and do not necessarily represent the official position of the US CDC.