Foodborne and Waterborne Infections in Elderly Community and Long-Term Care Facility Residents, Victoria, Australia

LTCF residents had lower or similar rates of these infections, except salmonellosis, than community residents.

I nfectious disease incidence varies with age, and elderly persons are considered more vulnerable than younger persons to foodborne and waterborne infections (1,2). In many countries, elderly persons unable to care for themselves live in long-term care facilities (LTCFs) where they receive assistance with meals, daily living, and health care (3). Food preparation practices and various exposures in LTCFs may modify the risk of foodborne and waterborne infections in these residents (4,5) when compared with elderly persons living in the community, who may have less safe food preparation practices (6)(7)(8).
A variety of pathogens transmitted by food or water, including Campylobacter sp., Clostridium perfringens, Cryptosporidium sp., Legionella spp., and Shigella sp., and various serotypes of Salmonella enterica can infect humans (9,10). Foodborne and waterborne infections predominantly manifest in elderly persons as gastroenteritis but, depending on the infectious agent, can result in pneumonia, bacteremia, and meningitis (11,12). Elderly persons can become infected by ingesting contaminated water or food or, as with Legionella spp., inhaling contaminated aerosols (13). Some infections are predominantly foodborne; others can be acquired from infected persons or animals or through contact with contaminated environments (4).
These agents can manifest as outbreaks in facilities, leading to community concern about the safety of residents (14,15). Although most outbreaks of gastroenteritis in LTCFs are spread from person to person and are generally mild (16), such outbreaks do result in higher case-fatality rates (CFRs) among residents (17). As a result, regulatory agencies in many countries have mandated programs to manage food safety in facilities. To prevent legionellosis in residents, health agencies commonly provide advice about disinfection of hot water systems that can be reservoirs for Legionella spp. (13).
Few studies have compared the incidence of infections caused by agents that can be transmitted by contaminated food or water consumed by elderly persons living in LTCFs and in the community. One study in the United States estimated that the lower limit of the death rate for  (18). Little examination has been done of the incidence of sporadic foodborne or waterborne diseases in institutionalized elderly persons.
To address this gap, we estimated rates of reported infection in persons >65 years of age living in Victoria, Australia, infected with any of 7 different pathogens according to whether they lived in a governmentsubsidized LTCF or in the community, and we examined the effect of age on incidence of disease. These pathogens were Campylobacter sp., Cryptosporidium sp., Legionella spp., Listeria sp., Salmonella enterica, Shiga toxinproducing Escherichia coli (STEC), and Shigella sp.

Infectious Disease Surveillance
Victoria is 1 of 6 states and 2 territories in Australia. Public health legislation mandates that all physicians and pathology laboratories in Victoria report cases of notifi able conditions under the Public Health and Wellbeing Regulations 2009 (www.health.vic.gov.au/ideas/notifying/ whatto) to the state's Department of Health. Health department staff members enter details about reported cases into a database. Among the 64 conditions notifi able as of April 14, 2010, a total of 14 were enteric diseases can be transmitted by contaminated food or water and 1 was Legionella infection that could potentially be transmitted by inhalation of contaminated water. Surveillance for these diseases has remained essentially unchanged in Victoria since the early 1990s, except for cryptosporidiosis, for which reporting was voluntary until 2001, when notifi cation became mandatory by law.
We analyzed data on all cases of campylobacteriosis, cryptosporidiosis, legionellosis, listeriosis, shigellosis, salmonellosis, and STEC infection that were reported to Victoria's Department of Health during January 1, 2000-December 31, 2009. Surveillance offi cers recorded whether cases were part of an outbreak or occurred in persons who had traveled overseas during their incubation period, which varied for different diseases. In addition, to identify information that was not recorded by surveillance offi cers, we reviewed surveillance data for all LTCF residents. More specifi cally, we identifi ed where >2 cases of the same pathogen occurred within the same facility within 2 weeks, and we recoded these cases as outbreak associated. Surveillance offi cers also recorded whether case-patients had died of the disease or of other concurrent conditions within the weeks after infection during public health follow-up.

Data Analysis
We categorized reported cases by residential status of the patient. An LTCF resident was a person who had a residential address of a government-subsidized LTCF, and a community resident was a person living in a private residence in Victoria (online Technical Appendix, wwwnc. cdc.gov/EID/pdfs/11-0311-Techapp.pdf). We excluded from analysis cases in persons without a valid address because we assumed the address was missing completely at random. We also excluded case-patients residing in privately funded facilities (supported residential services) catering to elderly or disabled persons or persons with psychiatric illness or dementia because they were not included in the denominator of LTCF residents. We counted the annual number of foodborne and waterborne infections, including those from epidemiologically important serotypes and species, in LTCF and community residents during the 10-year period. We obtained age-specifi c annual denominator data for residents of government-subsidized LTCFs from annual reports prepared by the Australian Institute of Health and Welfare (19). Denominator data for community residents were calculated by subtracting annual age-specifi c estimates of LTCF residents from estimated resident populations prepared by the Australian Bureau of Statistics (www. abs.gov.au/). We calculated annual rates of notifi cation for different diseases by residential status and compared them with the total rate of notifi cations for the state. We calculated CFRs by dividing the number of deaths from the disease in the different groups by the total number of cases of disease, including cases for which death status was unknown, for the 10-year period. To account for age differences in LTCF and community residents, we calculated age-adjusted relative risks (RRs) for death from infection with different pathogens by using Mantel-Haenszel methods.
To estimate incidence rate ratios (IRRs), we used a negative binomial regression model of the annual count of foodborne and waterborne infections by the period of notifi cation (2000-2004 and 2005-2009), sex, 3 categories of age (65-74, 75-84, and >85 years), and residence (LTCF and community). The number of persons living in LTCFs or the community in each age group for each year was entered into the model as an offset. We used robust variance estimation suited to longitudinal or clustered data to account for possible clustering from outbreaks. We assessed model fi t by examining the distribution of standardized Pearson residuals. To assess the effect on incidence, we repeated regression models excluding travel-associated cases and including only the fi rst case for each known outbreak in LTCFs and the community. We analyzed data by using Stata version 11.2 (Stata Corp., College Station, TX, USA).
The Australian National University human research ethics committee approved this study. Victoria's Department of Health approved release of the data.

Incidence in Persons >65 Years of Age
During January 1, 2000-December 31, 2009, a total of 8,534 cases of the 7 diseases were reported in persons >65 years of age. During data cleaning, we excluded 238 (2.8%) persons without valid residential addresses and 19 who lived in a private institution caring for elderly or disabled persons. A total of 8,277 cases were available for analysis, including 132 (1.6%) case-patients living in retirement villages.

Infections in Institutional and Community Residents
Rates of all reported infections in LTCF residents were similar to or lower than those in community residents, except infections from S. enterica (Table 1). No LTCF residents were reported with L. longbeachae, STEC, or Shigella sp. infections during the surveillance period. Among persons >65 years of age infected with Legionella species other than L. longbeachae were more likely to live in an LTCF (crude RR 1.15, 95% CI 1.1-1.2; p = 0.27). The reported rate of Campylobacter spp. was lower in LTCF residents than in community residents for all years of surveillance ( Figure 1). In contrast, reports of S. enterica serotype Typhimurium peaked because of outbreaks during the study period, which resulted in higher rates for LTCF residents overall ( Figure 2).

Associated Death
The CFR was highest for listeriosis and L. longbeachae infection, from which 17.8% and 6.7% of persons died, respectively ( Table 2). The age-adjusted RR for death from L. monocytogenes infections was 3.5 (95% CI 1.2-10.4; p = 0.03) for LTCF residents. No L. pneumophila-associated deaths were recorded in LTCF residents. For salmonellosis, most deaths were considered to have resulted from other concurrent conditions. The ageadjusted RR for death in LTCF residents infected with any S. enterica serotype did not differ signifi cantly from that in community residents (adjusted RR 2.8, 95% CI 0.8-9.1; p = 0.87). Death status was not routinely ascertained for persons with campylobacteriosis, although 1 death from campylobacteriosis and 11 deaths from concurrent conditions were recorded. No deaths were recorded for shigellosis or cryptosporidiosis.

Rates of Foodborne and Waterborne Infections
In multivariable analysis, the incidence of S. enterica serotype Typhimurium was higher in LTCF residents than  Table 3). IRRs could not be estimated for L. longbeachae and STEC infections and for shigellosis because no cases occurred in LTCF residents during the surveillance period. We observed a trend of increasing rates of reported infections for cryptosporidiosis, salmonellosis, and campylobacteriosis over time during the surveillance period. From multivariable analysis, reported incidence rates were higher in older age groups for L. monocytogenes, Campylobacter sp., and S. enterica serotype Typhimurium infections than in the base age group of persons 65-74 years ( Table 3). The incidence of L. pneumophila and non-Typhimurium serotypes of S. enterica infections and cryptosporidiosis did not differ signifi cantly by age group.

Accounting for Travel and Outbreaks
During their incubation period, 105 (1.3%) of the 8,277 persons with notifi able infections traveled internationally; all were community residents. Among community residents, 16  When we repeated multivariable models, excluding travel-associated infections and including a single case for each identifi ed outbreak, the incidence rate for S. enterica serotype Typhimurium was similar in LTCF residents and community residents (IRR 0.91, 95% CI 0.64-1.29; p = 0.59). For infections with non-Typhimurium serotypes of S. enterica, the incidence was higher in LTCF residents than in community residents (IRR 1.4, 95% CI 1.0-2.0; p = 0.05). After adjustment for travel and outbreaks, the incidence rate was lower for LTCF residents with Campylobacter spp. (IRR 0.57, 95% CI 0.48-0.68; p<0.001). For L. pneumophila, the incidence rate was lower, but not signifi cantly, for LTCF residents (IRR 0.63, 95% CI 0.29-1.3; p = 0.23).

Discussion
Rates of foodborne and waterborne infections among LTCF residents were lower than or similar to rates among community residents, except for salmonellosis, which was higher. In particular, rates of campylobacteriosis in LTCF residents were consistently lower throughout the entire study period, which was unexpected because incidence of this infection is universally high (20  The lower incidence of campylobacteriosis might result from the highly regulated food hygiene system for LTCFs. In 1998, Victoria was the fi rst Australian state to implement mandatory food safety programs for food service settings, and those serving vulnerable populations require independent auditing (24). These programs might have resulted in better understanding and practices by LTCF staff about food storage, cooking, and cross-contamination than by elderly persons in their own homes. Campylobacter infections in elderly community residents have been associated with risk possibly from cross-contamination during food preparation (25). Even though Campylobacter infections are associated with travel, a case-control study in Australia found that only 23 (2.8%) of 833 persons >5 years of age with campylobacteriosis had traveled overseas during the week before illness (21). When we accounted for known travel history and outbreak-associated cases, the IRR for Campylobacter spp. infections in LTCF residents was lowered.
S. enterica is a common cause of foodborne and waterborne outbreaks in LTCFs (17,26,27), a fi nding that our study confi rmed. We found that outbreaks of S. enterica serotype Typhimurium infections accounted for the higher incidence of these infections, but not for non-Typhimurium serotype infections, in LTCF residents. Sources of outbreaks in LTCFs often are not identifi ed, although eggs are commonly suspected as the cause in S. enterica serotype Typhimurium-associated outbreaks (17). Although residents are at higher risk for outbreak-associated disease, ascertainment of cases is biased in the institutional setting because of the common living environment, centralized access to health care, and collection of specimens by public health staff. During outbreaks, public health investigators often collect fecal specimens from LTCF residents with diarrhea, which would not occur for elderly persons in the community. Because surveillance is well established in Victoria, LTCFs are more likely than persons in the community to report outbreaks (16).
We did not fi nd any evidence to suggest that living in an LTCF increased a person's risk for legionellosis, despite the occasional occurrence of outbreaks and sporadic cases in this setting (13,28). LTCF residents reported with legionellosis were more likely to be infected with Legionella species other than with L. longbeachae. L. longbeachae is associated with gardening and potting mix, so we expected the incidence of this infection to be low in LTCF residents (29).The incidence of listeriosis was similar in LTCF and community residents. Given the food safety program requirements in facilities in Victoria, LTCF residents plausibly could be exposed to lower concentrations of L. monocytogenes in food, compared with community residents who may keep food longer, have poorer food preparation practices, and eat foods considered higher risk for transmitting foodborne pathogens (8).
Different clinical investigative approaches for LTCF and community residents with potential foodborne and waterborne disease might account for some of our fi ndings. Although clinicians might elect not to collect specimens when LTCF residents have diarrheal illness, we think it more likely that reporting is more complete in LTCFs. Most of the diseases in our study are serious illnesses, and infected persons would have severe gastrointestinal and extraintestinal symptoms lasting for several days or weeks (3,4). In a case-control study of campylobacteriosis  in Australia, 41% of case-patients had bloody diarrhea, and 75% had fever; both of these symptoms are strong predictors for physicians ordering laboratory tests (30,31). We were unable to control for potential confounding factors, such as concurrent conditions and factors that might predispose for infection. Many elderly persons with concurrent conditions live in the community, but the health status of LTCF residents is likely to be lower, and they are likely to be more frail. We would have partly controlled for frailty through inclusion of age in our multivariable model because elderly persons in institutions are the oldest and the most frail in society (32). In addition, our study was underpowered to detect an effect for diseases where notifi cation rates were very low. The potential bias in the fi nal estimates from lack of control of confounding would be more likely to result in increased incidence rates in LTCF residents. However, except for salmonellosis, infection rates were higher in community residents.
Our fi ndings should not be overinterpreted because our study was a retrospective record-based study in which we manually coded surveillance data and were unable to validate case-patients' addresses. It is possible that we were unable to correctly identify residential status of case patients from addresses. In some instances, residents were recorded as living at addresses where an LTCF and retirement village were on the same grounds, making determining whether a person lived in the facility diffi cult. Similarly, some persons might have been infected after moving into an LTCF, but the address on a pathology report still recorded their residential address in the community where they had previously lived. However, in Victoria, physicians and laboratories were required to report these infections, making it unlikely that both sources of notifi cation would incorrectly report the residential address. For Campylobacter infections, however, physicians report only 50% of notifi cations, with the remainder coming from laboratories (33).
Elderly community residents might receive meals from organizations that provide community support. In addition, elderly residents of LTCF might eat food that has been prepared outside the facility during excursions or brought in by visitors, which could result in exposure to foodborne pathogens. For both groups, these alternative routes of exposure would modify the risk for infection so that it did not truly refl ect the risk in their place of residence.
The strength of our approach was that we consistently coded addresses without regard to disease-causing agent, yet we observed distinct differences in reported incidence from disease to disease. Our fi ndings were consistent with what we know about these diseases, such as increasing incidence in older persons for diseases such as listeriosis. The CFRs were consistent with reports in the literature for elderly persons, although we assessed deaths only short term (i.e., in the weeks after infection) (34,35). In general, elderly persons have more severe outcomes from foodborne infections than do younger persons (4,18). Large-scale studies that used population-based registers have demonstrated that enteric diseases contribute to more deaths than recognized from short-term follow-up, even when controlling for concurrent conditions (36,37).
In our study, rates of surveillance reports for most infections in persons >65 years of age were similar to or lower than for persons <65 years of age, a fi nding that contradicts the common statement that elderly persons are at higher risk for foodborne disease. However, we did fi nd that the CFRs were high for some infections and that LTCF residents were affected more severely. We believe that our fi ndings can be generalized to other Australian states and territories with similar rates of infection and methods of surveillance (16,22). Other investigators could repeat this study by using record-linkage to compare their fi ndings with our fi ndings.
We observed a lower incidence of reported Campylobacter spp. infection in LTCF residents, which provides some reassurance for food safety regulators and the aged care industry. Our study highlights that most foodborne and waterborne infections are rare in elderly residents of LTCF and the community, but that these infections do cause occasional deaths. Primary research is needed into the specifi c causes of foodborne and waterborne infections in elderly persons in the community and in institutional settings that particularly accounts for the effect of concurrent conditions. In our study, elderly LTCF residents had an incidence of foodborne and waterborne infections that was similar to or lower than that that for elderly persons living in the community, except for S. enterica infections. reports from doctors and laboratories, information from the treating doctor's report were entered onto the database. Infections were prioritized for follow-up by public health officers, with more serious or transmissible infections classified as urgent. On the basis of this, public health staff entered additional information on the surveillance database, such as exposure information, vaccination status, any treatment or prophylaxis, and whether the case was part of an outbreak.

Foodborne and Waterborne Infections in
4. If a Web search on the facility identified that there had been an LTCF of the same name and address at the time of the case-patient's infection but was not currently in operation in 2010.
If a typical name of facilities of any type was mentioned in the first 6 fields and the name was not identified in the LTCF facility lists, a Google search and Google map search were conducted to identify whether the name referred to a retirement village or Supported Residential Service. In instances where the case-patient's address mentioned a "retirement village" but was the same as a recognized LTCF, we searched Google to determine whether there was a retirement village onsite at the same address. If so, the case-patient was coded as living at a "retirement village." All case-patients living in "retirement villages" were considered "community residents" for the purposes of this study.

Exclusions
In this study, case-patients were excluded if they were not part of the study population according to the following criteria.
1. Nonresidents of Victoria or Australia according to the residential address or country of residence were excluded from the study.
2. Case-patients for which information in any of the first 6 address fields was insufficient to assess whether they lived in an LTCF or the community were not included. Case-patients with a residential address listed as "RMB" (Rural Mail Box) followed by a number were coded as having a valid address. RMB indicated that a person received mail in a rural area at a designated mail box where defined street numbers were not in use. Case-patients with a residential address of PO Box (Post Office Box) followed by a number were coded as missing a valid residential address because it gave no indication of where a case-patient lived.
3. Resident of a "supported residential service"-privately run facilities providing care for people who were old, disabled, or suffering from dementia or psychiatric illness (see www.health.vic.gov.au/srs/index.htm) were not included in the study.