Routine Surveillance Data Collection

The New York City (NYC) Health Code requires providers and laboratories to report all cases of legionellosis in city residents to the NYC Department of Health and Mental Hygiene (DOHMH); these reports include positive results for Legionella in cultures, urine antigen tests, direct fluorescent antibody stains, and serologic testing. In addition, all Legionella-positive cultures are required to be sent to either the DOHMH Public Health Laboratory or the New York State Department of Health at Wadsworth Center for confirmation, speciation, and serogrouping. For this analysis, residents of the city of New York who had confirmed legionellosis during 2002–2011 were identified by using the Council of State and Territorial Epidemiologists criteria for confirmed cases (14). These criteria include radiographic or clinical pneumonia and laboratory diagnosis made by urinary antigen, culture, or 4-fold rise in L. pneumophila serum antibody titer. Three cases of Pontiac fever were found during 2002–2011 but were excluded from this study; all legionellosis cases we analyzed were classified as Legionnaires’ disease.

Case Investigation

DOHMH investigates all urine antigen, culture, direct fluorescent antibody stain, or nucleic acid assay results positive for Legionella and all reports of 4-fold or greater rise in antibody titers between acute- and convalescent-phase serum specimens. Because single reports of elevated Legionella serum antibody titers are not diagnostic for legionellosis, those reports are investigated on a case-by-case basis. For the investigations, information from medical charts is abstracted, and patients undergo a standardized interview. Data collected are patient sex, age, race/ethnicity, pre-existing medical conditions, occupation, nights away from home, recreational water exposures, and other risk factors for acquiring Legionella infection.

All case-patients or case-patient proxies were asked about work, nights away from home, visits to and stays in health care facilities, exposure to water aerosols, and other possible exposures during the 14 days before symptom onset. For this analysis, cases were considered to be definitely health care facility–associated if the case-patient resided in a hospital or nursing home for the entire 10 days (for 2002–2007) or the entire 14 days (for 2008–2011) before onset. (This change in criteria was made in 2008 in consultation with the New York State Department of Health in light of the consensus at that time that the incubation period was 2–14 days. However, NYC DOHMH has since returned to a standard 2–10 day incubation period for this determination.) Cases were considered possibly health care facility–associated if the case-patient resided in a hospital or nursing home for part of the 2–9 days (for 2002–2007) or 2–13 days (for 2008–2011) before onset. All other cases were considered community acquired.

Death data were recorded by whether the case-patient had died at the time the investigation was closed. Investigations for confirmed cases are considered closed when diagnosis is confirmed and the patient or proxy interview is completed or it is determined that the interview cannot be completed. Death classification may therefore not have included some case-patients who died from legionellosis after the investigation was closed and may have included some who died of causes other than legionellosis.

Data Sources

Intercensal population estimates for 2002–2009 were produced by DOHMH on the basis of the US Census Bureau Population Estimate Program and housing unit data obtained from the NYC Department of City Planning, available as of November 2012 (15). For 2010 and 2011, 2010 US Census data were used.

The Community Health Survey, a yearly cross-sectional telephone survey conducted by DOHMH that provides citywide public health surveillance data (15), was used to estimate the prevalence of diabetes in the general population. The American Community Survey, a yearly demographic survey conducted by the US Census Bureau (16), was used to calculate population denominators for occupational data. Occupational data collected during case-patient interviews were used to categorize case-patients into American Community Survey–defined occupational classifications.

Neighborhood-level poverty was assessed by using census tract poverty data provided by the US Census. Neighborhood-level poverty was defined as the percentage of residents with household incomes <100% of the federal poverty level on the basis of US Census data from 2000. Census tracts were classified into 6 poverty level categories: very low, <5%; low, 5%–9%; medium, 10%–19%; high, 20%–29%; very high, 30%–39%; and highest, >40%.

Statistical Analyses

Crude and age-adjusted population-based incidence rates were calculated, with age-adjustments standardized to the US Census 2000 population. Relative risks and their 95% CIs were used to compare the demographic characteristics of legionellosis case-patients with those of the general population of the city of New York. Student t-tests were used to compare average incidences between 2 population subgroups. Logistic regression was used to compare death outcomes across sex and acquisition setting while adjusting for age.

Case-patient addresses were geolocated to census tracts by using ArcMap 10 (http://www.esri.com/software/arcgis). Because health care facility–associated cases may be associated with different risk factor patterns, we restricted occupational and neighborhood-level poverty analyses to community-acquired cases. To assess the relationship between race/ethnicity and socioeconomic status, we calculated crude and age-adjusted legionellosis incidence by race/ethnicity in each neighborhood-level poverty group. All analyses were conducted by using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA). Results were considered statistically significant at p<0.05.

Cases

Figure 1

Figure 1. Annual number and incidence (no. cases/100,000 population) of Legionnaires’ disease cases, New York, New York, USA, 2002–2011.

In the city of New York, a total of 1,449 confirmed legionellosis cases were reported through routine surveillance for January 1, 2002–December 31, 2011. Cases were predominantly diagnosed by urinary antigen testing (n = 1,409, 97.2%); 5.7% (n = 82) of all cases were confirmed by culture. The method of diagnosis remained relatively constant over the study period. The crude incidence rate of legionellosis increased from 0.83 cases/100,000 population in 2002 to the highest incidence of 2.74 cases/100,000 population in 2009, a 230% increase (Figure 1). In 2010 and 2011, crude incidence rates remained high at 2.02 and 2.64 cases/100,000 population, respectively. Average yearly incidence for 2002–2011 was 1.75 cases/100,000 population.

Incidence of legionellosis increased in all age groups during the analysis period, but the largest increase (826%) was for the 70–79 age group from 2002 to 2010. Incidence of community-acquired cases was consistently higher during the summer and early fall months; 71.6% (n = 1,038) of cases were diagnosed during June–October.

Demographic Variables

Figure 2

Figure 2. Legionnaires’ disease incidence (no. cases/100,000 population) by sex and age group, New York, New York, USA, 2002–2011.

The average incidence of legionellosis was higher for the male population than for the female population (2.29 cases/100,000 population vs. 1.26 cases/100,000 population; p = 0.0004) and was higher for older age groups (Figure 2). Median case-patient age was 61.0 years (mean 61.8 years, range 9 months to 103 years), and overall incidence rates increased for each age group, with the highest incidence among those >80 years of age (Figure 2). Cases were predominately acquired in the community (88.3%, n = 1,279) rather than definitely (7.0%, n = 102) or possibly (4.7%, n = 68) acquired in a health care facility.

Race/ethnicity was unknown for 279 (19%) case-patients. Average incidence per year for non-Hispanic blacks (2.15 cases/100,000 population) was higher than that for non-Hispanic whites (1.56 cases/100,000 population; p = 0.13) and significantly higher than that for Hispanics (1.02 cases/100,000 population; p = 0.003) or Asian/Pacific Islanders (0.41 cases/100,000 population; p = 0.0004).

Deaths

Overall, 12.8% (185/1,449) of legionellosis case-patients died before the DOHMH case investigation was closed. Death rates were significantly higher for case-patients with definitely health care facility–associated infections than for those with community-acquired infections (odds ratio [OR] 4.78, 95% CI 3.07–7.46). Overall, 35.3% (36/102) of case-patients with definitely health care facility–associated infections died, compared with 10.2% (131/1,279) of those with community-acquired infections. The crude odds for death for women were 1.48 times that for men for community-acquired cases (95% CI 1.09–2.02; p = 0.02), but after adjusting for age, the odds for death were not significantly different between men and women (OR 1.33, 95% CI 0.97–1.83; p = 0.07).

Medical Risk Factors

Of the 1,449 legionellosis cases, 1,278 (88.2%) patients had >1 underlying medical condition that was a known risk factor for legionellosis (9). Current or past smoking (n = 879, 60.7%) and diabetes mellitus (n = 506, 34.9%) were the most frequently reported underlying conditions. After stratifying patients by age, the risk for diabetes was higher for persons with legionellosis than for the general population in every age category; the risk for diabetes was 1.9 times higher for legionellosis case-patients >65 years of age, 2.5 times higher for those 45–64 years of age, 5.7 times higher for those 25–44 years of age, and 6.3 times higher for those 18–24 years of age (Table 1).

Neighborhood Poverty Level

To assess neighborhood poverty level, we restricted our analysis to community-acquired cases. Of the 1,279 community-aqcuired cases, 1,261 (98.6%) could be geolocated. After patient age was adjusted for, the incidence of legionellosis for community-acquired cases followed a gradient; incidence in the highest poverty areas (3.0 average yearly cases/100,000 population) was 2.5 times higher than that for the lowest poverty areas (1.2 average yearly cases/100,000 population) (Table 2). As shown in Table 3, the same gradient existed within each racial/ethnic group, with the highest incidence of disease in the highest poverty group. However, after age adjustment, rates of legionellosis among black non-Hispanics remained higher than those for other race/ethnic groups in each poverty group.

Occupation

Among the 1,279 community-acquired legionellosis cases, 375 (29.3%) case-patients reported working in the 2 weeks before disease onset. The average and median age of case-patients who worked was 53.0 years, and the average age of case-patients who did not work was 64.9 years (median 65.0 years), a mean difference of 11.9 years (95% CI 9.9–13.8 years; p<0.0001). A higher percentage of working case-patients were male than female (71.5% vs. 59.1%; p<0.0001). Compared with the general population, legionellosis case-patients were significantly more likely to work in transportation (crude relative risk [RR] 2.36, 95% CI 1.82–3.06), repair (crude RR 1.86, 95% CI 1.11–3.11), protective services (crude RR 1.77, 95% CI 1.15–2.71), cleaning services (crude RR 1.54, 95% CI 1.07–2.22), or construction (crude RR 1.49, 95% CI 1.03–2.16) (Table 4).