Legionnaires Disease Surveillance in US Department of Veterans Affairs Medical Facilities and Assessment of Health Care Facility Association

Key Points

Question  What is the Legionnaires disease burden in the US Department of Veterans Affairs medical facilities, a health care system that has prioritized Legionnaires disease prevention with policy?

Findings  In this cohort study, the number of Legionnaires disease cases was low (n = 491) and 91% of cases had no VA exposure or only outpatient VA exposure. Total rates of Legionnaires disease significantly increased from 2014 to 2016, but rates in cases with overnight health care system exposure significantly decreased.

Meaning  Although total Legionnaires disease rates increased, health care system–associated (overnight stay) rates decreased significantly, suggesting that prevention efforts may have contributed to improved patient safety in these settings.

Importance  Legionnaires disease (LD) incidence is increasing in the United States. Health care facilities are a high-risk setting for transmission of Legionella bacteria from building water systems to occupants. However, the contribution of LD in health care facilities to national LD rates is not well characterized.

Objectives  To determine the burden of LD in US Department of Veterans Affairs (VA) patients and to assess the amount of LD with VA exposure.

Design, Setting, and Participants  Retrospective cohort study of reported LD data in VA medical facilities in a national VA LD surveillance system from January 1, 2014, to December 31, 2016. The study population included total veteran enrollees and enrollees who used the VA health care system.

Main Outcomes and Measures  The primary outcome was assessment of annual LD rates, categorized by VA and non-VA exposure. Legionnaires disease rates for cases with VA exposure were determined on both population and exposure potential levels. Rates by VA exposure potential were calculated using inpatient bed days of care, long-term care resident days, or outpatient encounters. In addition, types and amounts of LD diagnostic testing were calculated. Case and testing data were analyzed nationally and regionally.

Results  There were 491 LD cases in the case report surveillance system from January 1, 2014, to December 31, 2016. Most cases (447 [91%]) had no VA exposure or only outpatient VA exposure. The remaining 44 cases had VA exposure from overnight stays. Total LD rates from January 1, 2014, to December 31, 2016, increased for all VA enrollees (from 1.5 to 2.0 per 100 000 enrollees; P = .04) and for users of VA health care (2.3 to 3.0 per 100 000 enrollees; P = .04). The LD rate for the subset who had no VA exposure also increased (0.90 to 1.47 per 100 000 enrollees; P < .001). In contrast, the LD rate for patients with VA overnight stay decreased on a population level (5.0 to 2.3 per 100 000 enrollees; P < .001) and an exposure level (0.31 to 0.15 per 100 000 enrollees; P < .001). Regionally, the eastern United States had the highest LD rates. The urine antigen test was the most used LD diagnostic method; 49 805 tests were performed in 2015-2016 with 335 positive results (0.67%).

Conclusions and Relevance  Data in the VA LD databases showed an increase in overall LD rates over the 3 years, driven by increases in rates of non-VA LD. Inpatient VA-associated LD rates decreased, suggesting that the VA’s LD prevention efforts have contributed to improved patient safety.

Legionnaires disease (LD) is an acute pneumonia caused by Legionella species, primarily L pneumophila serogroup 1 in the United States.¹ The bacteria are naturally present in water,¹ and infection is associated with exposure from engineered water systems that allow Legionella to proliferate. Several prominent outbreaks have increased attention on LD in the United States recently.^2-6 Legionellosis (including LD and the milder Pontiac fever) is reportable to the Centers for Disease Control and Prevention (CDC); 6079 cases were reported in 2015.⁷ Rates of reported legionellosis have been increasing for decades.^8-10 However, the incidence of LD in the United States and the sources of infection are not well characterized. Underdiagnosis, underreporting, and unknown follow-up of sporadic cases complicate the ability to ascertain a true national picture. The result is an underappreciation for the burden of LD in the United States and missed opportunities for prevention. This is particularly important for health care settings, which have occupants at risk for Legionella infection¹¹ and for which numerous outbreaks have been described.² The CDC recently reported health care–associated (HCA) LD surveillance data from 21 jurisdictions.¹¹ Of the 2809 confirmed LD cases reported in 2015, 3% were classified as definite HCA LD and 17% were classified as possible HCA LD, substantiating health care as a source for exposure. However, national data over multiple years were not reported, categorization of possible HCA LD by inpatient or outpatient exposure was not stated, and data completeness was not known. Furthermore, many hospitals do not perform whole-house surveillance for HCA pneumonia, and guidelines^12,13 do not specifically indicate Legionella diagnostic testing in many patients with HCA pneumonia when it is identified. Therefore, the general burden of HCA LD in the United States may not be well represented by passive surveillance systems.

The Veterans Health Administration (VHA) in the US Department of Veterans Affairs (VA) is the largest integrated health care system in the United States, with more than 1200 sites of care, serving about 6 million veterans annually.¹⁴ In federal fiscal year (FY) 2016, 91% of veterans using VA benefits were male with a median age of 64 years¹⁵ and with higher morbidity than in the rest of the United States,^16,17 which are population factors with higher LD risk.^1,18 Accordingly, VHA has a Legionella prevention policy required at medical facilities to limit Legionella growth in water distribution systems and validation of effectiveness using both environmental and clinical surveillance.¹⁹ The policy promotes LD testing of symptomatic patients and requires heightened awareness for LD when facility water has tested positive for Legionella. Concomitant to publication of the policy in 2014, the VHA Central Office implemented a national standardized LD reporting system. This article presents an analysis of data in the system from January 1, 2014, to December 31, 2016, and with no such comparable surveillance database in the United States, provides a unique insight into the rate of LD—both community-associated and HCA—on a national scale.

Veterans Affairs medical facilities offer a range of services to enrolled veterans across the country and US territories.²⁰ During this analysis, VA medical facilities were administratively arranged into 142 health care systems (HCSs), with 170 medical centers²⁰ and 132 long-term care sites.²¹

Dissemination of findings that result from review of national VHA operational data sets beyond programmatic needs has been reviewed and approved by the institutional review board at the University of Cincinnati, the institutional review board of record for the Cincinnati VA Medical Center. In this retrospective review of data in an operational surveillance system, there was no greater than minimal risk to patients included in the system, the rights and welfare of patients in the system were not adversely affected, and patient medical records were not altered as a result of this work; these criteria determined that patient consent was not required. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Beginning October 15, 2014, LD data were reported into 2 databases prospectively by personnel at each VA HCS, centrally maintained at the VHA Inpatient Evaluation Center (IPEC).²² Legionnaires disease data reporting was required.

The Legionella Case Report Module database is used for ad hoc reporting of laboratory-confirmed, community- and VA-associated cases (eAppendix 1 and eTable 1 in the Supplement). Determination of case reporting information provided to this central surveillance system is made at the local facility level. In addition to prospective reporting, retrospective reporting of cases was also required from October 1, 2013, to October 14, 2014.

Legionnaires disease case reports included whether there was exposure to a VA building within the 10 days prior to LD symptom onset. Classification of cases as definitely or possibly associated with VA exposure followed surveillance definitions on the 2014 CDC legionellosis case report form,²³ with the additional detail of the type of exposure (inpatient, outpatient, or both) if the case was possibly associated with VA. For this article, we assessed VA exposure using the current²³ and previous²⁴ CDC definitions for HCA legionellosis to determine the association of outpatient exposure, a new criterion for the current definitions, with attribution of cases to health care. Definite VA-associated LD cases reported to the database received extensive review and often follow-up with the facility to confirm the diagnosis and classification, and to provide consultative assistance for prevention.

The Legionella Clinical Information Module database collects monthly reporting of aggregate LD diagnostic information at VA medical facilities, including the number of various LD diagnostic tests performed (eTable 2 in the Supplement). Retrospective reporting of data prior to October 1, 2014, was not required.

Data entered into both IPEC databases for each HCS were centrally reviewed monthly to determine if information was complete and consistent across related data elements. Facilities were contacted to update data if necessary. In addition to this routine monthly assessment, system validation was undertaken annually to determine if LD cases were being reported in the Legionella Case Report Module as expected (eAppendix 2 in the Supplement).

Cases from January 1, 2014, through December 31, 2016, were reviewed to determine the number of cases reported and the percentage of LD cases associated with the VA. For cases classified as definitely VA associated and possibly VA associated with inpatient stay, the patients’ electronic medical records were reviewed to determine type of exposure (acute care and long-term care) (eAppendix 3 in the Supplement). Of note, cases were assessed for classification as VA associated rather than HCA because we could only reliably ascertain health care contact in the VHA system.

Total LD rates in the VHA patient population for each year were determined using the most recently available VA data on total veteran enrollees in²⁵ and enrollees who were users of²⁵ the VHA system as population denominators; the first denominator category is comparable with use of the US population by the CDC to determine national rates, while the second denominator category reflects a subset veteran population more likely to have been diagnosed at a VA facility.

Annual rates of community- and VA-associated LD were also calculated. The VA-associated LD rates were assessed separately for patients with overnight stays (inpatient or residential) and patients with only outpatient encounters in the 10 days before symptom onset. For each of these 2 categories, rates were calculated in 2 different ways to account for population-based rates (using unique inpatients or unique outpatients in the patient treatment file data sets and unique residents in the extended care records, as appropriate, for denominator data)²⁶ and exposure potential rates (using outpatient encounters in the patient treatment file,²⁶ or inpatient bed days of care and long-term care resident days in IPEC,^27,28 as appropriate, for denominator data). The exposure potential denominators were selected to account for length of stay or amount of contact with health care buildings in a similar manner as is done for other HCA infections.²⁹ The numbers of cases, classifications, and rates were also assessed on a regional level for prospectively reported case data (2015 and 2016) using US Census Bureau divisions to delineate US regions³⁰ (eAppendix 4 in the Supplement). Denominators for regional rate calculations were derived from state-level VHA data for veteran enrollees who used the VHA system in FY 2015 and FY 2016.^31,32

Monthly diagnostic tests performed and test results for Legionella urine antigen tests (UATs) and clinical cultures were extracted from the IPEC Legionella Clinical Information Module for 2015 and 2016. Regional percentage of positivity for each method was calculated using the same US Census Bureau division categories described earlier.

National LD rate trends were assessed using Poisson regression analysis. A log-linear regression model was calculated with enrollees, inpatients, outpatients, or bed days of care used as an offset variable. Overdispersion was accounted for using a quasi-likelihood method approach to estimate a dispersion parameter. Pairwise comparisons of regional LD rates were done by χ² analyses (.01 level of significance). The exact results for each pairwise comparison are provided to allow for assessment of the strength of the differences in the presence of multiple comparisons. Legionella UAT positivity rates were assessed by χ² analyses for regional and monthly pairwise comparisons. Case fatality rates were compared using the Fisher exact test.

Statistical analyses were performed using SAS version 9.3 (SAS Inc). All analyses were 2-sided and a P value of less than .05 was considered significant unless otherwise noted.

There were 491 LD cases in the case report surveillance system from January 1, 2014, to December 31, 2016, and the number of cases increased each year (Table 1). Most cases (447 [91%]) had no VA exposure or only outpatient VA exposure in the 10 days prior to symptom onset and the remaining cases (n = 44) had VA exposure with overnight stay. Total LD rates from January 1, 2014, to December 31, 2016, increased for all VA enrollees (from 1.5 to 2.0 per 100 000 enrollees; P = .04) and for users of VA health care (2.3 to 3.0 per 100 000 enrollees; P = .04). The LD rate for the subset who had no VA exposure also increased (0.90 to 1.47 per 100 000 enrollees; P < .001). In contrast, the LD rate for patients with VA overnight stay decreased on a population level (5.0 to 2.3 per 100 000 enrollees; P < .001) and an exposure level (0.31 to 0.15 per 100 000 enrollees; P < .001). One-third of the cases (163 of 491) had some VA exposure in the 10 days prior to symptom onset. Thirteen of the 491 LD cases (3%) were definitely associated with a VA facility. Most VA-associated cases were in the possible HCA category (150 of 163 [92%]), the majority (119 of 150 [79%]) with only outpatient exposure in the 10 days prior to symptom onset. Definite VA-associated LD cases primarily had exposure in long-term care settings (11 of 13 [85%]) and possible VA-associated cases with overnight stay primarily had acute care exposure (26 of 31 [84%]) (eTable 3 in the Supplement). Five facilities had clusters of LD cases with VA overnight exposure, occurring primarily in 2014 (eTable 4 in the Supplement).

The unadjusted 30-day fatality rate for cases with possible inpatient VA association was higher (8 of 31 [25.8%]; P = .004) than the fatality rate for patients with LD with definite VA association (1 of 13 [7.8%]), possible outpatient VA association (8 of 119 [6.7%]), or no VA association (18 of 327 [5.5%]; no death data for 1 patient). Grouped together, the fatality rate for VA patients with overnight stay exposure (definite or possible inpatient) was higher (9 of 44 [20.5%]; P = .002) than the fatality rate for patients with outpatient or no VA exposure (26 of 446 [5.5%]; no death data for 1 patient).

Because inclusion of outpatient exposure was new in the 2014 CDC definitions for HCA LD,^23,24 we examined the impact of the change on attribution of LD cases in the VA data set (Table 2). Including outpatient encounters in the new definition shifted the number of VA-associated cases from 31 to 150, a 384% increase in epidemiologic case attribution. The additional 119 cases classified as possibly VA associated from outpatient-only exposure corresponded with the 27% decrease in classification of cases as community associated.

Total LD rates significantly increased between 2014 and 2016 when calculated using either denominator type (Table 3). Most LD cases were not associated with VA exposure, and the rates significantly increased over the review period (Table 3). In contrast, the rates for the subset of LD cases with inpatient VA exposure significantly decreased when calculated for population level and accounting for exposure potential (Table 3). Outpatient-only VA-associated LD rates were low, and the change over time was not significant regardless of calculation on a population or encounter level (Table 3).

Regional differences in LD incidence in the United States have been reported.⁹ We categorized the 355 LD cases reported to the IPEC system in 2015 and 2016 by region using US census divisions and we also found differences (Figure). In all regions, non-VA cases were in the majority. The East North Central (ENC) region reported the most LD cases (including the most definite VA-associated cases) and had the second highest rate (Figure). While the South Atlantic (SA) region had almost the same number of total cases as the ENC region, it had one of the lowest total rates of LD and this rate was significantly lower than the ENC rate (eTable 5 in the Supplement). The Middle Atlantic had the highest rate among the regions, and this rate was significantly higher than 3 other regions (New England, SA, and Pacific) (eTable 5 in the Supplement).

The case reports from 2014 to 2016 indicated that 482 of 491 cases (98%) had a UAT performed, with the majority having a positive result (463 of 482 [96.1%]). For 338 of 491 cases (68.8%), the only Legionella diagnostic test performed was the UAT. Clinical culture was done for 109 of 491 cases (22%), and 39 (36%) were positive. Many of the 39 cases with a positive clinical culture result also had a positive UAT result (28 of 39 [72%]); the other cases either had a negative UAT result (6 of 11) or did not have a UAT performed (5 of 11). The few cases without a positive UAT or culture (n = 17) were diagnosed using serology, immunochemistry, or nucleic acid testing.

The Clinical Information Module data set was examined to assess the amount of Legionella diagnostic testing performed by VA medical facilities in 2015 and 2016, and these data were categorized by US region and month. In total, 49 805 UATs (Table 4; eTable 6 in the Supplement) and 12 004 clinical Legionella cultures were performed nationally in the 2 years with positivity rates of 0.67% (335 positive results) and 0.23% (28 positive results), respectively. All regions had more than 2900 Legionella UATs performed except New England (Table 4). Two regions, Middle Atlantic and SA, performed more than 10 000 UATs each (eAppendix 5 in the Supplement), but the UAT positivity rates for these regions were some of the lowest in the country. The highest UAT positivity rate was in the ENC region. See eTable 7 in the Supplement for pairwise χ² analyses of regional UAT positivity. Analysis of UAT data by month showed significantly higher positivity rates in warmer months (Table 4; eTable 8 in the Supplement), despite more UATs performed in cooler months. Clinical culture positivity was low for all regions (eTable 9 in the Supplement).

The VHA LD reporting database is unique for collecting information on LD in a nationally distributed US health care system. Reporting indicated that LD was a very infrequent diagnosis regardless of exposure classification. Nonetheless, the overall LD rate in VHA patients significantly increased during the review period, corresponding with the significant increase in the subset of LD cases that had no VA exposure and in alignment with increasing LD rates in the United States.^9,10 In contrast, the number of LD cases with VA overnight exposure was very low, with significantly decreasing rates over the 3 years. Taken together, these data strongly support that community sources contribute to most of the LD infections in the United States.¹⁰ The decrease in the LD rate in patients and/or residents with an overnight stay may be a result of intense efforts at VA facilities after publication of the most recent policy in 2014¹⁹ to prevent Legionella growth in building water distributions systems (eAppendix 6 in the Supplement), a supposition similar to a previous review of LD and VHA policy.³³ While determining a direct correlation between current policy implementation and the decrease in reported cases is beyond the scope of this article, the reduction in case clusters over the 3-year period further suggests improved prevention practices. The results also indicate the particular importance of Legionella prevention programs, including LD diagnostic testing, in long-term care settings where high-risk occupants may have prolonged exposure to facility water sources.

The 2015 total LD rate in the VHA enrollee population in this study (1.9 cases per 100 000 persons) was comparable to the 2015 LD rate in the US population for all ages (1.89 cases per 100 000 persons).⁷ To our knowledge, this is the first report of national active surveillance of HCA LD cases in the United States and the first estimation of national HCA LD rates. While there is no known comparator in the US population for HCA LD rates, it is notable that the 2015 rate for VA LD with overnight stay (3.9 cases per 100 000 patients and/or residents) is similar to that in the general population for persons aged 40 to 64 years (2.73 cases per 100 000 persons),⁷ the age range that corresponds with the median age of female (age 46 years) and male (age 64 years) users of VA benefits.³⁴ Furthermore, the percentage of LD cases that were definitely VA associated (3%) was similar to recent CDC reporting of definite HCA LD in 21 US jurisdictions.¹¹

We did not calculate an overall rate of VA-associated cases because of the variability in the extent and types of exposure to facility water sources between inpatient and/or residential contact and outpatient contact. Furthermore, inclusion of outpatient contact in the possible HCA LD CDC definition is relatively new. This work demonstrates that the definition change, while not affecting the overall number of LD cases, shifted the classification of cases and resulted in an increase in attribution of possible HCA LD with a compensatory decrease in community-associated cases. While perhaps helpful for surveillance, the actual contribution of outpatient and/or transient contact with a health care building to transmission of Legionella is uncharacterized.

It has been surmised that the number of LD cases is underdiagnosed in the United States in part because pneumonia cases are often empirically treated with antibiotics. We assessed LD diagnostic testing in the context of Legionella prevention policy¹⁹ that promotes testing beyond guidelines.^12,13 Increased use of Legionella UAT and clinical culture by VA facilities in 2015 and 2016 compared with previously reported testing levels in FY 2012 (15 169 UATs and 3091 clinical cultures)³⁵ did not result in the detection of large numbers of LD cases or a marked increase from previously reported numbers of HCA LD in VA.^33,35 Regions where more than 10 000 UATs were performed in 2015 and 2016 had significantly lower UAT positivity rates in general than other regions that performed less testing. Nonetheless, the data substantiated regional differences in LD rates reported by others⁹ with higher LD rates in general in the eastern part of the United States. While extensive use of UAT did not result in increased UAT positivity rates, the regions that did the most testing in 2015 to 2016 corresponded to regions with the most numbers of LD cases in those same years (compare Table 4 and Figure). The ranking of the regional LD rates in VA are consistent with CDC reporting of regional rates for 2009⁹ for regions with the highest (Middle Atlantic and ENC) and lowest (Pacific) rates. A notable exception is the SA region, which had one of the highest rates in the CDC report for 2009 but one of the lowest rates in the VA.

Increased LD testing occurred in cooler months, perhaps reflecting the higher incidence of respiratory infections in those months. However, as observed by others,^8,9 the UAT positivity rate for LD was significantly higher in the summer months. Overall, the testing data suggest that heightened awareness for LD in patients with pneumonia may be important on an individual case level for optimal treatment and in regions of the country known to have higher incidence of LD,³⁶ but oversensitivity for diagnosis uses resources without necessarily increasing case identification.

This study has limitations. First, facilities may have missed identifying patients for Legionella testing. However, because heightened awareness for LD is established by policy and extensive testing is occurring nationally across the system, it is unlikely that a large number of cases were missed. Overall, the VHA requirement for reporting LD cases, the system validation review, and educational outreach provide confidence in the general completeness of the data. Second, patients with LD who were not diagnosed or subsequently cared for at VA medical facilities and had no VA exposure in the 10 days prior to symptom onset are not captured by the reporting system. It is unknown how many cases this represents, and this report focuses on those cases with VA contact and/or care. Third, this work uses facility-reported data from Legionella surveillance systems; we did not conduct routine medical record reviews of all reported cases to assess accuracy of the LD diagnosis or to confirm that patients had signs or symptoms of pneumonia. Future work is planned to examine patient demographic characteristics and symptoms as a detailed review of the spectrum of LD. Finally, similar to other surveillance systems, the VHA LD reporting modules did not collect information on implementation of interventions to prevent cases; a comparison of actions at facilities that did and did not have HCA LD cases to determine effective prevention practices was outside the scope of this work.

This study provides insight into HCA LD in the United States. Perhaps most importantly, these findings from a novel, rigorous reporting program add to the body of knowledge about HCA LD that was previously largely gleaned from local outbreak reports or limited passive surveillance. The number of VA-associated LD cases with overnight exposure was low, with decreasing rates in this category over the study period despite increasing overall rates. This finding sets the stage for additional reviews to assess the contribution of VHA’s proactive Legionella prevention policy, including analysis of national environmental Legionella data, on the observed improvement in patient safety. Generalization of the results in this study to HCA LD in the United States may be affected by VA population biases. Nonetheless, this study is informative to other health care facilities that are considering implementation of Legionella prevention policies. This is particularly relevant because regulatory,³⁷ accreditation,³⁸ and standards³⁹ organizations have recently prioritized water safety in health care.

Accepted for Publication: March 5, 2018.

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2018 Gamage SD et al. JAMA Network Open.

Corresponding Author: Shantini D. Gamage, PhD, MPH, National Infectious Diseases Service, Department of Veterans Affairs, 205 W Fourth St, Ste 1020, Cincinnati, OH 45202 (shantini.gamage@va.gov).

Author Contributions: Dr Gamage had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Gamage, Ambrose, Kralovic, Roselle.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Gamage, Ambrose, Simbartl.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Simbartl.

Administrative, technical, or material support: Gamage, Ambrose, Kralovic, Roselle.

Supervision: Gamage, Roselle.

Conflict of Interest Disclosures: None reported.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the Department of Veterans Affairs.

Meeting Presentation: A portion of this article was presented as a poster at IDWeek 2015; October 8, 2015; San Diego, California.

Additional Contributions: We thank the staff at the Veterans Health Administration Inpatient Evaluation Center. We thank the staff at Veterans Affairs health care systems across the nation and territories for the routine input of data into the Legionnaires disease reporting tools and monitoring of our veteran patients with Legionnaires disease. Gregory S. Roth, BA, Veterans Health Administration Inpatient Evaluation Center, assisted with developing the Legionnaires disease reporting tools and maintaining the system. Patricia A. Lye, BA, MT(ASCP), Veterans Health Administration National Infectious Diseases Service, provided LD case validation support. Ms Lye and Mr Roth did not receive compensation for their contributions.