Background
Legionnaires’ disease (LD) is an atypical pneumonia caused by the
Legionella bacterium. The majority of human infections are caused by
Legionella pneumophila, mainly serogroups 1 and 6 (
1). Other species such as
L. longbeachae can also cause disease (
2). Although LD is generally an uncommon and sporadic infection with a low attack rate, case fatality rate is high, typically 10% in Europe and between 15 and 34% for nosocomial cases (
3). The incubation period for LD is 2–10 days with a median of 6–7 days (
3). Worldwide, 75–80% of the reported cases are over 50 years and 60–70% are male (
4).
Legionella bacteria are ubiquitous in nature. The
pneumophila serotype is primarily found in man-made freshwater reservoirs, especially standing water where biofilms can develop (
5,
6). Legionella bacteria are also found elsewhere in the environment. For example,
L. longbeachae is often found in soil, compost, and potting-mixes (
2,
7). The main infection route for
Legionella is through inhalation of contaminated aerosols. Outbreaks of LD in Norway have been linked to cooling towers, hot tubs, and an industrial air scrubber (
8).
A urine antigen test (UAG) has become the most commonly used diagnostic method in Norway, as in most other countries (
3,
9). This test only detects
Legionella pneumophila serogroup 1 with an overall (pooled, weighted) sensitivity of 74% (from 54 to 91% depending on brand) and overall specificity of 99% (
10). The sensitivity is higher in community acquired and travel-associated cases compared to nosocomial cases (
11). Sensitivity is also higher in severe clinical illness and when urine is analysed after concentration methods (
12). Ideally, a positive UAG result should be confirmed by culture and isolation. The reference standard for diagnosis of LD is culture and isolation from bronchoalveolar lavage (BAL), sputum or biopsy or fine needle aspiration from lung tissue (
3). Nucleic acid detection in BAL, sputum or lung tissue is also carried out by hospital laboratories in Norway. Diagnosis by serology is another alternative, although seroconversion in most culture-positive patients is not detectable until at least 3 weeks after infection, and never detectable in up to 25% of culture positive patients (
1). Challenges with LD surveillance include under-diagnosis due to the UAG only detecting one serogroup of
L. pneumophila (
10). If no other test is applied, LD caused by other species or serogroups may go undetected. If the patient is successfully treated, LD may never become established. This is also true for patients not tested for legionella at all, which could include those with less severe clinical presentation. If notification for surveillance is carried out by laboratories, bedside testing, which is possible with UAG, may reduce the notification rate.
In Norway, LD has been mandatorily notifiable to the Norwegian Surveillance System for Communicable Diseases (MSIS) since 1980. The majority of patients are infected abroad, but major local outbreaks of LD occurred in 2001 (
13) and 2005 (
14). In the 2005 outbreak, 10 out of 56 registered patients died (
14). Before 2001, the annual number of reported cases was fewer than five cases most years. After the two large outbreaks, awareness increased. From 2006 to 2017, the average annual notified incidence was 42 cases, with an increasing trend.
In the US, a marked increase in incidence of notified cases was seen from year 2003 (
15), and a three-fold increase from year 2000 to 2009 (
16). In the EU/EES area, the age-standardised incidence rate of notified cases increased from 0.97 cases/100000 in 2011 (
17) to 1.8 per 100,000 in 2017, which was a 30% increase compared to 2016 (
18). Notified LD incidence normally peaks during the summer months. This was reflected in the summer months of 2017, when the EU/EES experienced the highest notification rate in five years. However, there was a decrease in notified cases in Norway during the same time period (
18,
19), but no known interventions that could explain a decrease in LD incidence. This difference in Norway raised concerns of under-reporting to MSIS, and warranted an evaluation of the national LD surveillance. We carried out an evaluation of the surveillance system to determine whether it accurately detects cases and outbreaks and describes trends, in order to be able to give recommendations for improvement of the surveillance.
Discussion
We evaluated the national surveillance of LD through MSIS with regard to outbreak and case detection and capacity to detect changes in incidence by time, place, and person. Our results suggest that the system overall functions well but with some room for improvement. The results from the linkage of MSIS to another data source with hospital treated LD cases suggest possible under-reporting to MSIS, and that the sensitivity of the system can improve. The survey results suggested differences in representativeness and acceptability of the system, which supports that MSIS does not capture all cases of LD. For cases notified to MSIS, the timeliness and the data quality were good for key variables for the response. If under-reporting is consistent with regard to time, place and person, the system would allow for changes in incidence to be detected.
The estimated external completeness reflects that not all cases are notified to MSIS and smaller outbreaks may not be reported. However, the system has proven to be able to detect smaller LD outbreaks, for example one with five cases in 2008 (
8). However, the under-reporting to MSIS suggested by our results may be overestimated. The NPR is an administrative register, and it is possible that some of patients found only in NPR had a tentative LD diagnosis that was later rejected. In addition, we found evidence of incorrectly recorded LD diagnoses in patients only found in NPR. The correctness, as well as the completeness, of diagnostic coding in NPR varies between diagnoses (
26‐
28), and limitations were reported also for infectious diseases in NPR (
29,
30). In addition, it is also possible that some patients found only in NPR fell ill and sought medical care abroad and were only admitted to the hospital in Norway for follow-up care. These groups of patients are not notified in MSIS, meaning the external completeness of MSIS is likely higher than estimated here. Further, one assumption of the capture-recapture method is that the two data sources are independent. However, both cases notified in MSIS and recorded in NPR were treated in the same hospital, and the assumption does not hold. The estimate of total number of cases is an over-estimate, which adds to a false low external completeness estimate for the system. One alternative to explore, in order to facilitate surveillance, may be real-time data linkage of NPR and MSIS.
Patients only found in MSIS reduce the external validity of MSIS. None of these patients with a Norwegian personal identification number could be manually matched to patients only in NPR. Some of them might be recorded with other diagnoses in NPR, for example pneumonia. It was beyond the scope of this study to investigate the LD diagnosis of patients found only in NPR or MSIS in more depth through a review of medical journals or full data from NPR including every diagnosis in the study period, and this would be interesting to assess further.
The internal completeness and internal validity were high for key variables, meaning the data quality was high for cases that were notified to MSIS. These attributes are important to be able to produce accurate statistics for different subgroups of patients, and the long-term effect of any interventions. Validating patient data against the population registry and contacting clinician and lab directly about empty data fields are routines which contribute to good data quality. Overall, the timeliness estimates for notified cases were fair. Because the date for immediate reporting is not always recorded in MSIS, but the date of notification from the clinician and laboratory is always included, the true timeliness of the surveillance is likely better than our estimates suggest.
Our survey suggested that both the representativeness (who is tested for LD and how) and acceptability (knowledge of notification criteria, routines for notification, use of MSIS data), were fair, but that the system is not used to its full potential. Improved representativeness and acceptability would increase the sensitivity of the system. Because the MSIS system is common to all notifiable diseases, a lack of awareness of notification criteria could potentially affect the surveillance of other conditions that are notifiable to MSIS. Several survey responses suggested lack of awareness of the “immediate reporting” component of the surveillance, which needs to be prompt to prevent further cases. The survey results also indicated a lack of standardised procedures for Legionella testing in hospitals. If case-ascertainment varies between hospitals and hospital units, this reduces the representativeness, as well as the sensitivity of the surveillance. Case-ascertainment was not explicitly part of our evaluation and we would need another study design to be able to assess this in depth.
Globally, it is assumed that LD is under-diagnosed and that
Legionella is an under-recognised cause of pneumonia (
31‐
34) and it may be beneficial to carry out a multi-country evaluation to identify common obstacles for surveillance. At this time, it is currently not possible to directly compare LD incidence in Norway to other countries as the prevalence of
Legionella sources and travel patterns may differ. However, possible reasons that may also be relevant to Norway include that LD is considered a severe disease and patients with less severe illness may not be tested for
Legionella. This is supported by our finding that standardised criteria for whom to test is lacking. Further, the commonly used UAG only detects
L. pneumophila serogroup 1. Our survey suggested another test methodology is commonly applied if LD is suspected, but if treatment which covers also
Legionella is initiated, and the patient recovers, the cause of the pneumonia may never be identified. This was mentioned in the free text comments of the survey (data not shown). In the US, surveillance was biased towards more severe LD cases, who were more likely to be tested for LD, missing those empirically treated with antibiotics active against
Legionella spp. and/or not requiring hospitalization (
35). Moreover, patients with travel-history were more likely to be tested for LD in the US (
16), which our survey results also suggested. The internal completeness for the variables that define a case as travel-associated was in our evaluation good. For gastro-intestinal infections, the data quality of these variables was questionable, as completion depends on what the General Practitioner (GP) knows or assumes (
36). The majority of LD cases diagnosed in Norway are reported as associated with (international) travel, and it is possible that the proportion travel-associated cases is over-estimated, if illness is assumed to be associated with any recent travel. However, if this increases the test activity for legionella, it will improve the sensitivity of the surveillance.
As expected, the UAG was the most frequently used diagnostic test for
Legionella. The test sensitivity has limitations (
10), meaning that although the survey suggested that another test method is commonly applied upon a negative UAG, it is possible that some cases caused by both
L. pneumophila serogroup 1 and other serogroups or species go undiagnosed, which reduces surveillance system sensitivity. Moreover, not more than 64% stated that positive UAG results are confirmed with culture and isolation “always” or “usually”. The MSIS notification criteria do not require a confirmatory test to be carried out. In order to ensure the representativeness of the system, use of common (national) guidelines on both which patients to test for
Legionella, and on confirmation of positive as well as negative results would be ideal. There are national guidelines for treatment of LD which also include diagnostic procedures (
37) but those who in the survey stated that they do have routines referred to internal guidelines.
Before the study, one of our hypotheses was under-reporting to MSIS due to UAG carried out in the hospital units. Since the hospitals’ microbiological laboratories report all positive test results on any notifiable disease daily, one could assume that notification rates would benefit from tests carried out by the laboratories. However, because a high proportion of survey respondents stated the UAG analysis is carried out by the hospital laboratory, this under-reporting is likely not extensive. However, in this study we did not assess the completeness, timeliness, and routines for MSIS notification by the primary microbiological laboratories.
The demographics of notified cases were consistent with what one would expect for LD based on known risk factors, and so was the reported county, meaning no demographic group appeared over- or under-represented.
Limitations of the evaluation
We cannot know to what extent the survey answers are representative of medical doctors (and other health care staff) in Norwegian hospitals in general. The roles and responsibilities varied between the respondents. We did not have access to names or contact details to individual doctors, and there was no way we would be able to reach every eligible doctor. We also did not receive responses from every hospital, or from every unit in each hospital that we wanted to reach. Nevertheless, the overall response rate was better than anticipated. The survey had six replies from doctors in units that would not be expected to treat LD patients, such as a cancer or women’s health clinic. We suspect that they might belong to a larger unit that we asked the survey to be forwarded to, such as general medicine units. However, the answers from such units did not stand out as having many “don’t know” answer and were retained in the data.
Although the clinical criteria for notification of LD to MSIS is pneumonia, it is theoretically possible that a patient with Pontiac fever, a milder non-pneumonic form of legionellosis, could be notified to MSIS. However, such patients would in Norway visit their GP who is highly unlikely to request a test for legionellosis. For the linkage of MSIS and NPR we assumed any patients with LD would be treated in a hospital, not by a GP whose reimbursement claims are not in NPR.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.