This study provides a comprehensive analysis of the epidemiology of legionellosis over the 21-year period of 2000–2020 in New Zealand by using notifications and national laboratory-based surveillance of Legionella cases. The study period saw a large increase in disease incidence driven by several factors that we investigated.

A major finding is the marked increase in the reported incidence of legionellosis from 2010 onward (Figure 1; Appendix Table 1). This increase is associated with improved case ascertainment, likely driven by increased clinical awareness of legionellosis and increased availability of specific laboratory testing for legionellosis. The marked increase in legionellosis notifications during 2015 and 2016 was caused by the LegiNZ prospective study, which provided a 12-month period (May 2015‒May 2016) of intensified surveillance (11). During that study, all lower respiratory samples from hospitalized notified-case patients who had suspected pneumonia were tested for Legionella spp. by PCR. An increase in case detection in 17 regions (Figure 5) was expected, with the national 86% increase more likely caused by historical underdiagnosis of the disease, rather than an increase in disease burden (12). For that reason, the legionellosis rate of 3.9 cases/100,000 during 2010–2020 probably provides a more valid estimate of the true population rate than seen previously; the higher rate of 5.4 cases/100,000 population detected by the LegiNZ study is likely to be particularly robust. Those rates put New Zealand above an estimated global mean rate of 2.8 cases/100,000 population (95% CI 2.7–2.9 cases/100,000 population) derived from the reported contribution of Legionella species to community-acquired pneumonia in multiple countries (1). The relatively small increase in the mortality rate of legionellosis during this period (from 0.6 deaths/100,000 population in 2000–2009 to 0.7 deaths/100,000 population in 2010–2020) is also consistent with the conclusion of greater case ascertainment of less severe cases being the main driver of the apparent increase in disease incidence during this period.

Unlike jurisdictions outside New Zealand that observed a temporary decrease in legionellosis at the beginning of the COVID-19 pandemic (13), rates for New Zealand did not appear to have been affected in 2020 (Figure 1). This finding suggests that environmental exposures to Legionella species may not have changed in New Zealand during this period, which could reflect the COVID-19 elimination strategy that enabled ordinary life to continue for most of that year, with only a few weeks under lockdown (14). The pandemic and its response had complex effects on the epidemiology of many infectious diseases and their surveillance. For example, studies undertaken in other jurisdictions outside New Zealand have identified greatly increased Legionella microbial contamination in building water systems (cooling towers) linked to extreme water stagnation caused by prolonged closures of commercial buildings, reinforcing the need for monitoring water and air conditioning systems (15).

The findings of this study have shown the dramatic shift in legionellosis diagnosis during a period when traditional techniques were largely replaced by molecular methods and UAT. However, a key limitation of UAT is that it cannot detect organisms other than L. pneumophila serogroup 1 (16); some authors have suggested that a total dependence on this diagnostic assay may miss up to 40% of legionellosis cases (17). Another limitation of UAT is that it does not generate material that can be used for typing methods. In New Zealand, the Legionella UAT has been used by several laboratories since 1998 but has decreased utility because of the high proportion of legionellosis caused by non–L. pneumophila species, such as L. longbeachae (Figure 3; Appendix Table 3). In this setting, a negative UAT result does not exclude legionellosis and necessitates further testing to elucidate either exclusion or inclusion. Because only 20.3% of the 2,675 cases diagnosed during the study period were caused by L. pneumophila serogroup 1 infection (Appendix Table 2), potentially 80% of cases could be missed if only UAT were used. A recent evaluation of the UAT for the diagnosis of L. longbeachae infection indicated a sensitivity of 59.1% and specificity of 82.2% (18). Further development of the assay should improve sensitivity to strengthen its application as a useful diagnosis tool, particularly in laboratories in which there is limited molecular testing capacity and because of the ease of specimen collection and rapidity of diagnosis (18).

During the 1990s, the drive for better diagnostic methods led to development of several PCRs. However, a combination of factors, including test cost, reagent quality issues, contamination problems, and the lack of trained and skilled staff, resulted in the initial slow adoption of molecular diagnostics for legionellosis (19). The technology has now matured, and since 2010, when some laboratories in New Zealand began routine molecular diagnostic testing for legionellosis, it has now become the method of first choice. This change was caused by the availability of more robust and sensitive assays for the detection of many difficult to diagnose diseases, in addition to legionellosis, and the overall reduction in test costs. Molecular testing for legionellosis has also been driven by its superior diagnostic utility compared with traditional methods because it enables detection of all Legionella species and can obtain a result within hours of sample collection (20).

The shift in laboratory methods during 2000–2020 has influenced the ability of routine surveillance to detect the contribution of species and serogroups. We observed a marked increase in L. longbeachae cases compared with other species, such as L. pneumophila, since 2010 (Figure 2; Appendix Table 2). We also observed a decrease in the identification of the serogroup for Legionella species caused by reductions in the use of the traditional methods, namely culture and serology, which risks gaps in surveillance information and can hinder cluster analysis and source tracing (Appendix Table 2) (21). This trend was caused by increased use of PCR testing alone that identifies the species, but not the serogroup of the Legionella species. No single laboratory test combines both optimal diagnostic accuracy with the ability to epidemiologically type the causative agent. To achieve this feature, a combination of molecular testing supported by culture, serologic testing, or both is required (19). In recent years, whole-genome sequencing has emerged as a major tool to support epidemiologic investigation (suspected clusters and outbreaks) of Legionnaires’ disease and for characterization of new strains, but this method still requires the culture isolation of the bacterium.

The results of this study show that the rates of legionellosis were highest in adults >60 years of age and in male notified case-patients, consistent with previously reported research (6). As the population ages in New Zealand, the burden of legionellosis is likely to continue to increase in the absence of effective measures to prevent or adequately control the risk for infection. Legionellosis rates were higher in persons from Europe compared with Māori, Pacific Peoples, and persons of other ethnicities during 2000–2009, even after age standardization. Those differences largely disappeared during 2010–2020, corresponding with increased case ascertainment. This pattern is different from that seen for other serious infectious diseases, for which rates are markedly higher for Māori and Pacific Peoples (22). Those unexpected differences need further investigation to see if there is systematic underdiagnosis of legionellosis across ethnic groups or if differences in exposure might explain the pattern seen.

This analysis found that infections caused by L. longbeachae increasingly dominated over those caused by L. pneumophila (Figure 2; Appendix Table 2). The largest contribution to this increase in L. longbeachae cases came from persons >60 years of age (Table 1). Early spring‒summer clusters of L. longbeachae infections are seen each year and might be linked to increased gardening activity in warmer months, which has been shown to provide several psychological, physical, and social benefits for older persons (23). In contrast, L. pneumophila infections appear to be spread evenly throughout the year, and transmission by aerosols containing contaminated water from cooling towers was the most identified source from outbreak investigations in New Zealand. Decreased incidence rates observed across the 3 Auckland DHBs between decades might reflect introduction of a bylaw in 2015 requiring owners to register their industrial wet cooling tower systems annually and monitor Legionella bacteria levels (24). In contrast, higher incidence rates in the Northland region might reflect the readiness of clinicians to consider testing for Legionella species in response to the national surveillance study (11).

Our study also provides data on major risk factors, exposures, and potential transmission settings. Few cases were classified as hospital acquired. A small percentage (6.2%) were classified as travel-associated, based on having a history of overseas travel during the incubation period. That percentage is lower than that for Europe, where 14.7% of detected Legionnaires’ disease cases in 2019 were linked to travel abroad, of which 79% were linked to overnight stays in hotels (25). This difference is observed despite New Zealand having among the highest per capita international travel rates in the world, with >3 million residents departing New Zealand in 2019 (26).

Our study has several limitations associated with the use of routinely collected surveillance data. The most critical limitation is the long-term underascertainment of legionellosis. This limitation has been partially corrected by using more sensitive molecular testing, resulting in a marked increase in measured rates of legionellosis during the 2010‒2020 period. A further limitation is the incomplete reporting of some variables.

More research on the epidemiology of legionellosis in New Zealand is warranted. The high percentage of hospitalizations (82.0%) reported during the study period means that those data can be analyzed to provide a useful basis to identify emerging issues and determine priorities for prevention. For example, the discharge data could be used to estimate the economic cost of hospitalized cases of legionellosis in New Zealand. It would also be useful to investigate the contribution of Legionella infection to the burden of community-acquired pneumonia of mild-to-moderate severity, which will often be treated empirically outside the hospital setting without any etiologic diagnosis. Proposed interventions to reduce the effect of legionellosis in New Zealand should also be evaluated after they are implemented to determine their efficacy.