Epidemiology of Legionnaires’ Disease in Italy, 2004–2019: A Summary of Available Evidence

Legionnaires’ disease (LD) incidence has been increasing in several European countries since 2011. Currently, Italy is experiencing high notification rates for LD, whose cause still remains scarcely understood. We sought to summarize the available evidence on the epidemiology of LD in Italy (2004–2019), characterizing the risk of LD by region, sex, age group, and settings of the case (i.e., community, healthcare, or travel-associated cases). Environmental factors (e.g., average air temperatures and relative humidity) were also included in a Poisson regression model in order to assess their potential role on the annual incidence of new LD cases. National surveillance data included a total of 23,554 LD cases occurring between 2004 and 2019 (70.4% of them were of male gender, 94.1% were aged 40 years and older), with age-adjusted incidence rates increasing from 1.053 cases per 100,000 in 2004 to 4.559 per 100,000 in 2019. The majority of incident cases came from northern Italy (43.2% from northwestern Italy, 25.6% from northeastern Italy). Of these, 5.9% were healthcare-related, and 21.1% were travel-associated. A case-fatality ratio of 5.2% was calculated for the whole of the assessed timeframe, with a pooled estimate for mortality of 0.122 events per 100,000 population per year. Poisson regression analysis was associated with conflicting results, as any increase in average air temperature resulted in reduced risk for LD cases (Incidence Rate Ratio [IRR] 0.807, 95% Confidence Interval [95% CI] 0.744–0.874), while higher annual income in older individuals was associated with an increased IRR (1.238, 95% CI 1.134–1.351). The relative differences in incidence between Italian regions could not be explained by demographic factors (i.e., age and sex distribution of the population), and also a critical reappraisal of environmental factors failed to substantiate both the varying incidence across the country and the decennial trend we were able to identify.


Introduction
Legionellosis is a collective term encompassing two distinctive clinical syndromes caused by Gram-negative bacilli from the genus Legionella, i.e., Legionnaires' Disease (LD) and Pontiac Fever (PF) [1][2][3]. Both disorders result if the bacteria are inhaled or aspirated but have strikingly different clinical characteristics. On the one hand, LD is a severe, sometimes fatal interstitial pneumonia, that shares several clinical features with pneumococcal and other bacterial pneumonia, with characteristic common extrapulmonary manifestations (i.e., renal failure, encephalopathy, pericarditis). On the other hand, PF is afebrile and generally flu-like illness persisting from two to six days, usually self-limiting and deprived of pulmonary parenchymal disease [2,3].
However, it must be stressed that our awareness of the actual incidence of LD remains largely unknown, for several reasons. Firstly, most of our understanding comes from symptomatic patients, who present with CALD/TALD or HALD. As LD cannot be clinically or radiologically distinguished from pneumonia cases of different etiology, a low suspicion index means that a significant share of all diagnoses is simply missed because specific diagnostic testing is not performed [6,23,24]. As several studies have shown antibody levels in healthy populations ranging from less than 1 to 45.1% [1], the overwhelming majority of all infections usually occur unnoticed, with only a small fraction of all cases developing either PF or LD. Second, when testing is done in acute disease, clinicians mostly rely on urinary antigen tests (UAT), that represents 82 to 97% of diagnosis in Europe and in the United States, respectively [6,24]. UAT can be performed rapidly and with very high specificity for Lp1 but has a very low sensitivity for non-Lp1 antigens [3,8,25]. Alternative diagnostic tests and laboratory procedures (i.e., culturing, identification of the bacterium using paired serology, detection of the bacterium in tissue or body fluids by immunofluorescent microscopy, and genotypic polymerase chain methods), are in turn affected by specific shortcomings including higher costs and increased turnaround times, being therefore inconsistently deployed because of economic and/or practical reasons, and suggesting that a large share of all cases may remain simply undiagnosed.
In other words, LD is an important public health problem in terms of morbidity and mortality, particularly in certain high-risk groups, but our incomplete understanding of its epidemiology impairs an appropriate implementation of response measures, including environmental interventions, preventive campaigns focusing on the safety of potential sources of the infection, and risk communication. As a consequence, an updated and comprehensive reappraisal of surveillance data may be particularly useful to national and local health authorities.
In Italy, epidemiological surveillance for LD started in 1983, when the Legionellosis National Registry was established and managed by the Italian National Institute of Health (Istituto Superiore di Sanità, ISS). Notification of LD became mandatory in 1990 [8], and only in 2000, the Italian Institute of Health (Istituto Superiore di Sanità, ISS) issued its first guidelines for the prevention and control of legionellosis. This document was followed in 2005 by instructions targeted to microbiology laboratories, environmental control, tourist accommodation, and spas. Later, in 2015, all national recommendations, including those for hospitals, were included in a single updated document [8,26].
Even though earlier data suggest a significant increase in notification rates, an updated and comprehensive review of Italian data has not been reported [8,24,27,28]. Our aim was therefore to collect available evidence on the temporal and spatial patterns of LD in Italy, assessing the characteristics of incident cases in the subsequent timeframe of 2011-2019. Eventually, we performed a comprehensive analysis on the possible influence of environmental factors on the epidemiology of LD, focusing on climate and demographics factors.

Settings
Italy is administratively divided into 19 ordinary regions and 2 autonomous provinces (i.e., AP of Trento and Bolzano). With a surface of 301,340 km 2 (116,350 sq mi), a total population of approximately 60 million inhabitants, and a population density of 201 people per square kilometer (520/sq. mi.), Italy is a densely but unevenly populated country in southern Europe. In fact, northwestern (Piedmont, Aosta Valley, Liguria, Lombardy) and northeastern regions (Veneto, AP of Trento and Bolzano, Friuli Venezia Giulia, and Emilia Romagna) encompass around 40% of the total area, but more than 45% of the total population, with low unemployment rates and high GDP per capita. Central (i.e., Lazio, Tuscany, Umbria, and Marche; 19% of total area), southern (i.e., Abruzzo, Apulia, Basilicata, Calabria, Campania, Molise), and Main Islands (i.e., Sicilia and Sardinia) are characterized by lower population density, lower economic development and higher unemployment rates that, in the past decades, were determinative of lower living standards for the inhabitants of southern regions, causing an intensive internal migration towards northeastern and northwestern Italy.
After nearly two decades characterized by stagnating population growth, in the time period 2001-2011, almost three million more Italians were registered, nearly all being young immigrants who therefore contributed to slow down the aging of the Italian population. According to available figures, people aged 65 and over increased steadily during past decades, now representing approximately 20% of the total population. Around 3% of them are assisted in the three main different kinds of Italian residential/institutional services: nursing homes, protected homes, and social health residential structures (i.e., structures focusing on older people or people with disabilities who require special care and support of medical, nursing and/or rehabilitation services).
According to the present EU case definition of LD, and to the national guidelines for Legionella spp. control and prevention [13,24,26,29], a confirmed case of LD is a patient presenting clinical and/or radiological signs of pneumonia, with at least one of three laboratory criteria including: isolation of Legionella spp. from respiratory secretions or any normally sterile site; detection of Lp1 antigen in UAT; rise in specific antibody level to Lp1 in paired serum samples. A probable case is a patient presenting clinical and/or radiological signs of pneumonia associated with a single high level of specific antibodies to Lp1 (≥1:256), or a positive direct immunofluorescence test, or a positive PCR. Until 2012, cases were reported according to the guidelines of the European Working group for Legionella Infections (EWGLI) [30], with confirmed cases including subjects having a clinical or radiological diagnosis of pneumonia with laboratory evidence of one of the followings: isolation of Legionella species from clinical specimens, a positive UAT for Lp1 using validated reagents or kits, seroconversion (i.e., 4-fold or greater increase in titre) at indirect immunofluorescent antibody test for Lp1. A presumptive case was defined by either a clinical or radiological diagnosis of pneumonia with laboratory evidence of: detection of Legionella spp. nucleic acid in a clinical specimen; a positive direct fluorescence on a clinical specimen using validated L. pneumophila monoclonal antibodies; a single high titre for Lp1 antigen (≥1:128); a 4-fold increase in antibodies against other Legionella spp. or non-Lp1 infections.
Annual reports on LD include: • the total number of diagnosed cases during the calendar year by gender, age group, and region of origin; • the total number of probable/presumptive vs. confirmed cases; • share of HALD, defined as patients residing in a hospital or nursing home for the entire 10 days before onset; • subset of cases occurring in nursing homes for the entire 10 days before onset (available for 2007-2019 only); • total number of TALD (either Italian and foreigner having being infected in Italy), defined as cases linked to tourist recreational facilities and tourist turnout; • share of CALD, i.e., cases where no specific risk factor was identified in the 10 days before the onset of clinical symptoms, and therefore considered sporadic/community acquired; • pre-existing medical conditions; • diagnostic procedures performed in notified cases (i.e., urinary antigenic testing; isolation of the pathogen, PCR analysis) For each study year, data on the Italian population, including year income for general population and renters, were obtained at a regional level from the Italian National Statistical Institute (ISTAT; http://demo.istat.it/, (accessed on 1 September 2021)) [31].
Pooled year meteorological data were obtained for each regional capital from the competent Regional Environmental Protection agencies (in Italian, ARPA). The Italian ARPA are the Italian environmental agencies, one for each region of Italy (excluding Trentino-Alto Adige/Süd Tirol, which has been split for the two Autonomous Provinces of Trento and Bolzano). When more weather stations were available at the same geographical level for the same timeframe, mean values were calculated.

Statistical Analysis
We performed a descriptive analysis of the surveillance data, i.e., geographical and temporal distribution of LD cases, with their respective demographic characteristics (age, sex), and clinical outcome, where available. Crude incidence rates (CIR) were calculated both at a national and regional level, with subsequent estimates of case fatality ratio (CFR), and crude mortality rates. Age-adjusted incidence rates (ASR) were calculated at a national level assuming the European standard population as reference [32]. In order to ascertain the impact of HALD over the total of cases locally reported, the share of HALD reported by the index region (i) was compared to the share of all HALD cases at a national level (I HALD = i HALD /Total HALD ), with a similar estimate for the whole of reported LD by the same region (I LD = i LD /Total LD ). The regional HALD Risk Ratio was then calculated as I HALD /I LD .
We analyzed then the corresponding annual incidence rates between 2004 and 2019, dichotomized by convenience in 2004 to 2011 vs. 2012 to 2019. Such cut-off was identified as, since 2012, European countries belonging to the European Legionnaires' Disease Surveillance Network (ELSDNet), including Italy, in accordance with the 2012 EU/EEA case definition, reported only cases with acute pneumonia [8,30,33], allowing a more appropriate comparison with European data. Annual trends for Italy and other European Countries [34][35][36] were initially compared through the Pearson's correlation test, and then assessed by normalizing the annual rates by the maximum number of new cases.
The relationship between the number of LD cases and both meteorological (i.e., average daily temperature, average precipitation rates) and demographic factors (i.e., share of population aged 50 years or more, annual income in the general population and in renters) was initially investigated through the Pearson's correlation test. Incident Rate Ratios (IRR) with their correspondent 95% CI were calculated in a Poisson regression model that included as the outcome variable either the yearly incident rates for LD (assessed at regional level) and the healthcare-related cases. The explanatory variables were represented by meteorological and demographic factors. In the analyses, meteorological data of the regional capital were assigned to all LD cases that occurred in all municipalities of the province.

Ethical Approval
No ethical approval was needed for this study, as no individual data were identifiable, and since we analyzed and presented only aggregated data.

Demographics
A total of 23,554 cases of LD were reported in Italy between 2004 and 2020 (Table 1) The main diagnostic option was represented by UAT (95.3%), while only 3.2% of all cases were identified by serology, whose share decreased from 5.7% in 2004-2011 to 1.9% in 2012-2019 (p < 0.001). A total of 17,193 cases (73.0%) were classified as CALD, as no specific risk factors were identified in the 10 days before the onset of symptoms, and their corresponding share increased from 65.8% in 2004-2011 to 76.6% in 2012-2019 (p < 0.001). On the contrary, HALD accounted for 1394 (5.9% of the total sample), and while their share substantially halved from 8.6% to 4.6% during the overall timeframe (p < 0.001), raw figures remained quite consistent (i.e., 84.4 cases/years 2004 to 2011, compared to 89.4 cases/years, t = −0.443 and p = 0.666 in Student's t-test). Unfortunately, data on residential homes were available only for the time period 2007-2019, with 523 total diagnoses, i.e., 2.5% of 21,158 total cases diagnosed in the respective timeframe. Moreover, a total of 209 cases were linked to dental care (5.9%), with no Eventually, 4967 cases were identified as TALD (21.1%), and again the corresponding share decreased from 25.6% in 2004-2011 to 18.8% in 2012-2019 (p < 0.001). Among TALD cases, 2188 occurred in foreigner travelers who presumptively were infected in Italy (9.3%), and 2779 cases in Italian travelers (11.8%). Of them, 257 (1.1%) were exposed to Legionella spp. abroad. Since 2013, involved structures increased from 47 to 81, and around one-third of them were previously characterized by incident cases (i.e., 30.3%, range 22.9% in 2019, 35.6% in 2013).
Unfortunately, reliable estimates for the nights spent in Italian Hotels and hospitality facilities for the assessed timeframe were not available, while a similar estimate was provided from the ISTAT for Italian international travelers. As a consequence, a crude incidence was calculated for this specific subgroup of travelers, with estimates ranging from 0.092 per 1,000,000 nights spent abroad in 2005 to 0.266 in 2018, and an average of 0.148 (95% CI 0.110-0.185) ( Table A1).
As reported in Table 2, crude incidence rates for LD increased from 1.048 cases per 100,000 (95% CI 0.958-1.139) in 2004 to 5.343 per 100,000 (95% CI 5.131-5.555) in 2019. The average yearly incidence was 2.464 cases per 100,000 (95% CI 1.880-3.048). AIR showed the very same trend, increasing from 1.053 cases per 100,000 (95% CI 0.869-1.237) in 2004 to 4.669 per 100,000 (95% CI 4.251-5.088) in 2019 (pooled estimate: 2.285 cases per 100,000, 95% CI 1.798-2.772). Table 2. Incidence of Legionnaires Disease (LD) in Italy (2004-2019). Crude incidence rates (CIR) and corresponding 95% confidence intervals (95%CI) were estimated assuming the total Italian population as the reference. Age-adjusted Incidence Rates (AIR) were calculated by means of the corresponding year census. Healthcare-associated cases of LD (HALD) were reported as percent value over the total of incident cases. AIR of Italy and EU-EEA countries were compared for the timeframe 2004-2019 and were well correlated (r = 0.871, 95% CI 0.647 to 0.956) (Figure 1). However, when the trend was assessed either as annual incidence estimates and incidence rates normalized by the maximum value of the assessed timeframe, a quite distinctive pattern was identified, with Italy exhibiting a surge of notified cases that substantially outmatched that of EU-EEA since 2012 ( Figure A1).

Geographic Distribution
Focusing on the geographic origin of reported cases, the large majority them were from the northern region, i.e., 43.2% from northwestern Italy, and 25.6% from northeastern Italy. During the assessed timeframe, the share of cases reported from northeastern regions increased from 21.9% to 27.4%, while the corresponding proportion of cases from northwestern Italy decreased from 47.2% to 41.1% (p < 0.001).
As shown in Table 3 Interestingly enough, AP of Trento was also characterized by the highest share of cases occurring as HALD (i.e., 18.0%), followed by Basilicata (16.2%), Calabria (9.7%), and Piemonte (9.1%), while in Lombardy and Emilia Romagna it was estimated to 7.2% and 6.3%, respectively.
The   CFR, including both CALD and HALD, was estimated to be 5.2% for the time period 2004-2019. As shown in Table 4, the majority of deaths occurred in non-healthcare settings cases (1002 vs. 221 HALD), with respective CFRs of 4.6% and 11.5% for CALD and HALD, respectively (p < 0.001). When national data were broken down by the timeframe (i.e., 2004-2011 vs. 2012-2019), CFR decreased from 5.8% to 4.9%, with a similar trend in non-healthcare (i.e., 5.0% vs. 4.5%), and in healthcare-related cases (13.3% vs. 10.2%). Table 3. Incidence of Legionnaires' Disease (LD) in Italy (2004-2019) by region. Crude incidence rates (CIR) and corresponding 95% confidence intervals (95%CI) were estimated assuming the regional Italian population as the reference. HALD Risk Ratio was calculated as the ratio between the share of healthcare-associated LD cases (HALD) on the whole of cases reported at national level (No. 1394) and the share of all cases occurring in the same region compared to all notified cases (No. 23,532).

Analysis of Risk Factors
Consistently with previous estimates, by assuming the timeframe 2004-2011 as the reference (  Regarding the geographic origin of reported cases, by assuming northwestern Italy as a reference, a substantially reduced risk for new diagnoses was identified in all other areas, from northeastern Italy (RR 0.819, 95%CI 0.793-0.845), to central Italy (RR 0.643, 95%CI 0.622-0.665), and particularly in southern regions (RR 0.209, 95%CI 0.199-0.220), and major islands of Sicily and Sardinia (RR 0.071, 95%CI 0.064-0.080).
Eventually, when focusing on the LD-associated deaths, cases occurring in healthcare facilities and residential homes were associated with a nearly doubled risk for eventual death compared to CALD (RR 2.489, 95% CI 2.169-2.858), which was substantially decreased in 2012-2019 timeframe compared to 2003-2011 (RR 0.835, 95%CI 0.746-0.934).
When meteorological factors were taken into account alongside the share of the population aged 50 years or more, and the year income (both for renters and general population), the latter factors were significantly correlated with incidence rates for LD (r = 0.405, 95% CI 0.175-0.594, p < 0.001; r = 0.695, 95% CI 0.540-0.804, p < 0.001; r = 0.643, 95% CI 0.470-0.768, p < 0.001; respectively) ( Table 6). Interestingly enough, the occurrence of HALD was in turn negatively correlated with air temperatures (r = −0.309, 95% CI −0.518 to −0.067), income of renters (r = −0.416, 95% CI −0.602 to −0.188, p < 0.001), and income of the general population (r = −0.259, 95% CI −0.476 to 0.012, p = 0.041). Table 6. Correlation of outcome variables total incident cases of Legionnaires' Disease (model 1) and healthcare-related cases (model 2), with average temperatures, daily precipitation rates, share of total population aged 50 years or more, and income (both for renters and for general population). In fact, when climate data were included in a Poisson regression model with the share of older residents over the total population (i.e., aged 50 years or more), air temperatures, daily precipitation rates, and annual income (both for general population and renters), a more distinctive pattern was identified. As shown in Table 7, a decreased estimate for the average air temperature was identified for both incident cases as a whole, and HALD cases (+1.0 • C; IRR 0.807, 95%CI 0.744-0.874 and IRR 0.884, 95% CI 0.783-0.999, respectively). On the contrary, while year income of renter (+1000 €/year) was associated with an increased occurrence of LD cases (IRR 1.238, 95% CI 1.134-1.351), it was characterized as a negative explanatory variable for healthcare-related cases (IRR 0.888, 95% CI 0.816-0.966). Table 7. Incidence rate ratios (IRR) for Legionnaire's Disease (LD) cases in Italy (2004-2019) by meteorological factors, and main characteristics of the index areas. IRR were calculated by means of a Poisson logistic regression assuming as outcome variable the annual incident cases of LD, both in total (total incident cases) and healthcare-related ones.

Discussion
Our review of surveillance data suggests that the notification rates of LD in Italy have substantially increased across the assessed timeframe 2004-2019. In other words, not only do our analyses mainly confirm the surge of LD cases reported by Rota et al., since 2013 [8], but our estimates are also quite consistent with available European data [34][35][36]. More precisely, the pooled incidence rate has increased 4. Collectively, such results point towards an increased risk for LD in Italy compared to most European countries, and particularly in the last decade.
To date, this upsurge is a common and not well-understood feature of nearly all high-income countries. Several explanations have been proposed [3,8,20,34,35], but none of them is either satisfying nor explicative [2,3,9,10,38]. In fact, we are unable to rule out that this increase in case numbers could be nothing more than an artifact. Better awareness of medical professionals towards a respiratory disorder otherwise deprived of clinical diagnostic features, associated with improved detection systems and changes in clinical diagnostic methods may have led to the increased testing frequency in high-risk individuals with clinical signs of pneumonia, particularly through UAT, and hence to an increase in detected cases [8,17,24,[26][27][28]. In other words, we cannot rule out that our data could have been affected by a "diagnostic epidemic" too, rather than representing an actual epidemic of the pathogen. As a corollary, we could also speculate that the lack of appropriate testing in the previous decades could have impaired our understanding of the true LD burden of disease, which is progressively ascertained [17,39,40]. However, this explanation is quite unlikely. Not only were urinary diagnostic tests introduced more than 20 years ago [41], well before the surge in LD notification rates (around 2013), but the proportion of positive cases on the whole of collected specimens remained quite consistent, and also the share of LD diagnosed by means of urinary assays has been substantially stable in the last years [1,41]. Moreover, it should be kept in mind that UAT, despite its convenience and the relatively reduced turnaround times when compared to alternative diagnostic methods (i.e., culturing, serology, immunofluorescent microscopy, and genotypic polymerase chain reaction) [1], is reliable only for Lp1 strains [3,10,21,25]. As 95.3% (and 96.8% since 2012) of reported cases were detected through UAT, we cannot rule out that official estimates might be somewhat underestimating the actual burden of disease.
Another possible explanation has been found in the ongoing demographic transition. Since the earlier reports, older age has been identified as the main risk factor for LD, and also in our data people aged 40 to 49 years old had an RR of developing LD that was 4.320 (95% CI 4.051 to 4.606) higher than in people younger than 40 years old, and it progressively increased to 6.259 (95% CI 5.894 to 6.647) in the age group 50 to 59 year-olds, 8.145 (95% CI 7.676 to 8.462) in the age group 60 to 69, 9.610 (95% CI 9.058 to 10.197) among individuals aged 70 to 79 years, and 9.447 (95% CI 9.361 to 10.565) in older subjects. In other words, the steady increase in the age of European, and particularly of Italian populations, has progressively overstretched the share of individuals being at high risk to develop LD, then contributing to the overall surge of incident cases. However, even though the proportion of subjects aged 50 years or more in the general Italian population has increased from 38.7% in 2004-2011 to 42.5% in 2012-2019, the occurrence of LD in these age groups has increased even more steadily (i.e., 77. More recently, climate change and several environmental features have been advocated as a possible contributor to the upsurge of notification rates for LD and other pathogens. This is particularly intriguing as Italy, and particularly its northern regions, has been severely hit by climate change in past decades [42][43][44][45][46][47], with well-documented consequences on the ecology of several infectious diseases [48,49]. Although Legionella spp. is common in the environment, dry environments do not support them, and the pathogen is particularly sensitive to drying conditions [50]. Even though available source data lacked the monthly occurrence of reported cases, and we were therefore unable to perform an accurate analysis of the seasonality of LD in Italy, a striking and well-known epidemiologic feature of LD, and particularly of CALD, is seasonality, as more cases are reported during the summer. Some evidence indicates that high environmental temperatures (and more precisely, large temperature excursions), and high relative humidity during the warm season, together with increased rainfall during the late spring months, would drive the summer spike in incidence [51][52][53]. As hinted by Hicks et al., it is reasonable that heavy rain can both favor the proliferation of host microorganisms or increase organic sediments in the water network which, in turn, support the growth of Legionella.
Still, multivariable analysis by means of a Poisson regression model identified air temperature as a negative effector (IRR 0.807, 95% CI 0.744 to 0.874), while precipitation rate was actually unrelated to the notification rate. In other words, our estimates were hardly comparable with available evidence but, contradictory as they were, they may find several explanations.
To begin with, it should be stressed that our analyses were forcibly coarse. Retrieved data were based on the region of origin: indeed, Italian regions represent a secondary administrative level, including quite large and heterogeneous areas. The reliability of the multivariable analysis may thus have been affected by other factors, including the economic development of assessed areas, and the appropriate notification of incident cases. For instance, being central and southern regions of Italy characterized by a warmer climate and lower economic development, the characterization of environmental factors and the annual income of renters as a potential risk factor (IRR 1.238, 95%CI 1.134 to 1.351) should be cautiously assessed.
In fact, a diffuse under-reporting in central and southern regions had already been addressed [8] and advocated as an explanation for the extensive heterogeneity among prevalence data we retrieved. In other words, the geographic trend in notification rates we identified, with estimates significantly decreasing from northwestern regions to northeastern, central and southern regions, and lower rates in Sicily and Sardinia, rather than stressing the relevance of environmental factors could be nothing more than an artifact.
The potential impact of misreporting on overall estimates is also hinted at by analysis of HALD cases. On the one hand, HALD cases are usually unrelated to environmental features; in fact, despite the substantial reduction of their proportion on all reported cases from 2004 (15.9%) to 2019 (4.5%), the raw number of actual reported cases remained quite consistent across the study period (84.4 cases/years 2004 to 2011, compared vs. 89.4 cases/years). On the other hand, calculation of HALD Risk Ratio mirrors the estimates for the overall incidence rates, suggesting that mostly in northern regions (i.e., 1.687, 95% CI 1.500 to 1.894 for northeastern regions, 1.244, 95% CI 1.055 to 1.432 for northwestern regions) their occurrence has been disproportionally high-or, conversely, that in most the central Italy, in southern Italy, as well as in Sicily and Sardinia, the occurrence of HALD cases was disproportionally low during the assessed timeframe.
Interestingly, official data on the consumption of antimicrobial drugs suggest a mixture of misreporting and heterogeneous management of pneumonia cases as another reliable explanation for retrieved data. According to the Italian Medicines Agency (i.e., Agenzia Italiana del Farmaco, AIFA in Italian), during the last decade, northern regions have experienced a sustained decrease in daily consumption of antimicrobial drugs having intracellular activity such as Fluoroquinolones ( Table A2) [54,55]. As clinical features of LD are quite indistinctive, it is reasonable that a large share of cases occurring in central and southern Italy are still treated with broad-spectrum antimicrobial drugs: as Legionella spp. remains quite sensitive to Fluoroquinolones, without early microbiological testing, the clinical recovery may hinder a proper diagnosis, with consequent under-reporting [9][10][11]. On the contrary, where guidelines have promoted the use of drugs such as macrolides and penicillin, less effective against Legionella spp., clinicians may more frequently perform diagnostic testing in order to identify the pathogen, properly tailoring the antimicrobial treatment, but also leading to an increased number of diagnosed cases of LD [56,57]. As current Italian guidelines do recommend the systematic referral to diagnostic tests for community-acquired pneumonia, including the UAT, the increasing share of CALD may be otherwise explained through an increased sensitivity of the clinicians towards LD.
Notwithstanding, climate change and economic development may have influenced the increasing occurrence of LD in a more subtle way, i.e., through the ever-increasing spread of air-conditioning systems, particularly in the community, in order to cope with the rising summer temperatures, and the progressive shift from heating systems based on steam and hot water to heat pumps. Air-conditioning systems and convector radiators are (alongside drinking fountains, hot tubs, sinks, toilets, sprinklers, and showers) well understood as common sources of Legionella. In fact, the pathogen tends to flourish when the facilities are not maintained in proper terms, e.g., water is sufficiently warm to allow the proliferation of Legionella, stagnant, deprived of chemical disinfectants, and/or pipes are progressively corroded, with a resulting abundance of nutrients. Moreover, Europe, and most prominently Italy, is not only an aging continent-but also its urban infrastructures are aging as well. In large part, the water distribution system is very antiquated; as water travels through the distribution system and enters a building, it can lose disinfectant as well as interact with the materials, temperatures, and design of the building's plumbing, with subsequent risk of its contamination by Legionella spp., that are then dispersed in the environment.
The increasing use of air-conditioning systems in a rapidly aging population may also explain the increasing occurrence of CALD compared to HALD. Focusing on our data, while in 2004 HALD cases accounted for 15.9% (95% CI 13.5 to 18.3), the corresponding share decreased to 4.5% (95% CI 4.4 to 4.8) in 2019, with a nadir of 4.0% (95% CI 3.5 to 4.5) in 2018. In fact, Italian health authorities have progressively implemented a series of preventive measures to avoid HALD cases, including HEPA (High-Efficiency Particulate Air) filters, that show only a limited pairing on the legal framework for domestic plants. An indirect confirmation may be found through the analysis of the corresponding trend in TALD.
TALD cases have played a central role in LD history since the first description of this disorder, and travel is a well-known risk factor for LD for a variety of reasons. First, water systems of accommodation sites are often complex, with a large number of outlets (e.g., showers) [58]. Therefore, not only are water systems of accommodation sites sometimes not regularly flushed, with water stagnation that favors the growth of Legionella, but the length of pipe systems impairs an adequate control of water temperatures. Moreover, accommodation sites are likely to have facilities such as swimming pools and whirlpool spas, which are in turn associated with an increased risk of LD and that, particularly in seasonal accommodation structures, are not periodically managed and sanitized [58]. Last but not least, hotels may host a large number of visitors, who might be exposed to the same source during their stay, with resulting clusters and even outbreaks [30,53,59]. As a consequence, specifically targeted interventions have been promoted not only at the Italian level, but more broadly at the European level, including the implementation of an integrated surveillance network in all EU-EEA countries [30,34,38]. Not coincidentally, even though the number of national and international tourists and tourism-related facilities has more than doubled during the last decade (according to the estimates of ISTAT, the number of international travelers in Italy skyrocketed from 165 million in 2010 to nearly 400 million in 2019), the raw number of notified cases between 2004-2011 and 2012-2019 exhibited a far more limited increase, from 2024 to 2943, with a substantial decrease for the proportion of TALD on the total of LD (i.e., 25.6 to 18.8%). It is, therefore, reasonable to think that the aimed interventions of health authorities on water and climatization systems of accommodation sites may have been instrumental in limiting the increase of new TALD cases.
The possible relevance of factors other than demographic and environmental ones are also suggested by the small but significant difference that was appreciated across the assessed timeframe regarding the gender of reported cases. Individuals of male gender were consistently more often affected than females (RR 2.522, 95% CI 2.453 to 2.594), but they accounted for 71.7% of total cases in 2004-2011 compared to 69.7% in 2012-2019. The higher occurrence of LD in males compared to females has been repetitively but unsatisfying inquired [8,17]. While it is quite unlikely that differences in health-seeking behavior may have resulted in such differences [17], other factors such as smoking habits and occupational exposures hardly explain both the higher risk of males and the progressively increasing share of female cases [60,61].
Limitations. Albeit interesting, our data are affected by significant shortcomings. First, we have drawn our estimates from National Bulletins, and therefore we were able to summarize and analyze only information preventively reported by National Authorities. As a consequence, not only do we lack significant data about the demographics of LD cases, with resulting uncertainties in eventual estimates, but significant information about the reported signs and symptoms were not available to our analysis. This is particularly interesting, as an accurate analysis of clinical features of reported cases may either confirm or dismiss the hypothesis that the upsurge of LD cases should be more properly retained as an intensification of testing for Legionella [2,3,25,38,62]. Second, National Bulletins obviously included reported cases, but both previous remarks and the significant heterogeneity in incidence estimates hint for an inconsistent and often inappropriate notification of incident cases to the competent health authorities [8,27,39,40]. Such heterogeneities, coupled with the potential unreliability of some source data necessarily recommend a cautious generalization of our results. Third, we are totally deprived of information about the seasonality of reported cases. As a consequence, an appropriate analysis of environmental factors-including climate factors such as air temperature, and humidity, but also the availabilities of appropriate facilities (i.e., air-conditioning, thermal convectors, etc.), is to date either difficult or substantially biased from its roots [50,62]. Eventually, it must be stressed that our estimates only included LD, while the occurrence of PF was not ascertained as it is not included in the current case definition for LD by the European Union [38].

Conclusions
In conclusion, the present summary of Italian national reports on LD suggests that the annual notification rates of LD have substantially increased across the assessed timeframe. Available estimates hint for substantially higher notification rates in northern regions (mostly, highly developed regions from northwestern Italy) compared to southern ones and main islands. Interestingly, the upsurge in new LD cases mainly affected CALD and TALD, while HALD remained quite consistent in terms of raw number, being more frequently reported from northern regions than from southern ones. At the moment, no comprehensive explanation may be inferred, rather suggesting the interaction of several different factors. Even though most of the aforementioned epidemiological features appear very similar to those observed in other European countries and in the USA as well, suggesting a significant role for demographic transition and climate change (the latter both directly and indirectly), quite heterogeneous regional policies in terms of the testing of individuals with clinical signs and symptoms of pneumonia, and the suspected under-reporting of LD cases to competent health authorities may have contributed to the conflicting epidemiological features we have collected and reported. Therefore, in regions with persistent low notification rates, ad hoc studies should be designed and performed in order to assess reasons for under-ascertainment.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.