JCP Online First, published on June 17, 2013 as /jclinpath Original article

JCP Online First, published on June 17, 2013 as 10.1136/jclinpath-2012-201209 Original article Diagnostic, therapeutic and economic consequences of a positive urinary antigen test for Legionella spp. in patients admitted with community-acquired pneumonia: a 7-year retrospective evaluation M F Engel, 1 L van Manen, 1 A I M Hoepelman, 1 S Thijsen, 2 J J Oosterheert 1 1 Department of internal medicine and infectious diseases, University Medical Centre Utrecht, Utrecht, The Netherlands 2 Department of medical microbiology, Diakonessen hospital, Utrecht, The Netherlands Correspondence to Dr M F Engel, Department of internal medicine and infectious diseases, University Medical Centre Utrecht, Post box 85500, Utrecht 3508 GA, The Netherlands; m.f.engel-2@umcutrecht.nl Received 17 September 2012 Revised 17 March 2013 Accepted 18 April 2013 To cite: Engel MF, van Manen L, Hoepelman AIM, et al. J Clin Pathol Published Online First: [ please include Day Month Year] doi:10.1136/jclinpath-2012-201209 ABSTRACT Aims A positive urinary antigen test for Legionella spp. (Legionella urinary antigen test; LUAT) allows an early switch from empiric to targeted treatment (TT) in hospitalised, community-acquired pneumonia (CAP) patients. We aimed to evaluate the diagnostic, therapeutic and economic consequences of this frequently used test 7 years after its implementation. Methods We retrospectively evaluated LUATs performed between 2005 and 2011 in two teaching hospitals. All tests performed in hospitalised CAP patients were used in the economic evaluation and positive tests were included in the treatment evaluation. Data on patient characteristics, admission and outcome were retrieved from the patients files. The number of days gained by making a rapid aetiological diagnosis, the number of days TT could be provided and their costs were calculated. Results Of 4485 LUATs, 2504 (56%) were performed for CAP including 55 (1%) positive tests ( 1041/positive test). In 26 (60%) of the 43 included positive tests, LUAT was the only test showing Legionella spp. Subsequently, earlier TT was possible in the remaining cases during 209 cumulative admission days ( 274/TT day). LUAT led to detection of Legionella spp. 13 days earlier per case ( 203/day) as compared with culture/ serology alone. Conclusions Timely LUAT use in accordance with current guidelines allows early detection and treatment of CAP caused by Legionella spp. at considerable expense. INTRODUCTION Legionella species (spp.) make up 1% 8% of the detected causes of community-acquired pneumonia (CAP) in hospitalised patients in the Netherlands. 1 With a mortality rate of 5% 30% 2 in hospitalised CAP patients and up to 50% in CAP patients admitted to the intensive care unit (ICU), this is considered a serious infection. Therefore, empirical treatment including coverage of Legionella spp. (ie, β-lactam combined with a macrolide or fluorochinolone monotherapy; if Legionella spp. is not suspected, β-lactam monotherapy is considered adequate empirical treatment as well by Dutch guidelines) is often recommended for CAP patients with severe disease 3 4 or those with risk factors for Legionella pneumonia. 1 5 Early detection of Legionella spp. allows early notification of public health services, which will conduct source research even in sporadic cases, and a timely switch from broad spectrum antibiotics to targeted treatment (TT) in individual patients. It may thereby lead to a reduced pressure on the development of antibiotic resistance. For CAP patients, a reliable prediction of the causative microorganism is not possible based on clinical presentation, laboratory data or radiological findings and microbiological testing is indispensable to establish an aetiological diagnosis. 6 7 Cultures for Legionella spp. take at least 7 days and generally cultures yield no result in up to 46% of cases. 1 The Legionella urinary antigen test (LUAT) is a rapid test that has shown high specificity and modest sensitivity in the detection of Legionella spp. serogroup one and is often the only indicator of Legionella infection. 8 Therefore, international guidelines advice Legionella urinary antigen testing for all CAP patients with severe disease and/or risk factors (eg, recent travelling, during outbreaks of Legionnaires disease, treatment failure on β-lactam antibiotics). 3 4 The 2012 Dutch guidelines advise the LUAT for all hospitalised CAP patients within 12 h of admission. 1 As with all diagnostic tests, the clinical consequences of the LUAT are dependent on its application in clinical practice. Patient factors (eg, comorbidities) or physician factors (eg, reluctance to stop β-lactam treatment or not being aware of guideline advice) may impede a switch to TT in case of a positive LUAT result. 9 To date, it is unknown to what extent current use of the LUAT leads to timely detection of Legionella spp. and whether physicians switch to TT based on positive results. We assessed the diagnostic, therapeutic and financial consequences of the current use of the LUAT in hospitalised CAP patients. METHODS Design and setting A retrospective cohort study was performed in two Dutch teaching hospitals in the city of Utrecht, the Utrecht University Medical Centre (UMCU, 1042 beds) and the Diakonessen Hospital (DH, 627 beds). Clinical consequences of LUATs performed in CAP patients between the introduction of the test in 2005 and 30 September 2011 were evaluated. Tests Tests that were performed in adult patients hospitalised because of CAP were eligible for inclusion. CAP was defined as the physicians diagnosis being pneumonia as stated in admission and discharge Copyright Article author (or their employer) 2013. Produced by BMJ Publishing Group Ltd under licence. Engel MF, et al. J Clin Pathol 2013;00:1 6. doi:10.1136/jclinpath-2012-201209 1

letters in cases that did not fit the criteria for hospital-acquired pneumonia or healthcare associated pneumonia as defined by the American Thoracic Society. 10 Tests were considered ineligible for analysis if they were not performed for CAP, were performed in children (ie, <18 years of age), were test runs or had intermediate results. In the economic evaluation, all tests in CAP patients with positive and negative results were used. For the analysis of diagnostic and therapeutic consequences of a positive LUAT, we excluded tests if the outcome was negative. We excluded positive tests as well if recording of clinical data was insufficient to evaluate the primary outcome (ie, patient not traceable, no treatment data), if patients were transferred to or from another hospital and if patients had visited the emergency room without being admitted. If a test was performed twice within 1 week, the second test was excluded. The latter exclusion criteria were assumed to be non-differential (eg, not leading to selection bias). Materials and methods In both hospitals the imunnochromatographic assay was used (BinaxNOW Portland, Maine, USA) in accordance with the manufacturer s guidelines with unconcentrated urine, which provides results within 15 min. This test has shown high specificity (0.99) and modest sensitivity (0.74) for the detection of L pneumophilia serogroup one, a subtype responsible for 85% 90% of pneumonia cases caused by Legionella spp. 11 13 The sensitivity of this test is slightly lower during the first days of infection and in mild pneumonia and higher in severe 14 15 pneumonia. During this study, local CAP protocol in the participating centres dictated LUAT use in a slightly different subset of patients. The UMCU protocol advocated, according to the national 2005 protocol, 5 LUAT use in CAP patients with moderate-severe disease (ie, Pneumonia Severity Index (Fine score) 16 3 or confusion-urea-respiratory rate-blood pressureage score (CURB-65) 17 of 2). In the DH, the test is advised in CAP patients with a CURB-65 score of 3. A list of all LUATs performed during the study period was extracted from the microbiology department s electronic database (General Laboratory Information Management System 2008 UMCU and SERIMBA 1.8.0.1 DH) and provided to our research group. For all included tests, the following clinical patient data were retrieved from the electronic patients file: demographics, comorbidities, severity scores on admission (CURB-65 score, Fine score), admission data (length of hospital stay, intensive care admission), details on treatment (antibiotics), microbiological testing (cultures, serology; positive if IgM is positive for Legionella spp. or if IgG titres rise four times within 4 weeks) and data on clinical outcome (outpatient post visits, readmission and mortality within 30 days of admission). In line with clinical practice, severity scores were calculated with the available data only. Data analysis and outcome The primary outcome measure was the diagnostic yield of the LUAT which was expressed as the additional L pneumophilia spp. cases detected by adding the LUAT to regular microbiological tests (ie, cultures of blood, sputum and other respiratory tract samples; serology) only in the UMCU. The number of days gained by making a rapid aetiological diagnosis by detection with the LUAT as opposed to the delayed detection using cultures was calculated as well. Results of microbiological testing were considered to be available to the treating physician as soon as the results were recorded in the electronic patients file or when the physician was informed about the definite results by phone as recorded in the microbiology consult registration. Diagnostic delay was expressed as the number of days between admission and LUAT application and as the number of days between LUAT application and the test result. Secondary outcomes included the number of cases switched to TT based on the LUAT and the number of days the use of empirical therapy was prevented due to the test. TT was defined as targeted antibiotic treatment of L pneumophilia (ie, fluoroquinolone (ciprofloxacin, moxifloxacin, levofloxacin), azitromycin or erythromycin not combined with empiric treatment) in accordance with (inter-) national guidelines. 4 18 In order to determine if the TT was applied based on the LUAT, the timing of the initiation of the TT was related to the timing of the test. If TT was initiated after performing the LUAT and cultures yielded no results or if TT was initiated after performing the LUAT but before culture results were available, it was assumed that a switch to TT was based on the LUAT. The number of days on which the use of empirical therapy was prevented was determined in each case receiving TT based on the LUAT. The days on which TT was administered were counted until the end of a standard course of empirical treatment (ie, admission day 10) or until cultures grew L pneumophilia. In the analysis we corrected numbers if the TT was halted before one of those endpoints. Therapeutic delay was expressed as the number of days between LUAT result and start of TT (ie, patients still on empiric treatment not directed at Legionella spp.). Cost calculation We evaluated the costs of performing the LUAT in terms of cost per positive result (C*N(LUAT)/N( positive LUAT)), costs for every extra diagnosed case (C*N(LUAT)/N(extra diagnosed cases)), costs per day TT could be administered (C*N(LUAT)/N (TT days)) and costs per day public healthcare services could be informed earlier as opposed to regular microbiological testing (C*N(LUAT)/N(days gained by early detection)). Direct healthcare cost were calculated from a hospital perspective using 2012 national tariffs (in Euros) in order to achieve a high interhospital comparability. Final amounts were rounded off to whole Euros (exchange rate: 1 is US$1.3, May 2012). The study was executed according to Good Clinical Practice principles. As this study was performed using retrospective reviews of patients medical files, participating patients were not inconvenienced. Therefore, approval of the Medical Ethics Committees was not necessary. RESULTS Between 2005 and September 2011, 4485 LUATs were performed in both hospitals. Overall, 56% (2504/4485) of all LUATs were performed for CAP, and 55 (1%) of these tests yielded a positive result (figure 1). The remaining 1981 (44%) tests were ineligible for analysis. If the test indication was not CAP, the LUAT was performed for conditions such as healthcare associated pneumonia or fever with unknown origin in case of blood malignancy. In all, 12 (22%) of these 55 tests were excluded because they were performed twice in the same patient within 1 week (3), recording of clinical data was insufficient (4), the admission was after a CAP episode (2), the patient was transferred to another hospital (2) or visited the emergency room without being admitted (1). This left 43 cases (unique patients) for detailed analysis of treatment decisions. Patient characteristics and outcome The mean age was 59 years (SD 13) and there were less women (n=16, 37%) then men (table 1). A total of 15 (35%) patients 2 Engel MF, et al. J Clin Pathol 2013;00:1 6. doi:10.1136/jclinpath-2012-201209

Figure 1 Flow of tests through the study. * Based on a random sample of 10% out of all patients with a negative Legionella urinary antigen test (LUAT) using SPSS. CAP, community-acquired pneumonia. had 1 comorbidities; the most frequent comorbidities were lung disease (n=7, 16%) and congestive heart failure (n=6, 14%). The median Fine score on admission was 80 (IQR 63 104) and the median CURB-65 score was 1 (IQR 1 2). In 35% of cases (n=15), initial antibiotic treatment was a combination of a β-lactam antibiotic with a fluoroquinolone (table 1). Data on age, sex, comorbidity, microbiological testing and patient outcome were complete for all cases. In 22 (51%) cases, severity scores were calculated while one or more variables were missing. In seven patients, a second pathogen was cultured from respiratory specimens and/or detected through serological testing; these included S aureus (1), Candida spp. (1), Influenza virus (2) and other viruses (3). In four cases (9%), patients had a non-respiratory co-infection during admission which needed antibiotic treatment. In eight cases (19%), four of which within 48 h of admission, microbiological testing on respiratory material showed a second pathogen besides Legionella spp. The median duration of admission was 8 days (IQR 6 20). In 16 cases (37%), patients were (partially) treated on the ICU with a median duration of stay of 11 days (IQR 5 23). A total of 15 (35%) patients visited the outpatient clinic after discharge and 2 (5%) were readmitted within 30 days after admission due to respiratory tract pathology. In only one case, the patient died during follow-up due to respiratory tract pathology. Notably, the 13 patients in whom the initial therapy did not cover Legionella spp. were comparable with the remaining patients with respect to the number of ICU admissions (46% vs 33%, p=0.429), mortality rates (8% vs 0% p=0.129) and duration of admission (8 days IQR 7 30 vs 10 IQR 6 19 days, p=0.573). Diagnostic consequences In all, 55 (2%) of all 2504 tests performed for CAP yielded a positive result, resulting in a number needed to test (NNT) of 46 (UMCU 84, DH 31). In 26 (60%) of the 43 included cases, LUAT was the only microbiological test that detected Legionella spp. In the remaining 17 (40%) cases detected through cultures and/or serological testing in addition to the LUAT, cumulatively, 282 days were gained by rapid detection with the LUAT as opposed to detection using cultures or serological testing (median 13 IQR 9 21 for these 17 cases). In 32% (7/22) of patients from whom sputum samples were cultured, samples were positive for Legionella spp. (50% (6/12) for lower respiratory samples and 31% (4/13) for detection of Legionella spp through seroconversion). Notably, in one case, Legionella spp. was detected through serological testing 54 days after admission. There was no diagnostic delay ( 1 day) resulting from the Engel MF, et al. J Clin Pathol 2013;00:1 6. doi:10.1136/jclinpath-2012-201209 3

Table 1 Baseline characteristics of community-acquired pneumonia patients with a positive Legionella urinary antigen test (n=43) Characteristics Mean/median/number Age 59 (± 13) Female 16 (37%) Comorbidity 1 15 (35%) Malignancy* 0 (0%) Liver disease* 0 (0%) Congestive heart failure* 6 (14%) Cerebrovascular disease* 4 (9%) Kidney disease* 2 (5%) Lung disease 7 (16%) Pneumonia severity** CURB-65 1 (IQR 1-2) >2 6 (14%) Fine score 80 (IQR 63-104) Risk class IV 11 (26%) Risk class V 5 (12%) Initial antibiotic Beta-lactam 12 (28%) treatment Fluoroquinolone 7 (16%) Beta-lactam + Fluoroquinolone 15 (35%) Beta-lactam + macrolide 8 (19%) Other 1 (2%) Culture positive for Legionella 17 (40%) Co-infection (non-respiratory) 4 (9%) Admitted to UMCU 10 (23%) All statistical analyses were performed using SPSS version 15.0. * As defined in the Fine score ** Severity scores were calculated with the available data only timing or processing of the LUAT in 88% (n=38/43) and 98% (n=42/43) of cases, respectively. Overall, in 84% (n=36/43) of cases, the diagnosis was not delayed (ie, LUAT result was available on first or second admission day). Therapeutic consequences Overall, 36 (84%) of the 43 cases were treated with TT for Legionella spp. Two cases were excluded from further analysis because the culture for Legionella spp became positive simultaneously with the LUAT (n=1) or TT was administered before the LUAT result (n=1). In the 34 remaining cases, including all four patients with a non-respiratory co-infection, the LUAT led to TT; in 21 of these cases cultures were negative for Legionella spp and TT was administered after LUAT, and in the 13 remaining cases TT was administered after the LUAT but before positive culture results. Due to the LUAT, TT was administered during 209 cumulative admission days in the aforementioned 34 cases, with a median duration of 6 days (IQR 5 9) per case. The median time between the LUAT result and the administration of TT (ie, therapeutic delay) was 1.4 days (IQR 0 2). Notably, our sample contained two negative LUATs in patients with cultures and/or serology showing Legionella spp. Patient groups in which a switch to TT (ie, no combination or empirical therapy) may be impeded are patients with lung pathology (eg, COPD patients potentially colonised with P aeruginosa), patients admitted to the ICU, patients with a second infection or a second pathogen isolated from respiratory samples. For these patient groups, we evaluated if they received fewer days of TT as compared with the remaining patients. Surprisingly, patients with lung pathology received TT during more days than the remaining patients (median 9 IQR 7 10 vs 6 IQR 5 8 p=0.05). Patients admitted to the ICU (median 7 IQR 5 9 vs remaining patients 6 IQR 5 9 p=0.48), with a second infection (median 6 IQR 3 9 vs remaining patients 6 IQR 5 9 p=0.77) and those with a second bacterium cultured from respiratory tract samples (median 5 IQR 2 9 vs remaining patients 6 IQR 5 9 p=0.26) did not receive less days of TT as compared with the remaining patients. Last, pneumonia severity scores (ie, CURB-65 and Fine score) on admission of patients not receiving TT based on the LUAT result did not differ significantly from the remaining patients ( p>0.4). Costs The cumulative costs of performing the LUAT for CAP were 57 267 ( 22.87 per test*2504 tests). The cost per positive result was 1041 ( 57 267/55) and 2203 ( 57 267/26) per extra diagnosed case. The costs per day TT could be administered were 274 ( 57 267/209). The cost per day public healthcare services could be informed earlier as opposed to regular microbiological testing was 203 ( 57 267/282). Notably, 20 sputum cultures and seven serological tests for Legionella spp were negative in patients with a positive LUAT at a cost of 630 (20* 21+7* 30), tests which may have been redundant if LUAT had been applied in a timely fashion. DISCUSSION We showed that only 2% of all LUATs performed in hospitalised patients with CAP were positive at the cost of 2203 per identified case. In 60% of these cases, the LUAT was the only test showing Legionella spp The benefits of positive test results were timely detection of Legionella pneumonia, on average 13 days earlier as compared with culture and/or serology, and subsequent TT of individual patients during 209 cumulative admission days. The major strength of this study is that it provides a complete overview of benefits and costs of current LUAT implementation in secondary as well as tertiary care. First we evaluated the diagnostic consequences. As detection of Legionella spp through seroconversion was part of daily practice in the participating hospitals, we incorporated data on serological tests in our analysis. However, because detection of Legionella spp through seroconversion often leads to a delayed diagnosis, as was shown by our study data, the diagnostic delay may have been overestimated using this approach. Second, while assessing physicians treatment decisions, we evaluated the effect of positive LUAT results only because it is unlikely that negative LUAT results influence patient management (national guidelines advice continuous antibiotic coverage of Legionella spp in case of a negative LUAT in CAP patients with severe disease 1 ) and false negative LUAT results are rare (our sample contained a least two false negative results in CAP patients, although additional tests for Legionella spp were not performed in all included patients). Notably, 28% of patients were initially treated with β-lactam monotherapy which does not cover Legionella spp. This shows that, while guidelines provide tools for a solid assessment of CAP aetiology during the initial assessment of patients, determining the causative pathogen in CAP prior to microbiological testing remains a subjective matter. In order to determine if TT was applied based on the LUAT, we assumed that changes in antibiotic regimen to TT in the absence of other microbiological tests positive for Legionella spp were made based on the positive LUAT result. However, it may have been that some of these antibiotic changes were a consequence of general streamlining of antibiotic therapy based on 4 Engel MF, et al. J Clin Pathol 2013;00:1 6. doi:10.1136/jclinpath-2012-201209

clinical observations and would have occurred irrespective of the LUAT result. Furthermore, some patients included in the DH did not meet the DH guideline criteria (ie, CURB-65 score 3). While numbers are low, this suggests that hospital protocol was not consistently followed and strict adherence to protocol would have led to different outcome measures. Therefore, the effect of the LUAT on antibiotic treatment decisions presented in this paper may have been overestimated or underestimated. The cost and benefits presented in this study only partially reflect the true effect of LUAT implementation in practice. The LUAT was applied as an addition to the diagnostic arsenal of physicians treating hospitalised CAP patients and might thereby have more extensive benefits than solely guiding antibiotic therapy. Possibly, the early aetiological diagnosis obtained through positive LUAT results led to a decrease in the use of other microbiological tests, a shorter duration of hospital stay or even a decrease in mortality. Due to the retrospective design of this study, which covered 7 years, we were unable to acquire data on clinical parameters during the admission for a majority of patients. Therefore, we are unable to assess the aforementioned aspects because cannot control for confounding factors such as illness severity. In addition, the economic value of timely detection of Legionella spp. and the exact advantage for public health are hard to asses. Due to the low incidence of Legionella pneumonia (1% 8% 1 ), only 55 tests were positive during the 7-year inclusion period of this study. The low number of included tests impeded analysis of the effect of the different approaches in the two participating hospitals. This analysis would further be limited by the fact that a comparison of guideline effects between hospitals would require a complex analysis of physician behaviour. Still, our data showed that the protocol dictating the application of the LUAT in all admitted CAP cases (UMCU) led to an NNT of 84 as opposed to 31 when testing hospitalised patients with severe CAP only (DH). Several other studies have evaluated the clinical use of the LUAT. Two mono-centre, retrospective cohort studies showed that a positive LUAT led to adequate treatment alterations in 22% 60% of cases with the NNT ranging from 34 to 128. 8 19 In this study, a higher percentage (ie, 84%) of adequate treatment changes was found and a comparable NNT of 46. As in our study, a low number of positive tests were included, 6 and therefore no definite conclusions can be drawn. Dionne et al 19 calculated the costs per positive LUAT which were at 4439 much more costly than the 1041 found in this study. A recent randomised controlled trial evaluated if therapy based on the LUAT result as compared with empirical therapy affected patient outcome and healthcare costs. 20 No substantial benefits of LUAT based therapy were identified in the 177 randomised CAP patients. Contrastingly, a study on the effect of timely TT including 141 hospitalised patients with Legionella pneumonia showed that ICU-free survival was higher if patients were treated adequately within 24 h after admission. 21 Policymakers have essentially three choices when it comes to LUAT use: non-restrictive use (ie, using only guideline advice to regulate LUAT use), restrict its use to moderate-severe CAP patients only (eg, through requiring preauthorisation) or not using the test at all. Effects on healthcare costs (ie, average annual expenditure of 2664 per hospital) and antibiotic treatment of the first approach are described in this paper. The second approach would theoretically lead to a reduced NNT and inherently to a lower number of Legionella cases identified of the 26 patients in this two centre study in whom LUAT was the only test showing Legionella spp, 13 (50%) did not fit the national guideline criteria for moderate-severe pneumonia (Fine score 3 or CURB-65 2). With an annual admission rate of moderate-severe CAP in the Utrecht region of approximately 72 per hospital, 22 costs for restrictive use of the test would add up to 1647 per centre annually (72* 22.87). The effect of using other algorithms like reserving Legionella urinary antigen testing for CAP patients admitted to ICUs and/ or those failing on β-lactam monotherapy is yet to be explored. Finally, the most economic approach is to discard the LUAT completely. Combining this approach with standard empirical treatment which covers Legionella spp (ie, β-lactam combined with macrolide or fluorochinolone monotherapy) would warrant individual patient safety. However, timely notification of public healthcare services in case of an outbreak would be impeded, less cases of Legionella pneumonia would be identified and the therapeutic benefits of the test as described in earlier would be completely lost. In conclusion, timely LUAT use in accordance with current Dutch national guidelines allows early detection and treatment of CAP caused by Legionella spp at considerable expense. Policy makers should consider test implementation using the benefits and costs described in this paper. Based on the data presented, we advice restricted (ie, in CAP patients with severe disease and/ or risk factors only) and timely LUAT use in accordance with current international guidelines. Future research should be aimed at evaluating economic and therapeutic consequences of such a strategy. What this paper adds To date, it is unknown to what extent current use of the Legionella urinary antigen test (LUAT) leads to timely detection of Legionella spp in patients hospitalised with community-acquired pneumonia (CAP) and whether physicians switch to targeted therapy based on positive test results. We are the first to assess the diagnostic, therapeutic and financial consequences of the current use of this test in hospitalised CAP patients. Take-home messages Due to the low incidence of CAP caused by Legionella spp. the LUAT has a high number needed to test which results in high costs for every case identified by this test, 2203. Over 40% of all 4485 LUATs performed in the last 7 years were not performed for CAP, which is virtually the only evidence based indication for this test. LUAT implementation led to timely detection of Legionella pneumonia; on average 13 days earlier as compared with culture and/or serology and subsequent targeted treatment of individual patients during 209 cumulative admission days. Contributors MFE: study design, conduct, analysis, writing/editing manuscript. LvM: study conduct, writing/editing manuscript. AIMH: study design, editing manuscript. ST: study conduct, analysis, editing manuscript.jjo: study design, analysis, editing manuscript. Competing interests None. Engel MF, et al. J Clin Pathol 2013;00:1 6. doi:10.1136/jclinpath-2012-201209 5

Ethics approval This study was performed using retrospective reviews of patients medical files; participating patients did not experience any disadvantage or discomfort. Obtaining patients informed consent was not feasible over the study period of 7 years. The study was executed according to Good Clinical Practice principles. Provenance and peer review Not commissioned; externally peer reviewed. REFERENCES 1 Wiersinga WJ, Bonten MJ, Boersma WG, et al. SWAB/NVALT (Dutch Working Party on Antibiotic Policy and Dutch Association of Chest Physicians) Guidelines on the management of community-acquired pneumonia in adults. Neth J Med 2012;2:90 101. 2 Centres for Disease Control and Prevention; National Center for Immunization and Respiratory Diseases. Patient Facts: Learn More about Legionnaires disease. Jun 2011. http://www.cdc.gov/legionella/patient_facts.htm (accessed 30 Dec 2012). 3 Lim WS, Baudouin SV, George RC, et al. BTS guidelines for the management of community acquired pneumonia in adults: update 2009. Thorax 2009;64(Suppl 3):1 55. 4 Mandell LA, Wunderink RG, Anzueto A, et al. Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults. Clin Infect Dis 2007;44(Suppl 2):27 72. 5 Schouten JA, Prins JM, Bonten MJ, et al. Revised SWAB guidelines for antimicrobial therapy of community-acquired pneumonia. Neth J Med 2005;8:323 35. 6 Farr BM, Kaiser DL, Harrison BD, et al. Prediction of microbial aetiology at admission to hospital for pneumonia from the presenting clinical features. British Thoracic Society Pneumonia Research Subcommittee. Thorax 1989;44:1031 5. 7 Sopena N, Sabria-Leal M, Pedro-Botet ML, et al. Comparative study of the clinical presentation of Legionella pneumonia and other community-acquired pneumonias. Chest 1998;113:1195 200. 8 Garbino J, Bornand JE, Uckay I, et al. Impact of positive legionella urinary antigen test on patient management and improvement of antibiotic use. J Clin Pathol 2004;57:1302 5. 9 Engel MF, Postma DF, Hulscher ME, et al. Barriers to an early switch from intravenous to oral antibiotic therapy in hospitalised patients with communityacquired pneumonia. Eur Respir J 2013 Jan;41(1):123 30. doi: 10.1183/ 09031936.00029412. Epub 2012 May 31. 10 American Thoracic Society, Infectious Diseases Society of America. Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcare-associated pneumonia. Am J Respir Crit Care Med 2005;4:388 416. 11 Blazquez RM, Espinosa FJ, Martinez-Toldos CM, et al. Sensitivity of urinary antigen test in relation to clinical severity in a large outbreak of Legionella pneumonia in Spain. Eur J Clin Microbiol Infect Dis 2005;24:488 91. 12 Muder RR, Yu VL. Infection due to Legionella species other than L. pneumophila. Clin Infect Dis 2002;35:990 8. 13 Shimada T, Noguchi Y, Jackson JL, et al. Systematic review and metaanalysis: urinary antigen tests for Legionellosis. Chest 2009;136:1576 85. 14 von Baum H, Ewig S, Marre R, et al. Community-acquired Legionella pneumonia: new insights from the German competence network for community acquired pneumonia. Clin Infect Dis 2008;46:1356 64. 15 Yzerman EP, den Boer JW, Lettinga KD, et al. Sensitivity of three urinary antigen tests associated with clinical severity in a large outbreak of Legionnaires disease in The Netherlands. J Clin Microbiol 2002;40:3232 6. 16 Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med 1997;336:243 50. 17 Lim WS, van der Eerden MM, Laing R, et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax 2003;58:377 82. 18 Wiersinga WJ, Bonten MJ, Boersma WG, et al. SWAB/NVALT (Dutch Working Party on Antibiotic Policy and Dutch Association of Chest Physicians) guidelines on the management of community-acquired pneumonia in adults. Neth J Med 2012;70(2): 90 101. 19 Dionne M, Hatchette T, Forward K. Clinical utility of a Legionella pneumophila urinary antigen test in a large university teaching hospital. Can J Infect Dis 2003;2:85 8. 20 Falguera M, Ruiz-Gonzalez A, Schoenenberger JA, et al. Prospective, randomised study to compare empirical treatment versus targeted treatment on the basis of the urine antigen results in hospitalised patients with community-acquired pneumonia. Thorax 2010;65:101 6. 21 Lettinga KD, Verbon A, Weverling GJ, et al. Legionnaires disease at a Dutch flower show: prognostic factors and impact of therapy. Emerg Infect Dis 2002;8:1448 54. 22 Endeman H, Schelfhout V, Voorn GP, et al. Clinical features predicting failure of pathogen identification in patients with community acquired pneumonia. Scand J Infect Dis 2008;40:715 20. 6 Engel MF, et al. J Clin Pathol 2013;00:1 6. doi:10.1136/jclinpath-2012-201209