Legionella pneumophila pneumonia in South Eastern Sydney: diagnosis, prognosis and public health outcomes

Abstract

Background

Legionellosis causes more severe disease than other aetiologies of community‐acquired pneumonia (CAP). Early recognition and treatment of Legionella pneumonia reduce mortality. Urinary antigen testing (UAT) has facilitated rapid and reliable diagnosis of Legionella pneumonia.

Aims

The aims of this study were as follows: (i) identify characteristics of CAP patients more likely to be associated with a positive Legionella UAT, (ii) identify clinical features of Legionella pneumonia associated with higher risk of intensive care unit (ICU) admission and (iii) describe public health interventions resulting from positive Legionella UATs.

Methods

We conducted a case–control study of CAP patients admitted at two South‐Eastern Sydney hospitals between January 2015 and March 2024 who underwent Legionella UAT. Multivariable logistic regression was used to identify characteristics associated with a positive Legionella UAT and ICU admission.

Results

There were 13 697 Legionella UATs during the study period, and 51 (0.4%) were positive. A total of 100% of the Legionella UAT‐positive cases could be detected by conducting Legionella UAT in CAP patients with at least one of the following characteristics: >20 standard alcoholic drinks/week, confusion, failure of outpatient antibiotic therapy, diarrhoea, myalgias, temperature >40°C, serum sodium <137 mmol/L and CRP >300 mg/L. CURB‐65 score, CRP and heart rate were independent factors associated with ICU admission. Positive Legionella UAT results led to 11 regulatory actions under the Public Health Act.

Conclusion

Our study identified characteristics of CAP patients more likely to be associated with a positive Legionella UAT. Future prospective studies with greater power and in different populations may assist in refining this model to create a clinical decision‐making tool to determine the appropriateness of Legionella UAT in CAP.

Introduction

Legionella accounts for 5%–15% of community‐acquired pneumonia (CAP) cases in Australia, and the incidence of Legionella pneumonia is increasing. ¹ Legionella pneumophila serotype 1 (Lp1) accounts for over 80% of cases of Legionella pneumonia. ² A total of 928 Legionella pneumophila cases were reported in New South Wales (NSW) between 2010 and 2022 including 784 (84.5%) confirmed and 144 (15.5%) probable cases. ³ The total number of cases increased from 35 in 2010 to 114 in 2022. ³

Legionella pneumonia is associated with more severe disease than other causes of CAP, as demonstrated by higher rates of intensive care unit (ICU) admissions and mortality. ⁴ With early recognition and appropriate treatment, mortality can be reduced to below 5%. ⁴ As Legionella is difficult to culture, Legionella pneumonia has historically been challenging to diagnose. ⁵ However, urinary antigen testing has greatly increased the ability to reliably test for Legionella. The urinary antigen assay for Lp1 has a sensitivity of 70% and a specificity that approaches 100% and a result turnaround time of as little as 15 min in some laboratories. ⁶ , ⁷ Each Lp1 urinary antigen test costs $28.65 on the Medicare Benefits Schedule. ⁸ A new urinary antigen assay which includes a target for Legionella longbeachae was not in use at the study sites at the time this study was conducted. The Australian Therapeutic Guidelines suggest that Legionella nucleic acid amplification testing (NAAT) on nasopharyngeal swab or sputum samples can also be performed for diagnosis of Legionella pneumonia. However, this test may be less accessible in some settings in NSW and more expensive than the Lp1 urinary antigen test. ⁹ Respiratory pathogen NAAT costs between $78.25 and $85.55 on the Medicare Benefits Schedule. ¹⁰ In the interest of diagnostic stewardship, it would be ideal to identify specific characteristics of patients with positive Lp1 urinary antigen results to help inform decisions regarding when it would be appropriate to order this investigation. ¹¹

The existing literature contains limited descriptions of the clinical characteristics of an Australian cohort of patients presenting with CAP with a positive Lp1 urinary antigen. Previous studies in the United Kingdom, Europe and the United States have found that recent travel, diarrhoea and temperature >38°C were history and examination findings associated with a positive Lp1 urinary antigen result. ⁵ , ¹² , ¹³ Risk factors for acquiring Legionella pneumonia included male sex, smokers and increasing age. ⁵ , ¹² , ¹³ Laboratory abnormalities suggestive of Legionella pneumonia include hyponatraemia, acute kidney injury, transaminitis and high C‐reactive protein (CRP). ⁷ , ¹⁴ , ¹⁵

Recommendations for clinical settings where Legionella urinary antigen testing is appropriate vary across different countries. The Australian Therapeutic Guidelines (TGs) recommend Legionella urinary antigen testing if a patient presents with high‐severity CAP requiring ICU admission and has risk factors for legionellosis. ⁸ The risk factors listed by the TGs are chronic lung disease, smoking, diabetes, advanced kidney disease, cancer and immune compromise (including chronic corticosteroid use). ⁸ The American IDSA/ATS guidelines recommend Legionella urinary antigen testing if a CAP patient requires ICU admission, has failed outpatient antibiotic therapy, drinks excessive alcohol, has recently travelled or has a pleural effusion. ¹⁶ The British Thoracic Society guidelines recommend Legionella urinary antigen testing in anyone with moderate or high severity CAP, that is, a CURB‐65 score of 2 or above. ¹⁷

A positive Lp1 urinary antigen result has significant public health implications. Legionellosis is a nationally notifiable condition in Australia given its potentially severe consequences. ¹⁸ Legionella outbreaks are typically associated with contamination of manufactured water systems including water cooling towers and spa baths. ¹⁸ In NSW, water cooling systems (including cooling towers) are regulated under the Public Health Act (PHA) to control legionellosis risk. ¹⁸ There is limited description in the literature about the public health interventions generated by positive Lp1 urinary antigen results in NSW.

Objectives

The three main objectives of our study were as follows:

1. Identify characteristics of patients with CAP which are more likely to be associated with a positive Lp1 urinary antigen result.

2. Identify clinical features of Legionella pneumonia which are associated with a higher risk of ICU admission.

3. Describe public health interventions resulting from positive Lp1 urinary antigen results in patients with CAP.

Methods

Ethics approval for this study was granted by the South‐Eastern Sydney Local Health District (SESLHD) Human Research Ethics Committee (2024/ETH00740). We conducted a retrospective review of CAP patients admitted at St George Hospital and Sutherland Hospital between January 2015 and March 2024 who underwent urinary antigen testing. These patients were identified through the NSW Health Pathology database. We then accessed each patient's electronic medical records (EMRs) to collect demographic, medical history and clinical data. These data were de‐identified and recorded in a REDCap database. ¹⁹ , ²⁰ We used the Notifiable Conditions Information Management System (NCIMS) via the SESLHD Public Health Unit (PHU) to collect information about the public health interventions which were generated from the positive Lp1 urinary antigen results.

To identify characteristics more likely to be associated with a positive Lp1 urinary antigen in CAP, we used statistical analysis to compare CAP patients with and without a positive Lp1 urinary antigen result. For the Lp1 negative group, we utilised data from a subset of 470 patients, which had formed a cohort for a separate pneumococcal urinary antigen research project, with a study period of January 2019 to March 2024. During this period, there were a total of 235 patients with positive pneumococcal urinary antigens. We then selected 235 matched (based on age and gender) controls with negative pneumococcal and Lp1 urinary antigens. For continuous variables, we compared means between the two groups using Welch's t test, and for categorical variables we performed a chi‐squared test of independence.

Power calculations were performed following data collection to ensure our analysis generated adequate power. For Welch's t test, a sample size of 470 in the non‐Legionella pneumonia group and 51 in the Legionella pneumonia group was sufficient to detect a standardised mean difference of 0.5, with 0.9 power and using a significance level of 0.05. For the chi‐squared test, a sample size of 470 in the non‐Legionella pneumonia group and 51 in the Legionella pneumonia group was sufficient to detect a Cramér V of 0.15, with 0.9 power and using a significance level of 0.05.

Multivariable logistic regression was performed to generate algorithms to identify patient characteristics which would warrant Lp1 urinary antigen testing. For comparison, we calculated the sensitivity, specificity, negative predictive value, number tested and number of correctly identified positive results for the Australian, American and British guideline recommendations regarding when Lp1 urinary antigen testing should be undertaken in CAP patients. Multivariable logistic regression was also performed to generate an algorithm to predict ICU admission in our cohort.

Results

During the study period of January 2025 to March 2024, 20 956 patients were diagnosed with CAP at the two study sites, and 14 361 pneumococcus and 13 697 Lp1 urinary antigens were sent. There were 51 positive Lp1 urinary antigen results. Baseline characteristics are presented in Table 1. Temporal changes in characteristics of the positive Lp1 urinary antigen group are in Table 2. Pneumonia history presentation characteristics are demonstrated in Table 3. Observation signs and pathology results upon presentation are demonstrated in Table 4. Chest X‐ray features are demonstrated in Table 5.

Table 1.

Baseline patient characteristics

	Positive (n = 51)	Negative (n = 470)	P‐value
Sex
Male	40 (78%)	258 (55%)	0.002
Female	11 (22%)	212 (45%)
Mean age (SD)	66.6 (14.8)	69.4 (16.7)	0.21
Smoking status
Current smoker	13 (25%)	51 (11%)	0.01
Ex‐smoker	15 (29%)	147 (31%)
Never smoked	12 (24%)	185 (39%)
Unknown	11 (22%)	87 (19%)
Alcohol intake
0–2 Standard drinks/week	19 (37%)	254 (54%)	0.006
3–10 Standard drinks/week	3 (6%)	23 (5%)
10–20 Standard drinks/week	0 (0%)	13 (3%)
>20 Standard drinks/week	7 (14%)	18 (3%)
Unknown	22 (43%)	162 (35%)
Penicillin/Cephalosporin allergy	7 (14%)	54 (11.5%)	0.81
Asplenia	1 (2%)	3 (1%)	0.86
Myocardial infarction	7 (14%)	74 (16%)	0.86
Congestive cardiac failure	1 (2%)	54 (11%)	0.06
Peripheral vascular disease	0 (0%)	26 (5%)	0.17
Cerebrovascular event	0 (0%)	36 (8%)	0.08
Dementia	1 (2%)	26 (5%)	0.45
Chronic lung disease	4 (8%)	127 (27%)	0.005
Connective tissue disease	2 (4%)	49 (10%)	0.22
Peptic ulcer disease	0 (0%)	14 (3%)	0.43
Liver disease
None	50 (98%)	459 (97%)	0.84
Mild	1 (2%)	8 (2%)
Moderate to severe	0 (0%)	3 (1%)
Diabetes
None or diet controlled	43 (84%)	391 (83%)	0.47
Uncomplicated	8 (16%)	66 (14%)
End‐organ damage	0 (0%)	13 (3%)
Hemiplegia	0 (0%)	1 (0.2%)	1
Chronic kidney disease	2 (4%)	17 (4%)	1
Solid tumour
None	46 (90%)	376 (80%)	0.07
Localised	1 (2%)	62 (13%)
Metastatic	4 (8%)	32 (7)
Leukaemia	3 (6%)	27 (6%)	1
Lymphoma	2 (4%)	18 (4%)	1
AIDS	0 (0%)	0 (0%)	N/A
Median Charlson Comorbidity Index (SD)	4.57 (2.29)	5.56 (2.79)	0.005

SD, standard deviation.

Table 2.

Temporal changes in positive Lp1 urinary antigen group

Year	No. of cases	% Male	% Female	Median age	Mean age	% ICU admission	% Deceased
2015	7	43%	57%	63	67	43%	0%
2016	6	100%	0%	80.5	76.2	17%	0%
2017	5	100%	0%	61	63.2	40%	0%
2018	3	67%	33%	62	55	33%	0%
2019	5	60%	40%	63	56.4	0%	0%
2020	11	91%	9%	67	60.3	18%	0%
2021	5	80%	20%	76	74	40%	0%
2022	4	75%	25%	84.5	81.25	25%	50%
2023	5	100%	0	70	67.8	40%	20%
2024	0	—	—	—	—	—	—

ICU, intensive care unit.

Table 3.

Presenting history characteristics

	Positive (n = 51)	Negative (n = 470)	P‐value
Confusion	17 (33%)	62 (13%)	<0.001
Prior use of antibiotics	18 (35%)	93 (20%)	0.02
Recent travel	8 (16%)	30 (6%)	0.03
Headache	11 (22%)	39 (8%)	0.005
Diarrhoea	16 (31%)	51 (11%)	<0.001
Nausea or vomiting	12 (24%)	63 (13%)	0.08
Myalgias	19 (37%)	44 (9%)	<0.001
Arthralgias	0 (0%)	16 (3%)	0.36

Table 4.

Vital signs, blood pathology results and CURB‐65 score

	Positive (n = 51), mean (SD)	Negative (n = 470), mean (SD)	P‐value
Vital signs
Temperature (°C)	38.7 (0.975)	37.8 (1.15)	<0.001
Respiratory rate (breaths per min)	26.8 (8.59)	26.4 (7.56)	0.74
Oxygen saturations (%) on room air	92.8 (4.52)	92.2 (6.63)	0.38
Heart rate (beats per min)	110 (24.5)	100 (22.0)	0.01
Systolic blood pressure (mmHg)	127 (22.2)	131 (25.3)	0.36
Diastolic blood pressure (mmHg)	75.7 (11.8)	73.9 (13.9)	0.32
Pathology results
Serum sodium	131 (5.19)	136 (5.09)	<0.001
Serum alanine transaminase	90.9 (186)	37.6 (59.4)	0.06
Serum aspartate transaminase	180 (516)	50.6 (129)	0.09
Serum creatinine	134 (104)	108 (86.9)	0.09
Haemoglobin	127 (26.8)	125 (20.9)	0.54
Total white cell count	11.5 (5.72)	12.4 (7.35)	0.32
C‐reactive protein	282 (115)	147 (137)	<0.001
CURB‐65 score	2.14 (1.27)	1.94 (1.24)	0.30

Table 5.

Chest X‐ray features

	Positive (n = 51)	Negative (n = 470)	P‐value
Chest X‐ray features
Single lobar consolidation	30 (59%)	250 (53%)	0.54
Multi‐lobar consolidation	20 (39%)	167 (36%)	0.71
Interstitial oedema	0 (0%)	23 (5%)	0.21
Pleural effusion	9 (18%)	80 (17%)	1
Lobe of consolidation
Right upper lobe	13 (25%)	57 (12%)	0.02
Right middle lobe	10 (20%)	64 (14%)	0.34
Right lower lobe	18 (35%)	219 (47%)	0.17
Left upper lobe	16 (31%)	55 (12%)	<0.001
Left lower lobe	19 (37%)	225 (48%)	0.20

Variable definitions are available in Appendix I. The variables significantly associated with Legionella pneumonia compared to non‐Legionella pneumonia on univariate analysis were male sex, current smoking, excess alcohol use (>20 standard drinks per week), confusion, failure of outpatient antibiotic therapy, recent travel, headache, diarrhoea, myalgias, higher temperature, elevated heart rate, lower serum sodium, elevated CRP and left or right upper lobe consolidation on chest X‐ray. Multivariable logistic regression concluded that the independent predictors of a positive Lp1 urinary test result were male sex, excess alcohol use (>20 standard drinks per week), confusion, failure of outpatient antibiotic therapy, diarrhoea, myalgias, higher temperature, lower serum sodium and higher CRP.

Using a multivariable logistic regression model, we tested the sensitivity, specificity and negative predictive value of testing for Lp1 urinary antigen only if there was the presence of any independent predictors (excluding male sex) and denoted this as Model 1. Male sex was excluded despite being an independent predictor to avoid gender biased testing. For the continuous variables of temperature, serum sodium and CRP, we modelled scenarios with different cutoffs until we identified cutoffs with an optimal level of sensitivity and a minimum number of tests. A grid search was performed over the range of temperature, CRP and sodium values in the dataset, and the cutoffs were chosen such that the specificity was the highest under the condition that sensitivity was close to 100%. The cutoffs identified were temperature >40°C, serum sodium <137 mmol/L and CRP >300 mg/L. We also tested a Model 2, which excluded serum sodium and CRP as criteria and only included the predictors which could be ascertained through history and physical examination. The goodness of fit and predictive power of each model were checked using receiver operating characteristic (ROC), area under the curve (AUC) and calibration curves. The results of the foregoing analysis and the comparison with the Australian, American and British guideline recommendations are presented in Table 6. We tested both Models 1 and 2 after excluding the 235 pneumococcal urinary antigen positive patients, and the models performed similarly, as demonstrated by Table 7.

Table 6.

Specificity, sensitivity and negative predictive value (NPV) of clinical recommendations and data models

	Sensitivity	Specificity	NPV	No. of positives (out of 51 cases)	No. of tests (out of 521)	Positives: tests
Australia	0.24 (95% CI: 0.14–0.37)	0.82 (95% CI: 0.81–0.89)	0.91	12	96	0.13
USA	0.76 (95% CI: 0.63–0.86)	0.35 (95% CI: 0.28–0.40)	0.93	39	343	0.11
UK	0.73 (95% CI: 0.59–0.83)	0.39 (95% CI: 0.38–0.50)	0.93	37	322	0.11
Model 1	1.00 (95% CI: 0.93–1.00)	0.24 (95% CI: 0.22–0.34)	1.00	51	406	0.13
Model 2	0.92 (95% CI: 0.82–0.97)	0.54 (95% CI: 0.47–0.59)	0.98	47	262	0.18

CI, confidence interval.

Table 7.

Specificity, sensitivity and negative predictive value of models 1 and 2 excluding pneumococcal urinary antigen positive patients

	Sensitivity	Specificity	NPV	No. of positives	No. of tests (out of 286)	Positives: tests
Model 1	1.00	0.28	1.00	51	221	0.23
Model 2	0.92	0.53	0.97	47	157	0.39

ICU admission and mortality data are presented in Table 8. We used logistic regression to generate a statistical model which demonstrated that CURB‐65 score, CRP and heart rate were independent factors associated with ICU admission in Legionella pneumonia patients. We have denoted this as P(ICU^LP) in the equation below. An increase in CURB‐65 by 1 point is associated with an increase in the log‐odds of ICU admission by 0.74 (95% CI: 0.04–1.61, P = 0.06). An increase in CRP by 1 point is associated with an increase in the log‐odds of ICU admission by 0.009 (95% CI: 0.001–0.019, P = 0.04). An increase in heart rate by 1 point is associated with an increase in the log‐odds of ICU admission by 0.05 (95% CI: 0.02–0.11, P = 0.02). We used DeLong's test to demonstrate that the performance of this model was superior to the CURB‐65 score alone in predicting ICU admission in our study cohort. For two correlated ROC curves, that the AUC of the P(ICU^LP) score logistic regression model (0.82) is significantly higher than the AUC (0.62) of the model with only CURB‐65 (95% CI of the difference: 0.025–0.387, P = 0.03).

$$P\left({\mathrm{ICU}}^{\mathrm{LP}}\right)=\text{CURB}-65\left(0–5\right)+{\mathrm{CRP}}^{*}+\text{Heart Rate}^$$

$$^0–100=0\text{points},101–120=1\text{point},121–140=2\text{points},\mathrm{etc}.$$

Table 8.

Outcomes of intensive care unit (ICU) admission, mortality and length of stay (days)

	Positive (n = 51)	Negative (n = 470)	P‐value
ICU admission	16 (31%)	90 (19%)	0.06
Mortality	3 (6%)	43 (9%)	0.61
Length of stay (SD)	10.8 (10.9)	9.7 (10.5)	0.48

SD, standard deviation.

Score correlation with ICU admission probability is demonstrated by the graph in Figure 1. Table 9 provides percentage risk of ICU admission based on a particular P(ICU^LP) score.

Figure 1.

Probability of intensive care unit (ICU) admission versus P(ICU^LP) score. A graph demonstrating the relationship between a calculated P(ICU^LP) score (based on CURB‐65 score, C‐reactive protein and heart rate) and probability of ICU admission in community‐acquired pneumonia patients.

Table 9.

Intensive care unit (ICU) admission risk (%) according to P(ICU^LP) score

Score	ICU admission risk (%)
0–3	<2
4	5
5	12
6	27
7	50
8	73
9	88
10	95
>10	>95

The other variables from baseline characteristics and pneumonia characteristics were not significant and thus not included in this model. Further logistic regression undertaken to assess whether a positive Lp1 urinary antigen was significantly associated with ICU admission generated a P‐value of 0.93 after controlling for CRP, CURB‐65 and heart rate. This suggests that a positive Lp1 urinary antigen is not a significant independent factor associated with ICU admission.

The public health interventions generated from the positive Lp1 urinary antigen results are demonstrated in the flowchart in Figure 2. Outcomes of related public health investigations were not obtained for 18 cases as they were residents of other health districts. The 31 cases investigated by the SESLHD PHU determined that 12 of the cases were part of outbreaks, and Lp1 was found in five environmental samples collected during the public health investigation. Investigations also led to 11 regulatory actions under the PHA. Regulatory action refers to any action taken under the PHA, including an improvement notice, prohibition order or warning letter. Further definitions of these regulatory action terms are provided in Appendix II.

Figure 2.

Public health interventions. A flowchart demonstrating the outcomes of public health investigations of the 51 positive Lp1 urinary antigen results.

Discussion

The results of our study provided an analysis of patient characteristics, outcomes and public health interventions associated with CAP cases with a positive Lp1 urinary antigen result at St George and Sutherland Hospitals in South‐Eastern Sydney.

Australian TGs recommend Legionella active antibiotics in empiric therapy for both moderate (doxycycline) and severe (azithromycin) pneumonia. ⁸ However, given the public health significance of Legionella pneumonia, it is still imperative that this diagnosis be identified in a timely manner. While conducting a Lp1 urinary antigen test on all patients presenting with CAP would maximise opportunities of diagnosing Legionella pneumonia, only 0.4% of tests (51 out of 13 697 tests) in our study returned positive Lp1 urinary antigen results. As such, this may be a low‐value pathology test with substantial impact on public health resources. Our data suggest that existing guideline recommendations for Legionella urinary antigen testing from Australia, the United States and the United Kingdom have poor sensitivity in our population and would miss a large proportion of Legionella pneumonia cases, further impeding public health efforts to manage and mitigate Legionella pneumonia. In addition, our results demonstrate that Legionella urinary antigen is undertaken more often in clinical practice than guidelines recommend.

We aimed to develop a model for clinicians with close to 100% sensitivity while also maintaining high specificity to ensure that no Legionella pneumonia cases are missed, while also minimising low‐yield testing. Our data suggest that the optimal combination to achieve 100% sensitivity is to perform Lp1 urinary antigen testing in CAP patients who had at least one of the following characteristics upon presentation to the emergency department: >20 standard alcoholic drinks/week, confusion, failure of outpatient antibiotic therapy, diarrhoea, myalgias, temperature >40°C, serum sodium <137 mmol/L or CRP >300 mg/L. Using this algorithm, 401 out of 521 patients would have been tested, and all 51 positive Legionella urinary antigen cases in our cohort would have been correctly identified. If we removed serum sodium and CRP from this algorithm, only 47 out of the 51 positive cases would have been identified, that is, four positive cases would have been missed. However, this would have also reduced the number of tests performed to 262, compared to 401. If each test cost $28.65, a reduction of 139 tests would equate to a monetary saving of $3982.35 in the subset of 521 patients analysed in this study. Further health economic evaluation would be useful in determining whether the public health benefit in detecting four positive Legionella pneumonia cases is worth the additional cost.

Our study determined that CURB‐65 score, CRP and heart rate were factors that may predict ICU admission in CAP patients with a positive Lp1 urinary antigen result and performed better than CURB‐65 alone in our cohort. A retrospective study in the United States demonstrated that CURB‐65 score had low sensitivity for critical care intervention. ²¹ This differs from an observational study conducted in Italy where multivariable analysis identified diabetes, bilateral pneumonia and cardiovascular events as factors independently associated with ICU admission. ²² Future prospective studies with greater power may be able to refine this clinical scoring tool to help identify those at highest risk of severe disease.

The main limitations of our study were the small sample size and the retrospective nature of data collection, which may have led to missed data variables. For example, we relied on the history documented on the EMR for collection of variables such as presenting symptoms, smoking status and alcohol intake. These may have been present but not documented. Future prospective studies may provide more complete and accurate data. The public health intervention data are also incomplete as we were unable to follow up on 18 cases, which were referred to overseas jurisdictions or other Australian jurisdictions based on their residence or any identified exposures outside of the SESLHD jurisdiction.

Another potential limitation of our study was that 50% of the negative Lp1 urinary antigen group consisted of pneumococcal urinary antigen positive patients, which may make this cohort less representative of the general CAP patient population. However, we tested both our models after excluding the 235 pneumococcal urinary antigen‐positive patients, and the results were very similar to those obtained with models where they were included, as demonstrated in Table 6. Therefore, we do not believe that this limitation had any significant impact upon our results.

As our model was designed using retrospective data, the next step is for external validation of the model in the real world with a prospective study. Prospective studies could include economic evaluation to analyse the costs of Legionella urinary antigen testing with respect to benefits such as early antibiotic rationalisation and outbreak detection. Other potential avenues for future research could include similar studies to assess the urinary antigen assay, which includes Legionella longbeachae, as well as the Legionella NAAT in the diagnosis of Legionella pneumonia. Comparisons between these different investigations could enhance clinician decision‐making when selecting diagnostic methods for CAP presentations, which would improve diagnostic accuracy and minimise unnecessary public healthcare expenditure.

Conclusion

Legionella pneumonia is associated with significant morbidity and mortality and thus rapid diagnostic tests such as Lp1 urinary antigen testing are valuable in early identification of cases and ensuring that appropriate clinical and public health management is implemented. It is important to improve diagnostic stewardship to optimise utilisation of limited public health resources. Our study has identified some characteristics of patients presenting with CAP, which are more likely to be associated with a positive Lp1 urinary antigen. Our data demonstrate that this model performs better in our population in terms of negative predictive value than current guideline recommendations from Australia, the United States and the United Kingdom. Future prospective studies with greater power and in different populations may assist in refining this model.

Variable definitions

• Current smoker: patient reported smoking 100 or more cigarettes in their lifetime and was actively cigarette smoking at the time of presentation to the emergency department.

• Former smoker: patient reported smoking 100 or more cigarettes in their lifetime but had stopped, that is, quit smoking prior to their presentation to the emergency department.

• Diarrhoea: patient reported diarrhoea since illness onset upon history taking at presentation to the emergency department.

• Confusion: new onset disorientation to person, place or time OR Abbreviated Mental Test (AMT) score of 8 or less.

• Recent travel: patient reported interstate or overseas travel within 2 weeks of illness onset.

Regulatory action definitions

• Improvement notice: a formal notice issued following an inspection that requires certain actions, for example, a system clean or repair, by a certain time.

• Prohibition order: a formal notice issued following an inspection to cease operation of a cooling water system until certain rectifications are made. Both improvement notices and prohibition orders attract a fee for the operators of the contaminated cooling water system.

• Warning letter: a letter that is issued with reference to the PHA following an inspection advising that certain measures should be taken otherwise an improvement notice or prohibition order will be issued.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.