The timely identification of Legionella in patients presenting with CAP to the emergency department is challenging because single clinical and laboratory parameters have shown low sensitivity and/or specificity. A reliable diagnostic score integrating different parameters is lacking . The results of this study indicate that six clinical and laboratory parameters, namely high body temperature, absence of sputum production, low serum sodium and platelet counts, and high LDH and CRP concentrations combined in a diagnostic score reliably estimate the likelihood of Legionella in patients who present with CAP to the emergency department.
Current guidelines on the management of patients with CAP recommend that empirical antibiotic coverage should be extended to Legionella in "suspicious cases", although no single parameter can reliably identify or rule out patients at risk for Legionella [8, 13]. Previous studies have addressed this dilemma and assessed potential predictors of Legionella CAP in different settings [1, 5–12]. A Spanish study found that Legionella should be suspected in middle-aged, alcohol drinking, healthy male patients, if patients lack response to previous beta-lactamic drugs, if headache, diarrhea, severe hyponatremia, and elevation in serum CK levels were present, or if no cough, expectoration, and thoracic pain were found . A recent critical review included 13 different studies that reported clinical details for the evaluation of Legionella CAP, however, concluded that using a syndromic approach cannot be recommended and, thus, an empiric antibiotic therapy covering for Legionella for all patients with CAP requiring hospitalization should be recommended . This, however, results in indiscriminate use of, mostly unnecessary, dual antibiotic therapy and thus, increased antibiotic exposure and emergence of resistance. Two previous studies have proposed a clinical prediction rule to identify Legionella CAP [5, 6]. The CBPIS score has been proposed with a maximum of 17 points based on the evaluation of temperature, serum creatinine, sodium and LDH concentrations, headache, vomiting and smoking. However, in a prospective validation study, this scoring system had a low sensitivity and/or specificity to diagnose or exclude Legionella CAP . Guidance of antibiotic therapy using this score would have left half of the patients with Legionella CAP without specific coverage and would have led to the administration of an unnecessarily broad antibiotic regimen in 14% of patients without Legionella. In addition, the model was unable to distinguish between Legionella pneumophila and Streptococcus pneumoniae, because the majority of patients were categorized into the "intermediate-probability" group. Similarly, the clinical criteria proposed by the Winthrop-University Hospital (WUH) for the identification of Legionellosis showed an inadequately low sensitivity and specificity of 78% . In addition, the WUH study included clinical data collected throughout the first 7 days of hospitalization limiting its use as a screening tool in an emergency department setting. Importantly, both studies based their scoring system on the comparison of Legionella CAP with pneumococcal CAP, which does not reflect the clinical situation in the emergency department, where the decision about empirical antibiotic therapy must be taken in all patients with CAP. In this study, a thorough and comprehensive 10 year retrospective data collection was conducted and data from two prospective studies presenting to the emergency department of the same institution were used to compare clinical and laboratory parameters. Thus we believe the proposed score reflects the clinical situation more realistically.
Simplicity of a diagnostic score is a major determinant for its future usefulness in daily practice. The WUH scale consists of 23 different criteria, which increases its complexity and may correlate with non-adherence as observed in clinical severity prediction scores [6, 24]. The herein proposed scoring system composes of only 6 routinely measured clinical and laboratory parameters that showed an independent predictive value in multivariate logistic regression analysis and a high overall diagnostic performance as demonstrated in ROC analysis.
As with any diagnostic score, the "optimal" cut off should be chosen based on the pretest probability and the clinical context, particularly the risk for adverse medical outcomes of patients. Supported by recent guidelines, treatment decision on the empirical antibiotic therapy should be based on both, the diagnostic probability and the risk of patients based on a prognostic assessment. Importantly, the PSI – but not the CURB65 score – showed a high prognostic accuracy to predict adverse medical outcomes in patients with Legionella CAP in this study. This finding is reassuring because prognostic CAP scores mainly depend on age and Legionella CAP patients tend to be younger. Thus, taking the diagnostic limitations of urinary antigen testing into account, empirical antibiotic therapy covering for atypical pathogens in patients with negative urinary antigen testing but high diagnostic probability (diagnostic score ≥ 4) may be advisable. Conversely, in low risk patients with low diagnostic probability (diagnostic score ≤ 1) and no clinical or diagnostic evidence of CAP caused by other atypical pathogens, macrolides may initially be withheld until the results of the urinary antigen tests become available.
Some limitations should be considered in interpreting our results. With a retrospective design, our results are preliminary and prospective validation is needed prior to a widespread use in clinical practice. The diagnosis of Legionella spp. in non-Legionella CAP patients was performed routinely by urinary antigen testing without PCR, culture or serology and we may have classified some patients incorrectly. Moreover there was no further routinely investigation for other atypical bacterial pathogens. We chose the 10 year retrospective design to provide the necessary number of Legionella cases to calculate multivariate analysis with adequate power. The proposed score consists of mainly objectively measurable parameters, which may robustly be ascertained in retrospect. Still, anamnestic parameters (i.e. absence of sputum production), even if highly available, may not be accurately assessed in retrospect and are thus underestimated in favour of laboratory parameters and clinical data.