A Novel Diagnostic Scoring System to Differentiate between Legionella pneumophila Pneumonia and Streptococcus pneumoniae Pneumonia.

Objective We investigated a novel diagnostic scoring system to differentiate Legionella pneumophila pneumonia from Streptococcus pneumoniae pneumonia. Methods We retrospectively reviewed the clinical data of 62 patients with L. pneumophila pneumonia (L-group) and 70 patients with S. pneumoniae pneumonia (S-group). Results The serum sodium (Na) levels tended to be lower according to the severity [age, dehydration, respiratory failure, orientation disturbance, low blood pressure (A-DROP)] score in the L-group. On a multivariate analysis, we found that four factors were independent predictive markers for inclusion in the L-group: relative bradycardia [hazard ratio (HR) 5.177, 95% confidence interval (CI): 1.072-24.993, p=0.041], lactate dehydrogenase (LDH) levels ≥292 IU/L (HR 6.804, 95% CI: 1.629-28.416, p=0.009), C-reactive protein (CRP) levels ≥21 mg/dL (HR 28.073, 95% CI: 5.654-139.462, p<0.001), and Na levels ≤137 meq/L (HR 5.828, 95% CI: 1.411-24.065, p=0.015). Furthermore, a total score [ranging from 0 to 4, the sum of the points for each factor (0 or 1)] ≥3 points indicated a higher probability of inclusion in the L-group than in the S-group. The diagnostic accuracy of a total score of 3 had a sensitivity of 36.3%, specificity of 100%, and area under the curve of 0.682 (95% CI: 0.558-0.806, p=0.004), and that of a total score of 4 had a sensitivity 27.4%, specificity of 98.2%, and area under the curve (AUC) of 0.627 (95% CI: 0.501-0.754, p=0.045). The diagnostic accuracy had low sensitivity but high specificity. Conclusions We found four markers that might be useful for differentiating L-group from S-group and created a novel diagnostic scoring system.

Introduction

Although Legionella pneumophila pneumonia and Streptococcus pneumoniae pneumonia can both be life-threatening diseases that present as lobar pneumonia, a simple method of differentiating them has yet to be reported. Regarding clinical findings, general physicians understand that L. pneumophila pneumonia causes hyponatremia and/or relative bradycardia; however, the diagnostic accuracy of such findings is relatively unknown. Therefore, we retrospectively reviewed the data of L. pneumophila pneumonia patients to generate a simple diagnostic scoring system to differentiate it from S. pneumoniae pneumonia.

Materials and Methods

Patients and study design

We retrospectively reviewed the medical records of patients diagnosed with L. pneumophila pneumonia between January 2001 and December 2016 and searched the Japanese medical literature using the database “Ichushi” (1-12). We included patients whose data were available in the literature. For comparison, we also reviewed the medical records of S. pneumoniae pneumonia patients during the period between January 2013 and September 2014. Patients co-infected with other pathogens were excluded from the study. This study was approved by the Ethics Board of Kyorin University.

Definition

L. pneumophila pneumonia was identified based on the following: 1) a new shadow visualized on chest radiography or chest computed tomography (CT); 2) sputum culture result positive for L. pneumophila; 3) polymerase chain reaction result positive for L. pneumophila; or 4) a positive result for the legionella urinary antigen test (BinaxNOWⓇ, Alere, Tokyo, Japan). Similarly, S. pneumoniae pneumonia was identified based on the following: 1) a new shadow visualized on chest radiography or chest CT; 2) sputum culture result positive for S. pneumoniae obtained from good sputum samples (Geckler 4 or Geckler 5); or a positive result for S. pneumoniae urinary antigen test (BinaxNOWⓇ). Respiratory failure was defined as oxygen saturation measured by pulse oximetry <90% or a respiratory status indicating that oxygen supply was required. Relative bradycardia was based on the definition by Cunha (13). All data were collected at the time of admission.

Statistical analyses

Categorical data are presented as percentages of the total or numerically, as appropriate. Statistical comparisons of nonparametric data were performed using the Mann-Whitney test or Wilcoxon's signed-rank test. Comparisons of categorical data were made using Pearson's chi-squared test. All tests were two-sided. A value of p<0.05 indicated statistical significance. Logistic regression modeling was used for uni- and multivariate analyses to identify predictive risk factors for L. pneumophila pneumonia compared with those for S. pneumoniae pneumonia. Receiver operating characteristic (ROC) curves were used to assess the differentiative powers of serum C-reactive protein (CRP), lactate dehydrogenase (LDH), and sodium (Na) levels. The cut-off point for the serum markers was determined as the minimum value of [(1-sensitivity)2+(1-specificity)2]. Data were analyzed using the SPSS version 20.0 software program for Windows (IBM Japan, Tokyo, Japan).

Results

We found a total of 62 patients with L. pneumophila pneumonia (24 cases from our hospital and 38 from the literature review) and 70 patients with S. pneumoniae pneumonia from our hospital.

Clinical characteristics of the patients with L. pneumophila pneumonia and S. pneumoniae pneumonia

A comparison of the data between the patients with L. pneumophila pneumonia (L-group) and those with S. pneumoniae pneumonia (S-group) groups showed that the median age was similar between the groups; however, the proportion of men was significantly higher in the L-group (Table 1) than in the S-group. The duration from the initial onset of symptoms to the first visit to a local health facility or to our hospital was significantly shorter in the L-group [median 3 days, interquartile range (IQR): 1.0-5.0, p=0.028] than in the S-group (median 3 days, IQR: 1.8-7.0). Interestingly, the body temperature and heart rate were significantly higher in the L-group (median, 38.8℃, IQR: 37.2-39.3℃, p<0.001 and median 108 beats/min, IQR: 94-120 beats/min, p<0.001, respectively) than in the S-group (median 38.0℃, IQR 37.0-38.9°C and 100 beats/min, IQR: 84-107 beats/min, respectively). The frequency of relative bradycardia was markedly higher in the L-group (44.4%, p=0.001) than in the S-group (15.6%). The age, dehydration, respiratory failure, orientation disturbance, low blood pressure (A-DROP) score proposed by the Japanese Respiratory Society [comparable result to that of CURB-65 (14)] showed similar trends in the L- and S-groups (Table 1).

Table 1.

Demographical and Clinical Details of Patients with Legionella Pneumophila Pneumonia and Streptococcus Pneumoniae Pneumonia.

		Legionella pneumophila pneumonia (n=62)		Streptococcus pneumoniae pneumonia (n=70)		p value
Age (years)		68.5 (55-80.5)		72.5 (43.8-83.0)		0.473
Sex (M/F)		54/8		38/32		<0.001
Underlying respiratory diseases		12/28 (42.9%)		31/43 (72.1%)		1.0
Inpatients		61 (98.4%)		34 (48.6%)		<0.001
Smoker		24/28 (85.7%)		31/45 (68.9%)		0.162
Initial onset to first visit to a local or our hospital (days)		3.0 (1.0-5.0)		3.5 (1.8-7.0)		0.028
Body temperature (°C)		38.8 (37.2-39.3)		38.0 (37.0-38.9)		<0.001
Heart rate		108 (94-120)		100 (84-107)		<0.001
Relative bradycardia		24/54 (44.4%)		10/64 (15.6%)		0.001
Respiratory failure#		11/45 (24.45)		22/54 (40.7%)		0.133
Diagnostic methods
Positive for urinary antigen test		58/60 (96.7%)		25/47 (53.2%)		<0.001
Culture positive		8/21 (38.1%)		70 (100%)		<0.001
PCR positive		5/6 (83.3%)		NA
A-DROP*
0		11 (20%)		20 (35.7%)		0.09
1		18 (32.7%)		14 (25%)		0.407
2		13 (23.6%)		6 (10.7%)		0.082
3		9 (16.4%)		12 (21.4%)		0.629
4		3 (5.5%)		3 (5.4%)		1.0
5		1 (1.8%)		1 (1.8%)		1.0

All data are expressed as median (25th-75th percentile) or number (%).

*A-DROP score was available in only 55 patients with Legionella pneumophila pneumonia and 56 patients with S. pneumoniae pneumonia

#Definition of respiratory failure is SpO2 less than 90% or respiratory status required oxygen supply

PCR: polymerase chain reaction, A-DROP: age, dehydration, respiratory failure, orientation disturbance, low blood pressure, NA: not available

A comparison of the serum laboratory data between the L- and S-groups

Serum laboratory results showed that the white blood cell count was comparable between the L- and S-groups; however, the levels of LDH, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and CRP were significantly higher in the L-group (median 378 IU/L, IQR 299-523, p<0.001; median 66.5 IU/L, IQR 36-132, p<0.001; median 43.5 IU/L, IQR 23.5-75.3, p<0.001; median 27.4 mg/dL, IQR 22.1-38.2, p<0.001, respectively) than in the S-group (median 224 IU/L, IQR 197-283; median 23 IU/L, IQR 16-33.5; median 15 IU/L, IQR 12-25; median 7.8 mg/dL, IQR 2.9-14.9, respectively) (Table 2). The serum sodium levels were significantly lower in the L-group (median 135.5 meq/L, IQR 132-139, p<0.001) than in the S-group (median 139 meq/L, IQR 137-141).

Table 2.

Comparison of the Serum Laboratory Results between the Patients with Legionella Pneumophila Pneumonia and S. Pneumoniae Pneumonia.

	Legionella pneumophila pneumonia (n=62)		Streptococcus pneumoniae pneumonia (n=70)		p value
WBC (×103/μL)	10,600 (7,600-13,800)		10,900 (8,350-14,525)		0.662
LDH (IU/L)	378 (299-523)		224 (197-283)		<0.001
AST (IU/L)	66.5 (36-132)		23 (16-33.5)		<0.001
ALT (IU/L)	43.5 (23.5-75.3)		15 (12-25)		<0.001
Na (meq/L)	135.5 (132-139)		139 (137-141)		<0.001
CRP (mg/dL)	27.4 (22.1-38.2)		7.8 (2.9-14.9)		<0.001

All data are expressed as median (25th-75th percentile).

WBC: white blood cell count, AST: aspartate aminotransferase, ALT: alanine aminotransferase, CRP: C-reactive protein, LDH: lactate dehydrogenase

Distribution of serum sodium levels in L. pneumophila pneumonia categorized by A-DROP score

The serum sodium levels seemed to be linked to the A-DROP score (Fig. 1). As shown in Fig. 1, the sodium levels in patients with an A-DROP score of 4 (median 140 meq/L, IQR 138.5-144.3) were significantly higher than those in patients with a score of 0 (median 138 meq/L, IQR 133-139.5, p=0.05), 1 (median 135 meq/L, IQR 132-138, p=0.002), or 3 (median 137 meq/L, IQR 134-140, p=0.043), but not in patients with a score of 2 (median 137 meq/L, IQR 131-140, p=0.074) or 5 (median 143.5 meq/L, p=0.180).

Figure 1.

The distribution of the serum sodium levels in Legionella pneumophila pneumonia categorized by the A-DROP score. All data are expressed as the median (25th-75th percentile). A-DROP: age, dehydration, respiratory failure, orientation disturbance, low blood pressure

When patients with L. pneumophila pneumonia were divided into groups according to their A-DROP scores (1, 2, 3, and 4), we found that the serum sodium levels were significantly lower in the A-DROP <3 group (median 136.5 meq/L, IQR 133-139, p=0.026) than in the A-DROP ≥3 group (median 139 meq/L, IQR 136-140.5) (Fig. 2). The trend was similar for the A-DROP <4 group (median 137 meq/L, IQR 133-139, p=0.001) versus the A-DROP ≥4 group (median 141 meq/L, IQR 139-143) (Fig. 2), implying that the serum sodium levels were likely to be lower in patients with lower A-DROP scores [mild to moderate severe community-acquired pneumonia (CAP)].

Figure 2.

A comparison of the serum sodium levels in groups categorized by the A-DROP score. A-DROP<1 means A-DROP 0. A-DROP<2 means A-DROP ≤0 plus A-DROP 1. A-DROP<3 means A-DROP ≤0 plus A-DROP 1 plus A-DROP 2. A-DROP<4 means A-DROP ≤0 plus A-DROP 1 plus A-DROP 2 plus A-DROP 3. All data are expressed as the median (25th-75th percentile). A-DROP: age, dehydration, respiratory failure, orientation disturbance, low blood pressure

Differentiation of L. pneumophila pneumonia from S. pneumoniae pneumonia

The most appropriate cut-off levels (sensitivity, specificity) for differentiation between the L-group and S-group were 21.5 mg/dL for CRP (78.6%, 91%), 292 IU/L for LDH (71.8%, 83.9%), and 137.5 meq/L for sodium (72.6%, 77.6%) (Fig. 3). Of these 3 factors, a univariate analysis showed that the cut-off level for serum CRP of 21.5 mg/dL had the highest power to predict L. pneumophila pneumonia [hazard ratio (HR) 36.667, 95% confidence interval (CI): 12.001-112.032, p<0.001] (Table 3). Diagnostic accuracy was also demonstrated by other predictive markers, such as the serum sodium levels ≤137 meq/L (HR 12.706, 95% CI: 5.267-30.562, p<0.001), LDH levels ≥292 IU/L (HR 11.667, 95% CI: 4.475-30.417, p<0.001), and relative bradycardia (HR 4.320, 95% CI: 1.824-10.231, p<0.001) (Table 3).

Figure 3.

Differentiating Legionella pneumophila pneumonia from Streptococcus pneumoniae pneumonia using the serum CRP, LDH, and sodium levels. AUC: area under the curve.

Table 3.

Univariate Analysis of Discrimination between Legionella Pneumophila Pneumonia and S. Pneumoniae Pneumonia.

		Hazard ratio		95% CI		p value
Age		1.002		0.981-1.022		0.881
Sex		5.684		2.360-13.689		<0.001
Smoker		2.71		0.790-9.292		0.113
Underlying respiratory diseases		0.944		0.390-2.285		0.898
Initial onset to visit to a local or our hospital (days)		1.163		1.007-1.344		0.04
Respiratory failure*		0.471		0.197-1.123		0.089
WBC		1		1.0-1.0		0.447
LDH		0.992		0.988-0.996		<0.001
LDH ≥292 IU/L		11.667		4.475-30.417		<0.001
Na		1.272		1.144-1.415		<0.001
Na ≤137 meq/L		12.706		5.267-30.652		<0.001
CRP		0.85		0.802-0.901		<0.001
CRP ≥21 mg/dL		36.667		12.001-112.032		<0.001
Heart rate		0.977		0.957-0.997		0.026
Body temperature		0.664		0.488-0.902		0.009
Relative bradycardia		4.320		1.824-10.231		0.001

#Definition of respiratory failure is SpO2 less than 90% or respiratory status required oxygen supply

WBC: white blood cell count, CRP: C-reactive protein, LDH: lactate dehydrogenase

Based on the multivariate analysis, the above 4 predictive factors (relative bradycardia, LDH ≥292 IU/L, CRP ≥21 mg/dL, and Na ≤137 meq/L) seem to be reliable markers for differentiating between L. pneumophila pneumonia and S. pneumoniae pneumonia (Table 4), even in the setting of A-DROP <4 (Table 5). All parameters showed strong, independent associations with L. pneumophila pneumonia.

Table 4.

Multivariate Analysis of Discrimination between Legionella Pneumophila Pneumonia and S. Pneumoniae Pneumonia.

		HR		95% CI		p value
Relative bradycardia		5.177		1.072-24.993		0.041
LDH ≥292 (IU/L)		6.804		1.629-28.416		0.009
CRP ≥21 (mg/dL)		28.073		5.651-139.462		<0.001
Na ≤137 (meq/L)		5.828		1.411-24.065		0.015

CRP: C-reactive protein, HR: hazard ratio, LDH: lactate dehydrogenase

Table 5.

Multivariate Analysis of Discrimination between Legionella Pneumophila Pneumonia and S. Pneumoniae Pneumonia, with Specific Reference to the Patients with A-DROP 0, 1, 2, and 3.

		HR		95% CI		p value
Relative bradycardia		5.797		1.103-30.474		0.038
LDH ≥292 (IU/L)		6.462		1.445-28.902		0.015
CRP ≥21 (mg/dL)		22.243		4.132-119.742		<0.001
Na ≤137 (meq/L)		4.812		1.059-21.868		0.042

CRP: C-reactive protein, HR: Hazard ratio, LDH: lactate dehydrogenase

The diagnostic accuracy of the total score for differentiating L. pneumophila pneumonia from S. pneumoniae pneumonia

To estimate the probability of L. pneumophila pneumonia, we assigned scores of 0 or 1 for serum LDH levels ≥292 IU/L, serum sodium levels ≤137 meq/L, serum CRP levels ≥21 mg/dL, and relative bradycardia, with a total score ranging from 0 to 4. The proportion of patients with L. pneumophila pneumonia compared to S. pneumoniae pneumonia for each total score (diagnostic score) increased markedly at 3 and 4 points (Fig. 4). The differentiative accuracy of the total score proved to be promising when the total score was 3 points [sensitivity 36.3%, specificity 100%, positive predictive value (PPV) 100%, negative predictive value (NPV) 72.7%, and AUC 0.682; 95% CI: 0.558-0.806, p=0.004] or 4 points (sensitivity 27.35%, specificity 98.2%, PPV 90%, NPV 69.6%, and AUC 0.627; 95% CI: 0.501-0.754, p=0.045) (Table 6).

Figure 4.

Patients with Legionella pneumophila pneumonia and Streptococcus. pneumoniae pneumonia according to the diagnostic score. The numbers in the boxes represent the total numbers of patients.

Table 6.

Diagnostic Accuracy of the Total Score in Differentiating Legionella Pneumophila Pneumonia from S. Pneumoniae Pneumonia.

Total score		Sensitivity (%)		Specificity (%)		PPV (%)		NPV (%)		AUC		95% CI		p value
0		0		57.1		0		49.2		0.286		0.182-0.389		0.001
1		6.1		61.1		8.7		51.6		0.336		0.223-0.449		0.011
2		30.3		81.5		50		65.7		0.559		0.432-0.686		0.301
3		36.3		100		100		72.7		0.682		0.558-0.806		0.004
4		27.3		98.2		90		69.6		0.627		0.501-0.754		0.045

AUC: area under the curve, PPV: positive predictive value, NPV: negative predictive value

Different clinical model specifications

The four parameters had different HRs for predicting L. pneumophila pneumonia. A combination of these parameters also differentiated L. pneumophila pneumonia from S. pneumoniae pneumonia, with AUCs between 0.653 and 0.828, even in the setting of A-DROP <4, with AUCs ranging from 0.654 to 0.827 (Table 7).

Table 7.

Specifications of Different Clinical Models.

Model				All cases*		Especially focused on the A-DROP<4**
		Independent variables		AUC (95% CI)		AUC (95% CI)
Model 1		Relative bradycardia		0.653 (0.532-0.775, p=0.015)		0.664 (0.535-0.794, p=0.017)
Model 2		LDH ≥292 (IU/L)		0.773 (0.669-0.878, p<0.001)		0.786 (0.676-0.896, p<0.001)
Model 3		CRP ≥21 (mg/dL)		0.828 (0.731-0.925, p<0.001)		0.827 (0.722-0.931, p<0.001)
Model 4		Na ≤137 (meq/L)		0.761 (0.656-0.865, p<0.001)		0.752 (0.637-0.867, p<0.001)
Model 5		LDH ≥292 (IU/L) and CRP ≥21 (mg/dL)		0.795 (0.683-0.907, p<0.001)		0.791 (0.673-0.908, p<0.001)
Model 6		LDH ≥292 (IU/L) and Na ≤137 (meq/L)		0.697 (0.570-0.825, p=0.003)		0.708 (0.575-0.840, p=0.004)
Model 7		LDH ≥292 (IU/L) and relative bradycardia		0.797 (0.684-0.910, p<0.001)		0.797 (0.679-0.915, p<0.001)
Model 8		CRP ≥21 (mg/dL) and Na ≤137 (meq/L)		0.664 (0.534-0.794, p=0.014)		0.654 (0.516-0.792, p=0.031)
Model 9		CRP ≥21 (mg/dL) and relative bradycardia		0.714 (0.588-0.840, p=0.001)		0.726 (0.595-0.856, p=0.002)
Model 10		Na ≤137 (meq/L) and relative bradycardia		0.790 ( 0.675-0.906, p<0.001)		0.791 (0.671-0.912, p<0.001)
Model 11		LDH ≥292 (IU/L) and CRP ≥21 (mg/dL) and Na ≤137 (meq/L)		0.690 (0.562-0.819, p=0.004)		0.684 (0.548-0.820, p=0.01)
Model 12		LDH ≥292 (IU/L) and CRP ≥21 (mg/dL) and relative bradycardia		0.674 (0.544-0.804, p=0.009)		0.684 (0.548-0.820, p=0.01)
Model 13		LDH ≥292 (IU/L) and Na ≤137 (meq/L) and bradycardia		0.640 (0.508-0.772, p=0.035)		0.649 (0.510-0.787, p=0.03)

*All cases consist of Legionella pneumophila pneumonia (n=62) and S. pneumoniae pneumonia (n=70)

**A-DROP<4 group consist of Legionella pneumophila pneumonia (n=42) and S. pneumoniae pneumonia (n=40).

LDH: lactate dehydrogenase, CRP: C-reactive protein

Change in the serum sodium levels during the clinical course

Of the 14 L. pneumophila pneumonia patients for whom data were available, the serum sodium levels at the time of discharge had increased significantly from those at the time of admission (Wilcoxon's signed-rank test: p=0.021) (Fig. 5).

Figure 5.

Change in the serum sodium levels during the clinical course.

A comparison of the diagnostic yield for L. pneumophila pneumonia between Rico's score and our new score

To examine the difference in the diagnostic accuracy between Rico's score, which is the most recently developed scoring system for L. pneumophila pneumonia, and our new scoring system, we applied the two scores to the data of enrolled patients (Table 8). The total number of patients in the L- and S-groups with available data to determine Rico's score and the new score were n=34 and n=68, and n=33 and n=56, respectively. A Rico's score of ≥4 showed a sensitivity of 47.1% and specificity of 95.6%, with an AUC of 0.713 (95% CI: 0.597-0.830, p<0.001); however, a score of ≥5 was not statistically significant for a diagnosis. In contrast, our new scoring system was reliable for both scores (≥3 and ≥4). In patients with a score of ≥3 in particular, the diagnostic accuracy for L. pneumophila pneumonia showed a sensitivity of 63.6% and specificity of 98.2%, with an AUC of 0.809 (95% CI: 0.703-0.916, p<0.001), which was a higher diagnostic yield than that obtained with a Rico's score of 4.

Table 8.

Comparison of the RICO’s Score and New Score.

		L-group		S-group		Sensitivity (%)		Specificity (%)		AUC (95%CI, p value)
RICO’s score (≥4)		16		3		47.1		95.6		0.713 (0.597-0.830, p<0.001)
RICO’s score (≥5)		6		2		17.6		97.1		0.574 (0.451-0.696, p=0.228)
New score (≥3)		21		1		63.6		98.2		0.809 (0.703-0.916, p<0.001)
New score (≥4)		9		1		27.3		98.2		0.627 (0.561-0.754, p=0.045)

Total number of patients (L-group, S-group) whose data can be available for RICO’s score and/or New score were (n=34, n=68), (n=33, n=56), respectively.

AUC: area under the curve, L-group: Legionella pneumophila pneumonia group, S-group: St reptococcus pneumoniae pneumonia group, 95%CI: 95% confidence interval

Discussion

This study demonstrated the application of a novel scoring system for the diagnosis of L. pneumophila pneumonia. Furthermore, we found that serum sodium levels tended to be low based on the A-DROP score, which is contrary to the common perception of general physicians. L. pneumophila pneumonia and S. pneumoniae pneumonia are the most common causes of CAP. The clinical presentation and chest radiographic findings in L. pneumophila pneumonia might not be specific in hospitalized patients; however, they present with lobar pneumonia, as is also seen in cases of S. pneumoniae pneumonia. Differentiation can be difficult, but no report has yet described simple predictive factors for distinguishing L. pneumophila pneumonia from S. pneumoniae pneumonia.

In this regard, our study also successfully generated a novel scoring system for predicting L. pneumophila pneumonia and differentiating it from S. pneumoniae pneumonia. L. pneumophila pneumonia can be easily diagnosed if the score is ≥3 using 4 simple factors: relative bradycardia, serum LDH levels ≥292 IU/L, CRP levels ≥21 mg/dL, and Na levels ≤137 meq/L. This will enable physicians to treat patients with appropriate empiric antibiotic regimens and facilitate early decision-making. In addition, we indicated the clinical utility of the scoring system in the specifications of the 13 different clinical models. Furthermore, a score of ≥3 with our novel system seemed to be a more reliable and powerful for detecting L. pneumophila pneumonia than the previous score known as “Rico's criteria” (≥4) (15,16). Of note, Rico's criteria were basically developed to rule out L. pneumophila pneumonia from cases of CAP due to all respiratory pathogens, whereas the new score in this study was applied to cases of lobar pneumonia suspected of having either L pneumophila or S. pneumoniae. Regardless of the radiological similarities, the differentiation of L. pneumophila pneumonia from S. pneumoniae pneumonia will play a pivotal role not only in choosing the proper antibiotic treatment but also in recognizing the possibility of life-threatening situations over a short period of time due to other pathogens.

Interestingly, our study was the first to show that the serum sodium levels were directly associated with the A-DROP score. Patients with less severe L. pneumophila pneumonia are likely to have lower values of serum sodium levels than those with severe and/or very severe L. pneumophila pneumonia of unknown pathophysiology.

Hyponatremia is common in Legionnaires' disease and is considered to be the result of inappropriate secretion of antidiuretic hormone. However, the precise underlying mechanisms are uncertain, and previous reports have suggested that an alternative explanation is needed (17-19). Recently, Schuetz et al. confirmed that no correlation exists between serum CT- provasopressin (precursor of ADH) and sodium levels in Legionnaires' disease (20); however, their data appeared to show a positive correlation between the serum sodium levels and the disease severity (CURB-65 score). In the present study, the sodium levels returned to the normal range in due course, suggesting a Legionella infection. Thus, hyponatremia might be a valuable marker, especially in less severe cases of pneumonia.

Several limitations associated with the present study warrant mention. First, it was a retrospective study. Second, it had a relatively small number of L. pneumophila pneumonia patients. Third, we specifically focused on the differentiation between L. pneumophila pneumonia from S. pneumoniae pneumonia as representative causes of lobar pneumonia showing similar radiological features and the potential to induce a life-threatening condition.

However, to our knowledge, this study included the largest number of Legionella-infected patients, and our simple predictive scoring system can be used in the setting of lobar pneumonia, which is a characteristic radiological feature of both S. pneumoniae pneumonia and L. pneumophila pneumonia. Our scoring system warrants further study to ascertain the diagnostic accuracy in all CAP patients with various etiologies.

Conclusions

This study was the first to show that patients with less severe L. pneumophila pneumonia are likely to have lower serum sodium levels than patients with S. pneumoniae pneumonia. We also generated a novel scoring system with high specificity using four simple predictive factors.

The current study was approved by the institutional review board of Kyorin University School of Medicine (IRB: H29-013), and written informed consent was waived.

The authors state that they have no Conflict of Interest (COI).