INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The etiology of adult community-acquired pneumonia (CAP) has been under constant study in different local settings during the past decade (1). The microbial patterns reported have differed considerably, depending on the epidemiologic area studied, the patient population included, and the extent and nature of the microbiologic techniques used. The knowledge of predominant microbial patterns in CAP, however, represents the basis for initial decisions about its empirical antimicrobial treatment (10). It therefore appears crucial that large tertiary care centers determine the peculiar microbial patterns prevalent in CAP at least in their hospitalized patients with the disease.

Instead of the traditional syndromic approach (i.e., the clinical differentiation of "typical" and "atypical" penumonia), recent guidelines of the American Thoracic Society (ATS) for the initial management of CAP proposed a strategy for its initial empirical treatment based largely on the criteria of age, comorbidity, and severity of pneumonia, which are thought to represent the main determinants of microbial etiologies of CAP (10). This concept was derived from a synopsis of studies from different areas and with different populations (2, 4, 11- 13). However, no study so far has comprehensively determined the impact of these three criteria on microbial etiologies of CAP in a distinct hospital setting.

We therefore conducted a prospective study of the etiology of CAP in hospitalized patients with such pneumonia, in order to determine the local microbial spectrum and the influence of age, comorbidity, and severity of pneumonia on microbial patterns in the disease. We also were interested in the potential capacity of classical clinical and radiographic features of typical pneumonia to pneumococcal pneumonia as opposed to nonpneumococcal pneumonia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population

From October 1996 to December 1997, we prospectively studied 395 consecutive patients with CAP who were admitted to our hospital (an 1,000-bed university teaching hospital). CAP was defined as the presence of symptoms of lower respiratory tract infection, along with a new infiltrate on chest radiography and no emerging alternative diagnosis during follow-up, in a patient not hospitalized for at least the previous 72 h. Patients were excluded for severe immunosuppression, such as manifested by neutropenia (< 1.0 × 10⁹/L), or from human immunodeficiency virus (HIV) infection, solid-organ or bone-marrow transplantation, or treatment with steroids in a dosage of > 20 mg prednisone-equivalent per day for 2 wk or more. Patients with pneumonia as an expected terminal event of a chronic disabling illness, tuberculosis, and postobstructive pneumonia due to lung cancer were also excluded.

Data Collection

Clinical, radiographic, and laboratory data were recorded on a data sheet and entered in a computer database. The following data were recorded on admission: age, gender, smoking and alcohol habits, comorbid illnesses, antimicrobial treatment prior to hospital admission, duration of symptoms before the diagnosis of pneumonia, clinical symptoms (body temperature, presence of chills, pleuritic chest pain, purulent sputum, confusion (i.e., disorientation with regard to person, place, or time that was not known to result from chronic stupor or coma), clinical presentation (presence of rhales, respiratory rate, heart rate, arterial systolic and diastolic blood pressures), results of blood gas analysis (Pa_O2, Pa_CO2, Pa_O2/FI_O2), chest radiographic pattern (alveolar, interstitial or mixed infiltrate, uni- versus bilateral involvement), leukocyte count, and serum creatinine concentration. At the clinical endpoints of hospital discharge or death, the following parameters were additionally retrieved: results of microbial investigations, admission to the intensive care unit (ICU), and in-hospital outcome.

Comorbidities were defined as follows: cardiac comorbid illness: treatment for coronary artery disease or congestive heart failure, or presence of valvular heart disease; pulmonary: treatment for asthma or chronic obstructive pulmonary disease (COPD), or presence of interstitial lung disorders; renal: preexisting renal disease with a documented abnormal serum creatinine level outside the period of the pneumonia episode; hepatic: preexisting viral or toxic hepatopathy; disorders of the central nervous system (CNS) presence of symptomatic acute or chronic vascular or nonvascular encephalopathy, with or without dementia; diabetes mellitus: diagnosis of intolerance to glucose and treatment with oral antidiabetics or insulin; neoplastic illness: any solid tumor active at the time of presentation or requiring antineoplastic treatment within the preceding year.

Alcohol abuse was defined as the ingestion of an estimated amount of  80 g alcohol per day for at least the year before presentation. Smokers were defined as current smokers of  10 cigarettes/d during at least the preceding year.

Microbiologic Evaluation

Regular sampling was done of sputum; blood, for two blood cultures; and serum for paired serology (at admission and within the 4th and 8th weeks thereafter). Pleural puncture, transthoracic needle puncture, and tracheobronchial aspiration (in mechanically ventilated patients), with flexible fiberoptic bronchoscopy for protected specimen brush (PSB) sampling and bronchoalveolar lavage (BAL) as additional diagnostic techniques, were used according to the clinical judgment of the physician in charge.

Sputum was Gram stained. Representative sputum originating from the lower respiratory tract was defined as that containing > 25 granulocytes and < 10 epithelial cells per low power field (lpf) (total magnification: ×100) (14). Such validated sputum, blood, culture samples, as well as pleural fluid, transthoracic needle aspiration samples, undiluted and serially diluted tracheobronchial aspirates (TBAS), and PSB, and BAL fluid (BALF) samples were plated on the following media: blood-sheep agar, CDC agar, chocolate agar and Sabouraud agar. Undiluted PSB and BALF samples were also cultured on charcoal-yeast-extract agar. Identification of microorganisms was done according to standard methods (15). Results of quantitative cultures were expressed as colony-forming units per milliliter (cfu/ml).

Diagnostic Criteria

The etiology of pneumonia was classified as presumptive if a valid sputum sample yielded one or more predominant bacterial strains. The etiology was considered definite if one of the following criteria was met: (1) blood cultures yielding a bacterial or fungal pathogen (in the absence of an apparent extrapulmonary focus); (2) pleural fluid and transthoracic needle aspiration cultures yielding a bacterial pathogen; (3) seroconversion (i.e., a fourfold increase in IgG titers for Chlamydia pneumoniae (IgG  1:512), Chlamydia psittaci (IgG  64), Legionella pneumophila  1:128, Coxiella burnetii  1:80, and respiratory viruses, (i.e. Influenzaviruses A and B, parainfluenza viruses 1 to 3, respiratory syncytial-virus, adenovirus; (4) a single increased IgM titer for C. pneumoniae  1:32, C. burnetii  1:80, and Mycoplasma pneumoniae (any titer); (5) a single titer  1:128 or a positive urinary antigen for L. pneumophila; (6) bacterial growth in cultures of TBAS  10⁵ cfu/ml, in PSB  10³ cfu/ml, and in BALF  10⁴ cfu/ml. Growth of fungi in respiratory samples was considered diagnostic only in the presence of a concomitant positive blood culture growing the same microorganism.

Independently of microbiologic results, a diagnosis of probable aspiration was made in cases of witnessed aspiration or in the presence of risk factors for aspiration (severely altered consciousness, abnormal gag reflex, or abnormal swallowing mechanism) (16).

Statistics

Results are expressed as means ± SD. Continuous variables were compared by using Student's t test; categorical variables were compared by using the chi-square test or Fisher's exact test where appropriate.

The most frequent pathogens, as well as grouped etiologies (atypical viral pathogens; atypical bacterial pathogens (excluding Legionella sp.); atypical viral and bacterial pathogens (again excluding Legionella sp.); nonpneumococcal, nonatypical pathogens; gram-negative enteric bacilli (GNEB) and Pseudomonas aeruginosa; and undetermined etiology (for definitions see Table 2), were tested for an association with age (less or more than 60 yr), smoking habit, alcohol abuse, comorbid illnesses as defined earlier, and severe pneumonia (requiring ICU admission) by using chi-square tests with odds ratios (ORs) and 95% confidence intervals (CIs). Multivariate analysis was applied by using stepwise forward logistic regression.

To discriminate pneumococcal and nonpneumococcal pneumonia (and excluding mixed etiologies), sensitivities and specificities of clinical and radiographic features of typical pneumonia (i.e., eight criteria overall, including acute onset < 72 h, fever  39° C, chills, pleuritic chest pain, purulent sputum, rhales, leukocytosis  12 × 10⁹/L, and alveolar infiltrates) were calculated according to standard formulas. The overall accuracy of combined criteria was evaluated with receiver operating characterictic (ROC) curves, which were constructed by plotting sensitivity against 1  specificity at different numbers of criteria fulfilled (from one of eight to eight of eight).

All reported p values are two-tailed. The level of significance was set at 5%.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Descriptives

The 395 patients (260 male and 135 female) had a mean age of 68 ± 18 yr. Twenty-three patients (6%) were admitted from a nursing home. Clinical symptoms and radiographic patterns, as well as comorbid illnesses, are summarized in Table 1. The mean duration of symptoms before the diagnosis of pneumonia was 5 ± 5 d. Sixty-two patients (16%) had received antimicrobial pretreatment prior to hospital admission for pneumonia. Seventy-six percent of patients (n = 301) had at least one comorbid illness, with pulmonary comorbidity the most frequent comorbid condition (45%). Vital-sign abnormalities at admission, as defined by Fine and coworkers (23), were present in 242 (61%) patients, and included respiratory rate  30/min in 187 patients (47%), systolic blood pressure < 90 mm Hg in nine (2%), heart rate  125 beats/min in 40 (10%), confusion in 96 (24%), and body temperature  40° C in 10 (3%) and < 35° C in one. Respiratory failure (Pa_O2/FI_O2 < 250) was present in 142 of 395 (36%) patients. Sixty-four patients (16%) required ICU admission, and 37 of these required mechanical ventilation (9%). Nineteen patients (5%) died.

Microbial Etiology of CAP

A microbial etiology could be determined in 182 of the 395 (46%) patients. It was presumptive in 65 of 395 (16%) and definite in 117 of 395 (30%) cases. Mixed infections were present in 41 (10%) patients (37 [9%] with two pathogens and four [1%] with three pathogens) (Table 2).

Overall, 227 pathogens were detected, of which 72 (32%) were detected by sputum examination (presumptively diagnostic) and 155 (68%) were detected by other techniques (definitely diagnostic). S. pneumoniae constituted 65 of the 227 pathogens (29%), and was the most frequent pathogen. The second to fifth most frequent pathogens were Hemophilus influenzae (n = 25, 11%), influenza viruses A and B (n = 23, 10%), Legionella sp. (n = 17, 8%), and C. pneumoniae (n = 15, 7%), respectively. With regard to grouped pathogens, atypical bacterial pathogens (excluding Legionella sp.) accounted for 41 of the 227 pathogens detected (18%), viral pathogens for 39 of the 227 (17%), and nonpneumococcal, nonatypical microorganisms for 65 of the 227 (29%). The last-named group included 13 (6%) GNEB and 12 (5%) P. aeruginosa, respectively (Table 2).

Mixed infections were present in 41 (23%) of the 182 patients in whom a microbial etiology was determined. Of these, only one apparent pathogen was covered by the initial empiric antimicrobial treatment in 19 instances (46%). Pathogens not covered included: viruses (n = 13), M. pneumoniae (n = 1), C. burnetii (n = 2), and P. aeruginosa (n = 3).

Twenty-eight patients (7%) had suspected aspiration pneumonia. The etiology in seven cases of this subgroup (25%) included Legionella sp., C. pneumoniae, and C. burnetii in one case each, Escherichia coli in three cases (including one case of mixed infection with Enterococcus faecalis), and Klebsiella pneumoniae in one case.

The etiology was determined in seven of 23 (30%) patients with nursing-home-acquired pneumonia, and included S. pneumoniae (n = 3), H. influenzae (n = 2), E. coli (n = 1), and E. coli and E. faecalis (n = 1). Both infections with E. coli were blood culture-positive. Aspiration pneumonia occured in four of the 23 (17%) patients, with a trend toward more frequent occurence than in patients with domestically acquired pneumonia (p = 0.07).

Diagnostic Yield of Different Techniques

Sputum was collected from 243 patients, of whom 130 provided a valid sample. Blood cultures were achieved for 261 patients, and serologic samples were taken from 363 patients (single samples in 159 cases and paired samples in 204). In the remaining patients, these regular investigations were not done because of inability to cooperate (sputum), early death, patient noncomplicance (paired serology), and collection errors. Serology was done significantly less often in patients with no comorbidity (p < 0.0001) and in the subgroup of patients with CNS disorders (p < 0.01). In addition, TBAS were obtained from 31 patients, pleural fluid from 29, and transthoracic puncture specimens from five patients. PSB and BAL were performed in 10 patients each.

Noninvasive techniques had yields of 53% (sputum), 13% (blood culture), 45% (serology), and 58% (TBAS). Pleural fluid produced a yield of 14%, transthoracic puncture of 20%. and PSB and BAL of 30% and 20%, respectively (Table 3).

A pneumococcal etiology of CAP was identified exclusively by sputum testing in 35 (54%) cases, by blood culture in 17 (26%), by pleural fluid in one (2%), by TBAS in six (9%), and by more than one technique in a further six cases (9%) (sputum and blood culture in three cases [5%]; and blood culture and pleural fluid; TBAS and PSB; and blood culture, pleural fluid, and transthoracic puncture in one case each). Thus, 22 cases (34%) were bacteremic. Legionellosis was diagnosed by serology in 15 of 17 (88%) cases (single results in three cases and paired results in 12), and by urine antigen testing in two cases (12%). Serologic results were confirmed by a positive urine antigen result in a further four cases. Most diagnoses of GNEB were based on techniques other than sputum examination (10 of 13, 77%), whereas P. aeruginosa was detected mainly by sputum examination (10 of 12 cases, 83%). Ten of 12 patients with P. aeruginosa had COPD, including one with definite bronchiectasis. Of these, only one patient had no comorbid illness, but was a heavy cigarette smoker (20 pack-yr). P. aeruginosa was the only pathogen detected in seven of the 12 patients in whom it was found, whereas five had mixed infections, with S. pneumoniae as the second pathogen in two cases, C. burnetii as the second pathogen in two cases, and E. coli as the second pathogen in one case).

The diagnostic yield was not significantly different in 62 patients who had antimicrobial pretreatment from those who did not. Sputum cultures were equally frequently positive in both groups (30% versus 29%, p = 0.88). On the other hand, blood cultures and pleural fluid were more often positive in patients without antimicrobial pretreatment (15% versus 7% [p = 0.18], and 18% versus 2% [p = 0.04], respectively). A converse trend was obvious with regard to serology (23% positive versus 32%, p = 0.17).

Impact of Age, Comorbidity, and Severity on Microbial Etiology

An age  60 yr was not associated with any discernible microbial etiology. However, patients aged < 60 yr significantly more frequently had CAP caused by atypical bacterial pathogens, especially M. pneumoniae. No association with microbial etiology could be determined for patients with nursing-home acquired pneumonia (Table 4).

Patients without comorbid illnesses were also more likely to have a viral or atypical bacterial etiology. Current cigarette smokers were more likely to have pneumonia caused by Legionella sp. and C. pneumoniae. Pulmonary and hepatic comorbid illness, as well as alcohol abuse, were significantly associated with CAP due to S. pneumoniae and to mixed infections. Pulmonary comorbidity predisposed to CAP caused by GNEB and P. aeruginosa, and particularly the latter. Infections due to atypical bacterial pathogens, especially C. pneumoniae, were more frequent in patients with alcohol abuse. CNS disorders favored aspiration pneumonia. Hepatic comorbidity and alcohol abuse were associated with bacteremic infections, and alcohol abuse was associated with bacteremia due to S. pneumoniae in particular. These associations are listed in Table 4.

In the population aged < 60 yr and without comorbid illnesses, an etiology of the group comprising atypical bacterial pathogens was more probable (10 of 40 cases [40%]) (OR: 3.5; 95% CI: 1.4 to 8.3; p = 0.002), and especially pneumonia due to M. pneumoniae (five of 13 cases [39%]) (OR: 5.9; 95% CI: 1.5 to 21.8; p = 0.001). In patients aged  60 yr and having any comorbidity, there was an association with GNEB (21 of 24 cases [88%]) (OR: 4.4; 95% CI: 1.2 to 23.4; p = 0.01), and particularly with P. aeruginosa (11 of 12 cases [92%]) (OR: 6.7; 95% CI: 1.0 to 291.4; p = 0.04).

Severe pneumonia (requiring ICU admission) was significantly associated with the pathogens S. pneumoniae and bacteremic S. pneumoniae in particular, and with mixed infections and bacteremic infections. The association with the GNEB group and with P. aeruginosa had a comparable ORs of 2.3 in each case, but failed to reach significance. However, in a multivariate analysis including these five etiologies, pneumonia due to S. pneumoniae, and to the GNEB group and P. aeruginosa remained independently associated with severe pneumonia (Table 5).

Association of Microbial Etiology with Death

Mortality was highest in cases of pneumonia due to GNEB (two of 13 cases, 15%) and P. aeruginosa (one of 12 cases, 8%). It was 6% (four of 65 cases) for S. pneumoniae, 5% (one of 22 cases for bacteremic pneumococcal pneumonia), 8% (one of 13 cases) for M. pneumoniae, 4% (one of 25 cases) for H. influenzae, and 6% (two of 34 cases) for bacteremia. Legionellosis, C. pneumoniae and viral pathogens had no associated mortality. Mortality was 4% (eight of 213 cases) in the group with CAP of undetermined etiology. One patient with mixed infection (S. pneumoniae and H. influenzae) died despite appropriate initial antimicrobial treatment.

Association of Features of Typical Pneumonia with Etiologies

The sensitivity of the six clinical features of typical pneumonia, leukocytosis  12 × 10⁹/L, and alveolar infiltrates for pneumonia caused by S. pneumoniae was low (ranging from 24% to 82%), and the specificities did not exceed 70% for legionellosis or the viral and atypical bacterial pathogen group (ranging from 7% to 67%) (see Table 6).

ROC curves showed that in the presence of four of eight criteria, sensitivity for pneumococcal pneumonia reached 80% and specificities for legionellosis, viral and atypical bacterial pathogens, nonpneumococcal, nonatypical pathogens, and undetermined etiologies ranged from 12 to 66%, whereas in the presence of seven of eight criteria, specificities reached 96 to 100% at the expense of a very low sensitivity (3%) (Figure 1).

View larger version (27K): [in this window] [in a new window]   Figure 1.   ROC curves for from one to eight classical clinical and radiographic criteria for typical CAP. The open square at the right top represents the presence of one criterion, the open square at the left bottom represents the presence of eight criteria.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study, one of the largest unicenter studies, including 395 consecutive hospitalized patients, provides a comprehensive insight into the impact of age, comorbidity, and severity on the etiology of CAP. The main results were that: (1) patients aged < 60 yr without comorbid illnesses were more likely to have an etiology comprising viral or atypical bacterial pathogens, especially M. pneumoniae; (2) pulmonary, hepatic, and CNS comorbidities, as well as current cigarette smoking and alcohol abuse, all had a significant bearing on distinct etiologic patterns; (3) severe pneumonia (requiring ICU admission) was independently associated with the pathogens S. pneumoniae and with the GNEB group and P. aeruginosa; and (4) whereas age, comorbidity, and severity were strongly associated with the etiology of pneumonia, classical clinical and radiographic features of typical pneumonia were neither sensitive nor specific for the differentiation of pneumococcal and nonpneumococcal etiologies.

With noninvasive diagnostic techniques used for the majority of patients, the percentage of patients with presumptive and definite etiologies (46%) was within the range of the expected percentages. This was also true for the diagnostic yields of the different noninvasive and bronchoscopic techniques. Likewise, the distributions of the main etiologic groups and the five most frequent pathogens were comparable to those in several previous series. S. pneumoniae was unanimously found to be the leading pathogen in CAP (1). In other recent series, H. influenzae was reported to account for 4 to 11% of episodes (1), influenza viruses A and B for 3 to 8% (1, 2, 6), Legionella spp. for 2 to 16% (1), and C. pneumoniae for 3 to 18% (4, 6), all of which are well comparable to our frequencies of 11%, 10%, 8%, and 7%, respectively. Nevertheless, we found an unusually high percentage of viral infections (21%, including 13 of 39 [33%] cases of mixed viral/bacterial infections, as compared with 9 to 16% in other series [1-3, 5-8]), which may be mainly attributable to the high number of paired serologies obtained. We also found a relatively high percentage of GNEB (6%) and of P. aeruginosa (5%) among the pathogens involved. Only four recent studies had a proportion of > 3% of CAP caused by GNEB (4, 6, 8, 18), with the highest proportion reaching 9% (18). P. aeruginosa was reported to account for only 1 to 2% in other series (4, 5, 8, 9), except for one recent study that reported 5% (6). Whereas in our study the former etiologies were definite in most (77%) cases, P. aeruginosa was identified exclusively by sputum examination on 85% of occasions, mostly in patients with COPD (and bronchiectasis in one patient). Since such patients may be colonized with P. aeruginosa (19), the etiologic role of this pathogen in the pneumonia episode may be questioned. However, all 12 patients in whom P. aeruginosa was identified in our study had a valid sputum sample, and six of the 12 had no evidence of other pathogens. Thus, it seems prudent to consider the pathogens named here as at least probable pathogens of pneumonia. Similar considerations are applicable to patients with pneumococcal pneumonia proven only by sputum testing, and with COPD (19).

The impact of age was restricted to an association of younger age (< 60 yr) with atypical bacterial pathogens, especially M. pneumoniae. This finding is in accord with the findings in several previous studies (7). An association of advanced age with gram-negative pathogens and P. aeruginosa has been claimed by several authors (10, 31). Although in our study an age  60 yr with any comorbid illness was significantly associated with these pathogens, advanced age alone was not, suggesting that comorbidity rather than age is the determinating condition predisposing to these etiologies. Accordingly, these etiologies were rare in two recent studies of CAP in the elderly, ranging between 0 and 5% (21, 22).

Patients without comorbid illnesses were more likely to have a viral or atypical bacterial etiology. However, given that serology was done significantly more often in these patients, this finding must be interpreted with caution.

Smokers were more likely to have CAP caused by Legionella spp. and C. pneumoniae. Whereas smoking has only recently been described as an independent risk factor for sporadic legionellosis (23), no such association has yet been shown for pneumonia due to C. pneumoniae.

Pneumococcal pneumonia was significantly associated with pulmonary and hepatic comorbidities as well as with alcohol abuse. Although COPD, hepatic cirrhosis, and high alcohol consumption have all repeatedly been described as risk factors for CAP (2, 3, 11, 13, 24), they have not been recognized as having an association with pneumococcal pneumonia. In accord with these findings of our study, hepatic comorbidity and alcohol abuse were also associated with bacteremia, and alcohol abuse with bacteremic S. pneumoniae pneumonia in particular, suggesting a continuum of local and systemic immunocompromise as underlying the pathogenetic mechanism of this association. Whereas COPD may predispose to pneumococcal pneumonia by previous tracheobronchial colonization (19), hepatopathy may cause an impairment of phagocyte clearance of invading pathogens. An impairment of several local and systemic host immune mechanisms in alcoholic individuals, and especially of lymphocyte functions, has been described (25, 26). Perhaps even more surprising, but consistent with the described immunologic alterations, was the finding that infections due to atypical bacterial pathogens, and especially C. pneumoniae, were also associated with alcohol abuse. This finding has not been reported previously. On the other hand, we could not confirm a higher incidence of GNEB reported in a smaller previous study of CAP using a similar definition of alcohol abuse (24).

Pulmonary comorbidity also predisposed to pneumonia due to GNEB and P. aeruginosa, probably owing mainly to previous tracheobronchial colonization in COPD (and/or bronchiectasis) (19). We were unable to establish associations of these pathogens with other conditions, such as diabetes mellitus or aspiration. As expected, suspected aspiration pneumonia was strongly associated with CNS disorders. However, this association may have included an overestimate of the true incidence of aspiration since the diagnosis of aspiration was largely based on criteria for likely aspiration as a consequence of CNS disorders. Conversely, because of significantly fewer serologies performed on these patients, atypical pathogens may have been underestimated.

Because there is no universally accepted definition of severe CAP, it was defined for our study as pneumonia requiring ICU admission. This definition is clearly not irrefutable, nor does it imply that all patients were severely ill to the same degree. Nevertheless, it reflects clinical judgment in a specialized center, and the results may therefore well be relevant to other settings. Severe pneumonia was associated with the pathogens S. pneumoniae, GNEB, and P. aeruginosa, as well as with mixed etiologies and bacteremic infections. In multivariate analysis, CAP due to S. pneumoniae, and to GNEB and P. aeruginosa, was in each case an independent predictor of severity. These etiologic associations were certainly not due to increased age, since none of these pathogens was associated with an age  60 yr. Although there are no studies of risk factors for severe pneumonia with which to compare our results, several studies of series of patients with severe CAP have shown S. pneumoniae (27, 28), GNEB (27, 29), P. aeruginosa (13, 30), and bacteremia (13, 27, 31) to be significant risk factors for death, in accord with our own findings.

In terms of mortality, GNEB had the highest associated mortality (15%). Pneumonia due to P. aeruginosa had a relatively low mortality (8%) as compared with that of ventilator-associated pneumonia, which may reach 70% (32). This inconsistency may be explained by differences in severity of comorbidity, bacterial load, and antimicrobial susceptibility. Surprisingly, the mortality from bacteremic pneumococcal pneumonia was equal to that from noninvasive pneumonia (5% versus 6%), and the mortality associated with bacteremia in general did not exceed 6%. Others have reported a mortality twice as high for blood culture-proven pneumococcal pneumonia (3). One might speculate that the low mortality rates in these bacteremic patients are attributable to a short interval between onset of symptoms and appropriate antimicrobial treatment. In any case, these mortality rates must be interpreted with caution, since both the overall and pathogen-related mortality rates were generally low.

Single or combined clinical and radiographic features of typical pneumonia failed to discriminate pneumococcal pneumonia from nonpneumococcal etiologies. This finding corroborates several previous reports (33). As pointed out by others, the nature of the host and the immunologic host response most probably have a more important bearing on the clinical features of pneumonia than does the underlying pathogen. A further important confounder, only rarely considered, may be that viral etiologies especially may often include undetected mixed infections with S. pneumoniae. Thus, the traditional classification of CAP into typical and atypical pneumonia in order to consistently guide initial empirical antimicrobial therapy in hospitalized patients is no longer adequate. A remarkable exception may be the rare patient who presents with the complete pattern of typical clinical and radiographic features of CAP.

In conclusion, given that age, comorbidity, and severity of pneumonia had strong associations with distinct etiologic patterns of CAP, whereas classical features of typical pneumonia were neither sensitive for pneumococcal pneumonia nor specific for other etiologies, our study clearly supports a management approach based on these criteria as recommended by the ATS, rather than on the traditional syndromic approach. This strategy, together with constant epidemiologic surveys in different hospital settings, and notably a regular, comprehensive diagnostic workup in the most severe cases of pneumonia, may offer the most beneficial perspective on the initial management of patients hospitalized with CAP.