Early- and Late-Onset Pneumonia: Is This Still a Useful Classification?

Abstract

The choice of empirical treatment of nosocomial pneumonia in the intensive-care unit (ICU) used to rely on the interval after the start of mechanical ventilation. Nowadays, however, the question of whether in fact there is a difference in the distribution of causative pathogens is under debate. Data from 308 ICUs from the German National Nosocomial Infection Surveillance System, including information on relevant pathogens isolated in 11,285 cases of nosocomial pneumonia from 1997 to 2004, were used for our evaluation. Each individual pneumonia case was allocated either to early- or to late-onset pneumonia, with three differentiation criteria: onset on the 4th day, the 5th day, or the 7th day in the ICU. The frequency of pathogens was evaluated according to these categories. A total of 5,066 additional cases of pneumonia were reported from 2005 to 2006, after the CDC criteria had been modified. From 1997 to 2004, the most frequent microorganisms were Staphylococcus aureus (2,718 cases, including 720 with methicillin [meticillin]-resistant S. aureus), followed by Pseudomonas aeruginosa (1,837 cases), Klebsiella pneumoniae (1,305 cases), Escherichia coli (1,137 cases), Enterobacter spp. (937 cases), streptococci (671 cases), Haemophilus influenzae (509 cases), Acinetobacter spp. (493 cases), and Stenotrophomonas maltophilia (308 cases). The order of the four most frequent pathogens (accounting for 53.7% of all pathogens) was the same in both groups and was independent of the cutoff categories applied: S. aureus was first, followed by P. aeruginosa, K. pneumoniae, and E. coli. Thus, the predictabilities of the occurrence of pathogens were similar for the earlier (1997-to-2004) and later (2005-to-2006) time frames. This classification is no longer helpful for empirical antibiotic therapy, since the pathogens are the same for both groups.

For a long time it was common to distinguish between an early onset (the first 4 days) and a late onset (after the 4th day) of ventilator-associated pneumonia (VAP) (18). Early-onset nosocomial pneumonia was believed to be due primarily to gram-negative bacteria, such as Haemophilus influenzae, and methicillin (meticillin)-sensitive Staphylococcus aureus (MSSA) and Streptococcus pneumoniae. For late-onset nosocomial pneumonia, the most commonly encountered causative pathogens reported were higher-level antibiotic-resistant gram-negative bacteria, such as Pseudomonas aeruginosa, Acinetobacter spp., or methicillin-resistant S. aureus (MRSA). This classification leads to different strategies for empirical antimicrobial treatment: monotherapy with narrow-spectrum antibiotics for the treatment of early-onset pneumonia but broad-spectrum therapy for Pseudomonas spp. or MRSA with late-onset infection (2).

Meanwhile, guidelines and articles that discuss this classification are using different criteria for distinguishing between early- and late-onset nosocomial pneumonia. Some authors regard the 4th day in the intensive-care unit (ICU) as the last day of an early onset (4, 6, 16, 20, 22), while others set the 5th (23) or even the 7th (11, 14, 24) day as the limit. Furthermore, it still remains uncertain from the relevant literature whether the given threshold refers to the number of days in the hospital or the number of days following intubation (23). The concept itself, though, still remains accepted in general (3, 23).

During our surveillance activities in several ICUs, however, we got the impression that this widely used classification may no longer be appropriate for determining the antimicrobial therapy required. We have therefore used data from the German National Nosocomial Infection Surveillance System, KISS (Krankenhaus-Infektions-Surveillance-System), to investigate this question, drawing on a huge database provided by German ICUs (10).

MATERIALS AND METHODS

The ICU component of KISS was established in 1997. Since then, the number of participating ICUs has increased continuously (10). The method used by KISS is almost identical to the surveillance method of the National Nosocomial Infections Surveillance (NNIS) System in the United States (8). The definitions from the Centers for Disease Control and Prevention (CDC) (9) are used for the diagnosis of nosocomial infections in KISS as well. The KISS methodology for calculating device-associated infection rates, comparing infection rates, and recording complications also derives from the NNIS system. Whenever a nosocomial infection occurs, the pathogens identified in the specimen are reported to the surveillance system. As many as four pathogens can be recorded for each infection site. For S. aureus, it is also recorded whether the isolate is MSSA or MRSA.

For the current analysis, we calculated the percentages of nosocomial pneumonia cases with MSSA, MRSA, P. aeruginosa, and other pathogens out of all the pneumonia cases identified on any given ICU day during the first 14 ICU days. In addition, we compared the pathogens identified according to three cutoffs (the 4th day, the 5th day, and the 7th day) to distinguish between early- and late-onset nosocomial pneumonia. These days are numbered from the day of admission of the patient to the ICU. The first day on which clinical signs or symptoms of pneumonia were noticed—or, alternatively, the day on which (positive) samples were drawn for microbiological testing (whichever happened first)—is considered the day of onset. To compare the two groups for the 7-day limit, relative risks with 95% confidence intervals were calculated.

Since the CDC definitions of nosocomial pneumonia were recently modified (15), we also looked separately at the time frames 1997 to 2004 and 2005 to 2006 in order to evaluate whether this modification has influenced the distribution of pathogens in our surveillance system.

Because the majority of pneumonia cases were diagnosed on the basis of clinical criteria and endotracheal aspirate cultures, we performed a distinct analysis for those pneumonia cases diagnosed by quantitative culture of bronchoalveolar lavage (BAL) specimens, because such samples are considered to yield more-accurate data than endotracheal aspirates (6). Uniform threshold values for cultured specimens in the diagnosis of pneumonia were ≥10⁴ CFU per ml.

RESULTS

The 1997-to-2004 time frame.

Up to the end of 2004, we had an overview of 11,285 pneumonia cases from 308 ICUs monitored during 12,879 observation months (KISS). Most of the ICUs are interdisciplinary ICUs (48.1%), followed by surgical (21.4%), medical (19.2%), neurosurgical (3.6%), pediatric (3.2%), neurological (2.6%), and cardiosurgical (1.9%) ICUs. The proportion of pneumonia cases considered to be ventilator associated was 87.4%. The overall VAP rate was 7.5 cases per 1,000 ventilator days (German National Reference Centre for the Surveillance of Nosocomial Infections, 2006 [http://www.nrz-hygiene.de]). In 1,305 cases of pneumonia (11.6%), no detectable pathogens were recorded. For the remaining 9,980 cases, a total of 14,911 pathogens were recorded, for an average of 1.5 pathogens per pneumonia case.

In 1,609 cases (14.3% of all cases), the pathogens were identified by quantitative culture of BAL specimens. The nine most frequent pathogens, on the basis of all the material sent in for microbiological investigation or on the basis only of BAL fluid quantitative cultures, can be found in Table 1. The most frequent pathogen was S. aureus (2,718 cases, of which 720 were MRSA infections), followed by P. aeruginosa (1,837 cases), Klebsiella pneumoniae (1,305 cases), Escherichia coli (1,137 cases), Enterobacter spp. (937 cases), S. pneumoniae (671 cases), Haemophilus spp. (509 cases), Acinetobacter spp. (493 cases), and Stenotrophomonas maltophilia (308 cases). The order of frequency is almost identical for the two groups (based on all kinds of specimens or on BAL fluid cultures only).

TABLE 1.

The nine pathogens most frequently identified in samples from patients with nosocomial pneumonia^a

Pathogen	No. (%) of isolates identified on the basis of:
	All materials sent for microbiological investigation (n = 11,285)	Quantitative cultures of BAL specimens only (n = 1,609)
S. aureus	2,718 (24.1)	450 (28.0)
P. aeruginosa	1,837 (16.3)	318 (19.8)
K. pneumoniae	1,305 (11.6)	159 (9.9)
E. coli	1,137 (10.1)	187 (11.6)
Enterobacter spp.	937 (8.3)	148 (9.2)
S. pneumoniae	671 (6.0)	97 (6.0)
Haemophilus spp.	509 (4.5)	72 (4.5)
Acinetobacter spp.	493 (4.4)	77 (4.8)
S. maltophilia	308 (2.7)	64 (4.0)

^aFrom KISS, 1997 to 2004.

The distribution of pneumonia cases according to the categories of early- and late-onset pneumonia can be found in Table 2. This distribution is also almost identical for all VAP cases and for only those cases that were confirmed by culture of BAL specimens.

TABLE 2.

Numbers of pneumonia cases according to the different limits for classification^a

Classification of pneumonia	No. of cases (% of all cases)	No. of VAP cases (% of all VAP cases)	No. of cases identified by culture of BAL fluids (% of all BAL fluid culture-identified cases)
“Early onset”
1-4 days	2,235 (19.8)	1,799 (18.2)	315 (19.6)
1-5 days	3,271 (29.0)	2,705 (27.4)	463 (28.8)
1-7 days	5,124 (45.4)	4,325 (43.8)	747 (46.4)
“Late onset”
>4th day	9,050 (80.2)	8,080 (81.8)	1,294 (80.4)
>5th day	8,041 (71.0)	7,174 (72.6)	1,146 (71.2)
>7th day	6,161 (54.6)	5,554 (56.2)	862 (53.6)

^aFrom KISS, 1997 to 2004.

Table 3 shows the number of isolates per 100 pneumonia cases for the same pathogens. The order of frequency of the four most frequent microorganisms is also shown in Table 3. There are no major differences in the order of frequency of pathogens, regardless of the definitions of “early” and “late”.

TABLE 3.

Isolates per 100 pneumonia cases for the most frequent microorganisms, together with the order of frequency of the four most frequent pathogens^a

Pathogen	No. of isolates per 100 pneumonia cases (order of frequency)
	“Early-onset” pneumonia			“Late-onset” pneumonia
	1-4 days	1-5 days	1-7 days	>4th day	>5th day	>7th day
S. aureus	25.7 (1st)	26.8 (1st)	26.9 (1st)	23.7 (1st)	23.0 (1st)	21.0 (1st)
MSSA	21.4	22.9	22.7	16.8	13.8	14.5
MRSA	4.3	4.0	4.3	6.9	6.5	6.5
P. aeruginosa	11.6 (2nd)	11.6 (2nd)	11.9 (2nd)	17.4 (2nd)	16.1 (2nd)	19.9 (2nd)
K. pneumoniae	10.8 (3rd)	10.7 (4th)	11.1 (3rd)	11.8 (3rd)	10.6 (3rd)	12.6 (3rd)
E. coli	10.0 (4th)	10.8 (3rd)	10.6 (4th)	10.1 (4th)	8.7 (4th)	10.1 (4th)
S. pneumoniae	9.3	8.9	8.3	5.1	4.2	4.3
Enterobacter spp.	6.4	6.7	7.5	8.8	7.9	9.4
Haemophilus spp.	6.9	6.9	6.7	3.9	3.1	2.9
Acinetobacter spp.	2.6	2.6	3.2	4.8	4.5	5.5
S. maltophilia	1.3	1.3	1.4	3.1	3.0	3.8

^aFrom KISS, 1997 to 2004.

The 7-day cutoff seems to be the most appropriate for evaluating differences between the two onset groups, because it yields groups with similar sizes and also corresponds to the median time of occurrence of pneumonia in our database (7 days). We therefore used this cutoff to calculate the relative risks of early- versus late-onset pneumonia caused by particular pathogens, by comparing the numbers of isolates per 100 pneumonia cases for the two groups (Table 4). After considering only those results with significant confidence intervals, we confirmed MSSA, S. pneumoniae, and Haemophilus spp. as early-onset pathogens and P. aeruginosa, MRSA, Enterobacter spp., Acinetobacter spp., and S. maltophilia as late-onset pathogens.

TABLE 4.

Relative risks of early- and late-onset pneumonia with particular pathogens^a

Pathogen	No. of isolates per 100 pneumonia cases		Relative risk (95% CI)b	Type of pathogen according to the time of onsetc
	“Early onset” (1-7 days)	“Late onset” (>7 days)
S. aureus	26.9	21.7	1.24 (1.16-1.32)	Early
MSSA	22.7	13.6	1.67 (1.54-1.81)	Early
MRSA	4.2	8.2	0.52 (0.44-0.60)	Late
P. aeruginosa	11.9	19.9	0.60 (0.55-0.65)	Late
K. pneumoniae	11.1	12.0	0.93 (0.84-1.03)
E. coli	10.6	9.6	1.11 (0.99-1.24)
S. pneumoniae	8.3	4.0	2.06 (1.77-2.40)	Early
Enterobacter spp.	7.5	9.0	0.83 (0.74-0.95)	Late
Haemophilus spp.	6.7	2.7	2.48 (2.07-2.98)	Early
Acinetobacter spp.	3.2	5.4	0.59 (0.49-0.71)	Late
S. maltophilia	1.4	3.8	0.37 (0.28-0.48)	Late

^aCalculated by comparing the numbers of isolates per 100 cases in early- and late-onset pneumonia, using the 7-day cutoff for distinguishing between the two groups. Data are from KISS, 1997 to 2004.

^b95% CI, 95% confidence interval.

^cRecorded for significant results only.

The 2005-to-2006 time frame.

When the CDC definitions of nosocomial pneumonia were changed, KISS adapted the diagnosis criteria accordingly. That is why, as mentioned in Materials and Methods, we evaluated pathogens from nosocomial pneumonia cases separately for this time period.

An additional 5,068 cases of nosocomial pneumonia were reported during these 2 years. For 5,066 of those, data on the time between ICU admission and the onset of infection were also available. A total of 5,969 pathogens were reported for the 5,068 pneumonia cases (1.47 pathogens per case on average). The order of the five most frequent pathogens was shown to be the same regardless of the length of the patient's ICU stay before infection. There were no differences from the earlier time frame (1997 to 2004) in the order of frequency of pathogens.

DISCUSSION

The administration of accurate and timely initial empirical antibiotic therapy has been shown to have a major impact on mortality from nosocomial pneumonia (1, 5, 7, 25). Because early-onset nosocomial pneumonia is most often reported as being due to antibiotic-sensitive pathogens, while late-onset nosocomial pneumonia is frequently caused by more-resistant pathogens, guidelines recommend monotherapy with narrow-spectrum antibiotics for early-onset infections and broad-spectrum therapy for late-onset infections (2, 6, 12).

To our knowledge, we present here the largest study of its kind ever published (>300 ICUs contributing; surveillance of >16,000 cases of VAP; >20,000 pathogens recovered). According to our data, the order of frequency of microorganisms is almost the same in the two groups; the differences are not distinctive enough for a decision about appropriate initial therapy, regardless of the modification of the CDC definitions. The cutoff day used for distinguishing between early- and late-onset pneumonia also has only a minor impact. Thus, the absolute duration of mechanical ventilation seems to be insufficient for choosing the kind of antimicrobial treatment. That is why the classification of nosocomial pneumonia as early- or late-onset pneumonia should, in our view, no longer be recommended. In particular, the high percentage of P. aeruginosa pneumonia cases in the early-onset group, together with the increasing percentage of MRSA pneumonia, could lead to undertreatment of patients if one follows this principle of classification too rigidly. The frequent identification of P. aeruginosa and MRSA in the early-onset pneumonia group may be due to hospitalization prior to ICU admission and previous exposure to antibiotics. Micek et al. showed that an earlier hospital stay within 12 months increases the risk for infection by gram negative pathogens (19). Kothe et al. identified age and certain comorbidities as additional risk factors for colonization by gram-negative bacteria (17). The latest guidelines of the American Thoracic Society have already been modified and note that patients with early-onset pneumonia who have received antibiotics or have been hospitalized within the past 90 days have a greater risk of being colonized and infected with multiresistant pathogens and should be treated in the same way as patients with late-onset pneumonia (3).

Other colleagues doubt the usefulness of this classification. Mosconi et al., comparing patients with early- and late-onset pneumonia, discovered similar risks of death in the two groups (21). This was also concluded by Heyland et al. for the 7-day limit (14) and by Ibrahim et al. for the 4-day limit (16). Giantsou et al. investigated this question, focusing on pneumonia cases diagnosed by BAL, and came to the same conclusion (11). Hedrick et al. did find some differences in the distribution of organisms in early-onset (predominantly gram-positive bacteria) and late-onset (predominantly gram-negative bacteria) VAP among trauma surgical patients, but they were unable to detect any significant differences among nontrauma surgical patients, and no differences in mortality between early- and late onset pneumonia were found either in trauma patients or in nontrauma patients (13). Verhamme et al. reported that pathogens potentially resistant to multiple drugs were isolated in more than half (52%) of cases of early-onset ICU-acquired pneumonia (26). However, all these authors used data from individual institutions, and their data may not be representative of all intensive-care patients.

Our analysis, too, has some potential limitations: We used the sensitive, but somewhat less specific, CDC criteria for diagnosing nosocomial infections (9), which can lead to classifying patients as pneumonia patients when there is actually no pneumonia. These criteria are based mainly on clinical signs and symptoms as well as on microbiological findings. Since 2005, radiological abnormalities and/or changes have always been required in order to diagnose any case of pneumonia. However, there is not always a proven link between the results from the microbiological laboratory and the infection. For example, S. maltophilia is detected quite frequently in specimens from the respiratory tract but has rather low pathogenicity and thus may not be the actual cause of infection. In interpreting the data, the diagnostic quality of the individual ICUs should also be taken into account. Depending on the type of ICU, the frequency of disease and the level of diagnostic effort may differ. Only a subgroup of ICUs routinely perform BAL when nosocomial pneumonia is suspected, while others use clinical criteria for diagnosing pneumonia in combination with the identification of pathogens from tracheal secretions. Only in 14.3% of the pneumonia cases in our database were the pathogens identified from bronchoalveolar secretions or blood. Some uncertainty remains, therefore, as to whether some patients included in our study did not in fact have infections but were only colonized by the pathogens. However, most probably this is true for early- as well as late-onset pneumonia. In addition, the order of frequency did not change substantially when only the cases diagnosed by quantitative culture of BAL specimens were considered.

Furthermore, as mentioned earlier, as many as four pathogens can be recorded and referred to the surveillance system for each pneumonia case. Thus, it is almost impossible to identify a single causative organism for a specific nosocomial infection. Instead, all microorganisms were considered equally for this analysis.

Moreover, usually many different microbiology laboratories are involved in national surveillance systems. Almost all the laboratories in our surveillance system apply standard procedures, although the diagnostic quality may still differ. However, differences in laboratory practice most probably influenced both categories in the same way.

Despite the limitations mentioned above, we conclude that our data do confirm the classification of the most important pathogens that cause pneumonia. However, in our findings, the distribution of pathogens remains unchanged regardless of the time of onset of symptoms. Thus, in our view, the concept of early- and late-onset pneumonia pathogens is no longer helpful for the management of empirical antibiotic therapy. Empirical antimicrobial therapy should be identical for both groups.

Since inadequate initial antimicrobial treatment increases the risk of a fatal outcome of VAP, we recommend starting therapy with broad-spectrum antibiotics at high dosages. However, as soon as a microbiological result is available that narrows the spectrum of potential causative agents (which usually happens after approximately 3 days), the antimicrobial regimen should be modified accordingly. Other diagnostic tools, such as rapid Gram staining, urinary Legionella antigen testing, or a multiplex PCR test for viral respiratory agents, could also contribute to more-rapid diagnosis of the causative agent and thus restrict the use of antibiotics.

In addition, more randomized, controlled trials are necessary to address the issue of risk factors for the development of VAP and the impact of multidrug-resistant pathogens on morbidity and mortality due to this disease.