• Clinical failure;
  • community-acquired pneumonia;
  • management;
  • outcome;
  • prognosis;
  • risk-factors


  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

For patients with community-acquired pneumonia (CAP), clinical response during the first days of treatment is predictive of clinical outcome. As risk assessments can improve the efficiency of pneumonia management, a prospective cohort study to assess clinical, biochemical and microbiological predictors of early clinical failure was conducted in patients with severe CAP (pneumonia severity index score of >90 or according to the American Thoracic Society definition). Failure was assessed at day 3 and was defined as death, a need for mechanical ventilation, respiratory rate >25/min, PaO2 <55 mm Hg, oxygen saturation <90%, haemodynamic instability, temperature >38°C or confusion. Of 260 patients, 80 (31%) had early clinical failure, associated mainly with a respiratory rate >25/minute (n = 34), oxygen saturation <90% (n = 28) and confusion (n = 20). In multivariate logistic regression analysis, failure was associated independently with altered mental state (OR 3.19, 95% CI 1.75–5.80), arterial PaH <7.35 mm Hg (OR 4.29, 95% CI 1.53–12.05) and PaO2 <60 mm Hg (OR 1.75, 95% CI 0.97–3.15). A history of heart failure was associated inversely with clinical failure (OR 0.30, 95% CI 0.10–0.96). Patients who failed to respond had a higher 28-day mortality rate and a longer hospital stay. It was concluded that routine clinical and biochemical information can be used to predict early clinical failure in patients with severe CAP.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Community-acquired pneumonia (CAP) is a common disease that is associated with serious complications such as respiratory insufficiency, sepsis and death. CAP requires hospitalisation of 15–50% of patients, and 5–10% of patients require management in an intensive care unit (ICU) [1–7]. Despite advances in antimicrobial therapy, the mortality rate among hospitalised patients is still high, ranging from 2% to 30%[8]. Clinical response during the first 2–3 days of treatment appears to predict outcome [9–11]. Non-response is associated with increased morbidity and mortality, but once clinical stability has been achieved, clinical deterioration caused by pneumonia is rare [12]. Risk assessments could help physicians to improve the efficiency of pneumonia treatment, with optimal monitoring of high-risk patients preventing unnecessary complications, ICU admissions or deaths; patients at low risk for treatment failure may be switched from parenteral to oral antibiotics early in the treatment process, thereby potentially reducing the length of hospital stay and leading to more efficient use of available healthcare resources.

There is currently an absence of data to predict the response (or failure to respond) to therapy of patients treated for severe CAP. Recent studies have analysed risk-factors for early clinical failure in patient cohorts with mild-to-severe CAP, and found that independent factors associated with early clinical failure were multilobar pneumonia, pleural effusion, a pneumonia severity index (PSI) score >90, Legionella pneumonia, Gram-negative pneumonia, liver disease, leukopenia, dyspnoea and confusion [13–15]. Current criteria, such as the APACHE II and PSI scores, only predict mortality, and do not seem to be sufficiently accurate to guide clinical care. Moreover, these algorithms include >15 variables and are impractical to use routinely [16–20]. Therefore, the aim of the present study was to assess clinical, biochemical and microbiological predictors of early clinical failure in patients hospitalised with severe CAP.

Patients and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Patients and setting

This study analysed patients included in a multicentre, prospective randomised trial of the cost-effectiveness of an early switch from parenteral to oral therapy for severe CAP. The study was performed in two university medical centres and five teaching hospitals in The Netherlands between July 2000 and June 2003. The study was approved by the medical ethics committees of all participating hospitals, and all patients gave written informed consent to participate. Adult patients (aged ≥18 years) admitted with severe CAP were eligible for inclusion in the study. Pneumonia was defined as a new or progressive infiltrate on chest X-ray, plus at least two of the following criteria: cough; sputum production; rectal temperature >38°C or <36.1°C; auscultatory findings consistent with pneumonia; leukocytosis (> 10 000/mm3, or >15% bands); C-reactive protein more than three-fold the normal upper limit; and positive blood culture or positive culture of pleural fluid [21]. Severe pneumonia was defined as PSI >90 (Fine class IV or V) or as fulfilling the American Thoracic Society criteria for severe CAP [22]. Patients who had been treated for this episode of pneumonia with antibiotics in the outpatient setting were also included. Patients with cystic fibrosis, known colonisation with Gram-negative bacteria following structural damage to the respiratory tract, a life-expectancy of <1 month because of underlying disease, infections other than pneumonia requiring antibiotic treatment, or immunosuppression (neutropenia <0.5 × 109/L or a CD4 count <200/mm3), and patients admitted directly to an ICU, were excluded. Criteria for ICU admission included respiratory failure or haemodynamic instability not responding to adequate intravenous fluid therapy and requiring vasopressor support. All patients received intravenous treatment with antimicrobial agents for at least 3 days.

Microbial analysis

Sputum and blood samples were collected, cultured and evaluated according to standard procedures. Microorganisms cultured from blood or sputum were recorded. In addition, NOW tests (Binax Inc., Portland, ME, USA) were used to detect urinary antigen for Legionella pneumophila and Streptococcus pneumoniae. Acute and convalescent sera were collected and tested for Mycoplasma pneumoniae, L. pneumophila and Chlamydia pneumoniae. Any non-contaminating microorganism cultured from a blood or sputum sample, or detected by urinary antigen testing, was considered to be a cause for the episode of pneumonia. For M. pneumoniae, a four-fold or greater increase in titre between paired sera, or a single titre of ≥1:40, in immune fluorescence tests (Serodia-MycoII; Fujirebio Inc., Malvern, PA, USA), was considered to be indicative of infection [23]. For L. pneumophila, a four-fold increase in the antibody titre to ≥1:128, or single titres of ≥1:256, were considered to be suggestive of Legionella pneumonia [24]. For C. pneumoniae, detection of IgM above established values, seroconversion of IgG between acute and convalescence samples, high amounts of IgG in single titres, or a combination of these factors, were considered to be serological evidence of infection (ELISA; Savyon Diagnostics Ltd, Ashdod, Israel).


Early clinical failure was assessed after 3 days and was defined as death, a need for mechanical ventilation, respiratory rate >25/min, oxygen saturation <90%, Pa02 <55 mm Hg, haemodynamic instability, altered mental state, or fever (<1°C decline in temperature if >38.5°C on admission) [25]. A changed mental state was defined as an acute alteration in the mental state of the patient, as observed by family or the treating physician. If none of these features was present, patients were considered to be responders. Inappropriate therapy was defined as treatment with antibiotics to which the pathogens identified were not susceptible, e.g., β-lactam monotherapy for atypical pathogens such as C. pneumoniae, M. pneumoniae and L. pneumophila.

Clinical characteristics and course

A clinical history was taken upon admission and a physical examination was performed. Demographical criteria and the medical history were recorded. Neoplastic disease was defined as any cancer, except basal or squamous cell cancer of the skin, that was active at the time of presentation or had been diagnosed during the previous year [16]. PSI and APACHE II were scored, arterial blood gas analysis on room air was measured, routine chemical and haematological laboratory tests were performed, and a chest X-ray was taken [16,18]. Antibiotic treatment instituted by the treating physician was recorded. Antibiotic prescriptions were based on the Dutch guidelines for CAP management and were not dictated by the study protocol, except with regard to a switch to oral treatment on the third day following admission [26]. Response to treatment was evaluated by the principal investigator and dedicated research nurses on the third day of hospitalisation. Early failure was assessed using vital parameters, arterial blood gas analysis on room air, routine laboratory tests and physical examination. In-hospital clinical data, such as temperature, blood pressure, O2 saturation on room air, heart rate and respiratory rate, were measured daily. All patients were followed for a maximum of 28 days.

Statistical analysis

To identify individual risk-factors for early clinical failure, descriptive statistics were calculated as proportions and means (SD) to describe baseline characteristics in the two outcome groups (early response and early failure), using SPSS for Windows, v.11.5 (SPSS Inc., Chicago, IL, USA). The construction of the prognostic model began with univariate assessment of the association of each characteristic and early failure by estimation of OR and the corresponding 95% CIs for the total study population. In the next stage, multivariate logistic regression modelling was applied to select those variables that were associated with early failure, with p <0.10 as criterion for entry [27]. Forward selection was also performed to verify whether any previously deleted potentially relevant characteristic had been eliminated incorrectly from the model. For each patient, the individual probability of the outcome was calculated from the final model (predicted probability) [28].

Model evaluation

The reliability of the multivariate logistic regression model was determined by the Hosmer-Lemeshow goodness-of-fit statistic [27]. The area under the receiver operator curve (ROC) was used to assess the model's discriminative ability [28,29]. The area under the ROC represents the probability that the logistic regression model will assign a higher probability of the outcome to a randomly chosen patient with an outcome than to a randomly chosen patient without an outcome. An area under the curve (AUC) estimate of 0.5 indicates no discrimination, whereas an estimate of 1.0 indicates perfect discrimination. Models with AUC values between 0.70 and 0.79 are generally considered as having moderate discriminative properties, and those ≥0.80 as good.

Development and applicability of the prediction rule

The regression coefficients of the derived multivariate model were used to construct the prediction rule. The presence or absence of specific characteristics was coded as 1 (present) or 0 (absent). To simplify interpretation, the coefficients of the final model were first divided by the lowest coefficient and the regression coefficients were rounded to develop an overall score value in the final prognostic scores. The scores for individual prognostic variables were added to form the prognostic score for clinical failure.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Between July 2000 and June 2003, 273 consecutive patients with severe CAP were included in the study. Thirteen patients were subsequently excluded from analysis: six patients withdrew informed consent; six patients had other infections outside the respiratory tract that required antibiotic treatment; and data concerning early failure were incomplete for one patient. Of the 260 remaining patients, 180 (69%) were responders and 80 (31%) had early clinical failure. Reasons for failure are shown in Table 1. A single failure criterion at day 3 was present in 46 (57%) patients, while multiple failure criteria were present in 34 (43%) patients.

Table 1.   Reasons for early clinical failure
Reasons for failure (evaluated at third day of hospitalisation)Failure n = 80 (31%)
Death5 (6%)
Need for mechanical ventilation5 (6%)
Respiratory rate >25/min34 (42%)
Oxygen saturation <90%28 (35%)
Fever19 (24%)
Altered mental state20 (25%)
Haemodynamically unstable4 (5%)
Multiple features34 (43%)

In univariate analysis, total PSI score, APACHE II score, lower Glasgow Coma Score, presence of a changed mental state, arterial pH <7.35 and arterial PaO2 <60 mm Hg (variables of the PSI score) were significantly more predictive of early failure. A history of chronic heart failure was found more often among early responders (Table 2). Causative microorganisms were identified in 134 (52%) patients, in 90 (50%) responders and in 45 (56%) patients with early failure (Table 3). The most frequently isolated microorganism in both groups was S. pneumoniae. Gram-negative microorganisms were identified more often among inpatients with early failure (OR 2.21, 95% CI 1.04–4.69). The initial therapy instituted most frequently was β-lactam monotherapy (Table 4). Initial therapy with macrolide/β-lactam combinations was associated with early clinical failure in univariate analysis.

Table 2.   Characteristics of patients included in the study
CharacteristicResponder n = 180Early failure n = 80OR 95% CIp
Age67.9 (16.0%)69.15 (16.8%)1.011.00–1.020.57
Pneumonia severity index score108.3 (23.9%)123.9 (26.2%)1.031.01–1.04<0.001
APACHE score13.3 (4.4%)14.8 (5.1%)1.071.01–1.130.03
Glasgow coma score14.7 (1.1%)14.4 (1.1%)0.800.62–1.020.07
Female gender57 (32%)23 (29%)0.870.49–1.550.69
Nursing home5 (3%)4 (5%)1.800.48–7.050.37
Mental state change39 (22%)36 (45%)2.961.68–5.21<0.001
Pleural effusion X-ray30 (17%)17 (21%)1.030.96–1.100.38
Temperature <35°C or >40°C24 (13%)11 (14%)1.000.95–1.060.93
Systolic BP <90 mm Hg4 (2%)3 (4%)1.030.95–1.110.48
Heart rate >125/min39 (22%)21 (26%)1.030.97–1.090.41
Respiratory rate >30/min55 (31%)31 (39%)1.030.99–1.050.20
Arterial PaH <7.35 mm Hg7 (4%)12 (15%)4.361.65–11.550.003
Arterial Pa02 <60 mm Hg56 (31%)35 (44%)1.721.00–2.960.05
PaO266.9 (19.5%)69.3 (27.9%)1.010.99–1.020.43
pH7.45 (0.06%)7.43 (0.08%)0.030.00–0.180.006
Systolic blood pressure134.3 (24.5%)138.5 (30.8%)1.011.00–1.020.24
Temperature38.6 (1.2%)38.3 (1.2%)0.850.68–1.060.15
Respiratory rate26.3 (8.1%)28.1 (9.8%)1.020.99–1.060.16
Heart rate105.5 (22.0%)106.7 (24.5%)1.000.99–1.010.71
Medical history
 Chronic heart failure27 (15%)4 (5%)0.300.10–0.880.03
 Neoplasm44 (24%)18 (23%)1.000.98–1.020.73
 Cerebrovascular incident13 (7%)7 (9%)1.020.93–1.120.67
 Kidney disease17 (9%)6 (8%)0.980.89–1.080.61
Table 3.   Microorganisms associated with severe community-acquired pneumonia
AetiologyResponders n = 180Failures n = 80OR 95% CIp
  • a

    Moraxella catarrhalis, Citrobacter freundii, Proteus mirabilis, Acinetobacter spp.

  • b

    Staphylococcus aureus, Aspergillus spp.

Streptococcus pneumoniae49 (27%)20 (25%)0.890.49–1.630.71
Haemophilus influenzae11 (6%)3 (4%)0.600.16–2.210.44
Other Gram-negative bacteria17 (9%)15 (19%)2.211.04–4.690.04
Escherichia coli36   
Enterobacter spp.22   
Klebsiella spp.51   
Pseudomonas spp.02   
Chlamydia pneumoniae10 (6%)8 (10%)1.890.72–4.980.20
Legionella pneumophila7 (4%)3 (4%)0.960.24–3.820.96
Mycoplasma pneumoniae4 (2%)4 (5%)2.320.56–9.500.24
Multiple pathogens20 (11%)9 (11%)1.010.44–2.340.97
Other pathogensb12 (7%)5 (6%)0.880.32–2.440.81
Unknown aetiology90 (50%)35 (44%)0.850.49–1.500.56
Table 4.   Initial therapy for patients with community-acquired pneumonia
TherapyResponders n = 180Failures n = 80OR 95% CIp
  • a

    Fluoroquinolones (ciprofloxacin), tetracyclines (doxycycline), penicillins, co-trimoxazole.

  • b

    A microorganism was detected in 135 patients.

β-lactam monotherapy109 (61%)44 (55%)0.800.47–1.360.40
Macrolide monotherapy3 (1%)1 (1%)0.750.08–7.290.80
Cephalosporin monotherapy33 (18%)13 (16%)0.860.43–1.750.69
Other monotherapya14 (8%)3 (4%)0.460.13–1.660.24
β-lactam/macrolide combination12 (7%)12 (15%)2.471.06–5.770.04
Cephalosporin/macrolide combination6 (3%)5 (6%)1.930.57–6.530.29
Macrolide/other combination3 (2%)2 (3%)1.510.25–9.230.65
Inappropriate treatmentb18 (20%)10 (22%)1.430.60–3.400.42

In multivariate analysis, the simplest predictive model with the highest predictive value included altered mental state, arterial PaH <7.35, PaO2 <60 mm Hg and an absence of a history of heart failure as independent predictors of early clinical failure (area under the ROC curve 0.70, 95% CI 0.63–0.77; Table 5). As the results of microbiological investigations of sputum are not available upon admission, those results were not included in the analysis.

Table 5.   Results of multivariate analysis
Altered mental status3.191.75–5.80<0.001
Arterial PaH4.291.53–12.050.006
Heart failure0.300.10–0.960.04
Arterial PaO21.750.97–3.150.063

From the multivariate model, a prediction rule was derived in which a score was assigned to the presence of each variable. The predicted probability of outcome was determined as 1/(1 + e-LP), where the linear predictor (LP) = −1.41 + (1.16 altered mental status) + (1.46 × pH <7.35) + (−1.19 × heart failure) + (0.56 × arterial PaO2 <60 mm Hg). A prognostic score for each patient (minimum −2 to maximum 6 points), reflecting the probability of early failure, was calculated by adding the scores of relevant characteristics. Patients with a cut-off of <1 point had an 18% possibility of early clinical failure, whereas patients with a cut-off point of >3 points had a 75% possibility of failure (Table 6).

Table 6.   Cut-off points of prediction rule for early failure in patients hospitalised with community-acquired pneumonia
Points scoreaResponders (%)Failures (%)Total (patients)
  1. a Scores: altered mental status, 2 points; pH <7.35, 3 points; PaO2 <60, 1 point; heart failure, −2 points; maximum score 6 points.


Inappropriate therapy

Ten (22%) patients with early clinical failure and 18 (20%) patients who responded to treatment received inappropriate therapy, based on the results of microbiological culture (OR 1.43, 95% CI 0.60–3.40). All patients who received inappropriate therapy remained alive at day 28 (Table 7). Of ten patients who experienced early failure and received inappropriate therapy, four had Gram-negative bacteria (two Pseudomonas aeruginosa and two Enterobacter cloacae) isolated from sputum samples, in one case in combination with S. pneumoniae. The other six patients had serological evidence of infection with an atypical pathogen, including two cases with S. pneumoniae isolated from sputum samples (one in combination with pneumococcal bacteraemia). In all patients, empirical therapy was appropriate for S. pneumoniae as a probable cause of CAP.

Table 7.   Characteristics of patients receiving inappropriate therapy for community-acquired pneumonia (CAP)
PatientEarly clinical failureMicrobial cause of CAPBasis of diagnosisCo-infection (if any)Therapy at day 1Criteria of early clinical failure at day 3Mortality at day 28
  1. UAT, urinary antigen test.

 1YesPseudomonas aeruginosaSputum AmoxycillinFeverNo
 2YesChlamydia pneumoniaeSerology AmoxycillinFeverNo
 3YesC. pneumoniaeSerology AmoxycillinFeverNo
 4YesC. pneumoniae/Mycoplasma pneumoniaeSerologyEscherichia coli/Streptococcus pneumoniae (sputum)CeftriaxoneRespiratory rate > 25/minNo
 5YesEnterobacter cloacaeSputum AmoxycillinOxygen saturation < 90% + respiratory rate > 25/minNo
 6YesE. cloacaeSputum AmoxycillinRespiratory rate > 25/minNo
 7YesC. pneumoniaeSerologyS. pneumoniae (blood culture + sputum)AmoxycillinOxygen saturation < 90%No
 8YesP. aeruginosaSputumS. pneumoniae (sputum)AmoxycillinOxygen saturation < 90% + PaO2 < 55No
 9YesM. pneumoniaeSerology AmoxycillinFeverNo
10YesM. pneumoniaeSerology CeftriaxoneOxygen saturation < 90% + PaO2 < 55No
11NoLegionella pneumophila/C. pneumoniaeSerology Amoxycillin No
12NoC. pneumoniae/M. pneumoniaeSerology Amoxycillin No
13NoE. cloacaeSputumS. pneumoniae (sputum)Amoxycillin No
14NoC. pneumoniaeSerologyHaemophilus influenzae (sputum)Amoxycillin No
15NoC. pneumoniaeSerology Ceftriaxone No
16NoCitrobacter freundiiBlood culture Amoxycillin No
17NoL. pneumophilaSerologyCorynebacterium spp. (blood culture)Ceftriaxone No
18NoC. pneumoniaeSerology Amoxycillin No
19NoAspergillus spp.SputumS. pneumoniae (sputum)Ceftriaxone No
20NoC. pneumoniaeSerology Ceftriaxone No
21NoL. pneumophilaSerology Amoxycillin No
22NoL. pneumophilaSerology + UATS. pneumoniaePenicillin No
23NoM. pneumoniaeSerology Amoxycillin No
24NoL. pneumophilaSerologyS. pneumoniae (blood culture)Amoxycillin No
25NoP. aeruginosaSputumE. coli (sputum + blood culture)Cefuroxime No
26NoAcinetobacter spp.Sputum Amoxycillin No
27NoM. pneumoniaeSerology Cefuroxime No
28NoC. pneumoniaeSerologyS. pneumoniae (blood culture)Ceftriaxone No

Of 18 patients who responded and received inappropriate therapy, three had Gram-negative bacteria (P. aeruginosa, E. cloacae and Acinetobacter spp.) isolated from sputum samples, in one case in combination with S. pneumoniae. One patient had a blood culture positive for Citrobacter freundii, with no other cause for CAP identified, but responded clinically after receiving amoxycillin at day 3. Another patient had Aspergillus isolated from sputum samples (together with S. pneumoniae), but had no apparent risk-factors for yeast infection. One patient had a positive urinary antigen test for L. pneumophila, and had S. pneumoniae isolated from tracheal aspirates. This patient showed a rapid clinical response with penicillin therapy. The remaining 12 patients had serological evidence of infection with an atypical pathogen, but all showed a good clinical response with β-lactam therapy at day 3. None of the patients receiving inappropriate therapy had been switched to appropriate therapy at day 3.


During the follow-up period of 28 days (inpatients and outpatients), patients with early clinical failure had a significantly higher chance of all-cause death and had longer lengths of hospital stay. Nine (12%) patients with early clinical failure died, not including the five patients who died before day 3, whereas eight (4.4%) patients who responded early died (OR 2.93, 95% CI 1.09–7.92). The length of hospital stay of patients with early clinical failure was 13.4 ± 5.3 days, compared with 9.6 ± 4.7 days for early responders (mean difference 3.8 days, 95% CI 2.6–5.4 days).


  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Clinical and biochemical data that are assessed routinely for patients with CAP can be used to determine the likelihood of early clinical failure in patients hospitalised with severe CAP. Early clinical failure is associated with increased mortality and a longer stay in hospital. Close monitoring of patients at high risk for treatment failure may prevent unnecessary deaths, ICU admissions and associated costs.

The strengths of the present study were the prospectively collected data, strict criteria for inclusion, and the focus on patients with a high risk for early clinical failure. In addition, early failure was evaluated according to previously defined and generally accepted criteria. Until now, no studies have analysed prognostic factors for early clinical failure in patients with severe CAP. Arterial PaH <7.35 mm Hg, arterial Pa02 <60 mm Hg, altered mental state upon admission, and an absence of a history of chronic heart failure, were independent predictors of early clinical failure in patients treated for severe CAP. Isolation of Gram-negative pathogens was also a strong predictor for early clinical failure, but as culture results from sputum are not available at the time of admission, this variable was excluded from the final analysis.

Hypoxaemia, acidosis and confusion may all indicate tissue hypoperfusion caused by oxygen deficiency, and are associated with high mortality in patients with CAP [30]. As hypoxaemia and acidosis can be determined only with proper arterial blood-gas analysis, the importance of performing this diagnostic procedure at admission should be emphasised. In contrast to the Glasgow Coma Score, confusion, defined as an acute alteration in the mental state of patients, as observed by family or treating physicians, was an independent predictor for clinical failure. However, this definition does not use an objective scale, which limits the internal validity of the model and its applicability in other settings.

Remarkably, a history of chronic heart failure was related inversely to early clinical failure. Although it is possible that some patients with a history of heart failure were admitted because of congestive heart failure, which is sometimes difficult to differentiate from pneumonia on a chest X-ray, all patients had evidence of new or progressive infiltrates on chest X-ray, fulfilled the clinical criteria for pneumonia, and had a PSI score of >90, or fulfilled the ATS criteria for severe CAP. Although heart failure has been associated with an adverse clinical outcome in patients with CAP, it has also been associated with a high risk for viral pneumonia [31,32]. It is, therefore, possible that viral respiratory infections with a less severe course were more prevalent among patients with heart failure. Indeed, no pathogenic bacteria were isolated from 68% of patients with heart failure, compared with 45% of patients without heart failure (p 0.02).

In univariate analysis, treatment with a combination of β-lactams and macrolides was associated with early failure. In The Netherlands, most patients hospitalised with severe CAP on regular wards are treated with β-lactam monotherapy, although combination therapy is sometimes used for more severe cases of CAP, as judged by the treating physician. There was no statistically significant difference in the prevalence of inappropriate treatment prescribed for failing and responding patients. An atypical pathogen was detected in the majority of patients who received inappropriate treatment and who were not treated with a macrolide or fluoroquinolone. The most common pathogens isolated from these patients were C. pneumoniae and M. pneumoniae, both of which usually cause mild infections. Moreover, diagnosis of atypical infections by means of serological investigations is difficult, especially for C. pneumoniae, with high rates of false-positives [33]. The possible influence of penicillin resistance in S. pneumoniae on early clinical failure could not be investigated, as pneumococcal resistance to penicillins in The Netherlands is <1%.

The present study has a number of limitations. First, evaluation of patients included in a clinical trial risks selecting patients who differ from those encountered in daily clinical practice. However, analysis of a sample of 56 consecutive patients meeting the inclusion criteria who were unable or unwilling to provide informed consent showed a similar disease severity (average PSI score of 106.45) and age (average age 68.4 years) compared with the study population. Therefore, generalisation of the results to immunocompetent patients admitted to non-ICU wards because of severe CAP seems justified. Second, while explicit definitions of clinical response are absent, criteria for clinical stability usually include normalisation of heart rate, systolic blood pressure, respiratory rate, temperature, oxygenation and mental status [12,34]. Therefore, these criteria were used to define clinical failure, although the use of other criteria may lead to different outcomes. This is illustrated by the lower rates of failure in other studies in which different criteria for early failure were used, including lack of response or worsening of clinical and/or radiological status after 48–72 h, a requirement for changes in therapy, and/or performance of invasive procedures for diagnostic and therapeutic purposes [14,35]. Furthermore, the higher failure rate in the present study was probably also related to the strict selection of severe CAP. Third, the definition used for severity of disease can be questioned. Severe CAP was defined on the basis of higher PSI scores and ATS criteria. In the present cohort, only 19% of patients were in Fine class V, which probably explains the relatively low mortality rate in the study in comparison with the original derivation cohort. In addition, patients who were admitted directly to an ICU were not included in the analysis, which probably influenced mortality rates in the cohort when compared with Fine's original derivation cohort [16].

In patients who do not respond to initial therapy, the possibility of an incorrect diagnosis, an inadequate host-related response, and drug-related or pathogen-related problems, e.g., antibiotic-resistant pathogens, should be considered. However, the feature that is most important in early clinical failure has not yet been determined. In the present study, strict diagnostic criteria for CAP were used for inclusion, discordant therapy appeared not to be associated with early clinical failure, and all isolates of S. pneumoniae were susceptible to β-lactam antibiotics. Nevertheless, 31% of patients still fulfilled the criteria for early clinical failure. Therefore, an inadequate host response was probably the most important factor in early treatment failure in the study cohort. Whether genetic predisposition, as suggested recently [36–38], contributes to early failure remains to be determined.

In summary, clinical and biochemical data that are usually assessed routinely in patients with CAP can be used to indicate the possibility of early clinical failure in patients hospitalised with severe CAP. Close monitoring of patients at high risk for treatment failure may prevent unnecessary deaths, ICU admissions and associated costs. Early identification of patients at low risk for early failure may assist physicians in scheduling treatment strategies, e.g., whether or not to switch to oral antibiotics early in the treatment programme. The prognostic variables identified in the present study are easy to assess and might be helpful in reaching these treatment decisions. The PSI score predicted early clinical failure in multivariate analysis nearly as accurately as the combination of the four variables mentioned above. Therefore, these variables seem to be more practical for daily use. Nevertheless, the prediction rules derived from the data require validation in external cohorts of severe CAP to evaluate their general applicability.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References
  • 1
    Marrie TJ. Epidemiology of community-acquired pneumonia in the elderly. Semin Respir Inf 1990; 5: 260268.
  • 2
    Almirall J, Morato I, Riera F et al. Incidence of community-acquired pneumonia and Chlamydia pneumonia infection: a prospective multicenter study. Eur Respir J 1993; 6: 614618.
  • 3
    Lave JR, Fine MJ, Sankey SS et al. Hospitalized pneumonia. Outcomes, treatment patterns, and costs in urban and rural areas. J Gen Intern Med 1996; 11: 415421.
  • 4
    Woodhead MA, Macfarlane JT, McCracken JS et al. Prospective study of the aetiology and outcome of pneumonia in the community. Lancet 1987; i: 671674.
  • 5
    Guest JF, Morris A. Community-acquired pneumonia: the annual cost to the National Health Service in the United Kingdom. Eur Respir J 1997; 10: 15301534.
  • 6
    British Thoracic Society Research Committee and Public Health Laboratory Service. The aetiology, management and outcome of severe community-acquired pneumonia on the intensive care unit. Respir Med 1992; 86: 713.
  • 7
    Torres A, Serra-Batlles J, Ferrer A et al. Severe community-acquired pneumonia: epidemiology and prognostic factors. Am Rev Respir Dis 1991; 144: 312318.
  • 8
    Fine MJ, Smith MA, Carson CA et al. Prognosis and outcomes of patients with community-acquired pneumonia. A meta-analysis. JAMA 1996; 275: 134141.
  • 9
    Ramirez JA, Vargas S, Ritter GW et al. Early switch from intravenous to oral antibiotics and early hospital discharge: a prospective observational study of 200 consecutive patients with community-acquired pneumonia. Arch Int Med 1999; 159: 24492454.
  • 10
    Ramirez JA, Bordon J. Early switch from intravenous to oral antibiotics in hospitalized patients with bacteremic Streptococcus pneumoniae community-acquired pneumonia. Arch Int Med 2001; 161: 848850.
  • 11
    Ramirez JA. Switch therapy in adult patients with pneumonia. Clin Pulm Med 1995; 2: 327333.
  • 12
    Halm EA, Fine MJ, Marrie TJ et al. Time to clinical stability in patients hospitalized with community- acquired pneumonia: implications for practice guidelines. JAMA 1998; 279: 14521457.
  • 13
    Menéndez R, Torres A, Zalacain R et al. Risk factors of treatment failure in community-acquired pneumonia: implications for disease outcome. Thorax 2004; 59: 960965.
  • 14
    Rosón B, Carratala J, Fernandez-Sabe N et al. Causes and factors associated with early failure in hospitalized patients with community-acquired pneumonia. Arch Intern Med 2004; 164: 502508.
  • 15
    Menéndez R, Torres A, Rodríguez de Castro F et al. Reaching stability in community-acquired pneumonia: the effects of the severity of disease, treatment, and the characteristics of patients. Clin Infect Dis 2004; 39: 17831790.
  • 16
    Fine MJ, Auble TE, Yealy DM et al. A prediction rule to identify low-risk patients with community acquired pneumonia. N Engl J Med 1997; 336: 243250.
  • 17
    Lim WS, Van Der Eerden MM, Laing R et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax 2003; 58: 377382.
  • 18
    Knaus WA, Draper EA, Wagner DP et al. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13: 818829.
  • 19
    Angus DC, Marrie TJ, Obrosky DS et al. Severe community-acquired pneumonia: use of intensive care services and evaluation of American and British Thoracic Society diagnostic criteria. Am J Respir Crit Care Med 2002; 166: 717723.
  • 20
    Oosterheert JJ, Bonten MJ, Hak E et al. Severe community-acquired pneumonia: what's in a name? Curr Opin Infect Dis 2003; 16: 153159.
  • 21
    Chow AW, Hall CB, Klein JO et al. General guidelines for the evaluation of new anti-infective drugs for the treatment of respiratory tract infections. Clin Infect Dis 1992; 15 (suppl 1): s62s88.
  • 22
    Official Statement of the American Thoracic Society. Guidelines for the management of adults with community-acquired pneumonia. Am J Resp Crit Care Med 2001; 163: 17301754.
  • 23
    Weingarten SR, Riedinger MS, Varis G et al. Identification of low-risk hospitalized patients with pneumonia. Implications for early conversion to oral antimicrobial therapy. Chest 1994; 105: 11091115.
  • 24
    Woodhead MA, Macfarlane JT, American Thoracic Society. Comparative clinical and laboratory features of legionella with pneumococcal and mycoplasma pneumonias. Br J Dis Chest 1987; 81: 133139.
  • 25
    Stout JE, Yu VL. Legionellosis. N Engl J Med 1997; 337: 682687.
  • 26
    Kasteren ME, Wijnands WJ, Stobberingh EE et al. Optimization of the antibiotics policy in The Netherlands. II. SWAB guidelines for the antimicrobial therapy of pneumonia in patients at home and as nosocomial infections. The Netherlands Antibiotic Policy Foundation. Ned Tijdschr Geneeskd 1998; 142: 952956.
  • 27
    Hosmer DW, Lemeshow S. Applied logistic regression. New York: Wiley, 1998; 135175.
  • 28
    Hanley JA, McNeill BJ. The meaning and use of the area under the receiver operating characteristic curve (ROC). Radiology 1982; 143: 2936.
  • 29
    Hak E, Wei F, Nordin J et al. Development and validation of a clinical prediction rule for hospitalisation due to pneumonia or influenza or death during influenza epidemics among community-dwelling elderly persons. J Infect Dis 2004; 189: 450458.
  • 30
    Mortensen EM, Coley CM, Singer DE et al. Causes of death for patients with community-acquired pneumonia: results from the Pneumonia Patient Outcomes Research Team cohort study. Arch Intern Med 2002; 162: 10591064.
  • 31
    Ruiz M, Ewig S, Torres A et al. Severe community-acquired pneumonia. Risk factors and follow-up epidemiology. Am J Respir Crit Care Med 1999; 160: 923929.
  • 32
    De Roux A, Marcos MA, Garcia E et al. Viral community-acquired pneumonia in nonimmunocompromised adults. Chest 2004; 125: 13431351.
  • 33
    Tuuminen T, Palomaki P, Paavonen J. The use of serologic tests for the diagnosis of chlamydial infections. J Microbiol Meth 2000; 45: 265279.
  • 34
    Rhew DC, Tu GS, Ofman J et al. Early switch and early discharge strategies in patients with community- acquired pneumonia: a meta-analysis. Arch Intern Med 2001; 161: 722727.
  • 35
    Ortqvist A, Kalin M, Lejdeborn L et al. Diagnostic fiberoptic bronchoscopy and protected brush culture in patients with community-acquired pneumonia. Chest 1990; 97: 576582.
  • 36
    Roy S, Knox K, Segal S et al. MBL genotype and risk of invasive pneumococcal disease: a case-control study. Lancet 2002; 359: 15691573.
  • 37
    Waterer GW, Quasney MW, Cantor RM et al. Septic shock and respiratory failure in community-acquired pneumonia have different TNF polymorphism associations. Am J Respir Care Med 2001; 163: 15991604.
  • 38
    Waterer GWEI, Bahlawan L, Quasney MW et al. Heat shock protein 70–2+1267 AA homozygotes have an increased risk of septic shock in adults with community-acquired pneumonia. Crit Care Med 2003; 31: 13671372.