ORIGINAL ARTICLE /j x

ORIGINAL ARTICLE 10.1111/j.1469-0691.2007.01915.x Factors impacting on length of stay and mortality of community-acquired pneumonia J. Garau 1, F. Baquero 2,E.Pérez-Trallero 3, J.-L. Pérez 4, A. M. Martín-Sánchez 5, C. García-Rey 6, J. E. Martín-Herrero 6 and R. Dal-Ré 6, on behalf of the NACER Group 1 Department of Internal Medicine, Hospital Mútua de Terrassa, Terrassa, 2 Department of Clinical Microbiology, Hospital Ramón y Cajal, Madrid, 3 Department of Clinical Microbiology, Hospital Donostia, San Sebastián, 4 Department of Clinical Microbiology, Hospital Son Dureta, Palma de Mallorca, 5 Department of Clinical Microbiology, Hospital Insular, Gran Canaria and 6 Medical Department, GlaxoSmithKline S.A., Tres Cantos, Spain ABSTRACT A 1-year retrospective multicentre study was performed to identify factors influencing hospital length of stay (LOS) and mortality of patients (n = 3233) admitted to hospital because of community-acquired pneumonia (CAP). Pneumonia severity index (PSI) high-risk classes (IV and V), positive blood culture, admission to an intensive care unit (ICU), multi-lobar involvement and alcohol consumption were associated independently with prolonged LOS. Tobacco smoking was associated with a reduced LOS. The LOS varied markedly among centres. Only PSI high-risk class, admission to ICU and multi-lobar involvement were associated with early, late and global mortality. Positive blood cultures, antimicrobial therapy according to treatment guidelines and the establishment of an aetiological diagnosis were linked to reduced late and global mortality. These data suggest that early mortality associated with CAP is highly dependent on the clinical status of the patient at presentation. Conversely, late mortality seems to be associated more closely with clinical management factors; hence, an aetiological diagnosis and compliance with appropriate therapeutic guidelines have a significant influence on outcome. Keywords Community-acquired pneumonia, guidelines, hospitalisation, microbiology, outcome, therapy Original Submission: 11 April 2007; Revised Submission: 17 August 2007; Accepted: 12 October 2007 Clin Microbiol Infect 2008; 14: 322 329 INTRODUCTION Community-acquired pneumonia (CAP) remains a significant cause of hospital admission and death. Mortality rates have stabilised at c. 12% since 1950, despite advances in knowledge of this illness, improvements in the management of severe infectious diseases, the availability of intensive care, and the use of potent antimicrobial agents and effective vaccines [1,2]. Although there are a large number of international guidelines and recommendations concerning the management of CAP, there are also wide variations among centres and clinicians concerning reasons for admission, microbiological testing and therapy, Corresponding author and reprint requests: R. Dal-Ré, Glaxo- SmithKline S.A., Medical Department, c Severo Ochoa 2, Tres Cantos, 28760 Madrid, Spain E-mail: rafael.dal-re@gsk.com which suggests that physicians do not follow these recommendations uniformly. This is important, since some studies have shown that compliance with guidelines and or treatment recommendations is associated with lower mortality in patients with CAP [3 5]. Besides treatment, other factors related to patient characteristics and or clinical practice have been associated with a poor prognosis and mortality. With some of these factors, several scoring systems have been developed to aid physicians in the assessment of the severity of illness, which should always be used in association with good clinical judgement [6,7]. Fine et al. [8] identified several risk-factors that were associated independently with mortality, including demographical factors, co-morbidities, findings from the initial physical examination, and laboratory and radiographical findings at the initial examination, from which a predictive outcome Journal Compilation Ó 2007 European Society of Clinical Microbiology and Infectious Diseases

Garau et al. Factors impacting on LOS and mortality in CAP 323 rule, based on a pneumonia severity index (PSI), was developed. This 20-variable rule categorised patients into five risk classes, according to their 30-day risk of death, ranging from 0.1% for class I to 27% for class V. In addition to the PSI, other scoring systems, e.g., the five-variable CURB65 and the CRB65 systems (without the requirement for laboratory investigations), have also been shown to be useful tools in predicting the severity of CAP [9]. Although many factors related to a patient s characteristics cannot be modified, some factors related to clinical practices can be addressed. Therefore, the identification of factors influencing mortality, especially those factors related to clinical management that could be modified, should remain a priority. The aim of the present study was to identify factors influencing hospital length of stay (LOS) and overall, early and late mortality in patients admitted to hospital because of CAP. MATERIALS AND METHODS Study subjects Patients were identified using the International Classification of Diseases, 9th revision (ICD-9), for discharge diagnoses consistent with CAP: a primary or secondary discharge diagnosis of pneumonia (codes 480.0 483.99 and 485 487.0). To ensure the accuracy of this CAP diagnosis, one or two physicians at each centre, who were specially trained for this study, reviewed the medical records of every potentially eligible patient to ascertain the presence of: (i) a new infiltrate on a chest radiograph, compatible with pneumonia and not attributable to another cause, and (ii) an acute onset of signs and symptoms that were suggestive of a lower respiratory tract infection. Patients aged <18 years, those discharged from any hospital in the 14-day period before the current admission or who were transferred from another hospital, those previously discharged directly from the emergency room or transferred to another hospital before the resolution of the episode, or those who had a diagnosis of tuberculosis, were excluded. Study design This was a retrospective, multicentre study of patients with CAP who required hospital admission during a 1-year period in ten Spanish university tertiary-care hospitals. All patients admitted to hospital because of CAP between 1 November 2001 and 31 October 2002 were included in the study, and their data were recorded on a standardised case report form. All forms were checked thoroughly by independent reviewers for the accuracy of the data. This audit-like methodology ensured that no data were missed. Clinical information was obtained from the medical charts. Microbiological data were obtained from medical charts and or laboratory records. All study data were identified by study subject number and were doubleentered into a computerised database. Accuracy of data entry was verified and discrepancies were corrected by reference to the primary record. As recommended by the International Guidelines for Ethical Review of Epidemiological Studies (http://www.cioms.ch/frame_guidelines_nov_2002), this study was approved by the Research Ethics Committee of the Hospital Mútua de Terrassa, Barcelona, Spain. Analysis LOS was defined as the time (days) spent in hospital. Excess LOS was defined as the number of days (upper limit of the corresponding 95% CI hazard ratio) over the mean LOS. Early mortality was defined as death from any cause within 2 days of hospital admission, and late mortality as death >2 days after admission. Overall mortality was defined as death from any cause during the hospitalisation period. Pneumonia severity upon hospital admission was estimated using the validated prediction rule, calculated according to the PSI scores, developed by the Pneumonia Outcomes Research Team, as described previously [8]. An aetiological diagnosis was considered positive (definitive or presumptive) in the following situations: isolation of a respiratory pathogen from a normally sterile specimen; isolation of a respiratory pathogen from a protected specimen brush or bronchoalveolar lavage fluid (bacterial growth of 10 3 or 10 4 CFU ml, respectively); isolation of Legionella pneumophila from sputum samples; a serological test with paired serum samples revealing a four-fold increase in antibody levels; urinary antigen tests positive for L. pneumophila or Streptococcus pneumoniae; and isolation of a predominant microorganism from a validated sputum, bronchial aspirate or tracheobronchial aspirate. Patients for whom no microbiological tests were performed, and patients with negative microbiogical results, were considered to have disease of an unknown aetiology. For the multivariate analysis, antibiotic regimens considered to be appropriate by the Infectious Diseases Society of America American Thoracic Society (IDSA ATS) included the use of a b-lactam in combination with a macrolide or a respiratory fluoroquinolone, or a respiratory fluoroquinolone alone at admission [2,10]. The use of either a b-lactam or a macrolide in monotherapy was considered to be inappropriate. Univariate associations for proportions were performed using a two-sided Pearson correlation coefficient. Univariate analysis of LOS was performed using the Kaplan Meier method. Independent prognostic factors for LOS were identified by a stepwise forward Cox regression model, in which all covariates of the univariate analyses were entered. For analysis of mortality, variables were analysed using a forward logistic regression model that included all the variables tested in univariate analyses. A two-sided p value <0.05 was considered to be significant in all analyses. The statistical analyses were performed using SPSS v.15.0 (SPSS Inc., Chicago, IL, USA). RESULTS In total, 3233 subjects with CAP who were admitted to the study hospitals met the inclusion criteria. The mean age was 66.6 ± 18.5 years (range 18 100 years) and 63.8% of the patients

324 Clinical Microbiology and Infection, Volume 14 Number 4, April 2008 were male. Table 1 shows patient outcomes, grouped according to PSI categories. With the exception of the establishment of an aetiological diagnosis, the variables in Table 1 were more frequent in patients in high-risk classes (IV and V). LOS was longer for patients in high-risk classes (IV and V) than it was for patients in lowrisk classes (I III). The proportion of patients belonging to each PSI class also varied widely among hospitals, as did the proportion of patients with an aetiological diagnosis, the LOS and the mortality rate (Table 2). Presumptive or definitive diagnoses were established for 23.6% (762 3233) of the patients. The pathogens identified most frequently were S. pneumoniae (n = 425, 13.1%), L. pneumophila (n = 68, 2.1%), Haemophilus influenzae (n = 46, 1.4%) and Mycoplasma pneumoniae (n = 38, 1.2%). Significant differences in the diagnostic techniques used were found between patients who died and those who survived in terms of the proportion of sputum samples cultured (40.0% vs. 18.6%), blood samples cultured (48.4% vs. 41.4%), Legionella urinary antigen tests (39.0% vs. 18.9%) and S. pneumoniae urinary antigen tests (31.6% vs. 14.3%) administered and other respiratory samples cultured (8.7% vs. 13.9%). Paired serological tests and culture of sputum samples were the diagnostic techniques yielding the highest rate of positive results (60.6% and 38.6%, respectively; Table 3). Overall, 22.5% (63 280) of the patients who died had received antibiotics before presentation to the emergency department. This percentage was similar to that for patients who survived (25.1%; 741 2953). No differences among the antibiotics administered were found between the two groups. The most frequent antibiotics administered before presentation to the emergency department were penicillins (279 patients), macrolides (199 patients), quinolones (182 patients) and cephalosporins (128 patients). Analyses of outcomes LOS was assessed for all patients except those who died in hospital. Mean LOS was 11.5 days (range 1 111 days), with a median of 9 days (inter-quartile range, 7 14 days). Table 4 shows the mean and median LOS analysed for each variable, Outcome measure Patients in PSI risk class (%) a I II III IV V Total Table 1. Outcomes grouped according to pneumonia severity index (PSI) risk class Patients 403 (12.5) 492 (15.2) 711 (22.0) 1145 (35.4) 482 (14.9) 3233 (100) Admission to the ICU 10 (2.5) 18 (3.7) 28 (3.9) 57 (5.0) 49 (10.2) 162 (5.0) Mechanical ventilation 6 (1.5) 6 (1.2) 20 (3.8) 40 (3.5) 28 (5.8) 100 (3.1) Aetiological diagnosis 118 (29.3) 123 (25.0) 174 (24.5) 252 (22.0) 95 (19.7) 762 (23.6) LOS, mean days 8.6 9.4 11.2 13.0 14.4 11.5 Early death b 0 1 (0.2) 2 (0.3) 15 (1.3) 36 (7.5) 54 (1.7) Late death c 2 (0.5) 10 (2.0) 18 (2.5) 98 (8.6) 98 (20.3) 226 (7.0) Total deaths 2 (0.5) 11 (2.2) 20 (3.8) 113 (9.9) 134 (27.8) 280 (8.7) LOS, length of stay (calculated excluding patients who died or were discharged home); ICU, intensive care unit. a Data are numbers of patients (%), unless otherwise indicated. b Died 48 h after admission to the hospital. c Died >48 h after admission to the hospital. Table 2. Characteristics and outcomes of study population grouped according to site of care Characteristics Sites a 1(n = 410) 2 (n = 712) 3 (n = 231) 4 (n = 235) 5 (n = 255) 6 (n = 376) 7 (n = 187) 8 (n = 336) 9 (n = 288) 10 (n = 203) Total (n = 3233) p value PSI risk class Low-risk b 164 (40) 392 (55) 90 (39) 125 (53) 122 (48) 190 (51) 108 (58) 214 (64) 112 (39) 89 (44) 1606 High-risk c 246 (60) 320 (45) 141 (61) 110 (47) 133 (52) 186 (49) 79 (42) 122 (36) 176 (61) 114 (56) 1627 <0.001 Aetiological 30 (7) 297 (42) 73(32) 48 (20) 43 (17) 61 (16) 44 (24) 93 (28) 43 (15) 30 (15) 762 <0.001 diagnosis LOS (days) 10.8 ± 6.5 11.0 ± 7.5 9.2 ± 5.3 11.1 ± 7.5 14.7 ± 8.1 13.3 ± 8.3 10.0 ± 5.8 7.8 ± 7.4 12.8 ± 7.6 17.3 ± 16.3 11.5 ± 8.4 <0.001 Deaths 45 (11.0) 59 (8.3) 10 (4.3) 19 (8.1) 34 (13.3) 35 (9.3) 2 (1.1) 7 (2.1) 29 (10.1) 40 (19.7) 280 (8.7) <0.001 PSI, pneumonia severity index risk class; LOS, length of stay (calculated excluding patients who died or were discharged home). a Data are presented as n (%) or mean ± SD unless otherwise indicated. b PSI classes I III. c PSI classes IV and V.

Garau et al. Factors impacting on LOS and mortality in CAP 325 Table 3. Diagnostic techniques used, showing the association between mortality and the number of positive results (n = 3233) Mortality (%) No (n = 2953) Yes (n = 280) p value No. of patients with procedures performed no. with positive results Gram s stain of sputum sample 987 (33.4) 50 (17.9) <0.001 1037 429 Culture of sputum sample 1179 (40.0) 52 (18.6) <0.001 1231 475 Culture of blood sample pre-antibiotic treatment 1429 (48.4) 116 (41.4) 0.02 1545 174 Paired serological test a 203 203 123 Legionella urinary antigen detection 1152 (39.0) 53 (18.9) <0.001 1205 61 Streptococcus pneumoniae urinary antigen detection 933 (31.6) 40 (14.3) <0.001 973 213 Culture of other respiratory samples b 258 (8.7) 39 (13.9) 0.003 297 85 a Most patients who died did not survive long enough for a second sample to be obtained. b One hundred and sixty-eight positive results in 123 patients from bronchoalveolar lavage, bronchial aspiration, protected-specimen brush sampling, trans-thoracic needle aspiration, and pleural fluid samples. Table 4. Variables associated with length of stay in patients who were discharged from hospital a Variable LOS (days) Mean Median Univariate p value Multivariate hazard ratio (95% CI) Multivariate p value PSI risk class Low-risk (classes I, II, III) 9.9 8 High-risk (classes IV, V) 13.3 11 <0.01 1.51 (1.40 1.63) <0.01 Blood cultures Negative 11.2 9 Positive 13.3 11 <0.01 1.22 (1.00 1.49) 0.056 ICU admittance No 11.2 9 Yes 21.1 18 <0.01 2.25 (1.82 2.78) <0.01 Empirical antibiotic No IDSA ATS first choice 11.6 9 IDSA ATS first choice 11.2 9 0.34 NS 0.09 X-ray multi-lobar No 11.2 9 Yes 13.4 11 <0.01 1.24 (1.11 1.39) <0.01 Aetiological diagnosis No 11.4 9 Yes 11.8 9 0.207 NS 0.40 Active tobacco user No 11.8 10 Yes 10.6 8 0.001 0.85 (0.77 0.94) <0.01 Regular consumption of alcohol No 11.4 9 Yes 13.1 11 0.008 1.34 (1.15 1.56) <0.01 Hospital centre <0.001 NA <0.01 PSI, pneumonia severity index; LOS, length of stay; IDSA ATS, Infectious Diseases Society of America American Thoracic Society; NS, not significant; NA, not applicable; ICU, intensive care unit. a Total cases = 3233. Total cases dropped in the Cox regression model = 492 (including 280 patients who died and 34 patients who were discharged home). Cases available for analysis = 2741. together with the univariate Kaplan Meier p value and the results of the multivariate Cox regression (hazard risk, 95% CI and p value). A PSI high-risk classification (classes IV V), admission to an intensive care unit (ICU), multi-lobar pneumonia and regular consumption of alcohol were associated independently with a more prolonged LOS in the multivariate Cox regression analysis. Excess LOS was 7.0 days for patients in a high-risk class, 19.9 days if admitted to an ICU, 4.4 days if there was multi-lobar involvement, and 6.3 if the patient consumed alcohol regularly. In contrast, being an active smoker of tobacco was associated with a reduced LOS. Overall, 280 patients died (8.7% mortality) during hospitalisation. By day 30, there were 269 deaths, i.e., 96.0% of the total mortality of the series (Fig. 1). Several variables showed an association with mortality in the univariate analysis (Table 5). For early mortality, the multivariate analysis indicated that only high-risk PSI classes (IV and V), admission to an ICU and multi-lobar involvement were associated directly with excess mortality. For late mortality, high-risk PSI classes, positive blood cultures, admission to an ICU and multi-lobar involvement were associated with higher mortality. More importantly, following IDSA ATS recommendations for empirical treatment of CAP and establishing the aetiology of the CAP episode were both associated with reduced late and overall mortality. Active use of tobacco was also linked to reduced mortality.

326 Clinical Microbiology and Infection, Volume 14 Number 4, April 2008 No. deaths 40 35 30 25 20 15 10 5 0 0 10 20 30 40 50 60 70 80 90 100 110 Fig. 1. Mortality over time. Days In patients with an aetiological diagnosis, changes in antimicrobial therapy as a consequence of the microbiological report were significantly more frequent than in patients without an aetiological diagnosis. No differences were found for other factors resulting in a change of antibiotic therapy, e.g., failure of treatment or the occurrence of adverse events (Table 6). DISCUSSION The present study supports the hypothesis that establishing an aetiological diagnosis for CAP and using empirical antibiotic treatment according to clinical practice guidelines for CAP are associated with reduced late mortality in hospitalised patients with CAP, even after adjusting for potential confounders. An aetiological diagnosis is also a factor favourable for overall mortality. In the present study, the diagnostic procedures were not the same for patients who died and Table 5. Variables associated with early ( 2 days), late ( 3 days) and total mortality a % mortality Multivariate OR (95% CI) Multivariate p value Variable Early Late Total Early Late Total Early Late Total PSI risk class Low-risk (classes I, II, III) 0.2 1.9 2.1 13.0 (4.0 42.6) 6.1 (4.0 9.3) 7.3 (4.9 10.9) <0.01 <0.01 <0.01 High-risk (classes IV, V) 3.1 12.1 15.2 Blood cultures Negative 1.7 6.6 8.2 3.8 (1.9 7.6) 3.4 (1.7 6.5) NS <0.01 <0.01 Positive 1.9 15.5 17.4 ICU admittance No 1.4 5.9 7.3 4.6 (2.1 9.9) 6.7 (4.4 10.5) 7.7 (5.0 11.7) <0.01 <0.01 <0.01 Yes 6.8 28.4 35.2 Empirical antibiotic No IDSA ATS first choice 1.2 8.1 9.3 0.6 (0.5 0.8) 0.7 (0.5 0.9) NS <0.01 0.02 IDSA ATS first choice 1.7 5.8 7.4 X-ray multi-lobar No 1.2 6.0 7.3 2.8 (1.4 5.5) 2.0 (1.4 2.9) 2.3 (1.6 3.3) <0.01 <0.01 <0.01 Yes 4.1 12.4 16.5 Aetiological diagnosis No 1.8 7.2 8.9 0.5 (0.3 0.8) 0.5 (0.3 0.8) NS <0.01 <0.01 Yes 1.3 6.4 7.7 Active tobacco user No 1.9 7.9 9.9 0.5 (0.3 0.9) 0.5 (0.3 0.8) NS 0.01 <0.01 Yes 0.6 3.9 3.9 Regular consumption of alcohol No 1.7 7.0 8.6 NS NS NS Yes 1.9 7.0 8.9 Hospital centre NA NS 0.048 NS PSI, pneumonia severity index; IDSA ATS, Infectious Diseases Society of America American Thoracic Society; NS, not significant; NA, not applicable; ICU, intensive care unit. a No. of selected cases = 3233. No. of rejected cases in the multivariate analysis = 212 (196 cases that received different antibiotics from those considered in the study and 18 cases in which radiological extension of the chest X-ray was not recorded). No. of cases included in the analysis = 3021. Table 6. Reasons for a change in antibiotic therapy, showing the association with an aetiological diagnosis and antimicrobial therapy at admission (n = 3021) a Aetiological diagnosis Empirical antibiotic IDSA ATS first choice Cause of change in antibiotic therapy No (n = 2322) Yes (n = 699) p value No (n = 1293) Yes (n = 1728) p value Failure of treatment 176 (7.6) 62 (8.9) NS 103 (8.0) 135 (7.8) NS Microbiological report 41 (1.7) 213 (30.5) <0.001 106 (8.2) 148 (8.6) NS Adverse events 57 (2.5) 11 (1.6) NS 25 (1.9) 43 (2.5) NS IDSA ATS, Infectious Diseases Society of America American Thoracic Society; NS, not significant. a Number of patients included in the multivariate analysis.

Garau et al. Factors impacting on LOS and mortality in CAP 327 those who survived. In accordance with IDSA and ATS guidelines [2,10], two pre-treatment blood cultures, as well as a Gram s stain and culture of expectorated sputum (the ATS recommends a Gram s stain and culture of sputum only if a drug-resistant pathogen or an organism not covered by the usual empirical therapy is suspected), should be performed. Both guidelines recommend that Legionella urinary antigen should be determined, at least for patients with severe CAP. In the present study, only 48.6% of the patients had pre-treatment blood cultures taken. However, this proportion was the same for those who died and those who survived. Other microbiological studies, including Gram s stain and culture of expectorated sputum, pneumococcal urinary antigen determination and Legionella urinary antigen determination, were performed more frequently for survivors than for patients who died, despite the fact that, in the case of Legionella urinary antigen, guidelines recommend this procedure, especially for patients with severe CAP. A large inter-hospital variability in LOS has been demonstrated previously, and this has been reported to be associated with factors such as PSI risk class, complications during hospitalisation, admission to an ICU, need for oxygen administration, and transfer to a long-term care facility [11,12]. In the present study, LOS varied markedly among centres, with a minimum of 7.8 days in centre 8 and a maximum of 17.3 days in centre 10. In the multivariate Cox regression model, LOS increased significantly with a PSI high-risk class, the presence of positive blood cultures, admission to an ICU, multi-lobar X-ray involvement, and regular alcohol consumption. Likewise, the individual hospital was an independent factor associated with LOS. In contrast, and surprisingly, active use of tobacco was associated with a reduced LOS. It has been reported previously that alcohol abuse increases the incidence and severity of bacterial pneumonia by depressing the inflammatory response to infection [13,14], but not the rate of mortality [15]. There are very few studies that have assessed the influence of smoking tobacco on LOS, but a recent study by Marrie et al. [16] found that cigarette smoking was a factor predictive of ICU admission. There are several factors that influence mortality among hospitalised patients with CAP. Most of these can be defined as patient factors that cannot be modified. In the present study, PSI high-risk classification, bacteraemia, ICU admission and multi-lobar X-ray involvement were all associated with increased overall mortality, whereas the establishment of an aetiological diagnosis, antimicrobial therapy according to IDSA ATS guidelines, and active use of tobacco were associated with decreased overall mortality. When early mortality and late mortality were analysed separately, PSI high-risk classification, admission to an ICU and multi-lobar X-ray involvement were associated independently with both early and late mortality. In contrast, empirical antibiotic treatment according to IDSA ATS guidelines, an aetiological diagnosis and active use of tobacco were associated with reduced late mortality, whereas a positive blood culture was associated with increased late mortality. Several studies have revealed that the use of antimicrobial therapy in accordance with guidelines seems to reduce 30-day mortality among patients hospitalised with pneumonia, especially severe CAP [17 19]. In the present study, empirical antibiotic treatment at admission according to IDSA ATS guidelines was associated with reduced late mortality, although no association was found for early and overall mortality, probably because of the diluting effect of early mortality. This agrees with the results of Marrie et al. [5], who reported that treatment with levofloxacin alone or with cefuroxime plus azithromycin was associated with decreased late mortality, but not with early mortality. Similarly, it has been suggested [20] that antibiotic therapy is less likely to have an impact in the first few days. The present study was retrospective, so it cannot be stated conclusively that the use of empirical antimicrobial therapy according to IDSA ATS guidelines has a significant impact on survival. However, there is no reason to believe that adherence or non-adherence to guidelines is based on the severity of the illness or on some other factor that may influence mortality. A major finding in the present study was the association of an aetiological diagnosis as an independent negative factor with overall and late mortality, but not with early mortality. To our knowledge, this is the first evidence for the potential impact of an aetiological diagnosis on overall and late mortality. Other retrospective studies have shown that outcome in patients with severe illness is not improved by establishing a

328 Clinical Microbiology and Infection, Volume 14 Number 4, April 2008 specific aetiological diagnosis [21 24]. However, these studies were not designed specifically to test this hypothesis. Some prospective studies have concluded that identification of the pathogen does not affect clinical outcome [25 28], although it might allow better patient management [29]. In contrast, other studies have shown the benefit of microbiological diagnosis [30 32]. Hence, the finding of Gram-positive diplococci following a Gram s stain correlated with a more rapid response to single-agent antibiotic therapy [31]. The taking of blood cultures within 24 h of admission has been associated with a significant reduction in 30-day mortality [32]. Among cases without an aetiological diagnosis, 274 (11.8%) patients had changes made to their initial antibiotic therapy as a consequence of failure of the antibiotic treatment, the presence of adverse events or the receipt of a microbiological report, as compared with 286 (40.9%) patients with an aetiological diagnosis. When the causes for the change in antibiotic therapy were analysed separately, only the receipt of a microbiological report was statistically significant (1.7% vs. 30.5%; p <0.0001). Overall, 31.2% of patients with an aetiological diagnosis had changes made in their treatment as a result of a microbiological report, which appears to be a higher proportion of patients with antibiotic treatment adjustment when compared with those without an aetiological diagnosis. Unfortunately, the design of this study did not allow conclusions to be drawn regarding the protective effect on mortality of the establishment of an aetiological diagnosis, as this could simply be an indicator of a better process of care in these patients. Paradoxically, current use of tobacco had a protective effect on mortality. Menendez et al. [33] reported that the incidence of treatment failure in patients with CAP was significantly lower for individuals with chronic obstructive pulmonary disease, and considered that concomitant treatment with steroids in these patients might be an explanation, since some studies have reported a beneficial effect following initial steroid treatment [34]. A retrospective study of patients with bacteraemic pneumococcal pneumonia found that tobacco exposure was a strong independent riskfactor for invasive disease among immunocompetent, non-elderly adults. Further studies are needed to clarify the influence of tobacco use on mortality and LOS in CAP. The main strengths of the present study were the large sample size, the large number of variables collected from the clinical records, and the accuracy of the data following the efforts of independent reviewers, who monitored 100% of the data of all patients charts. However, the study has several limitations. First, as with any retrospective study, there is potential for residual confounding; the results should therefore be extrapolated with caution, since retrospective studies are not suitable for hypothesis testing, but only for exploratory analyses. Second, the quality of the data analysed is naturally related to the quality of the medical and nursing records. Nevertheless, the study enabled an audit-like revision to be performed in each participating hospital. The main finding of this study was that a number of factors were associated independently with prolonged hospital stay and mortality. More specifically, compliance with the IDSA ATS guidelines for empirical therapy had a significant impact on late and global mortality, as well as the establishment of an aetiological diagnosis. In conclusion, the present study suggests that early mortality in CAP is associated mostly with a poorer baseline clinical status of the patient, which also determines the failure to respond to antibiotic therapy. Conversely, late mortality seems to be related closely to microbiological variables, and therefore an aetiological diagnosis and compliance with appropriate therapeutic guidelines are more important in determining the outcome in such cases. ACKNOWLEDGEMENTS This study was supported by a grant from GlaxoSmithKline S.A., Tres Cantos, Spain. J.G. has acted as a paid consultant to GlaxoSmithKline, Bayer, MSD, Wyeth Lederle and Sanofi Aventis. He has received funding for research from Wyeth Lederle and GlaxoSmithKline (unrelated to this article). C.G-R., J.E.M-H. and R.D-R. were employees of GlaxoSmith- Kline at the time of this study. E.P-T. and J.-L.P. declare that they have no conflicts of interest in relation to this work. No information has been provided by F.B. concerning the presence or absence of potential conflicting or dual interests. Other members of the NACER Group are: C. Ferrer, C. García, E. Letang, R. Cantón, M. Molina, A. González, C. Lloret, M. Dolores, E. Ruiz, M. Segura, F. Cañas, J. Torres, F. J. Vasallo, J. Ruiz, E. Sierra, J. M. Nogueira, J. J. Camarena, A. Mora, A. Macía,, R. Landínez, M. Moreno, J. Fernández, M. de la Rosa, M. C. González, C. Fernández, L. Aguilar, M. J. Giménez, A. Pedromingo, J. Lahuerta, Y. Carpintero, M. Díaz and J. Dorado.

Garau et al. Factors impacting on LOS and mortality in CAP 329 REFERENCES 1. Gilbert K, Fine MJ. Assessing prognosis and predicting patient outcomes in community-acquired pneumonia. Semin Respir Infect 1994; 9: 140 152. 2. Niederman MS, Mandell LA, Anzueto A et al. Guidelines for the management of adults with community-acquired pneumonia. Diagnosis, assessment of severity, antimicrobial therapy, and prevention. Am J Respir Crit Care Med 2001; 163: 1730 1754. 3. Menendez R, Ferrando D, Valles JM, Vallterra J. Influence of deviation from guidelines on the outcome of community-acquired pneumonia. Chest 2002; 122: 612 617. 4. Mortensen EM, Restrepo M, Anzueto A, Pugh J. Effects of guideline-concordant antimicrobial therapy on mortality among patients with community-acquired pneumonia. Am J Med 2004; 117: 726 731. 5. Marrie TJ, Wu L. Factors influencing in-hospital mortality in community-acquired pneumonia: a prospective study of patients not initially admitted to the ICU. Chest 2005; 127: 1260 1270. 6. Feldman C. Prognostic scoring systems: which one is best? Curr Opin Infect Dis 2007; 20: 165 169. 7. Labarere J, Stone R, Obrosky DS et al. Comparison of outcomes for low-risk outpatients and inpatients with pneumonia. Chest 2007; 131: 480 488. 8. Fine MJ, Smith MA, Carson CA et al. Prognosis and outcomes of patients with community-acquired pneumonia. A meta-analysis. JAMA 1996; 275: 134 141. 9. Yan Man S, Lee N, Ip M et al. Prospective comparison of three predictive rules for assessing severity of communityacquired pneumonia in Hong Kong. Thorax 2007; 62: 348 353. 10. Mandell LA, Bartlett JG, Dowell SF, File TM, Musher DM, Whitney C. Update of practice guidelines for the management of community-acquired pneumonia in immunocompetent adults. Clin Infect Dis 2003; 37: 1405 1433. 11. Cabre M, Bolivar I, Pera G, Pallares R. Factors influencing length of hospital stay in community-acquired pneumonia: a study in 27 community hospitals. Epidemiol Infect 2004; 132: 821 829. 12. Menendez R, Cremades MJ, Martinez-Moragon E et al. Duration of length of stay in pneumonia: influence of clinical factors and hospital type. Eur Respir J 2003; 22: 643 648. 13. Adams HG, Jordan C. Infections in the alcoholic. Med Clin North Am 1984; 68: 179 200. 14. Goss CH, Rubenfeld GD, Park DR, Sherbin VL, Goodman MS, Root RK. Cost and incidence of social comorbidities in low-risk patients with community-acquired pneumonia admitted to a public hospital. Chest 2003; 124: 2148 2155. 15. de Roux A, Cavalcanti M, Angeles MM et al. Impact of alcohol abuse in the etiology and severity of communityacquired pneumonia. Chest 2006; 129: 1219 1225. 16. Marrie TJ, Shariatzadeh MR. Admission to an intensive care unit: a descriptive study. Medicine 2007; 86: 103 111. 17. Mortensen EM, Restrepo M, Anzueto A, Pugh J. Effects of guideline-concordant community-acquired pneumonia requiring antimicrobial therapy on mortality among patients with community-acquired pneumonia. Am J Med 2004; 117: 726 731. 18. Menendez R, Torres A, Zalacain R et al. Guidelines for the treatment of community-acquired pneumonia: predictors of adherence and outcome. Am J Respir Crit Care Med 2005; 172: 757 762. 19. Dean NC, Silver MP, Bateman KA, James B, Hadlock CJ, Hale D. Decreased mortality after implementation of a treatment guideline for community-acquired pneumonia. Am J Med 2001; 110: 451 457. 20. Austrian R, Gold J. Pneumococcal bacteremia with especial reference to bacteremic pneumococcal pneumonia. Ann Intern Med 1964; 60: 759 776. 21. Leroy O, Santre C, Beuscart C et al. A five-year study of severe community-acquired pneumonia with emphasis on prognosis in patients admitted to an intensive care unit. Intens Care Med 1995; 21: 24 31. 22. Trevisani L, Putinati S, Sartori S, Ballerin L, Potena A. The value of routine microbial investigation in communityacquired pneumonia. Respir Med 1992; 86: 174 175. 23. Levy M, Dromer F, Brion N, Leturdu F, Carbon C. Community-acquired pneumonia. Importance of initial noninvasive bacteriologic and radiographic investigations. Chest 1988; 93: 43 48. 24. Woodhead MA, Arrowsmith J, Chamberlain-Webber R, Wooding S, Williams I. The value of routine microbial investigation in community-acquired pneumonia. Respir Med 1991; 85: 313 317. 25. Ewig S, Bauer T, Hasper E, Marklein G, Kubini R, Luderitz B. Value of routine microbial investigation in communityacquired pneumonia treated in a tertiary care center. Respiration 1996; 63: 164 169. 26. Sanyal S, Smith PR, Saha AC, Gupta S, Berkowitz L, Homel P. Initial microbiologic studies did not affect outcome in adults hospitalized with community-acquired pneumonia. Am J Respir Crit Care Med 1999; 160: 346 348. 27. Theerthakarai R, El-Halees W, Ismail M, Solis RA, Khan MA. Nonvalue of the initial microbiological studies in the management of nonsevere community-acquired pneumonia. Chest 2001; 119: 181 184. 28. Waterer GW, Wunderink RG. The influence of the severity of community-acquired pneumonia on the usefulness of blood cultures. Respir Med 2001; 95: 78 82. 29. Pachon J, Prados MD, Capote F, Cuello JA, Garnacho J, Verano A. Severe community-acquired pneumonia. Etiology, prognosis, and treatment. Am Rev Respir Dis 1990; 142: 369 373. 30. Rello J, Bodi M, Mariscal D et al. Microbiological testing and outcome of patients with severe community-acquired pneumonia. Chest 2003; 123: 174 180. 31. Boerner DF, Zwadyk P. The value of the sputum gram s stain in community-acquired pneumonia. JAMA 1982; 247: 642 645. 32. Arbo MD, Snydman DR. Influence of blood culture results on antibiotic choice in the treatment of bacteremia. Arch Intern Med 1994; 154: 2641 2645. 33. Menendez R, Torres A, Zalacain R et al. Risk factors of treatment failure in community acquired pneumonia: implications for disease outcome. Thorax 2004; 59: 960 965. 34. Skerrett SJ, Park DR. Anti-inflammatory treatment of acute and chronic pneumonia. Semin Respir Infect 2001; 16: 76 84.