Antibiotic Therapy and 48-Hour Mortality for Patients with Pneumonia

The American Journal of Medicine (2006) 119, 859-864 CLINICAL RESEARCH STUDY AJM Theme Issue: Pulmonology/Allergy Antibiotic Therapy and 48-Hour Mortality for Patients with Pneumonia Eric M. Mortensen, MD, MSc, a,b Marcos I. Restrepo, MD, MSc, a,c Antonio Anzueto, MD, c Jacqueline A. Pugh, MD a,b a Veterans Evidence Based Research Dissemination and Implementation Center, and Divisions of b General Internal Medicine and c Pulmonary/Critical Care Medicine, South Texas Veterans Health Care System, San Antonio, Tex. ABSTRACT PURPOSE: Although numerous articles have demonstrated that recommended empiric antimicrobial regimens are associated with decreased mortality at 30 days, there is controversy over whether appropriate antibiotic selection has a beneficial impact on mortality within the first 48 to 96 hours after admission. Our aim was to determine whether the use of guideline-concordant antibiotic therapy is associated with decreased mortality within the first 48 hours after admission for patients with pneumonia. METHODS: A retrospective cohort study was conducted at two tertiary teaching hospitals in San Antonio, Texas. A propensity score was used to balance the covariates associated with the use of guideline-concordant antimicrobial therapy. A multivariable logistic regression model was used to assess the association between mortality within 48 hours and the use of guideline-concordant antibiotic therapy, after adjusting for potential confounders including the propensity score. RESULTS: Information was obtained on 787 patients with community-acquired pneumonia. The median age was 60 years, 79% were male, and 20% were initially admitted to the intensive care unit. At presentation 52% of subjects were low risk, 34% were moderate risk, and 14% were high risk. Within the first 48 hours, 20 patients died. After adjustment for potential confounders, the use of guideline-concordant antimicrobial therapy (odds ratio 0.37, 95% confidence interval, 0.14-0.95) was significantly associated with decreased mortality at 48 hours after admission. CONCLUSION: Using initial empiric guideline-concordant antimicrobial therapy is associated with decreased mortality at 48 hours. Further research needs to investigate methods to ensure that patients with community-acquired pneumonia are treated with appropriate antimicrobial therapies. 2006 Elsevier Inc. All rights reserved. KEYWORDS: Antibacterial agents; Community-acquired infections; Pneumonia Dr. Mortensen was supported by a Department of Veteran Affairs Vertically Integrated Service Network 17 new faculty grant and a Howard Hughes Medical Institute faculty start-up grant 00378-001. Dr. Pugh was supported by Department of Veteran Affairs grants REA 05-129 and RCD 04-297. This material is the result of work supported with resources and the use of facilities at the South Texas Veterans Health Care System. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs. Manuscript submitted on September 27, 2005, and accepted in revised form on April 5, 2006. Requests for reprints should be addressed to Eric Mortensen, MD, MSc, VERDICT, ALMD/UTHSCSA, Ambulatory Care (11C6), 7400 Merton Minter Boulevard, San Antonio, TX 78284. E-mail address: mortensene@uthscsa.edu. The data suggest that antimicrobial therapy has little or no effect upon the outcome of infection among those destined at the onset of illness, to die within 5 days. Robert Austrian and Jerome Gold, 1964 1 Community-acquired pneumonia is the seventh leading cause of death in the United States and is the leading infectious cause of death. 22 Although mortality decreased precipitously with the advent of antimicrobial therapy, since 1950 mortality has been stable or gradually increased. 3 Because of this substantial mortality, numerous professional societies including the American Thoracic Society (ATS), Infectious Diseases Society of America (IDSA), British Thoracic Society, and others have published clinical practice guidelines for community-acquired pneumonia. 4-10 0002-9343/$ -see front matter 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2006.04.017

860 The American Journal of Medicine, Vol 119, No 10, October 2006 Although there is considerable evidence that the use of empiric guideline-concordant antimicrobials is associated with improved survival at 30 days, 11-15 controversy still exists about whether the use of antibiotics has an impact on survival in the first 48 to 96 hours after admission. 1,16 In addition, randomized clinical trials of antibiotics for communityacquired pneumonia frequently exclude patients who are expected to die within the first 48 to 72 hours because of the concern that these patients will die no matter what antimicrobial therapy they receive. 17 Our aim was to determine whether the use of guideline-concordant antibiotic therapy is associated with decreased mortality within the first 48 hours after admission for patients with community-acquired pneumonia. CLINICAL SIGNIFICANCE METHODS This was a retrospective cohort study of patients hospitalized with community-acquired pneumonia at two academic tertiary care hospitals in San Antonio, Texas. Both hospitals are teaching affiliates of the University of Texas Health Science Center at San Antonio. The Institutional Review Board of the University of Texas Health Science Center at San Antonio approved the research protocol with exempt status. Study Sites/Inclusion and Exclusion Criteria We identified all patients admitted to the study hospitals between January 1, 1999, and December 1, 2002, with a primary discharge diagnosis of pneumonia (International Classification of Diseases-Ninth Revision codes 480.0-483.99 or 485-487.0) or secondary discharge diagnosis of pneumonia with a primary diagnosis of respiratory failure (518.81) or sepsis (038.xx). Subjects were included if they were aged more than 18 years, had an admission diagnosis of community-acquired pneumonia, and had a radiographically confirmed infiltrate or other finding consistent with community-acquired pneumonia on chest x-ray or computed tomography obtained within 24 hours of admission. Exclusion criteria included discharge from an acute care facility within 14 days of admission, transfer after being admitted to another acute care hospital, and receiving comfort measures only on this admission. If a subject was admitted more than once during the study period, only the first hospitalization was abstracted. Data Abstraction Chart review data included demographics, comorbid conditions, physical examination findings, laboratory data, and chest radiograph reports. In addition, data on important processes of care measures for patients hospitalized with community-acquired pneumonia also were abstracted: first Pneumonia is the leading infectious cause of death in the United States. Prompt antibiotic therapy consistent with national practice pneumonia guidelines is associated with decreased mortality at 48 hours after admission. dose of antibiotics within 4 hours of admission, collection of blood cultures before antibiotic administration and in the first 24 hours, and oxygen saturation measurement within 24 hours of presentation. 18 Mortality was assessed using information from the Texas Department of Health and the Department of Veteran Affairs clinical database. Mortality status was assessed through December 31, 2002. Antimicrobial Therapy We obtained information on antimicrobial therapy given within the first 48 hours of admission. Antimicrobial therapy was considered guideline-concordant if it agreed with either the 2003 IDSA or 2001 ATS clinical practice guidelines. 7,19 Therapies considered to be guideline-concordant for patients on the medical wards included a beta-lactam with doxycycline or a macrolide, or an antipneumococcal fluoroquinolone alone. Regimens considered guideline-concordant for patients hospitalized in the intensive care unit (ICU) included a beta-lactam with a macrolide or antipneumococcal fluoroquinolone, antipseudomonal beta-lactam aminoglycoside macrolide, or an antipneumococcal fluoroquinolone with clindamycin, vancomycin, or an aminoglycoside (for penicillin-allergic patients). Guideline-appropriate beta-lactam antibiotics included ceftriaxone, cefotaxime, cefepime, ampicillinsulbactam, ampicillin (high dose), piperacillin-tazobactam, imipenem-cilastatin, and meropenem. Guideline-appropriate fluoroquinolones included levofloxacin, gatifloxacin, and moxifloxacin. Appropriate macrolides included erythromycin, clarithromycin, and azithromycin. For a therapy to be considered guideline-concordant, the patient must have received at least one dose of each component of the recommended combinations within the first 48 hours of admission. Risk Adjustment The pneumonia severity index was used to assess severity of illness at presentation. 20 The pneumonia severity index is a validated prediction rule for 30-day mortality in patients with community-acquired pneumonia. This rule is based on 3 demographic characteristics, 5 comorbid illnesses, 5 physical examination findings, and 7 laboratory and radiographic findings from the time of presentation. Patients are classified into 5 risk classes with 30-day mortality ranging from 0.1% for class I to 27% for class V for patients enrolled in the Pneumonia Patient Outcomes Research Team cohort study. 20 Statistical Analyses Univariate statistics were used to test the association of sociodemographic and clinical characteristics with all-cause mortality at 48 hours. Categoric variables were analyzed

Mortensen et al 48-Hour Mortality in Pneumonia 861 Table 1 Subject Demographic and Clinical Characteristics by Mortality at 48 Hours* 48-hour mortality Variable Alive (n 767) Dead (n 20) P value Age, years standard deviation 60.6 16.4 59.8 17.3.40 Men 603 (79) 18 (90).2 Nursing home resident 49 (6) 5 (25).001 Received antibiotics before admission 135 (18) 5 (25).4 Admitted through emergency department 640 (83) 16 (80).7 Admitted to intensive care within 24 h 142 (18) 12 (60).001 Mechanical ventilation 77 (9) 11 (55).001 Preexisting comorbid conditions Congestive heart failure 117 (15) 6 (30).07 Chronic pulmonary disease 211 (28) 7 (35).5 History of stroke 100 (13) 5 (25).12 Chronic liver disease 92 (12) 2 (10).8 History of malignancy 75 (10) 3 (15).4 Renal insufficiency 83 (11) 4 (20).2 History, physical, laboratory, and radiographic data Altered mental status 76 (10) 9 (45).001 Respiratory rate 30 breaths/min 80 (10) 2 (10).9 Systolic blood pressure 90 mm Hg 18 (2) 3 (15).001 Heart rate 125 beats/min 99 (13) 6 (30).03 Temperature 95 or 104 20 (3) 1 (5).5 Arterial ph 7.35 44 (6) 5 (25).001 Arterial oxygenation saturation 90% 167 (22) 9 (45).01 Hematocrit 30% 69 (9) 3 (15).4 Serum blood urea nitrogen 30 mg/dl 160 (21) 8 (40).04 Serum glucose 250 mg/dl 74 (10) 2 (10) 1.0 Serum sodium 130 meq/l 111 (15) 5 (25).2 Pleural effusion on chest radiograph 184 (24) 5 (25).9 Multilobar infiltrates 264 (35) 8 (40).6 Pneumonia Severity Index Class I-III 406 (53) 3 (15) Class IV 255 (33) 11 (55) Class V 106 (14) 6 (30).003 *Data are presented as number (%) or mean standard deviation. using the chi-square test, and continuous variables were analyzed using the Student t test. A propensity score technique was used to balance covariates associated with antimicrobial therapy between groups. 21-23 The propensity score was derived from a logistic regression model. The covariates included in the propensity score model were the pneumonia severity index, admission through the emergency department, ICU admission within 24 hours, receiving initial antibiotics within 4 hours, having positive blood cultures, and hospital assignment. We then created an ordered categoric variable based on a quartile stratification of the propensity score to include in the Cox and regression models. Stratifying by the propensity score there was not a significant difference between the dead versus live at 48 hours for patients who were hospitalized in the ICU (P.7) or medicine wards (P.2). We used a Cox proportional hazard model to estimate and graph the baseline survivor functions after adjusting for the propensity score. We used logistic regression to assess the impact of empiric antimicrobial therapy on mortality at 48 hours. A dichotomous indicator variable indexing whether a patient received a guideline-concordant therapy was our predictor variable. Covariates included in the model were the use/ non-use of guideline-concordant antimicrobial therapy, and the ordered categoric variable based on quartile stratification on the propensity score. Model fit was assessed using the Hosmer-Lemeshow goodness-of-fit test. 24 All analyses were performed using STATA version 8 (Stata Corporation, College Station, Tex). RESULTS Data were abstracted on 787 patients at the 2 hospitals. The mean age was 60 years with a standard deviation of 16 years. Of the population, 79% were male, 84% were admitted through the emergency department, and 20% were admitted to the ICU within the first 24 hours after admission. Mortality was 2.5% at 48 hours, 9.2% at 30 days, and 13.6% at 90 days. By pneumonia severity index, 52% were low risk (pneumonia severity index classes I-III), 34% were moderate risk (pneumonia severity index class IV), and 14%

862 The American Journal of Medicine, Vol 119, No 10, October 2006 Table 2 Subject Demographic and Clinical Characteristics by Use of Guideline-Concordant versus Non Guideline-Concordant Antibiotics* Variable Antimicrobial regimen Concordant (n 625) Nonconcordant (n 162) P value Age, years standard deviation 59.2 16.2 65.3 16.3.0001 Men 473 (76) 148 (91).0001 Nursing home resident 37 (6) 17 (11).04 Received antibiotics before admission 111 (18) 29 (18).9 Admitted through emergency department 527(84) 129 (80).2 Admitted to intensive care within 24 h 106 (17) 48 (30).0001 Mechanical ventilation 70 (11) 18 (11).9 Preexisting comorbid conditions Congestive heart failure 90 (14) 33 (20).06 Chronic pulmonary disease 162 (26) 56 (35).03 History of stroke 75 (12) 30 (19).03 Chronic liver disease 77 (12) 17 (10).5 History of malignancy 50 (8) 28 (18).001 Renal insufficiency 66 (11) 21 (13).4 History, physical, laboratory, and radiographic data Altered mental status 61 (10) 24 (15).07 Respiratory rate 30 breaths/min Systolic blood pressure 90 mm Hg 13 (2) 8 (5).04 Heart rate 125 beats/min 90 (14) 15 (9).09 Temperature 95 or 104 18 (3) 3 (2).5 Arterial ph 7.35 43 (7) 6 (4).14 Arterial oxygenation saturation 90% 140 (22) 36 (22).9 Hematocrit 30% 49 (8) 23 (14).01 Serum blood urea nitrogen 30 mg/dl 131 (21) 37 (23).6 Serum glucose 250 mg/dl 61 (10) 15 (9).8 Serum sodium 130 meq/l 93 (15) 23 (14.2).8 Pleural effusion on chest radiograph 150 (24) 39 (25).9 Multilobar infiltrates 226 (36) 46 (29).07 Pneumonia Severity Index Class I-III 347 (55) 62 (38) Class IV 194 (31) 72 (44) Class V 84 (13) 28 (18).0001 *Data are presented as number (%) or mean standard deviation. were high risk (pneumonia severity index class V). In regard to pneumonia-related processes of care, 28% of patients received the initial dose of antibiotics within 4 hours of presentation, 76% of patients had blood cultures obtained within 24 hours and before antibiotics, and 91% of patients had oxygenation assessed at presentation. Table 1 shows the demographic factors and clinical characteristics for this population by mortality at 48 hours. In the univariate analysis, variables significantly associated with mortality at 48 hours (P.05) included living in a nursing home, admission to the ICU within 24 hours, need for mechanical ventilation, altered mental status on presentation, arterial hypoxia, and arterial ph 7.35 or less. In addition, increasing pneumonia severity index risk class was associated with increased mortality at 48 hours. Of the patients hospitalized in the ICU, 12 of 20 died within 48 hours, and 70% of patients who were in the ICU received guideline-concordant antibiotics. Of patients who died within 48 hours, 67% of patients in the ICU received guideline-concordant antibiotics. There were no significant differences between the mortality rates for patients in the ICU who received guideline-concordant therapy versus non guideline-concordant (67% vs 69%, P.9). Table 2 shows the demographics, comorbid conditions, and clinical characteristics by whether or not a patient received guideline-concordant therapy. Variables significantly associated with the use of guideline-concordant therapy included admission to the ward, no history of chronic pulmonary disease, no history of stroke, no history of malignancy, hematocrit greater than 30%, and systolic blood pressure greater than 90 mm Hg. All variables with significant associations with the use of guideline-concordant therapy were included as part of the propensity score. There were 63 patients with positive blood cultures (excluding those patients who had only one blood culture positive for coagulase-negative Staphylococcus). The most frequently isolated organism by blood culture was Streptococcus pneumoniae in 34 patients (54%).

Mortensen et al 48-Hour Mortality in Pneumonia 863 1 Guideline-Concordant Antibiotics, n= 625 0.95 0.9 Non-Guideline Concordant Antibiotics, n= 162 0.85 0.8 0.75 0 1 2 3 4 5 6 7 Days Figure 1 Proportion of surviving patients hospitalized with community-acquired pneumonia by use of guideline-concordant antibiotics versus non guideline-concordant antibiotics after adjusting for the propensity score (P.05). The initial empiric antimicrobial therapy was guidelineconcordant in 79.4% of patients: 82% (519/633) of ward patients and 69% (106/154) of ICU patients. Of patients who died within 48 hours, 67% of ICU patients received guideline-concordant antibiotics compared with 38% of ward patients. The most common appropriately prescribed empiric antimicrobial regimens included the use of a betalactam and macrolide (n 156), a beta-lactam and fluoroquinolone (n 126), a beta-lactam fluoroquinolone macrolide (n 125), or an antipneumococcal fluoroquinolone only in ward patients (n 200). The most common non guideline-concordant antimicrobial regimen was the use of a beta-lactam alone (n 97), which was received by 35% (7/20) of patients who died within 48 hours compared with 12% (90/767) who survived that time period (P.002). Figure 1 demonstrates the results of the survival analysis after adjusting for the propensity score. It demonstrates statistically significant difference in survival starting at the first day after admission for those who received guidelineconcordant antibiotics versus non guideline-concordant antibiotics (P.05). At 7 days after hospital admission there was a 2.7% difference in survival. In the regression analysis, after adjusting for potential confounders using the propensity score, use of guidelineconcordant antibiotics was associated with decreased mortality at 48 hours (odds ratio 0.37; 95% confidence interval [CI], 0.15-0.95, P.04). DISCUSSION Community-acquired pneumonia continues to be an acute medical problem with substantial mortality and morbidity. 2 Our study calls into question the concept that mortality within the first 48 to 96 hours after admission is not modifiable, and provides further evidence for the beneficial effect of the use of the empiric antimicrobial regimens recommended by the IDSA/ATS guidelines. Our results strengthen the previous body of research addressing what antimicrobial therapies are appropriate for patients with community-acquired pneumonia. Previous studies have found that the use of a beta-lactam plus a macrolide is associated with significantly lower mortality. 11-14 These studies demonstrated that monotherapy with beta-lactam is associated with worse outcomes including increased mortality or increased length of stay. 12-14,25 Several other studies demonstrated that the use of empiric antimicrobial therapy concordant with national guidelines is associated with decreased mortality. 13,15,26 All of these studies examined in-hospital, or 30-day mortality, and did not examine the impact of antimicrobial therapy on mortality within the first 48 to 96 hours. To our knowledge, our study is one of the only studies to examine early mortality for patients with community-acquired pneumonia since the ground-breaking article by Austrian and Gold. 1 These results also call into question the research practice of excluding patients who physicians believe may not survive the first 48 to 72 hours from randomized clinical trials

864 The American Journal of Medicine, Vol 119, No 10, October 2006 of antibiotic therapy. These patients may have the most to benefit from these trials, especially because the control groups frequently receive guideline-concordant antibiotic combinations. 17 If patients at high risk of 48- to 72-hour mortality are included in antibiotic trials, one would expect that the mortality rates reported by randomized controlled trials may be closer to the rates reported in observational studies of community-acquired pneumonia. 27 Our study has several limitations that should be acknowledged. First, our sample was predominantly male because of the inclusion of a Veterans Affairs hospital, and it is possible, but unlikely, that females may have differential responsiveness to antibiotics from males. Also, because of the study design there were a number of variables associated with use versus non-use of guideline-concordant therapy. We used the propensity score to adjust for these differences; however, there is the potential that we were not able to fully adjust for these differences or other unmeasured differences. Finally, there were only a small number of deaths within 48 hours in this cohort so we are unable to fully explore other potential predictors of death or the impact of specific antimicrobial therapies on mortality. CONCLUSION This study demonstrates that receiving guideline-concordant antimicrobial therapy is associated with lower 48-hour mortality for patients hospitalized with community-acquired pneumonia. This finding provides further support for the use of empiric antimicrobial therapies consistent with guidelines from the IDSA and ATS. 7,19 Further research is needed to determine how to promote the use of guidelineconcordant antimicrobial regimens for patients hospitalized with community-acquired pneumonia. References 1. Austrian R, Gold J. Pneumococcal bacteremia with special reference to bacteremic pneumococcal pneumonia. Ann Intern Med. 1964;60:759-776. 2. Freid VM, Prager K, MacKay AP, Xia H. Health, United States, 2003 with Chartbook on Trends in the Health of Americans. Hyattsville, MD: National Center for Health Statistics; 2003. 3. Gilbert K, Fine MJ. Assessing prognosis and predicting patient outcomes in community-acquired pneumonia. Semin Respir Infect. 1994; 9(3):140-152. 4. Society BT. Guidelines for the management of community-acquired pneumonia in adults admitted to the hospital. Br J Hosp Med. 1993; 49(5):346-350. 5. Gialdroni Grassi G, Bianchi L. Guidelines for the management of community-acquired pneumonia in adults. Italian Society of Pneumology. Italian Society of Respiratory Medicine. Italian Society of Chemotherapy. Monaldi Arch Chest Dis. 1995;50(1):21-27. 6. Niederman M, Bass JB, Campbell GD, et al. Guidelines for the initial management of adult with community-acquired pneumonia: diagnosis, assessment of severity, and initial antimicrobial therapy. Am Rev Respir Dis. 1993;148:1418-1426. 7. 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(7):1730-1754. 8. Bartlett JG, Dowell SF, Mandell LA, File TM Jr, Musher DM, Fine MJ. Practice guidelines for the management of community-acquired pneumonia in adults. Infectious Diseases Society of America. Clin Infect Dis. 2000;31(2):347-382. 9. British Thoracic Society guidelines for the management of community acquired pneumonia in adults. Thorax. 2001;56(Suppl 4):IV1-64. 10. Mandell LA, Marrie TJ, Grossman RF, et al. Summary of Canadian guidelines for the initial management of community-acquired pneumonia: an evidence-based update by the Canadian Infectious Disease Society and the Canadian Thoracic Society. Can Respir J. 2000;7(5): 371-382. 11. Gleason PP, Meehan TP, Fine JM, Galusha DH, Fine MJ. Associations between initial antimicrobial therapy and medical outcomes for hospitalized elderly patients with pneumonia. Arch Intern Med. 1999; 159(21):2562-2572. 12. Houck PM, MacLehose RF, Niederman MS, Lowery JK. Empiric antibiotic therapy and mortality among Medicare pneumonia inpatients in 10 Western States: 1993, 1995, and 1997. Chest. 2001;119: 1420-1426. 13. Battleman DS, Callahan M, Thaler HT. Rapid antibiotic delivery and appropriate antibiotic selection reduce length of hospital stay of patients with community-acquired pneumonia. Arch Intern Med. 2002; 162:682-688. 14. Waterer GW, Somes GW, Wunderink RG. Monotherapy may be suboptimal for severe bacteremic pneumococcal pneumonia. Arch Intern Med. 2001;161(15):1837-1842. 15. Mortensen EM, Restrepo M, Anzueto A, Pugh J. Effects of guidelineconcordant antimicrobial therapy on mortality among patients with community-acquired pneumonia. Am J Med. 2004;117(10):726-731. 16. Wunderink RG, Waterer GW. Community-acquired pneumonia: pathophysiology and host factors with focus on possible new approaches to management of lower respiratory tract infections. Infect Dis Clin North Am. 2004;18(4):743-759, vii. 17. Marrie T. Community-acquired Pneumonia. New York, NY: Kluwer Academic; 2001. 18. Meehan TP, Fine MJ, Krumholz HM, et al. Quality of care, process, and outcomes in elderly patients with pneumonia. JAMA. 1997; 278(23):2080-2084. 19. Mandell LA, Bartlett JG, Dowell SF, File TM Jr, Musher DM, Whitney C. Update of practice guidelines for the management of community-acquired pneumonia in immunocompetent adults. Clin Infect Dis. 2003;37(11):1405-1433. 20. 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(4):243-250. 21. Stone RA, Obrosky DS, Singer DE, Kapoor WN, Fine MJ. Propensity score adjustment for pretreatment differences between hospitalized and ambulatory patients with community-acquired pneumonia. Pneumonia Patient Outcomes Research Team (PORT) Investigators. Med Care. 1995;33(4 Suppl):AS56-AS66. 22. Klungel OH, Martens EP, Psaty BM, et al. Methods to assess intended effects of drug treatment in observational studies are reviewed. J Clin Epidemiol. 2004;57(12):1223-1231. 23. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41-55. 24. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley and sons; 1989 (Wiley series in probability and mathematical statistics). 25. Brown RB, Iannini P, Gross P, Kunkel M. Impact of initial antibiotic choice on clinical outcomes in community-acquired pneumonia: analysis of a hospital claims-made database. Chest. 2003;123:1503-1511. 26. Malone DC, Shaban HM. Adherence to ATS guidelines for hospitalized patients with community-acquired pneumonia. Ann Pharmacother. 2001;35:1180-1185. 27. Shefet D, Robenshtok E, Leibovici L. Empirical atypical coverage for inpatients with community-acquired pneumonia. Arch Intern Med. 2005;165:1992-2000.