Figures
Abstract
Background
While the majority of healthcare in the US is provided in community hospitals, the epidemiology and treatment of bloodstream infections in this setting is unknown.
Methods and Findings
We undertook this multicenter, retrospective cohort study to 1) describe the epidemiology of bloodstream infections (BSI) in a network of community hospitals and 2) determine risk factors for inappropriate therapy for bloodstream infections in community hospitals. 1,470 patients were identified as having a BSI in 9 community hospitals in the southeastern US from 2003 through 2006. The majority of BSIs were community-onset, healthcare associated (n = 823, 56%); 432 (29%) patients had community-acquired BSI, and 215 (15%) had hospital-onset, healthcare-associated BSI. BSIs due to multidrug-resistant pathogens occurred in 340 patients (23%). Overall, the three most common pathogens were S. aureus (n = 428, 28%), E. coli (n = 359, 24%), coagulase-negative Staphylococci (n = 148, 10%), though type of infecting organism varied by location of acquisition (e.g., community-acquired). Inappropriate empiric antimicrobial therapy was given to 542 (38%) patients. Proportions of inappropriate therapy varied by hospital (median = 33%, range 21–71%). Multivariate logistic regression identified the following factors independently associated with failure to receive appropriate empiric antimicrobial therapy: hospital where the patient received care (p<0.001), assistance with ≥3 ADLs (p = 0.005), Charlson score (p = 0.05), community-onset, healthcare-associated infection (p = 0.01), and hospital-onset, healthcare-associated infection (p = 0.02). Important interaction was observed between Charlson score and location of acquisition.
Conclusions
Our large, multicenter study provides the most complete picture of BSIs in community hospitals in the US to date. The epidemiology of BSIs in community hospitals has changed: community-onset, healthcare-associated BSI is most common, S. aureus is the most common cause, and 1 of 3 patients with a BSI receives inappropriate empiric antimicrobial therapy. Our data suggest that appropriateness of empiric antimicrobial therapy is an important and needed performance metric for physicians and hospital stewardship programs in community hospitals.
Citation: Anderson DJ, Moehring RW, Sloane R, Schmader KE, Weber DJ, Fowler VG Jr, et al. (2014) Bloodstream Infections in Community Hospitals in the 21st Century: A Multicenter Cohort Study. PLoS ONE 9(3): e91713. https://doi.org/10.1371/journal.pone.0091713
Editor: Paul J. Planet, Columbia University, United States of America
Received: August 22, 2013; Accepted: February 13, 2014; Published: March 18, 2014
Copyright: © 2014 Anderson et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by the Robert Wood Johnson Foundation [#66327 to DJA] and the National Institute of Health/National Institute of Allergy and Infectious Diseases [K23 AIO95357 to DJA; K24 AI093969 to VGF]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: DJA has received research support from the CDC, AHRQ, and Merck, Inc., and receives royalties from UpToDate, Online. KES receives research support from Merck, Inc. DJW has consulting relationships with Merck, Pfizer, ASP, and Clorox and is on speaker's bureaus for Merck and Pfizer. VGF has consulting relationships with Merck, Pfizer, Novartis, Galderma, Novadigm, Durata, Achaogen, Affinium, Cerex, and MedImmune, and receives royalties from UpToDate, Online. DJS receives royalties from UpToDate, Online. RWM, RS, and ES report no other relationships or activities that readers could perceive to have influenced or give the appearance of potentially influencing the submitted work. These potential competing interests do not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
Introduction
Bloodstream infections (BSIs) are a leading cause of suffering and death in the US. As many as 250,000 BSIs occur each year [1], with a mortality rate of 35% and costs of up to $37,000 per case [2], [3]. In fact, BSI was one of the top causes of death in the US in 2008, leading to more than 35,000 deaths [4].
The majority of healthcare in the US is performed in smaller, non-teaching community hospitals._ENREF_4 The mean size of hospitals in the US was 160 beds in 2009, and 72% of hospitals had fewer than 200 beds [4]._ENREF_4 Of the 39 million hospital discharges in the US in 2010, 19.9 million (51%) were from non-teaching facilities [5]. Our understanding about the causes and risk factors for BSI in these community hospitals, however, is alarmingly inadequate. Prior studies on BSIs in community hospitals have been limited to specific organisms [6], [7], single institutions, intensive care units [8] and/or patients admitted prior to the emergence and spread of epidemiologically important multidrug-resistant organisms [9]–[14]. Even less is known about the impact of antibacterial resistance in community hospitals [15].
Limited available data suggest that patients with BSI are less likely to receive appropriate therapy in community hospitals as compared to tertiary care hospitals [15]. These treatment trends are important, as inappropriate empiric therapy leads to a 60% increase in mortality [16].
We undertook this study in order to 1) describe the epidemiology of BSIs in a network of community hospitals and 2) determine risk factors for inappropriate therapy for BSIs in community hospitals.
Materials and Methods
Ethics Statement
This study was reviewed and approved by the Institutional Review Board (IRB) of Duke University Health System. Participating community hospitals deferred to the Duke IRB (n = 5), or reviewed and approved the study via their local IRB (n = 4). Written patient consent was waived by all sites.
Study Design and Participating Hospitals
This retrospective cohort study included adult subjects admitted to nine community, non-academic hospitals in North Carolina and Virginia from January 1, 2003, through December 31, 2006 (hereafter, the “study period”). The median bed size of participating hospitals was 151 (range 102–355) beds. All hospitals were members of the Duke Infection Control Outreach Network (DICON) [17]. In brief, DICON provides infection control consulting, data feedback, education, and quality improvement services to 42 hospitals in the southeastern US.
Patient Selection
We reviewed data from all consecutive positive blood cultures from participating hospitals' microbiology laboratory databases during the study period. Patients were randomly selected from the overall cohort. Our goal enrollment was 1,400 patients. Assuming that 35% of patients would receive inappropriate therapy [15], [18], the study had >80% power to identify a risk factor for inappropriate therapy that was 8% prevalent among patients who received inappropriate therapy compared to 4% prevalent among patients who received appropriate therapy (alpha = 0.05). Random sampling by random number generator was performed to ensure that subjects were included equally from all four years of the study period.
Trained data abstractors collected all patient data by chart review, including detailed clinical, demographic, microbiologic, treatment data, and outcomes. A standardized data collection tool, data dictionary, and standard operating procedure were created prior to data abstraction. Any patient with a bloodstream infection was considered for inclusion. Selected patients were included in the database once. That is, if a patient's second cases of BSI was randomly selected for inclusion, it was excluded and a new patient was randomly selected.
Definitions
BSIs were defined using modified Centers for Disease Control and Prevention (CDC) criteria: ≥ 1 positive blood culture for all bacterial pathogens except common skin contaminants including Enterococci which require ≥ 2 positive blood cultures within 48 hours. Onset of infection was defined as the time of the first positive blood culture. “Appropriate empiric antimicrobial therapy” was defined as receipt of an antimicrobial agent with in vitro activity against the infecting organism within 24 hours after the onset of infection. Location of acquisition was defined using CDC criteria [19] as follows: 1) “community-onset, healthcare-associated” was defined as a BSI occurring <48 hours after admission plus the presence of ≥1 of the following healthcare risk factors: prior hospitalization, surgery, dialysis, or residence in a long-term care facility in the 12 months preceding the BSI, or the presence of an invasive device; 2) “community-acquired” was defined as a BSI occurring <48 hours after admission without one of the above healthcare risk factors; and 3) “hospital-onset, healthcare-associated” was defined as a BSI that occurred ≥ 48 hours after hospital admission.
Multidrug resistance was defined using consensus definitions [20]. Charlson comorbidity index and McCabe score were used to measure baseline severity of illness at the time of admission [21], [22]._ENREF_6 APACHE II score was calculated to measure severity of illness at the time of infection [23]. Functional status was measured as independent or not independent for 5 activities of daily living (ADLs) using the Katz criteria [24]._ENREF_22 A binary variable for functional status was created to measure severe disability which was defined as lack of independence with ≥3 ADLs. Finally, a secondary BSI was defined as a BSI that occurred as a result of a microbiologically-diagnosed infection from another body site, excluding central venous catheters.
Statistics
Rates were calculated as number of BSIs/1,000 patient-days. Standard descriptive statistics were used for categorical and continuous variables. Bivariable analyses were performed using the Student's t-test or Wilcoxon rank sums test, as appropriate, for continuous variables and the χ2 test for categorical variables. All tests were two-tailed; a p-value ≤0.05 was considered to be significant for all tests. Statistical analyses were performed using SAS v9.3 (Cary, NC).
The main outcome of interest was “failure to receive appropriate empiric antimicrobial therapy.” For simplicity, patients who experienced the outcome will hereafter been labeled as “cases.” Our a priori covariate of interest was location of acquisition. Logistic regression models were created to identify independent variables associate with the main outcome. For each model, candidate variables were included in the full model if p<0.2 in bivariable analysis. No correction for multiple comparisons was made to ensure that our models were inclusive of all appropriate candidate variables. Variables considered for inclusion in the models were assessed for missing data. All evaluated variables were missing fewer than 5% of the time. Thus, missing data for these variables were imputed using unconditional imputation: imputation of the mean for continuous variables or the mode for categorical variables [25]. Effect measure modification was evaluated using interaction terms with the co-variate of interest. The Likelihood ratio test was used to determine the most parsimonious model if group variables (e.g., Charlson score) and specific components of the group variable (e.g., diabetes) were candidate variables. Candidate variables were assessed for co-linearity, and co-linear variables were removed, as necessary. The final model was created using manual backwards selection. During this process, removed variables were assessed for confounding. A variable was considered to be a potential confounder if the β-estimate for any variable changed >10% after its removal. Finally, a class variable for “hospital” was included in the model, and the generalized estimating equation method was used to account for clustering of the outcome of interest within individual hospitals.
The first and primary model was created using only variables available to treating clinicians at the onset of infection. The second model was created in a similar fashion but candidate variables also included variables available after the onset of infection (e.g., multidrug resistance and polymicrobial infection).
Results
Epidemiology of BSIs in Community Hospitals
A total of 5,124 patients had a BSI during 1,371,467 patient days during the study period. The median rate of BSI per hospitals was 3.5 BSIs/1000 patient days (IQR 3.0–4.2). The study cohort consisted of a sample of 1,470 (29%) unique patients who were randomly selected for inclusion in the study. Patients randomly selected for inclusion had similar organism distribution to patients not selected.
Patients were generally older (mean age 65.3 ± 17.2) and Caucasian race (n = 759, 52%) but a high proportion of patients were African-American race (n = 680, 46%) (Table 1). A total of 1,091 (74%) patients were admitted from home, and 312 (21%) were admitted from a nursing home; 745 (51%) patients had been hospitalized in the prior 12 months.
The majority of BSIs were community-onset, healthcare associated (n = 823, 56%; Table 2); 432 (29%) patients had community-acquired BSI, and 215 (15%) had hospital-onset, healthcare-associated BSI. BSIs due to multidrug-resistant pathogens occurred in 340 patients (23%). A total of 1,514 separate pathogens were identified during the 1,470 BSIs (Table 3). Overall, the most common pathogen was S. aureus (n = 428, 28%), though type of infecting organism varied by location of acquisition (Figure 1). The most common multidrug-resistant pathogens were methicillin resistant Staphylococcus aureus (MRSA) (n = 203, 13%), E. coli (n = 79, 5%), coagulase-negative staphylococci (n = 51, 3%), and Pseudomonas aeruginosa (n = 8, 0.5%).
[FOOTNOTE] * CA = community-acquired, CO-HCA = community-onset, healthcare-associate, HO-HCA; hospital-onset, healthcare-associated, CoNS = coagulase negative Staphylococci.
The in-hospital mortality rate was 18% (n = 264; Table 4). Patients with any prior healthcare exposure were almost 3-fold more likely to die than patients without prior healthcare exposure (OR = 2.78 95% CI 1.94–4.00). The in-hospital mortality rate was 21% (n = 170) for patients with community-onset, healthcare-associated infection and 26% (n = 55) for hospital-onset, healthcare-associated infection.
Predictors of Failure to Receive Appropriate Empiric Antimicrobial Therapy
Bivariable analysis.
Inappropriate empiric antimicrobial therapy was given to 542 (38%) patients (hereafter, “cases”). The proportion of patients who recieved inappropriate therapy varied by hospital (median = 33%, range 21–71%). In bivariable analysis, cases were more likely than non-cases to require assistance with any ADL (62% v. 55%, p = 0.01), ≥3 ADLs (34% v. 22%, p<0.001), and had a higher median Charlson score (p = 0.02) (Table 1).
Cases were more likely to have prior healthcare exposure than non-cases (p = 0.0002). More specifically, cases were more frequently admitted from a nursing facility (p<0.001) or hospitalized in the prior 12 months (p = 0.002). Location of acquisition differed between the two groups (p<0.001); cases had more hospital-onset, healthcare-associated and less community-acquired BSI than non-cases. Cases were more likely to have an infection with a multidrug-resistant Gram positive organism (p<0.001).
Unadjusted in-hospital mortality was higher among cases compared to non-cases, but the difference was not statistically significant (p = 0.18; Table 4). Cases had longer total duration of hospitalization (p = 0.04) and were more likely to require PICC placement for outpatient IV antimicrobial therapy (p = 0.02). In addition, cases had different discharge destinations than non-cases (p = 0.01). The reasons for this difference may have been related to previously noted differences in admission sources (ie, more cases were admitted from facilities). In order to investigate this further, we performed a sub-analysis limited only to patients admitted from home (n = 1,091). In this group, a smaller proportion of cases were discharged back to home compared with non-cases (52% v. 62%, p = 0.003).
Multivariate Analysis.
Multivariate logistic regression identified several factors independently associated with failure to receive appropriate empiric antimicrobial therapy, including hospital where the patient received care (p<0.001), assistance with ≥3 ADLs (p = 0.005), and Charlson score (p = 0.05; Table 5). In addition, community-onset, healthcare-associated infection (p = 0.01) and hospital-onset, healthcare-associated infection (p = 0.02) were associated with failure to receive appropriate empiric antimicrobial therapy using community-acquired infection as reference. Important interaction was observed between Charlson score and location of acquisition.
We then created a three part variable for Charlson score to better describe this interaction: few or no co-morbidities (Charlson score ≤1), moderate co-morbidities (2–4), and severe co-morbidity (≥5). Among patients with community-onset, healthcare-associated BSI, the most likely to fail to receive appropriate empiric antimicrobial therapy had no or few co-morbidities (OR = 1.69, 95% CI 1.12–2.57; p = 0.006) despite having lower acuity of illness than patients with moderate or severe co-morbidities (APACHE II score [IQR]: 13 [10]–[16] vs. 15 [12]–[19] vs. 15 [12]–[18], p<0.0001). All patients with hospital-onset, healthcare-associated infection were at high risk of receiving inappropriate empiric antimicrobial therapy, though patients with severe co-morbidities were at highest risk (OR = 3.39, 95% CI 1.34–8.54; p = 0.01).
A second model was created to determine if inclusion of the presence of drug-resistance, a variable not available to clinicians when they prescribed empiric antimicrobial therapy, altered any of the independent predictors identified in Model 1. Although infection due to a multidrug-resistant organism was strongly associated with failure to receive appropriate empiric antibiotic therapy (OR = 2.17, 95% CI 1.48–3.18, p<0.001), the majority of the predictors identified in Model 1 remained independently associated with failure to receive appropriate empiric antibiotic therapy even with adjustment for MDR (Table 5). The exception to this trend was Charlson score, for which the p-value changed from 0.05 to 0.07.
Discussion
The paucity of data on BSIs in community hospitals occurs despite the fact that the majority of health care in the US is provided in this setting. In the current study, we used a large cohort of patients with BSI from 9 community hospitals to identify key aspects of the epidemiology of BSIs and risk factors for inappropriate therapy in patients in community hospitals. Key findings from our study included the following: 1) BSIs in community hospitals are typically community-onset, healthcare-associated infections; 2) the location of acquisition had a major impact on the causative organisms; 3) empiric antimicrobial treatment for patients with BSI in community hospitals is incorrect in approximately 1 of 3 patients; and 4) groups with the highest risk of receiving inappropriate empiric therapy include a) patients with hospital-onset, healthcare associated infections b) patients with community-onset, healthcare-associated infections and few co-morbidities, and c) patients with impaired functional status.
Healthcare exposure preceded the onset of BSI in almost 3 of every 4 patients in our cohort. For example, the majority of patients in our study cohort had central venous lines, had invasive devices present at the time of BSI, were elderly, and/or required assistance with activities of daily living. In fact, the majority of BSIs were community-onset, healthcare associated infections (56%). Authors of a recent multicenter study of 7 academically-affiliated hospitals concluded that community-onset, healthcare associated BSIs were 2-times more frequent than community-acquired BSIs and associated with 3-fold higher mortality [26]. Similarly, authors of a review of BSI data from 59 hospitals demonstrated that 62% of BSIs followed some type of healthcare exposure and the majority of BSISs (55%) were community-onset, healthcare-associated [27]. These findings and ours suggest that prevention efforts aimed at hospital-acquired BSI are less likely to be as effective or impactful as efforts to promptly and correctly provide effective empiric therapy for healthcare-exposed patients from the community.
The microbiology of BSIs in community hospitals has changed over the last few decades. S. aureus was the most common cause of BSI in our study cohort. In contrast, E. coli was the most common cause of BSI in community hospitals prior to 2000 [10], [12], [14]. Further analysis of our data showed that S. aureus was the most common pathogen in patients with community-onset, healthcare-associated BSI and hospital-onset, healthcare-associated BSI, but E. coli remained the most common pathogen among patients with community-acquired BSIs. Thus, the observed emergence of S. aureus may be related to the increasing frequency of complex outpatient medical care and the presence of indwelling devices prior to BSI.
In addition, patients in our study cohort more frequently had infections due to multi-drug resistant (MDR) pathogens than in previous studies. For example, three prior studies showed that the frequency of MRSA as a cause of BSI in community pathogens was 5-10 times lower (13% vs 1–3%) [9], [11], [12]. Overall, 23% of patients had BSIs due to MDR pathogens in our cohort. Similar changes in organisms and antimicrobial resistance have been observed in tertiary care centers previously [28]–[30], suggesting that BSIs in community hospitals are increasingly similar to BSIs seen in tertiary care centers.
Approximately 1 of every 3 patients with BSI in our study failed to receive appropriate empiric antimicrobial therapy. This proportion of inappropriate therapy is higher than in previously published studies from community hospitals. Approximately 20% of patients with BSI failed to receive appropriate empiric antimicrobial therapy prior to 2000 [11], [12]. In contrast, results from this study are similar to our recently published cohort of patients in community hospitals with MRSA bacteremia [15] and data published from tertiary care centers [18], [31]–[33]. Thus, this finding may be due to the higher prevalence of multidrug-resistant pathogens in participating community hospitals and/or suggestive of increasingly complex patients presenting to community hospitals.
Recent healthcare exposure, the need for assistance with ADLs, and infection due to a multidrug-resistant pathogen were independent predictors for failure to receive appropriate empiric therapy in this study. One counterintuitive predictor in our study cohort, however, deserves special emphasis: patients with few co-morbidities and community-onset, healthcare-associated BSI were at an increased risk of receiving inappropriate empiric antimicrobial therapy. We observed that this group had lower APACHE II scores and was less acutely ill at the time of onset of their bloodstream infection. Reasons for this finding are uncertain, although we theorize that treating clinicians were less suspicious of a severe illness such as BSI. Thus, they may have been less aggressive in administering any therapy or broad-spectrum antimicrobial therapy.
Our study has several important limitations. First, our retrospective design may have led to selection and/or misclassification bias. We addressed these issues by randomly selecting patients from the overall cohort and by training data abstractors in the use of a standard case review form with variable definitions. Second, our study involves patients admitted from 2003 to 2006. Our study, however, represents the largest, most detailed study on BSIs in community hospitals to date and provides a much needed update. In addition, we believe the issue of inappropriate empiric antimicrobial therapy has likely worsened over the last few years, as resistance rates and the number of overall BSIs have increased [34]; thus, our study may actually underestimate the problem of inappropriate empiric antimicrobial therapy. Third, our randomization approach may have led to selection bias. Our cases, however, had similar organism distribution as patients not selected by randomization. Fourth, we did not correct for multiple statistical comparisons. Thus, some of our “significant” statistical tests may be due to Type I error. The intent of this analysis, however, was to be exploratory and inclusive. Our findings will require confirmation in further studies. Finally, our multicenter study may not be generalizable to tertiary care centers.
In summary, our large, multicenter study provides the most complete picture of BSIs in community hospitals in the US to date. The types of BSIs seen in community hospitals have changed. Community-onset, healthcare-associated BSI is most common, and S. aureus is the most common cause. This shift has led to important ramifications. One of every 3 patients with a BSI receives inappropriate empiric antimicrobial therapy. Thus, our data confirm that appropriateness of empiric antimicrobial therapy is an important and needed performance metric for physicians and hospital stewardship programs [35]._ENREF_24 Risk of inadequate therapy is highest among patients with healthcare exposure and the disabled. Clinicians in community hospitals must identify these important risk factors when choosing antibiotic therapy, particularly given the adverse outcomes associated with inadequate therapy. We believe that most risk factors could be easily discernible using electronic data; efforts must be made to ensure that others, such as ADLs, are routinely included in electronic data. Ultimately, we believe an intervention whereby physicians are automatically notified of these risk factors when choosing empiric antimicrobial therapy is needed, particularly among community hospitals in the US.
Acknowledgments
The authors would like to thank Lauren Knelson and Joyce Frederick for assistance with data collection.
Disclaimer: Data were presented in part at IDWeek 2012 in San Diego, CA, USA, on October 20, 2012.
Author Contributions
Conceived and designed the experiments: DJA RWM KES DJW VGF DJS. Performed the experiments: DJA RWM ES. Analyzed the data: DJA RWM RS. Contributed reagents/materials/analysis tools: RS KES DJW VGF DJS. Wrote the paper: DJA RWM RS KES DJW VGF ES DJS.
References
- 1. Pittet D, Li N, Woolson RF, Wenzel RP (1997) Microbiological factors influencing the outcome of nosocomial bloodstream infections: a 6-year validated, population-based model. Clin Infect Dis 24: 1068–1078.
- 2. Anderson DJ, Kirkland KB, Kaye KS, Thacker PA 2nd, Kanafani ZA, et al. (2007) Underresourced hospital infection control and prevention programs: penny wise, pound foolish? Infect Control Hosp Epidemiol 28: 767–773.
- 3. Cosgrove SE, Sakoulas G, Perencevich EN, Schwaber MJ, Karchmer AW, et al. (2003) Comparison of mortality associated with methicillin-resistant and methicillin-susceptible Staphylococcus aureus bacteremia: a meta-analysis. Clin Infect Dis 36: 53–59.
- 4.
National Center for Healthcare Statistics (2012) Health, United States, 2011 with special feature on socioeconomic status and health. Hyattsville, MD.
- 5.
Agency for Healthcare Research and Quality (2013) Healthcare Cost and Utilization Project - Statistics on Hospital Stays.
- 6. Haddy RI, Cecil ML, Norris LL, Markert RJ (1991) Enterobacter bacteremia in the community hospital. J Fam Pract 32: 601–606.
- 7. Haddy RI, Mann BL, Nadkarni DD, Cruz RF, Elshoff DJ, et al. (1996) Nosocomial infection in the community hospital: severe infection due to Serratia species. J Fam Pract 42: 273–277.
- 8. Warren DK, Zack JE, Elward AM, Cox MJ, Fraser VJ (2001) Nosocomial primary bloodstream infections in intensive care unit patients in a nonteaching community medical center: a 21-month prospective study. Clin Infect Dis 33: 1329–1335.
- 9. Elhanan G, Raz R, Pitlik SD, Sharir R, Konisberger H, et al. (1995) Bacteraemia in a community and a university hospital. J Antimicrob Chemother 36: 681–695.
- 10. Haddy RI, Klimberg S, Epting RJ (1987) A two-center review of bacteremia in the community hospital. J Fam Pract 24: 253–259.
- 11. Mylotte JM, Kahler L, McCann C (2001) Community-acquired bacteremia at a teaching versus a nonteaching hospital: impact of acute severity of illness on 30-day mortality. Am J Infect Control 29: 13–19.
- 12. Scheckler WE, Bobula JA, Beamsley MB, Hadden ST (2003) Bloodstream infections in a community hospital: a 25-year follow-up. Infect Control Hosp Epidemiol 24: 936–941.
- 13. Setia U, Gross PA (1977) Bacteremia in a community hospital: spectrum and mortality. Arch Intern Med 137: 1698–1701.
- 14. Javaloyas M, Garcia-Somoza D, Gudiol F (2002) Epidemiology and prognosis of bacteremia: a 10-y study in a community hospital. Scand J Infect Dis 34: 436–441.
- 15. Kaye KS, Anderson DJ, Choi Y, Link K, Thacker P, et al. (2008) The deadly toll of invasive methicillin-resistant Staphylococcus aureus infection in community hospitals. Clin Infect Dis 46: 1568–1577.
- 16. Paul M, Shani V, Muchtar E, Kariv G, Robenshtok E, et al. (2010) Systematic review and meta-analysis of the efficacy of appropriate empiric antibiotic therapy for sepsis. Antimicrob Agents Chemother 54: 4851–4863.
- 17. Anderson DJ, Miller BA, Chen LF, Adcock LH, Cook E, et al. (2011) The network approach for prevention of healthcare-associated infections: long-term effect of participation in the Duke Infection Control Outreach Network. Infect Control Hosp Epidemiol 32: 315–322.
- 18. Ibrahim EH, Sherman G, Ward S, Fraser VJ, Kollef MH (2000) The influence of inadequate antimicrobial treatment of bloodstream infections on patient outcomes in the ICU setting. Chest 118: 146–155.
- 19. Klevens RM, Morrison MA, Nadle J, Petit S, Gershman K, et al. (2007) Invasive methicillin-resistant Staphylococcus aureus infections in the United States. Jama 298: 1763–1771.
- 20.
Magiorakos AP, Srinivasan A, Carey RB, Carmeli Y, Falagas ME, et al.. (2011) Multidrug-resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clinical microbiology and infection: the official publication of the European Society of Clinical Microbiology and Infectious Diseases.
- 21. Bion JF, Edlin SA, Ramsay G, McCabe S, Ledingham IM (1985) Validation of a prognostic score in critically ill patients undergoing transport. Br Med J (Clin Res Ed) 291: 432–434.
- 22. Charlson ME, Pompei P, Ales KL, MacKenzie CR (1987) A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40: 373–383.
- 23. Knaus WA, Draper EA, Wagner DP, Zimmerman JE (1985) APACHE II: a severity of disease classification system. Crit Care Med 13: 818–829.
- 24. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW (1963) Studies of Illness in the Aged. the Index of Adl: a Standardized Measure of Biological and Psychosocial Function. Jama 185: 914–919.
- 25. Landerman R, Land K, Pieper C (1997) An Empirical Evaluation of the Predictive Mean Matching Method for Imputing Missing Values. Sociol Method Res 26: 3–33.
- 26. Kollef MH, Zilberberg MD, Shorr AF, Vo L, Schein J, et al. (2011) Epidemiology, microbiology and outcomes of healthcare-associated and community-acquired bacteremia: a multicenter cohort study. J Infect 62: 130–135.
- 27. Shorr AF, Tabak YP, Killian AD, Gupta V, Liu LZ, et al. (2006) Healthcare-associated bloodstream infection: A distinct entity? Insights from a large U.S. database. Crit Care Med 34: 2588–2595.
- 28. Fowler VG Jr, Miro JM, Hoen B, Cabell CH, Abrutyn E, et al. (2005) Staphylococcus aureus endocarditis: a consequence of medical progress. Jama 293: 3012–3021.
- 29. Fowler VG Jr, Olsen MK, Corey GR, Woods CW, Cabell CH, et al. (2003) Clinical identifiers of complicated Staphylococcus aureus bacteremia. Arch Intern Med 163: 2066–2072.
- 30. Murdoch DR, Corey GR, Hoen B, Miro JM, Fowler VG Jr, et al. (2009) Clinical presentation, etiology, and outcome of infective endocarditis in the 21st century: the International Collaboration on Endocarditis-Prospective Cohort Study. Arch Intern Med 169: 463–473.
- 31. Centers for Disease Control (1998) National Nosocomial Infections Surveillance (NNIS) System report, data summary from October 1986-April 1998, issued June 1998. Am J Infect Control 26: 522–533.
- 32. Micek ST, Roubinian N, Heuring T, Bode M, Williams J, et al. (2006) Before-after study of a standardized hospital order set for the management of septic shock. Crit Care Med 34: 2707–2713.
- 33. Schramm GE, Johnson JA, Doherty JA, Micek ST, Kollef MH (2006) Methicillin-resistant Staphylococcus aureus sterile-site infection: The importance of appropriate initial antimicrobial treatment. Crit Care Med 34: 2069–2074.
- 34. Ammerlaan HS, Harbarth S, Buiting AG, Crook DW, Fitzpatrick F, et al. (2013) Secular trends in nosocomial bloodstream infections: antibiotic-resistant bacteria increase the total burden of infection. Clin Infect Dis 56: 798–805.
- 35. Society for Healthcare Epidemiology of America, Infectious Diseases Society of America, Pediatric Infectious Diseases Society (2012) Policy statement on antimicrobial stewardship by the Society for Healthcare Epidemiology of America (SHEA), the Infectious Diseases Society of America (IDSA), and the Pediatric Infectious Diseases Society (PIDS). Infect Control Hosp Epidemiol 33: 322–327.