Epidemiology of early-onset bloodstream infection and implications for treatment Richard S. Johannes, MD, MS Marlborough, Massachusetts Health care-associated infections: For over 35 years, infections have been divided into hospital acquired or community acquired. In 2002, in a study of bloodstream infections (BSIs), Friedman et al first suggested creating a new classification: health careassociated BSIs. Kollef et al furthered the concept of health care-associated infection in a 2005 population-based study of culture-positive pneumonia cases. Although the site of infection differed, Kollef et al s results supported Friedman et al s original concept. Then in 2006, Kollef et al reported a population-based study focused specifically on BSIs. Of 6697 reported cases, 468 (7%) had hospital-acquired BSIs; 3705 (55.3%) health care-associated BSIs; and 2524 (37.7%) community-acquired BSIs. The clinical features of those with health care-associated BSIs differed from those with community-acquired BSIs. For several organisms, including Staphylooccus aureus, Streptococcus pneumoniae, and gram-negative organisms, the frequencies for health care-associated and hospital-acquired BSIs were similar to each other but significantly different from community-acquired BSIs. After controlling for several clinical features, methicillin-resistant Staphylococcus aureus had the largest odds ratio for predicting in-hospital mortality. Both hospital-acquired and health care-acquired cases were independent risk factors for in-hospital mortality. Implications for treatment: Is more aggressive, empiric, gram-positive therapy warranted for this potentially sicker patient group? Wunderink pointed out the potential unintended consequences of such an approach and the paucity of good tools for early recognition of sickest patients. A study by Shorr et al of systemic inflammatory response syndrome, organ dysfunction, and mortality suggested that there may be approaches that could be used to stratify cases into high-risk groups who may benefit from more aggressive therapy. Most recently, Micek et al found that in health care-associated pneumonia cases, inappropriate initial empiric antibiotic treatment is an independent predictor of mortality. Treatment recommendations are evolving. Summary: For pneumonia and BSIs, health care-associated infections appear to be distinct entities. However, operational definitions still vary. Compared with hospital-acquired cases, health care-associated cases have different clinical characteristics. The outcomes of health care-associated infections tend to be intermediate of the community-acquired and hospital-acquired groups. Further research is urgently needed on the implications of health care-associated infection for early therapy. (Am J Infect Control 2008;36:S171.e17-S171.e21.) In this brief report, the emerging concept of health care-associated infection will be explored as a clinical classification distinct from community-acquired infections and hospital-acquired infections. Several lines of evidence that bear on health care-associated infections will be reviewed, including a recent collaborative study involving Cardinal Health Clinical Research Services that focused on health care-associated bloodstream infection (BSIs). Implications for treatment will also be considered. From Clinical Research Services, Cardinal Health, MediQual Services, Marlborough, MA. Address correspondence to Richard S. Johannes, MD, MS, Cardinal Health, MediQual Clinical Research Services, Marlborough, MA 01752. E-mail: dick.johannes@cardinalhealth.com. Disclosures: Dr. Johannes is an employee of Cardinal Health. 0196-6553/$34.00 Copyright ª 2008 by the Association for Professionals in Infection Control and Epidemiology, Inc. doi:10.1016/j.ajic.2008.10.003 METHODS Health care-associated infections For more than 35 years, infections have been divided into 2 categories: hospital acquired (nosocomial infections), ie, those acquired during hospitalization, and community-based infections, ie, those acquired outside the hospital setting. In 2002, based on analysis of data from 3 North Carolina hospitals, Friedman et al 1 suggested that, for BSIs, the community-acquired infections might be split into 2 groups: the traditional community-associated group and a group termed health care-associated BSIs. Patients with health care-associated BSIs represented a new group characterized by close physical associations with hospitals and related health care settings. Friedman defined cases of health care-associated BSIs as patients with BSIs who within 30 days of hospitalization had selfadministered intravenous therapy, had received intravenous care at home, or wound care through a skilled nursing care or a home health care agency. This seminal study of 504 patients suggested that those with health care-associated infections shared many features S171.e13
S171.e14 Vol. 36 No. 10 Johannes with those with hospital-acquired infections with regard to the spectrum of organisms and outcomes. One of the largest population-based studies related to the concept of health care-associated infection was conducted by Kollef et al 2 in collaboration with Cardinal Health Clinical Research Services. Rather than focusing on BSIs, this study, published in 2005, examined 4543 culture-positive pneumonia cases across 59 hospitals that occurred during the first 5 days of hospitalization. Although the specific infection in Kollef et al s study (pneumonia) differed from that in Friedman et al s early hypothesis (BSIs), the findings were conceptually similar. In particular, the mortality rates for the health care-associated and hospitalacquired groups were similar to each other and higher than the rate for the community-acquired group. In addition, as in the Freidman et al study, the distribution of culture results showed an increased frequency of Staphylococcus aureus and methicillin-resistant S aureus (MRSA) in the hospital-acquired and health care-associated groups compared with the community-acquired group. These findings supported the concept of health care-associated infection but, because of its focus on pneumonia, did not shed new light on health care-associated BSIs. Health care-associated BSIs The same group of investigators then used the same population-based data set, the Cardinal Health Outcomes Research database, to examine early-onset, health care-associated BSIs. 3 As in the pneumonia study, the cases from the same 59 hospitals, from January 1, 2002, to December 31, 2003, formed the study population. Although the concept of health care-associated infection has gained acceptance, firm rules for case definition have not yet been established. In this study, culture results were known for the first 5 days, thus, positive cultures represent early-onset infection. To put this into perspective, recently published results on the distribution of timing of acquisition of hospitalacquired infections, suggest that two thirds of cases occur before hospital day 6. 4 Using the case definitions shown in Table 1, 6697 cases of culture-positive BSIs were identified as hospital-acquired, health care-associated, or community-acquired BSIs. RESULTS Of 6697 cases of BSIs, 468 (7%) were hospital-acquired BSIs; 2524 (37.7%) were community-acquired BSIs, and 3705 (55.3%) were health care-associated BSIs, making health care associated BSIs the largest of the 3 groups. Clinical features differed across the 3 groups. The hospital-acquired BSI cases were younger than the health care associated or community-acquired BSI cases. The health care-associated BSI cases had the highest representation of Medicare payment. Compared with the community-acquired BSI cases, those with health-care associated BSIs were more likely to be insulin dependent; treated with immunosuppressive medications; and have comorbidities such as heart failure, renal failure, or cancer. On hospital admission, the patients in the health care-associated BSIs cases were more likely to have hypotension; hypothermia; alteration in mental status; and laboratory evidence of hypoalbuminemia, azotemia, and acidemia compared with the community-acquired BSI cases. Distribution of organisms The distribution of organisms isolated from blood cultures showed that, for several, organisms the frequencies for health care-associated and hospitalacquired BSIs were similar to each other and significantly different from community-acquired BSIs. As shown in Table 2, this pattern was seen for all S aureus, Streptococcus pneumoniae, and gram-negative organisms. Mortality rates The unadjusted mortality rates for communityacquired, health care-associated, and hospital-acquired BSIs were 10.0%, 14.9%, and 15.0%, respectively. Unadjusted median and mean lengths of stay for community-acquired, health care-associated, and hospitalacquired BSI were 6.0 and 7.7 days, 6.0 and 8.1 days, and 10.0 and 12.8 days, respectively. To control for severity, a multivariable logistic regression analysis was conducted. Age, 5 comorbidities, vital signs, alteration in mental status, and 11 laboratory results all showed statistical significance at the univariate level and were entered into a model, along with indicator variables for 6 of the pathogens and the infection categories hospital-acquired BSIs and health care-associated BSIs. After controlling for these clinical variables, the organism with the highest independent odds ratio for predicting in-hospital mortality was MRSA. Compared with community-acquired BSIs, the mortality risk for both health care-associated and hospital-acquired BSIs was significantly higher (Table 3). Implications for treatment This work, published in Critical Care Medicine in 2006, was accompanied by an editorial by Richard Wunderink, MD, 5 from Northwestern University. Wunderink cautioned that this sort of distinction, which suggests that patients with health care-associated BSIs are a
Johannes December 2008 S171.e15 Table 1. Case definitions Bloodstream infection category Hospital-acquired BSIs Health care-associated BSIs Definition Patients with a first positive blood culture.2 days and #5 days from admission* Patients with first positive blood culture #2 days of admission and any of the following*: Admission source indicates a transfer from another health care facility Receiving chronic dialysis Prior hospitalization within 30 days Metastatic cancer and currently on immunosuppressive medication Community-acquired BSIs Patients with first positive blood culture #2 days from admission* who do not meet hospital-acquired BSIs or health care-associated BSI definitions *With presence of either primary or secondary ICD9 code indicating a BSI. Table 2. Distribution of isolates CAB, n (%) HCAB, n (%) HAB, n (%) HCAB vs CAB, n (%)* HCAB vs HAB, n (%)* CAB vs HAB, n (%)* All Staphylococcus aureus 449 (17.8) 952 (25.7) 139 (29.7),.0001.1311,.0001 Gram negative 200 (7.9) 182 (4.9) 11 (2.9),.0001.1396,.0001 Fungal 1233 (48.9) 1607 (43.3) 179 (38.2),.0001.1678,.0001 Total No. of isolates 2524 3705 436 CAB, Community-acquired BSI; HCAB, health care associated BSI; HAB, hospital-acquired BSI. *Pairwise comparisons following Bonferroni adjustment. potentially sicker group with a high frequency of grampositive organisms and MRSA within the gram-positives, might lead to an overly aggressive, broader-spectrum, empiric antibiotic use with possible unintended adverse consequences. Wunderink also pointed to the paucity of good tools for early recognition of the sickest patients. Although this is clearly sage advice and caution is needed, is it possible that more aggressive, empiric, gram-positive therapy is warranted? Recent data pertain to this question. First, many clinical studies have used systemic inflammatory response syndrome and organ dysfunction to stratify cases by severity. At the 46th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICCAC) meeting in September 2006, our group presented on the relationship among systemic inflammatory response syndrome, organ dysfunction, and mortality. 6 This study also was based on the same population later used to subclassify the hospital-acquired BSI cases. The evidence showed that dysfunction in one or more major organs, with or without systemic inflammatory response syndrome, was associated with mortality rates equal to or greater than 15% (Fig 1). These findings suggest that there may be approaches that could be used to stratify cases into high-risk groups that might benefit from more aggressive empiric antibiotic therapy. In addition, our group presented a poster at ICAAC 7 in September 2007 showing that, between 2002 and 2005, the inpatient mortality rate for patients with gram positive-only BSIs stayed at nearly 20%, 7 a rate approximately twice that of gram-negative-only infection. Most recently, Micek et al 8 reported the results of a single center experience from Washington University in St. Louis that examined outcomes from health careassociated pneumonia. This study of 639 culturepositive pneumonia cases extended the time of prior hospitalization to 1 year and had the additional benefit of detailed clinical information available through chart review. The investigators found that the health careassociated pneumonia group was nearly twice the size of the community-acquired pneumonia group. A major contribution in this study was the inclusion of data regarding the appropriateness of initial empiric antibiotic treatment. For example, in the health careassociated pneumonia and MRSA group, nearly 30% of the initial empiric antibiotic regimens were characterized as inappropriate. The investigators went on to perform a regression analysis predicting inpatient mortality that included age, health care-associated pneumonia status, requirement for mechanical ventilation, bacteremia, race, Klebsiella species, and an indicator for inappropriate antibiotic therapy. The indicator for inappropriate empiric antibiotic therapy proved to be
S171.e16 Vol. 36 No. 10 Johannes Table 3. Risk factors associated with mortality in patients with BSIs Variable No. cases Crude mortality rate, % Odds ratio (95% CI) P value Organisms MRSA 422 22.5 2.70 (2.03-3.58),.0001 MSSA 1118 16.4 1.90 (1.53-2.36).0094 Fungal (non-candida) 64 21.9 2.66 (1.27-5.58).0116 Candida 69 26.1 2.32 (1.21-4.45),.0001 Polyorganism 400 20.0 1.69 (1.24-2.30).0009 Pseudomonas 189 18.0 1.60 (1.04-2.47).0340 BSI categories Hospital-acquired BSIs 468 15.0 1.81 (1.31-2.51).0004 Health care-associated BSIs 3705 14.9 1.24 (1.03-1.49).0218 MSSA, methicillin-susceptible S aureus. 40% 35% 37.2% 30% Mortality % 25% 20% 15% 15.6% 17.5% 28.6% 10% 12.8% 5% 2.4% 7.5% 0% All and No OD No OD 1 OD and 1 OD and > 1 OD > 1 OD n=6552 n = 679 (10.4%) n = 3311 (50.5%) n = 1460 (22.3%) n = 245 (3.7%) n = 70 (1.1%) n = 787 (12.0%) * p<0.0001 vs. and No OD Fig 1. The relationship among systemic inflammatory response syndrome, organ dysfunction, and inpatient mortality. an independent predictor of mortality with an odds ratio of 2.19 (95% confidence interval: 1.27-3.78). The combination of a large fraction of health careassociated cases, MRSA being a dominant organism, gram-positive organisms having a high mortality risk, and a substantial fraction of cases receiving inappropriate initial antibiotic therapy suggest that it might be reasonable to expand the aggressiveness of empiric antibiotic therapy in the health care-associated group. The proceedings of the Health Care-Associated Pneumonia Summit were recently published and provide a more detailed appraisal of therapy options. 8 DISCUSSION Although operational definitions of health care-associated still vary in the literature, health care-associated infections appear to be a distinct entity for both pneumonia and BSIs. Patients with this type of infection have different clinical characteristics yet similar outcomes compared with the traditional hospital acquired-infection population. In most cases, the outcomes of health care-acquired cases are intermediate of the traditional community-acquired and hospitalacquired groups. The implications of health careassociated infections for structuring recommendations for early therapy may become a pressing issue. Further research in this area is urgently needed. REFERENCES 1. Friedman ND, Kaye KS, Stout JE, McGarry SA, Trivette SL, Briggs JP, et al. Health care-associated bloodstream infections in adults: a reason to change the accepted definition of community-acquired infections. Ann Intern Med 2002;137:791-7. 2. Kollef MH, Shorr A, Tabak YP, Gupta V, Lui LZ, Johannes RS. Epidemiology and outcomes of health care-associated pneumonia: results from a large US database of culture-positive pneumonia. Chest 2005;128: 3854-62.
Johannes December 2008 S171.e17 3. Shorr AF, Tabak YP, Killian AD, Gupta V, Lui LZ, Kollef MH. Healthcareassociated bloodstream infection: a distinct entity? Insights from a large US database. Crit Care Med 2006;34:2588-95. 4. Kilgore ML, Ghosh K, Beavers CM, Wong DY, Hymel PA Jr, Brossette SE. The costs of nosocomial infections. Med Care 2008;46:101-4. 5. Wunderink RG. Healthcare-associated bacteremia: stirring the mud. Crit Care Med 2006;34:2685-6. 6. Shorr AF, Gupta V, Tabak YP, Killian AD, Johannes RS, Kollef MH. Epidemiology of systemic inflammatory response syndrome (SIRS) and organ dysfunction in culture positive bacteremia patients. Poster presentation at the 46th Interscience Conference on Antimicrobial Agents and Chemotherapy in San Francisco, CA, September 27-30, 2006. 7. Shorr AF, Peng MM, Tabak YP, Hyde CA, Spalding J, Gupta V. Trends of culture confirmed candidemia on admission from 2000-2005: analysis of 64,330 culture-positive blood stream infections. Poster presentation at the 47th Interscience Conference on Antimicrobial Agents and Chemotherapy in Chicago, IL, September 17-20, 2007. 8. Micek ST, Kollef KE, Reichley RM, Roubinian N, Kollef MH. Health careassociated pneumonia and community-acquired pneumonia: a single-center experience Epub August 6, 2007. Antimicrob Agents Chemother 2007;51:3568-73. 9. Kollef MH, Morrow LE, Baughman RP, Craven DE, McGowen JE Jr, Micek ST. Health care-associated pneumonia (HCAP): a critical appraisal to improve identification, management, and outcomes. Proceedings of the HCAP Summit. Clin Infect Dis 2008;46(Suppl):S296-334.