The Impact of Antimicrobial-Resistant, Health Care–Associated Infections on Mortality in the United States

The prevalence of antimicrobial resistance is increasing in the United States and Europe [1]. However, the impact of antimicrobial resistance on mortality is difficult to measure, because the relationship between infection and death is complex. Among the many factors that influence this relationship are patient age, comorbidities, severity of illness, adequacy of empirical therapy, history of antibiotic use, and treatment with invasive devices [2, 3]. Researchers who evaluate the impact of antimicrobial resistance on mortality are encouraged to restrict evaluations to a specific organism in a standard setting with use of a validated system [4].

The National Nosocomial Infection Surveillance System (NNIS) was used by the Centers for Disease Control and Prevention (CDC) and participating hospitals to monitor health care–associated infections (HAIs) in hospitals from 1970 through 2004 [5]. The most frequently reported HAI in the NNIS system was urinary tract infection. The mortality associated with urinary tract infection is relatively low [6]. In contrast, health care–associated pneumonias and bloodstream infections are less frequent, but they together caused 67% of the estimated 99,000 deaths related to HAI in 2002 [6]. In this study, we selected specific organisms associated with these 3 types of HAI to evaluate the proportion of HAI that caused or contributed to a patient's death by antimicrobial resistance status and to determine temporal changes in these proportions between the periods 1990–1994 and 2000–2004. We also compared the proportion of pathogens that were resistant to key antimicrobial agents associated with these infections in the 2 time periods.

Methods. We analyzed data from the NNIS, a voluntary surveillance collaboration between selected US hospitals and the CDC [5]. NNIS data were collected using standardized protocols called surveillance components. All infections were grouped into major and specific infection sites using standard CDC definitions that included laboratory and clinical criteria. For analyses of NNIS data, we grouped HAIs into 5 major categories: urinary tract infections, surgical site infections, pneumonias, bloodstream infections, and others. We included infections reported from intensive care units only, regardless of whether they were device associated. For each infection, the associated pathogen and antimicrobial susceptibility pattern were collected. Among patients with an HAI who died after hospital admission at a participating NNIS facility, infection-control professionals used the NNIS protocol to classify the relationship of the death to each infection as caused, contributed, not related, or unknown.

We selected data reported during 1990–1994 and during 2000–2004 regarding primary bloodstream infections due to Staphylococcus aureus, pneumonias due to Pseudomonas aeruginosa, and urinary tract infections due to Escherichia coli in participating intensive care units. We dichotomized these infections on the basis of the reported antibiogram: susceptible or not susceptible. For each infection, we selected the first organism tested and reported. We tested the proportion of infections associated with antimicrobial-resistant pathogens across the 2 time periods by use of Mantel-Haenszel Χ² analysis, and we stratified the data according to 3 categories of hospital size (<200 beds, 200–500 beds, and >500 beds) and according to 3 patient age groups (aged 0–18 years, 19–64 years, and ⩾65 years). We used the Mantel-Haenszel relative risk and 95% CIs to calculate the relative risk of death caused or contributed to by antimicrobial-resistant infection among cases for which this outcome was determined and reported.

Results. Data were available from 207 hospitals participating in the NNIS during the period 1990–1994 and 292 hospitals participating during the period 2000–2004. During the period 1990–1994, a total of 2035 isolates from patients with bloodstream infections due to S. aureus were tested for susceptibility to methicillin (or oxacillin or nafcillin) and reported in the NNIS, compared with 4627 such isolates in the period 2000–2004. During the period 1990–1994, a total of 2736 isolates from patients with pneumonia due to P. aeruginosa were tested for susceptibility to ceftazidime and reported in the NNIS, compared with 3882 such isolates in the period 2000–2004. Finally, during the period 1990–1994, a total of 2073 isolates from patients with urinary tract infections due to E. coli were tested for susceptibility to ciprofloxacin and reported in the NNIS, compared with 3112 such isolates in the period 2000–2004. During the period 2000–2004, a total of 397 deaths were associated with bloodstream infections due to methicillin-resistant S. aureus (MRSA), 123 deaths were associated with pneumonia due to ceftazidime-resistant P. aeruginosa, and 43 deaths were associated with urinary tract infections due to ciprofloxacin-resistant E. coli; these cases all involved a known relationship between the HAI and the death (i.e., caused or contributed).

The proportion of infections due to antimicrobial-resistant organisms was significantly greater in the period 2000–2004 than in the period 1990–1994 for the 3 selected sites and pathogens (table 1). Specifically, the proportion of S. aureus bloodstream infections due to MRSA increased significantly overall, from 27.0% in the period 1990–1994 to 54.1% in the period 2000–2004 (P<.001); increases were greatest among persons aged ⩾65 years (30.8% vs. 65.4%; P<.001) and in hospitals with <200 beds (18.8% vs. 59.2%; P<.001). The proportion of P. aeruginosa pneumonia cases due to ceftazidime-resistant P. aeruginosa increased overall from 16.6% in the period 1990–1994 to 22.7% in the period 2000–2004 (P<.001), and the increase also was greatest in hospitals with <200 beds (17.3% vs. 29.5%; P=.004). However, among children aged 0–18 years, the proportion of infections due to antimicrobial-resistant organisms did not change between the 2 periods. The proportion of E. coli urinary tract infections due to ciprofloxacin-resistant E. coli increased overall from 0.9% to 9.8% (P<.001), but the number of cases during the period 1990–1994 was too small in certain categories to draw conclusions according to patient age and hospital size.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 1

Percentages of selected health care–associated infections due to antimicrobial-resistant organisms, by patient age and hospital size, during the periods 1990–1994 and 2000–2004.

Among patients with an HAI who died for whom any association between the death and the HAI was reported, the proportion in whom the HAI caused or contributed to death was consistently lower in the period 2000–2004 than in the period 1990–1994 for each of the 3 antimicrobial-resistant infections studied (table 2). In contrast, among patients with antimicrobial-susceptible infections, the risk of an associated death was lower in 2000–2004 only for those with bloodstream infections due to methicillin-susceptible S. aureus. We observed no significant difference between the relative risk of death due to a resistant versus a susceptible pathogen in the period 2000–2004, compared with the period 1990–1994, for any of the selected infections.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 2

Among patients with selected health care–associated infections (HAIs) who died, the risk that death was caused or contributed to by the HAI during the periods 1990–1994 and 2000–2004, by antimicrobial susceptibility status, National Nosocomial Infection Surveillance System.

Discussion. We observed increases in the proportion of resistant organisms for 3 specific HAIs. In the case of bloodstream infections due to S. aureus, this finding confirms other reports. A prospective study of health care–associated bloodstream infections reported a similar increase in the proportion of S. aureus infections due to MRSA, from 22% in 1995 to 57% in 2001 [7]. However, our findings are unique with regard to P. aeruginosa respiratory infections due to ceftazidime-resistant P. aeruginosa and E. coli urinary tract infections due to ciprofloxacin-resistant E. coli. Reports of trends in ceftazidime-resistant P. aeruginosa infection not limited to pneumonias indicated no change [8] or a slight decrease in the proportion of resistant organisms [9]. The proportion of E. coli infection due to ciprofloxacin-resistant E. coli not limited to the urinary tract increased from <1% to 17.5% in a study of patients in US intensive care units [9].

We observed a decrease in the probability that any death was associated with the resistant HAI for the 3 selected infections during the period 2000–2004, compared with the period 1990–1994. Mortality associated with infections due to S. aureus among hospitalized patients, measured independently, decreased nationally from 7.1% in 1998 to 5.6% in 2003 [10]. However, we did not observe an improvement in the relative risk of death in the period 2000–2004, compared with the period 1990–1994. Comparable studies are lacking. Regarding deaths associated with bloodstream infections due to MRSA, results from a large meta-analysis found wide variability in estimates, with the ORs for death ranging from <1 to 13.5 [11]. Our measure of the relative risk of an associated death between these groups is within those boundaries.

We initiated this study to evaluate whether the impact of antimicrobial resistance on mortality was higher during the period 2000–2004. We observed no change in the relative risk of death. The lack of improvements reducing the risk of death due to antimicrobial-resistant organisms is of great concern. In this study, we cannot evaluate specific reasons for the observed lack of improvements; antimicrobial resistance and complications are multifactorial. Factors that might influence progress include the global nature of the problem, factors inherent to the microorganisms, widespread use of antibiotics, and the lack of novel treatments [12].

Our data have several limitations. Data from NNIS hospitals may not be representative of US hospitals [13]. We could not evaluate the role of the many factors that influenced death as an outcome, such as patient-specific severity of illness or comorbidities, both of which are critical to evaluating attributable mortality rates [14]. We also did not collect information related to treatment or device management. The actual determination of the relationship between an HAI and death is difficult; however, in the NNIS system, surveillance was conducted by well-trained infection-control professionals, and their judgment is likely to provide a reliable estimate of mortality associated with these infections. A strength of our method was the selection of 1 organism per site of infection for 3 types of HAIs across many institutions [4]. Future endeavors to assess the impact of resistance on mortality over time might include other factors associated with outcome, especially adequacy of treatment.

Controlling antimicrobial resistance requires both good infection control and judicious antimicrobial use. The framework of the Campaign to Prevent Antimicrobial Resistance in Healthcare Settings [15] can be useful in designing and implementing programs to prevent antimicrobial resistance [16]. Increased efforts to prevent the development and transmission of resistant organisms in all US hospitals are essential.

• Received April 14, 2008.
• Accepted May 29, 2008.

• © 2008 by the Infectious Diseases Society of America