The Relationship between Antimicrobial Resistance and Patient Outcomes: Mortality, Length of Hospital Stay, and Health Care Costs

Abstract

There is an association between the development of antimicrobial resistance in Staphylococcus aureus, enterococci, and gram-negative bacilli and increases in mortality, morbidity, length of hospitalization, and cost of health care. For many patients, inadequate or delayed therapy and severe underlying disease are primarily responsible for the adverse outcomes of infections caused by antimicrobial-resistant organisms. Patients with infections due to antimicrobial-resistant organisms have higher costs (∼$6,000–$30,000) than do patients with infections due to antimicrobial-susceptible organisms; the difference in cost is even greater when patients infected with antimicrobial-resistant organisms are compared with patients without infection. Strategies to prevent nosocomial emergence and spread of antimicrobial-resistant organisms are essential.

Awareness of the prevalence of antimicrobial resistance is growing among the medical community and the general public, and the impact of antimicrobial resistance on clinical and economic outcomes is the subject of ongoing investigation. An awareness of the effect of antimicrobial resistance on outcomes has several potential benefits. First, knowledge about the implications of resistance with regard to patient outcomes may prompt hospitals and health care providers to begin and support initiatives to prevent such infections (e.g., infection-control programs and antimicrobial agent management programs). Second, data can be used to influence health care providers to follow guidelines about isolation and to make rational choices with regard to the use of antimicrobial agents. Third, data can guide policy makers who make decisions about the funding of programs to track and prevent the spread of antimicrobial-resistant organisms. Fourth, such knowledge may stimulate interest in developing new antimicrobial agents and therapies. Finally, information about resistance may be important in defining the prognosis for individual patients with infection. In the present article, methodological issues that influence the results of studies of antimicrobial resistance outcomes will be acknowledged, and associations between resistance in specific pathogens and adverse outcomes, including increased mortality, length of hospital stay, and cost, will be reviewed.

Methodological Issues In Studies Of Antimicrobial Resistance Outcomes

Various methodological issues can influence the conduct and results of studies of antimicrobial resistance outcomes, as discussed in detail elsewhere [1–3]. The types of outcomes examined, the perspective of the study, the reference groups within the study, adjustments for confounding factors, and the type of economic assessment are among the factors that should be considered (table 1) [2].

With regard to outcomes, morbidity and cost, rather than mortality, may be the most sensitive measures with which to quantify the impact of antimicrobial resistance. The perspective of an outcome study determines the end points measured and affects how the economic impact of infection with resistant organisms is estimated. The cost for individual patients (relevant to the perspective of the hospital or third-party payers) pales in the face of the societal impact, which was estimated to be in the billions of dollars a decade ago [4]. Some of the most important influences on the patient and society, such as the gradual loss of efficacy of antimicrobial classes, are difficult to measure. It is essential to select the appropriate reference group (i.e., individuals infected with susceptible strains, colonized with resistant strains, or uninfected), control for the length of hospital stay, and adjust for the severity of the underlying illness and comorbidities before infection, because each of those factors can have a significant effect on outcomes measures.

Outcomes Of Infections With Antimicrobial-Resistant Gram-Positive Pathogens

Methicillin-resistantStaphylococcus aureus(MRSA). The impact of methicillin resistance on mortality rates among patients infected with S. aureus has been studied primarily in patients with bacteremia, and results of studies have varied [5–17]. To further address this issue, we conducted a meta-analysis of studies with relevant mortality data published between 1980 and 2000 [18]. When data from all studies (31 cohort studies including 3963 patients [34% of whom were infected with methicillin-resistant strains]) were pooled with a random-effects model, a significant increase in mortality associated with MRSA bacteremia, relative to methicillin-susceptible S. aureus (MSSA) bacteremia, was observed (OR, 1.93; P < .001). In subgroup analyses conducted to explore heterogeneity in the pooled analysis, mortality associated with MRSA infection was consistently higher, with minimal or no significant heterogeneity in each group. These analyses included studies adjusted for potential confounding variables, versus nonadjusted studies; studies with a high proportion of cases of nosocomial bacteremia (⩾70%) versus a low proportion (<70%); studies performed in an outbreak versus nonoutbreak setting; studies with a high proportion of catheter-associated infections (⩾40%) versus a low proportion (<40%); and studies with a high proportion of patients with endocarditis (⩾45%) versus a low proportion (<45%).

Length of hospital stay and costs related to MRSA bacteremia, compared with those related to MSSA bacteremia, were evaluated in 2 recently published cohort studies [19, 20]. In a study by our group, 346 patients admitted to the Beth Israel Deaconess Medical Center (Boston, MA) with clinically significant S. aureus bacteremia (96 case patients with MRSA infection and 252 control patients with MSSA infection) between 1997 and 2000 were evaluated. Among survivors, methicillin resistance was associated with significant increases in the median length of hospital stay after acquisition of infection (9 vs. 7 days for patients with MSSA bacteremia; P = .045) and hospital charges after S. aureus bacteremia ($26,424 vs. $19,212; P = .008). MRSA bacteremia was an independent predictor of increased length of hospitalization (1.3-fold increase; P = .016) and hospital charges (1.4-fold increase; P = .017). A second study prospectively evaluated 105 hemodialysis-dependent patients with S. aureus bacteremia who were admitted to Duke University Medical Center (Durham, NC) between 1996 and 2001 [20]. Thirty-four patients with MRSA infection were compared with 70 patients with MSSA infection. A propensity score for each patient's probability for having MRSA bacteremia, based on demographics, comorbidities, and APACHE II scores, was estimated using logistic regression and was used to adjust for confounding. Results for this population of patients undergoing hemodialysis were similar to those for the inpatient population at Beth Israel Deaconess Medical Center, with the adjusted median length of hospital stay longer (11 vs. 7 days; P < .001) and the adjusted median costs higher for the initial hospitalization ($21,251 vs. $13,978; P = .012) and after 12 weeks ($25,518 vs. $17,354; P = .015) for patients infected with MRSA.

Engemann et al. [21] evaluated clinical and economic outcomes attributable to methicillin resistance in a retrospective cohort study of patients with S. aureus surgical site infections primarily associated with cardiac or orthopedic procedures. During 1994–2000, 121 patients with a surgical site infection due to MRSA and 165 patients with a surgical site infection due to MSSA were identified, and another 193 uninfected patients, matched by type and year of surgical procedure, were selected. The investigators controlled for underlying severity of illness by use of the National Nosocomial Infection Surveillance risk index variables (American Society of Anesthesiologists score, duration of surgery, and wound class). The authors reported an independent contribution of methicillin resistance toward increased mortality, prolonged length of hospitalization, and increased hospital costs, which is consistent with the findings for bacteremia. The presence of MRSA in a surgical wound increased the adjusted 90-day postoperative mortality risk by 3.4-fold, compared with the presence of MSSA (P = .003), and by 11.4-fold, compared with the absence of infection (P < .001) (table 2). Patients with MRSA infection had mean attributable excess hospital charges of $13,901 and $41,274, compared with patients with MSSA infection and patients without infection, respectively. The design of this study shows the impact of choice of control groups on results.

Data on outcomes for patients with community-associated MRSA infection are limited. Martinez-Aguilar et al. [22] reported their findings from a retrospective study of 59 children with musculoskeletal infections (e.g., osteomyelitis, septic arthritis, and pyomyositis) caused by community-associated S. aureus (31 with MRSA infection and 28 with MSSA infection). The durations of fever (4.9 vs. 1.5 days; P = .001) and of hospital stay (14.5 vs. 12.7 days; P = .014) were significantly longer in the children infected with MRSA than in children infected with MSSA. The Panton-Valentine leukocidin gene, which encodes for a toxin of the same name that has been associated with leukocyte destruction and tissue necrosis, was found more frequently among the MRSA strains (87% of MRSA strains vs. 24% of MSSA strains; P < .001). Also of note, S. aureus isolates containing the Panton-Valentine leukocidin gene were associated with a greater proportion of complications (30.3% vs. 0%; P = .002).

Vancomycin-resistant enterococci (VRE). VRE were first isolated almost 2 decades ago [23] and have since become important nosocomial pathogens for which there are limited treatment options [24]. VRE infections have been shown to have a negative impact on mortality and cost of hospitalization [25–27]. For example, in a retrospective, cohort study of patients hospitalized between 1993 and 1997, Carmeli et al. [28] compared the health and economic outcomes of patients colonized or infected with VRE (n = 233; 42% had VRE in wounds, 31% had VRE in urine, 17% had VRE in intra-abdominal sites, and 9% has VRE in blood) with those of control subjects (percentages do not total 100 because of rounding; n = 647) without VRE colonization or infection who were matched for length of hospital stay, hospital ward, and calendar date (within 7 days). A propensity score was calculated to adjust for the risk of having VRE infection or colonization and was included in a multivariate analysis of each outcome. Compared with a similar but uninfected hospitalized cohort, patients with VRE infection had increased mortality (adjusted attributable mortality rate, 6%; adjusted relative risk [RR], 2.1; P = .04), length of hospital stay (attributable excess hospitalization, 6.2 days; multiplicative effect, 1.73; P < .001), and hospital costs (attributable cost, $12,766; multiplicative effect, 1.4; P < .001). Morbidity was also significantly higher among patients infected with VRE (e.g., 2.7-fold increased odds of undergoing a major surgical procedure and 3.5-fold increased odds of being admitted to an intensive care unit [ICU]).

Penicillin- and cephalosporin-resistantStreptococcus pneumoniae. In distinct contrast to the results of studies of staphylococci and enterococci, infection due to nonsusceptible S. pneumoniae has not been shown to adversely affect outcomes in most studies. For instance, in a prospective, international, observational study of 844 hospitalized patients with blood cultures positive for S. pneumoniae, discordant therapy (i.e., use of an antimicrobial agent that was classified as inactive in vitro) with penicillins, cefotaxime, and ceftriaxone did not result in a higher mortality rate, a longer time to defervescence, or more-frequent suppurative complications [29]. In studies of patients with cefotaxime-resistant pneumococcal meningitis [30] and bacteremic pneumonia [31], there were no differences in mortality, length of hospitalization, or need for admission to an ICU among case patients relative to matched control patients infected with susceptible isolates. Aggressive empirical use of vancomycin, favorable pharmacodynamics (i.e., most isolates had a cefotaxime MIC <4 µg/mL, a level reached or exceeded by cefotaxime in CSF and lung tissue), and community-acquired infection in otherwise healthy patients may explain these results.

Outcomes Of Infections With Resistant Gram-Negative Pathogens

Antimicrobial-resistantPseudomonas aeruginosa. Carmeli et al. [32] published one of the first studies to address outcomes associated with antimicrobial resistance in gram-negative pathogens. The study population included 489 patients with P. aeruginosa infection who were hospitalized between 1994 and 1996; at baseline, 144 (29%) had an isolate resistant to ceftazidime, ciprofloxacin, imipenem, and/or piperacillin, and 30 (6%) developed resistance during therapy. For 37% of the patients, the isolate recovered at baseline was nosocomially acquired. Culture specimens were obtained from wounds (41%), urine (23%), the respiratory tract (22%), effusion (5%), blood (5%), and tissue (4%). There were no differences in mortality or length of hospital stay between patients infected with a resistant isolate at baseline and those infected with a susceptible isolate at baseline (table 3). In contrast, the emergence of resistance was associated with a 3-fold greater risk of death (P = .02) and a 1.7-fold longer duration of hospital stay (P < .001). The estimated mean adjusted increase in duration of hospitalization was 5.7 days. The emergence of resistance was also associated with an increased risk of secondary bacteremia (14% vs. 1.4% in patients without emergence of resistance; RR, 9.0; P < .001). The investigators found no differences in hospital charges between any of the groups. The results of this study underscore the impact of emergence of resistance on patient outcomes.

Enterobacterspecies resistant to third-generation cephalosporins. Enterobacter species are common nosocomial pathogens, with almost one-third of strains causing third-generation cephalosporin—resistant infections in patients in the ICU [24]. In a nested, matched cohort study of patients admitted to Beth Israel Deaconess Medical Center between 1994 and 1997 [33], our research group evaluated the impact of emergence of resistance to third-generation cephalosporins on patient outcomes. Case patients (n = 46) had an initial culture that yielded Enterobacter species susceptible to third-generation cephalosporins and a subsequent culture from which a resistant strain was recovered. Reference patients from whom only susceptible Enterobacter strains were recovered (n = 113) were matched to case patients on the basis of the site of Enterobacter infection (including respiratory tract [44% of total], wounds [20%], effusion [18%], blood [13%], and urine [5%]) and length of hospitalization prior to isolation of the susceptible strain, with control patients required to have a length of hospital stay of at least the same duration as the time to isolation of a resistant strain for the matched case patients. Emergence of antimicrobial resistance in Enterobacter species resulted in significantly increased mortality (RR, 5.02), length of hospital stay (1.5-fold increase), and hospital charges (1.5-fold increase) (table 4). The median attributable duration of hospital stay due to emergence of resistance was 9 days, and the mean attributable hospital charge was $29,379.

Extended-spectrumβ-lactamase—producing (ESBL)Escherichia coliandKlebsiella pneumoniae. ESBL E. coli and K. pneumoniae were initially reported as causes of outbreaks [34, 35] and have become endemic in recent years (causing ∼7% of infections in ICU and non-ICU settings during 1998–1994) [24]. In a retrospective matched cohort study, Lautenbach et al. [36] evaluated outcomes in patients with E. coli or K. pneumoniae infection who were hospitalized at the University of Pennsylvania Medical Center (Philadelphia, PA) during 1997–1998. The study population included 33 case patients (i.e., patients infected with ESBL-producing isolates) and 66 control patients (i.e., patients infected with non–ESBL-producing isolates), who were matched to case patients on the basis of the species of the infecting organism (for case patients, 76% of isolates were K. pneumoniae), the anatomical site of infection (for case patients, 52% of infections were in the urinary tract, 15% were in wounds, 12% were in catheters, 9% were in blood, 9% were in the respiratory tract, and 3% were in abdominal sites), and the date of isolation. Exposure to antimicrobial agents was the only independent predictor of ESBL-producing E. coli or K. pneumoniae (OR for each additional day of antimicrobial therapy, 1.1; P = .006). Infection with ESBL-producing E. coli or K. pneumoniae was an independent predictor of higher median hospital charges subsequent to infection (1.7-fold increase), a higher mortality rate, and a longer length of hospital stay (table 5). Although mortality and length of hospitalization were greater for patients with ESBL-producing organisms, these 2 outcome measures did not reach the level of statistical significance because of the small sample size. These findings suggest that resistance need not increase mortality to have a dramatic impact on the cost of care.

Why Does Resistance Affect Outcomes?

Factors related to the host, the organism, and the treatment may contribute to increases in mortality, length of hospitalization, and costs associated with infection with resistant organisms. With regard to the host, severity of the underlying disease may be synergistic with infection with resistant organisms. Alternatively, some researchers argue that an inability to properly control for severity of the underlying illness may lead to the observed differences in outcomes.

Although increased virulence could explain the adverse impact of resistant pathogens on clinical outcomes, to date, no studies have demonstrated such an association, except for community-acquired MRSA. No existing evidence suggests that VRE strains are more virulent than vancomycin-susceptible strains, and resistance in gram-negative bacilli may actually reduce their fitness [37]. Similarly, in studies of health care—associated infection, MRSA has not been shown to be more virulent than MSSA [38, 39]. In contrast, there is some evidence to suggest that community-acquired MRSA is more virulent than health care—associated MRSA, on the basis of its shorter doubling time and the higher proportion of isolates with Panton-Valentine leukocidin gene and other exotoxin genes [38, 39]. Given the influence of antimicrobial resistance in the community on that in hospitals, the increased virulence of community-acquired MRSA is worthy of concern and certainly requires further study.

Treatment factors may contribute to adverse outcomes in patients infected with a resistant pathogen. These factors include (1) decreased effectiveness [40–42], increased toxicity [43], and/or improper dosing [44] of antimicrobial agents available for treatment; (2) a delay in treatment with or the absence of microbiologically effective antimicrobials; and (3) an increased need for surgery and other procedures as a result of these infections.

Mortality rates are higher among patients with ventilator-associated pneumonia who receive inappropriate empirical treatment (i.e., mismatch between the in vitro activity of the agent and the subsequent susceptibility results of the infecting pathogen) [45]. This association between inappropriate treatment and increased mortality has also been observed with other infections. For example, in a study of 167 patients with nosocomial S. aureus bacteremia during 1999 to 2001, Lodise et al. [46] found that, compared with prompt treatment, delayed treatment was an independent predictor of infection-related mortality (mortality rate, 33.3% vs. 19.3%; OR, 3.8; P = .01) and was associated with a longer duration of hospitalization after bacteremia (20 vs. 14 days; P = .05). Methicillin resistance was the most significant predictor of delayed appropriate treatment (OR, 8.3; P < .001). The same group of investigators found that receipt of inappropriate treatment also explained the increased length of hospital stay for patients with VRE bacteremia [47]. Similar associations have been observed for resistant gram-negative infections. In the study by Lautenbach et al. [36], time to effective therapy for infections due to ESBL-producing strains was ∼6-fold longer than that for infections caused by non–ESBL-producing strains (72 vs. 11 h). In addition, in a study of 85 episodes of ESBL-producing K. pneumoniae bacteremia, Paterson et al. [48] observed that failure to treat with an appropriate antimicrobial agent (i.e., one with in vitro activity against ESBL-producing K. pneumoniae) resulted in a significantly greater mortality rate (64% vs. 14% for patients who received an appropriate antimicrobial agent; OR, 10.7; P = .001).

Longer length of hospital stay and higher costs of care for patients infected with a resistant organism may also result from an increased frequency of surgical interventions required to control infection. Several groups of investigators have documented an increased need for surgery among patients infected with resistant organisms [28, 49, 50]. In a case series of 22 patients without cystic fibrosis who were infected with multidrug-resistant P. aeruginosa, Harris et al. [49] found that 89% of patients with clinical infection required surgery (e.g., debridement of infected tissue with or without revascularization), and 30% of patients required amputation. In the study by Carmeli et al. [28], patients with wound or abdominal infections caused by VRE were significantly more likely to require surgery, compared with patients without VRE infection (adjusted RR, 2.7; P = .001). A second study of patients infected with Enterococcus faecium also demonstrated that invasive interventions for intra-abdominal and intrathoracic infections were required more frequently in the cohort infected with a vancomycin-resistant strain (76% vs. 49% of the patients infected with a vancomycin-susceptible strain; P = .01).

In conclusion, there is an association between the development of resistance in S. aureus, enterococci, and gram-negative bacilli and increases in mortality, length of hospitalization, and costs of health care. This association is likely the result of inadequate or delayed therapy and may be related to the degree of severity of the underlying disease (with the exception of community-acquired MRSA). Patients with infections due to antimicrobial-resistant organisms have higher costs (∼$6,000–$30,000) than do patients with infections due to antimicrobial-susceptible organisms; the difference in cost is even greater when patients infected with antimicrobial-resistant organisms are compared with patients without infection. Thus, strategies to prevent the nosocomial emergence and spread of antimicrobial-resistant organisms are essential.

Acknowledgments

Potential conflicts of interest. S.E.C. has served on a scientific review panel for Cubist Pharmaceuticals.

References