An institutional perspective on the impact of recent antibiotic exposure on length of stay and hospital costs for patients with gram-negative sepsis

Knowledge of the clinical and economic impact of antimicrobial resistance is useful to influence programs and behavior in healthcare facilities, to guide policy makers and funding agencies, to define the prognosis of individual patients and to stimulate interest in developing new antimicrobial agents and therapies [1]. One of the most misunderstood issues is how to measure cost appropriately when analyzing the antibiotic treatment of infections. Although imperfect, existing data show that there is an association between antimicrobial resistance in Staphylococcus aureus, enterococci and Gram-negative bacilli and increases in mortality, morbidity, length of hospitalization and cost of healthcare [1]. Patients with infections due to antimicrobial-resistant organisms have greater hospital costs (US$6000-30,000) than do patients with infections due to antimicrobial-susceptible organisms [1, 2]. Given limited budgets, imperfect knowledge of the clinical and economic impact of antibiotic-resistant bacterial infections, coupled with the benefits of specific interventions targeted to reduce these infections or to improve their treatment, more optimal management strategies to control and improve antibiotic utilization are needed. The traditional way of accounting for antimicrobial costs in terms of pharmacy acquisition expenditures for antibiotics may not be appropriate when discussing the treatment of antibiotic resistant infections. Instead, increasing evidence suggests that the most cost-effective approach for the treatment of such infections is to provide appropriate therapy in a timely manner [3‐6].

Inappropriate initial antimicrobial therapy, defined as an antimicrobial regimen that lacks in vitro activity against the isolated organism(s) responsible for the infection, has been associated with excess mortality and lengths of stay in patients with severe sepsis and septic shock [5, 7‐11]. This is largely related to increasing bacterial resistance to antibiotics as a result of their greater use and limited availability of new agents [12, 13]. Escalating rates of antimicrobial resistance lead many clinicians to empirically treat critically ill patients with presumed infection with a combination of broad-spectrum antibiotics, which can perpetuate the cycle of increasing resistance [14]. Conversely, inappropriate initial antimicrobial therapy can lead to treatment failures and adverse patient outcomes [15]. Therefore, clinicians need to be aware of when it is clinically and economically reasonable to use newer, broad-spectrum antibiotics. Prior antibiotic exposure may represent an important marker for decision making, especially as new broad-spectrum antibiotics begin to enter the clinical setting.

The most recent Surviving Sepsis Guidelines recommend empiric combination therapy targeting Gram-negative bacteria, particularly for patients with known or suspected Pseudomonas infections, as a means to decrease the likelihood of administering inappropriate initial antimicrobial therapy [16]. The choice of an antimicrobial regimen that is active against the causative pathogen(s) is increasingly problematic as the treating physician usually does not know the susceptibilities of the pathogen(s) for the selected empiric antibiotics at the time of antibiotic prescription. We previously showed that recent antibiotic exposure is associated with increased hospital mortality in Gram-negative bacteremia complicated by severe sepsis or septic shock, primarily due to inappropriate initial treatment [17]. Therefore, we performed an investigation to determine whether recent antibiotic exposure also resulted in increased LOS and hospital costs among patients with severe sepsis or septic shock attributed to Gram-negative bacteremia.

Seven hundred fifty-four consecutive patients with bacteremic Gram-negative severe sepsis or septic shock were included in the study. The mean age was 59.3 ± 16.3 years (range, 18-99 years) with 394 (52.3%) males and 360 (47.7%) females (Table 1). There were 421 (55.8%) medical patients and 333 (44.2%) surgical patients. The mean duration of hospitalization was 10.2 ± 14.4 days (range, 1-96) at the time severe sepsis or septic shock occurred. The mean APACHE II score was 23.7 ± 6.7 (range, 4-43) with the majority of patients requiring vasopressors (58.5%) and mechanical ventilation (55.3%). Escherichia coli (30.8%), Klebsiella pneumoniae (23.2%), Pseudomonas aeruginosa (17.6%), Enterobacter species (10.1%), Acinetobacter species (8.4%), Proteus mirabilis (4.9%), Serratia marcescens (4.0%), and Stenotrophomonas maltophilia (1.9%) were the most common organisms isolated from blood cultures. Fifty-six (7.4%) patients had polymicrobial bacteremia.

Three hundred-ten (41.1%) patients had prior antibiotic exposure during the preceding 90 days. Of these, cefepime was the most common agent with previous exposure (50.0%) followed by ciprofloxacin (32.6%), imipenem or meropenem (28.7%), pipercillin-tazobactam (19.0%), and aminoglycosides (14.5%). Most prior antibiotic exposure was during the same hospitalization and within 21 days of severe sepsis onset for patients with healthcare-associated hospital-onset infections (77.9%). Among patients with healthcare-associated community-onset infections prior antibiotic exposure was either outpatient administered (56.7%) or administered during a hospitalization occurring within 90 days of infection onset (43.3%). When compared to cases with no prior antibiotic exposure, patients with prior exposure were significantly more likely to have healthcare-associated hospital-onset sepsis, sepsis occur in the intensive care unit setting, and a longer duration of stay prior to sepsis onset (Table 1). Patients with prior antibiotic exposure were also significantly younger, had lower Charlson co-morbidity scores, were more likely to have a pulmonary source of infection, and to require mechanical ventilation and vasopressor support.

Patients with prior antibiotic exposure had higher rates of inappropriate initial antimicrobial therapy (45.5% v. 21.2% p < 0.001) and hospital mortality (51.3% v. 34.0%, p < 0.001) compared to patients without prior antibiotic exposure. Inappropriate antibiotic therapy of severe sepsis was more likely to be associated with prior exposure to cefepime (37.4% v.12.9%; p < 0.001), ciprofloxacin (25.1% v. 8.1%; p < 0.001), ceftriaxone (7.2% v. 1.3%; p < 0.001), and imipenem (13.6% v. 5.4%; p < 0.001) compared to appropriate therapy. Similarly, prior exposure to cefepime, ciprofloxacin, and imipenem were associated with statistically longer hospital LOS.

Figure 1 reveals the probability of remaining hospitalized after the diagnosis of severe sepsis or septic shock as a function of prior antibiotic exposure. Subjects receiving prior antibiotic exposure had an unadjusted median increase of 5.0 days in LOS (13.0 days v. 8.0 days, p < 0.001). In a Cox model adjusting for multiple co-variates (see Table 2), prior antibiotic exposure remained linked with an increased probability of remaining hospitalized after the onset of Gram-negative sepsis. The adjusted hazard ratio related to prior antibiotic exposure equaled 1.473 (95% CI: 1.297-1.672). Other variables significantly correlated with remaining hospitalized included pre-infection length of stay, severity of illness, inappropriate antibiotic therapy, hospital-onset infection, and a pulmonary source of infection. After controlling for the other factors displayed in Table 2, we estimate that prior antibiotic exposure independently increased hospital length of stay by approximately 5.0 days.

Figure 2 depicts the hospital costs following the onset of Gram-negative sepsis for the main cost centers examined. Total median hospital costs were significantly greater for patients with prior antibiotic exposure compared to patients without prior antibiotic exposure (median values: $94,737 [25^th and 75^th percentiles: $53,941; $136,878] v. $21,329 [25^th and 75^th percentiles: $11,403; $41,461]; p < 0.001). LOS and total hospital costs following the onset of Gram-negative sepsis were strongly correlated (Spearman's rank correlation coefficient = 0.947; p < 0.001).

When only hospital survivors were examined, subjects receiving prior antibiotic exposure had an unadjusted median increase of 7.0 days in LOS (16.0 days v. 9.0 days, p < 0.001). Similarly, when only hospital survivors were examined total median hospital costs were significantly greater for patients with prior antibiotic exposure compared to patients without prior antibiotic exposure (median values: $93,764 [25^th and 75^th percentiles: $52,892; $134,310] v. $21,369 [25^th and 75^th percentiles: $11,981; $39,760]; p < 0.001).

Our study suggests that prior antibiotic exposure was associated with greater LOS following the onset of Gram-negative bacteremia complicated by severe sepsis or septic shock. This observation was confirmed in an adjusted analysis of LOS and in a multivariate analysis controlling for potential confounding variables. Not surprisingly, other important determinants of prolonged LOS in the multivariate analysis included pre-infection onset hospital LOS, inappropriate antibiotic therapy, severity of illness, and the lungs as the source of infection.

A likely explanation for the association we observed between hospital LOS and prior antibiotic exposure is the greater occurrence of antimicrobial resistance among the causative pathogen(s) associated with sepsis among patients having received prior antibiotic therapy. We previously demonstrated that prior antibiotic exposure was associated with a greater incidence of infection attributed to Gram-negative bacteria that were either antibiotic-resistant or multi-drug resistant [17]. Prior antibiotic exposure was also associated with greater administration of inappropriate initial antimicrobial therapy which has been linked with excess mortality and longer lengths of stay in septic patients [3‐7]. Shorr et al. evaluated the impact of a sepsis management protocol that emphasized identification of septic patients, aggressive fluid resuscitation, timely antibiotic administration, and appropriateness of antibiotics [6]. Use of the sepsis protocol was associated with lower hospital mortality and substantial institutional cost savings. The sepsis protocol significantly increased the use of appropriate initial antibiotic therapy, despite the initial use of potentially more costly antibiotic regimens, and represents a strategy for enhancing resource utilization while containing overall costs in hospitalized patients with sepsis [6].

The findings from our study are consistent with those observed by other investigators. Webster et al. recently described resistance trends in Escherichia coli and Klebsiella pneumoniae bloodstream infections in Oxfordshire, UK, over an 11 year period [22]. In their univariate analysis multidrug resistance (OR 2.77, 95% CI 1.23-6.22, p = 0.014), previous hospital admission (OR 1.27, 95% CI 1.01-1.60, p = 0.041) and isolation of Klebsiella pneumoniae (OR 1.53, 95% CI 1.05-2.22, p = 0.026) predicted a longer hospital LOS, as did age 7 years (OR 1.58, 95% CI 1.26-1.99, p < 0.0005). On multivariate analysis, multidrug resistance remained a strong independent predictor of increased LOS. Similarly, Lautenbach et al. examined the longitudinal trends in prevalence of imipenem-resistant Pseudomonas aeruginosa (IRPA) from 2 centers from 1989 through 2006 [23]. Isolation of IRPA was associated with longer hospital LOS after culture (p < .001) and greater hospital costs after culture (p < .001) than was isolation of an imipenem-susceptible strain. In multivariable analysis, IRPA infection or colonization remained an independent risk factor for both longer hospital LOS after culture and greater hospital costs after culture. Although these investigators did not specifically examine prior antibiotic exposure as a risk factor, antibiotic-resistant bacterial infection represents a good marker for having had such exposure occur [12, 13].

Our study is unique in identifying prior antibiotic exposure as an independent risk factor for prolonged LOS among patients with Gram-negative bacteremia complicated by severe sepsis or septic shock. This observation suggests that clinicians should search for and identify the presence of prior antibiotic exposure as an important consideration when prescribing empiric antibiotic therapy to patients with severe sepsis or septic shock. The identification of prior antibiotic exposure should result in specific therapeutic interventions. For example, clinicians should avoid antibiotics or antibiotic classes to which the patient was previously exposed. Additionally, severity of illness appears to play an important role in determining which patients are at greatest risk for adverse outcomes from inappropriate antibiotic therapy [17]. Patients with shock or neutropenia, especially those with prior antibiotic exposure, may benefit the most from initial combination therapy that includes an aminoglycoside in order to minimize the occurrence of inappropriate initial antimicrobial therapy [24, 25].

There are several important limitations of our study that should be noted. First, the study was performed at a single center and the results may not be generalizable to other institutions. However, the findings from other investigators corroborate the potential role of prior antibiotic exposure as an important determinant of outcome, including LOS, for patients with serious Gram-negative infections [5‐7, 24]. Second, the retrospective and observational nature of this study limits our ability to establish causality between prior antibiotic exposure and hospital LOS. The retrospective design also restricts our ability to identify all prior antibiotic exposure, especially outpatient exposure to antimicrobial agents. However, we are confident that the majority of all antibiotic exposure in the 90 days prior to severe sepsis onset was captured by our data base, especially as it captures cli9nical data from both inpatient and outpatient facilities. Third, our study may not have identified all of the factors contributing to length of stay in this patient population. Fourth, we did not evaluate the number of days preceding hospital admission when prior antibiotic exposure occurred nor did we assess the total number of days of prior antibiotic exposure. These could be other important variables that influence patient outcomes. Lastly, the presence of prior antimicrobial exposure may not be easily identified, especially in patients without easily accessible medical records describing recent outpatient and inpatient treatments. This is an important limitation for using the risk factor of prior antibiotic exposure in daily patient management decision making.

In conclusion, we observed that prior antibiotic exposure is relatively common and associated with adverse outcomes including greater hospital LOS among patients with Gram-negative bacteremia complicated by severe sepsis or septic shock. This observation suggests that clinicians treating patients with suspected Gram-negative bacteremia or sepsis should attempt to identify whether prior antibiotic exposure occurred. In clinical situations where recent antibiotic exposure is likely, but details concerning prior antibiotic exposure are unknown, combination empiric therapy directed against Gram-negative bacteria would be reasonable to increase the likelihood of appropriate therapy, until susceptibility data become available. Given the importance of prior antibiotic exposure as a risk factor for antibiotic resistance, inappropriate therapy, and increased mortality, and the availability of electronic medical records at many hospitals, institutions should try to formalize an approach for identifying prior antibiotic exposure in patients with serious infections.