Comparative effectiveness of antibiotic therapy for carbapenem-resistant Enterobacterales (CRE) bloodstream infections in hospitalized US veterans

Abstract

Background

Carbapenem-resistant Enterobacterales bloodstream infections (CRE-BSI) increase mortality three-fold compared with carbapenem-susceptible bloodstream infections. Because these infections are rare, there is a paucity of information on mortality associated with different treatment regimens. This study examines treatment regimens and association with in-hospital, 30 day and 1 year mortality risk for patients with CRE-BSI.

Methods

This retrospective cohort study identified hospitalized patients within the Veteran Affairs (VA) from 2013 to 2018 with a positive CRE blood culture and started antibiotic treatment within 5 days of culture. Primary outcomes were in-hospital, 30 day and 1 year all-cause mortality. Secondary outcomes were healthcare costs at 30 days and 1 year and Clostridioides difficile infection 6 weeks post culture date. The propensity for receiving each treatment regimen was determined. Multivariable regression assessed the association between treatment and outcomes.

Results

There were 393 hospitalized patients from 2013 to 2018 included in the study. The cohort was male (97%) and elderly (mean age 71.0 years). Carbapenems were the most prescribed antibiotics (47%). In unadjusted analysis, ceftazidime/avibactam was associated with a lower likelihood of 30 day and 1 year mortality. After adjusting, ceftazidime/avibactam had a 30 day mortality OR of 0.42 (95% CI 0.17–1.02). No difference was found in C. difficile incidence at 6 weeks post-infection or total costs at 30 days or 1 year post culture date by any treatments.

Conclusions

In hospitalized veterans with CRE-BSI, none of the treatments were shown to be associated with all-cause mortality. Ceftazidime/avibactam trended towards protectiveness against 30 day and 1 year all-cause mortality. Use of ceftazidime/avibactam should be encouraged for treatment of CRE-BSI.

Introduction

Carbapenem-resistant Enterobacterales (CRE) infections have been classified as an urgent threat by the CDC. CRE causes approximately 1100 infections each year with an 8% mortality rate and contribute approximately $130 million dollars in additional healthcare costs annually.¹ CRE bloodstream infections (CRE-BSI) mortality rates are much higher, estimated at 65%, owing in part to the limited treatment options.² A meta-analysis of bloodstream infections found that mortality was 3-fold higher among those with CRE-BSI compared with carbapenem-susceptible Enterobacterales bloodstream infections.³ Risk factors for CRE-BSI include hospital admission in the previous 90 days (particularly ICU admission), mechanical ventilation, indwelling catheter devices, previous antibiotic use, and chronic comorbidities such as chronic kidney disease or solid organ transplantation.^4,5

During the study time period there were no clinical guidelines that specifically addressed treatments for CRE infections. Two systematic reviews indicated that carbapenems, aminoglycosides, extended-spectrum cephalosporins and polymyxins were commonly used to treat these infections.^6,7 However, there were few data about the effectiveness of these treatments because of a lack of randomized controlled trials evaluating the outcome of treatment. Recently, IDSA recommended CRE-BSI be treated with carbapenem/β-lactamase combination antibiotics (meropenem/vaborbactam or imipenem/relebactam), or cephalosporin/β-lactamase combination antibiotics (ceftazidime/avibactam).⁸ The use of meropenem is also recommended in cases where the isolate is not carbapenemase producing and only ertapenem resistant. However, because CRE-BSI is a rare disease, there is a lack of real-world studies associating treatment regimens and mortality outcomes. The objective of this study was to assess treatment regimens prescribed for CRE-BSI and to compare the effectiveness of these therapies on in-hospital, 30 day and 1 year all-cause mortality. Secondary outcomes evaluated were cause-specific mortality, incidence of Clostridioides difficile infection (CDI) following treatment, and cost associated with infection and treatment.

Methods

Study design and setting

A retrospective cohort study was conducted of Department of Veterans Affairs (VA) hospitalized patients with a positive CRE blood culture between 1 January 2013 and 31 December 2018. Patients who met the following criteria were excluded: positive cultures from a specimen source other than blood, cultures collected outside of the inpatient setting, cultures that did not meet the VA definition of CRE (defined below), and cultures not associated with treatment within the defined window (2 days prior to 5 days post culture). Only the first CRE-positive culture collected during the study period was evaluated for each patient.

Data sources

The VA’s electronic data repository, the Corporate Data Warehouse (CDW), was used to obtain all healthcare data including hospitalization, pharmacy and microbiological culture information. Mortality data were obtained from both the joint Veterans Administration/Department of Defense Mortality Data Repository and the Centers for Disease Prevention and Control National Death Index. Cost data were obtained from the VA National Data Extracts Managerial Cost Accounting Office. Electronic health record reviews of 50 CRE-BSI patients (13% of cohort) were also conducted to verify the definitions of CRE and primary versus secondary BSI, and treatment(s) administered. Secondary BSI was defined as a positive culture from urine, respiratory or gastrointestinal tract, wound/skin/soft tissue, or catheter site within the 2 days prior to the positive blood culture. If there was no such previous culture, the positive blood culture was considered primary BSI.⁹

Covariates

Demographic variables including age, race and sex were extracted from the CDW. Pre-existing conditions noted in the 365 days before the culture date were collected to assess the Charlson comorbidity index, a predictor of 1 year mortality. The APACHE III scores were calculated to assess the severity of illness associated with the CRE-BSI event; clinical values obtained ±2 days from the culture date were used. The following factors were collected because of their association with CRE-BSI: urinary tract infection (UTIs), mechanical ventilation, steroid use and immunosuppressant agent used in the previous 90 days.⁵ UTIs were defined by a positive urine culture accompanied by a documented fever of at least 38.0°C, or an ICD9/10 code specific for UTIs. Hospitalization length of stay was collected for the 365 days prior to and after the culture date.

Microbiological cultures and CRE identification

Microbiological cultures that met the VA definition for CRE were included. CRE were defined as Escherichia coli, Klebsiella spp. or Enterobacter spp. isolates (a) resistant to imipenem, meropenem or doripenem,^10,11 and/or (b) resistant to ertapenem only and any tested third-generation cephalosporin.^11,12 Isolates that were only resistant to ertapenem were excluded. BSIs were defined as blood cultures that exhibited growth with an organism of interest with antibiotic susceptibilities consistent with CRE as described above. Bacterial species identification and antibiotic susceptibility testing was performed by each VA medical centre’s (VAMC’s) clinical laboratory according to local protocol and entered into the electronic medical record, from which CDW datasets are derived.

Antimicrobial treatment definition and timing

Antibiotics that were administered from 2 days before the culture date to 5 days post culture were recorded. This treatment window was defined based on previous literature that the median time to initiation of definitive therapy was 3 days post culture.¹³ Treatment administered from 2 days before to 2 days after the culture date was classified as empirical therapy. Definitive therapy was defined as treatment that was administered from 3 to 5 days post culture date. Antibiotics were grouped according to the American Hospital Formulary Service classification, which is a standard metric.¹⁴ The following classes of antibiotics were included: aminoglycosides, carbapenems, polymyxins, fluoroquinolones and extended-spectrum cephalosporins. Extended-spectrum cephalosporins were defined as third- and fourth-generation cephalosporins. Combination therapy was defined as treatment that included the use of two antimicrobials, one with intracellular and one with extracellular activity against the infectious organism based on the mechanism of action. Extended-spectrum cephalosporins, polymyxins, ceftazidime/avibactam and carbapenems were defined as having extracellular activity and aminoglycosides and fluoroquinolones were defined as having intracellular activity.

Adequacy of treatment and other covariates

Inadequate treatment was defined as agents administered despite the organism cultured being non-susceptible to the agent. Adequacy was categorized as either completely adequate, where all given therapy was appropriate for the organism cultured, partially adequate, where at least one of the antibiotics given was appropriate for the organism cultured, or inadequate therapy, where none of the agents given were appropriate given the organism cultured. In cultures where susceptibility testing for a particular antibiotic was not performed, treatment was presumed to be adequate.

Study outcomes

The primary study outcomes were in-hospital, 30 day and 1 year all-cause mortality. In-hospital mortality was defined as the number of patients who died in hospital within the associated hospital admission. Thirty-day and 1 year mortality was defined as the number of patients that died within 30 days and 365 days of the culture date, respectively. Secondary outcomes included cause-specific mortality, secondary CDI and cost of treatment. Cause-specific mortality was defined as those whose cause of death was listed as ‘infectious or parasitic’ via ICD-10 code, compared with those who survived their infection. Secondary C. difficile infection was defined as a positive laboratory (molecular or toxin assay) diagnostic CDI test that occurred within 6 weeks of the CRE culture date. Inpatient, outpatient and pharmacy costs associated with each treatment regimen were assessed separately and combined at both 30 days and 1 year post culture date to determine cost effectiveness associated with a particular treatment regimen.

Statistical analyses

Wilcoxon rank sum, chi-squared and Fisher’s exact tests were used to assess the relationship between each treatment regimen and medical, demographic and facility-level variables. Because the likelihood of receiving a certain antibiotic is not equal across all patients, the propensity of receiving a given treatment was calculated using logistic regression on the outcome of treatment, clustered within patient. Covariates used in propensity score calculations included demographics (gender, race, age etc.), culture characteristics (year, organism, source, previous culture), Charlson comorbidity score, and exposures in the 90 days prior to culture date (surgery, antibiotic, systemic corticosteroid, immunocompromised, inpatient admission, outpatient visit, ICU, mechanical ventilation, UTI). All-cause and cause-specific in-hospital mortality, 30 day mortality, and 1 year mortality were assessed using pooled logistic regression analysis. Unadjusted analyses and adjusted analyses were run for each of the antibiotic regimens included. Outcomes were adjusted for propensity score and adequacy of treatment. Cost analyses were conducted using generalized linear models with a gamma family and a log link function and robust standard errors clustered by patient. The number of CDI events 6 weeks post culture date were compared between treatment regimens and analysed using Wilcoxon rank sum tests. Data analysis was completed using SAS 9.4 and STATA 14.2.

Results

Overall cohort

There were 393 patients from 70 VAMCs included in the cohort. From the 50 charts randomly selected for review, there was 94% agreement on the definition of CRE, 90% agreement on treatment and 94% agreement on whether the infection was classified as primary versus secondary BSI. Overall, the number of CRE-BSI cultures decreased over time from 68 cultures in 2012 to 35 cultures in 2018.

The cohort was male (97%) and older [mean (SD) age 71.0 (12.1) years]. Hispanic persons were the largest racial/ethnic group, representing 37.7% of the cohort, followed by non-Hispanic white persons (33.1%) and non-Hispanic black persons (23.4%). The overwhelming majority of CRE were at high-complexity VAMCs (98%) and facilities outside the continental USA (35.6%), followed by the South (31.6%), Northeast (14.3%), West (10.4%) and Midwest (8.1%). Most cultures (68.2%) were from infections classified as primary BSI according to our study definition.

Treatments prescribed

Carbapenems were the most frequently prescribed (47.6%) antibiotics and ceftazidime/avibactam the least (9.9%). Figure 1 shows the percentage of usage for types and classes of antibiotics included in the 2020 IDSA guidelines for treatment of CRE. Ceftazidime/avibactam, which was FDA-approved and added to the VA formulary in 2015, was utilized for 9% of the positive cultures in 2015 and 46% of the cultures in 2018, a 5-fold increase over the 4 years.

Figure 1.

Percentage of antibiotics prescribed for CRE-BSI by year, 2012–18. Ceftazidime/avibactam was FDA-approved and added to the VA formulary in 2015.

Treatments prescribed varied by patient demographics (Table 1). Patients who received polymyxins were more likely to have an antibiotic exposure in the previous 30 days (Table 1) compared with those who did not receive polymyxins; this relationship was not observed with other treatment regimens. Most treatment regimens were defined as being partially adequate, ranging from 77.8% to 89.7%. Ceftazidime/avibactam was not considered inadequate treatment for any treatment regimen. Extended-spectrum cephalosporins had the highest rates of inadequacy of treatment (13.3%).

Table 1.

Demographics and medical characteristics of veterans with CRE-BSI by antibiotic treatment (n = 393)

	Antibiotic treatment, N (%)
Characteristics	Carbapenems n = 187 (47.6%)	Aminoglycosides n = 151 (38.4%)	Meropenem n = 115 (29.3%)	Ceftazidime/avibactam n = 39 (9.9%)	Polymyxins n = 130 (33.6%)	Fluoroquinolones n = 124 (31.6%)	Extended-spectrum cephalosporins n = 135 (34.4%)	Combination therapy n = 241 (61.3%)
Age, years (%)
18–49	3.7	2.6	6.1	5.1	4.5	4.0	6.7	3.3
50–64	30.5	31.1	29.6	17.9	22.0	24.2	18.5	26.6
65+	65.8	66.2	64.3	76.9	73.5	71.8	74.8	70.1
Race (%)
White	27.8	31.8	36.5	23.1	13.6	43.5	31.9	34.0
Black	23.5	29.1	28.7	23.1	15.9	29.8	22.2	27.8
Latino	43.3	33.8	28.7	51.3	67.4	21.8	40.0	32.4
Other/missing	5.4	5.3	6.0	2.6	3.0	4.8	5.9	5.8
Charlson, mean (SD)	3.8 (3.1)	4.9 (3.0)	3.9 (3.2)	4.5 (2.9)	3.7 (2.9)	3.7 (3.1)	4.0 (3.0)	3.8 (3.0)
APACHE III, mean (SD)	33.4 (11.1)	33.6 (10.9)*	33.2 (11.8)	33.2 (9.9)	32.1 (10.9)	34.1 (11.7)	33.6 (12.3)	33.4 (1.2)
ICU in <90 days (%)	61.0**	52.3	60.0	46.2	56.8	50.8	51.1	53.5
Antibiotic use <30 days (%)	85.0	80.1	84.3	79.5	93.9**	75.8*	81.5	79.7
Mechanical ventilation <30 days (%)	33.7**	24.5	29.6	25.6	31.8	25.8	18.5*	24.5
UTI <30 days (%)	13.9	17.2	13.0	0.0**	23.5***	10.5	11.9	15.4
Appropriateness of treatment (%)
Total adequacy	7.5	9.9	7.8	10.3	6.8	12.1	8.9	7.15
Partial adequacy	80.7	83.4	80.0	89.7	88.6	79.8	77.8	85.1
Inadequate	11.8	6.6	12.2	0.0	3.0	8.1	13.3	7.9

Presence for 1 day of the listed antibiotic during treatment was sufficient to be included in an antibiotic group. Each antibiotic group was compared with those who had none of that treatment during the observed treatment window; *significant at the 0.05 level; **P = 0.01; ***P ≤ 0.0001.

Mortality outcomes

In-hospital and 30 day mortality were comparable, at 37.4% and 35.9% of the cohort, respectively. African American, Hispanic and patients over the age of 65 years were all more likely to die than white patients and those under the age of 65 years (Table S1, available as Supplementary data at JAC Online). Two hundred and fifty patients (63.6%) of the cohort died within 1 year of the culture date. Unadjusted analysis found higher odds of in-hospital mortality for those receiving carbapenems and polymyxins compared with those without those treatments; however, after propensity score adjustment none of these associations remained statistically significant (Table 2).

Table 2.

Unadjusted and adjusted pooled logistic regression models for in-hospital mortality, 30 day mortality and 1 year mortality by antibiotic therapy (N = 393 patients)

	Unadjusted in-hospital mortality		Adjusted in-hospital mortality		Unadjusted 30 day mortality		Adjusted 30 day mortality		Unadjusted 1 year mortality		Adjusted 1 year mortality
Antibiotic therapya	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Carbapenems	1.69 (1.12–2.56)	0.012	1.29 (0.83–2.01)	0.260	1.78 (1.17–2.69)	0.007	1.48 (0.95–2.31)	0.086	1.71 (1.13–2.59)	0.012	1.26 (0.81–1.99)	0.309
Aminoglycosides	1.41 (0.93–2.14)	0.108	1.38 (0.39–4.91)	0.182	1.44 (0.94–2.19)	0.091	1.49 (0.95–2.34)	0.083	1.38 (0.90–2.13)	0.135	1.42 (0.89–2.25)	0.132
Extended-spectrum cephalosporins	0.77 (0.49–1.18)	0.228	0.99 (0.60–1.63)	0.978	1.13 (0.74–1.75)	0.570	1.25 (0.76–2.05)	0.380	0.83 (0.54–1.27)	0.392	1.27 (0.77–2.11)	0.349
Fluoroquinolones	0.80 (0.51–1.25)	0.326	1.05 (0.64–1.72)	0.850	0.54 (0.34–0.86)	0.009	0.72 (0.43–1.21)	0.219	0.58 (0.38–0.89)	0.015	0.62 (0.38–1.01)	0.055
Polymyxins	1.83 (1.19–2.82)	0.006	1.39 (0.78–2.51)	0.261	2.14 (1.39–3.30)	0.0006	1.77 (0.98–3.18)	0.058	2.45 (1.53–3.92)	0.0002	1.82 (0.98–3.39)	0.059
Meropenem	1.11 (0.71–1.73)	0.649	0.97 (0.60–1.57)	0.899	1.28 (0.82–2.01)	0.274	1.17 (0.73–1.89)	0.518	1.05 (0.67–1.65)	0.846	1.01 (0.62–1.64)	0.965
Ceftazidime/avibactam	0.47 (0.22–1.02)	0.056	0.46 (0.19–1.08)	0.075	0.43 (0.19–0.96)	0.039	0.42 (0.17–1.02)	0.055	0.45 (0.23–0.88)	0.019	0.56 (0.27–1.19)	0.135
Combination therapy	0.95 (0.62–1.44)	0.806	1.01 (0.65–1.58)	0.963	0.98 (0.64–1.49)	0.919	1.04 (0.66–1.62)	0.877	0.89 (0.59–1.37)	0.619	0.94 (0.60–1.47)	0.793

Bold cells represent models that are significant at the 0.05 level.

^aReference group for all comparisons are all antibiotics not associated with the respective treatment.

There was a 78% increase in all-cause 30 day mortality for those who were prescribed carbapenems, a 54% increase associated with fluoroquinolone use, and a more than 2-fold increase associated with polymyxins. However, after propensity score adjustment none of these associations remained significant. Ceftazidime/avibactam trended towards being protective for 30 day mortality after propensity score adjustment but ultimately did not reach statistical significance [OR = 0.42 (95% CI (0.17–1.02); P = 0.055]. Similar associations were seen between carbapenems, fluoroquinolones and polymyxins and 1 year mortality. However, again, after adjustment none of the associations remained significant.

Cause-specific mortality

Approximately 6% of the in-hospital deaths, 5% of the 30 day deaths and 7% of the 1 year deaths were attributed to infectious disease. After adjusting for treatment adequacy and propensity score, polymyxin use increased the odds of mortality due to infectious disease by 5.2 times for in-hospital mortality, 7.0 times for 30 day mortality and 6.5 times for 1 year mortality (Table 3). There were no other associations observed for cause-specific mortality.

Table 3.

Unadjusted and adjusted pooled logistic regression models for infectious disease-related mortality by antibiotic therapy

	Unadjusted in-hospital mortalitya		Adjusted in-hospital mortality		Unadjusted 30 day mortality		Adjusted 30 day mortality		Unadjusted 1 year mortality		Adjusted 1 year mortality
Antibiotic therapy	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Carbapenems	2.04 (0.87–4.66)	0.102	1.84 (0.76–4.45)	0.179	2.03 (0.80–5.15)	0.134	1.90 (0.70–5.18)	0.208	2.20 (0.98–4.96)	0.057	1.67 (0.69–3.98)	0.251
Aminoglycosides	0.86 (0.36–2.07)	0.737	1.01 (0.39–2.61)	0.986	1.22 (0.48–3.09)	0.674	1.43 (0.52–3.97)	0.451	1.04 (0.45–2.42)	0.924	1.24 (0.50–3.04)	0.646
Extended-spectrum cephalosporins	0.82 (0.34–1.97)	0.649	0.97 (0.35–2.70)	0.956	1.64 (0.65–4.10)	0.294	1.59 (0.55–4.61)	0.388	0.89 (0.39–2.07)	0.793	1.40 (0.51–3.87)	0.514
Fluoroquinolones	0.50 (0.18–1.38)	0.181	0.63 (0.21–1.89)	0.412	0.44 (0.14–1.36)	0.155	0.52 (0.15–1.77)	0.294	0.41 (0.16–1.06)	0.065	0.44 (0.15–1.25)	0.122
Polymyxins	2.72 (1.19–6.26)	0.018	5.16 (1.59–16.79)	0.006	2.65 (1.06–6.65)	0.038	7.00 (1.82–26.89)	0.005	3.37 (1.47–7.73)	0.004	6.52 (1.87–22.78)	0.003
Meropenem	0.98 (0.39–2.44)	0.962	1.17 (0.43–3.18)	0.753	1.14 (0.42–3.08)	0.801	1.12 (0.38–3.31)	0.841	0.95 (0.39–2.31)	0.906	1.01 (0.39–2.61)	0.982
Ceftazidime/avibactam	0.63 (0.14–2.79)	0.539	0.49 (0.09–2.54)	0.398	0.79 (0.18–3.58)	0.762	0.59 (0.11–3.14)	0.539	0.67 (0.19–2.42)	0.541	0.72 (0.17–2.98)	0.650
Combination therapy	0.41 (0.18–0.96)	0.039	0.54 (0.22–1.35)	0.188	0.42 (0.17–1.06)	0.065	0.49 (0.18–1.36)	0.174	0.48 (0.21–1.07)	0.073	0.18 (0.02–2.12)	0.284

Bold cells represent models that are significant at the 0.05 level.

^aReference group for all comparisons are all antibiotics not associated with the respective treatment.

Secondary outcomes

Only 25 (6.4%) CRE-BSI cultures were associated with a positive CDI test in the 6 weeks following antibiotic treatment for CRE-BSI. Almost half [48.0% (12 of 25 CDI cases)] were associated with fluoroquinolones (P = 0.067). None of the CRE-BSI cultures associated with positive CDI testing were associated with ceftazidime/avibactam treatment.

Although there was no statistical difference in unadjusted average total costs 30 days post culture date (Table S2) by treatment regimen, aminoglycosides had the lowest total average 30 day cost, ($43 695.50) and carbapenem the highest ($48 352.60). The total cost of patients receiving polymyxins was $47 231.00 and ceftazidime/avibactam was $45 242.90. However, there were differences in the components of total cost by antibiotic regimen. Fluoroquinolones had higher 30 day average outpatient costs ($954.80) compared with other regimens. Polymyxins and carbapenems had lower outpatient costs compared with other regimens ($420.20 and $497.70, respectively). CRE cultures associated with extended-spectrum cephalosporin treatment had the lowest total average cost of $205 383.20 and aminoglycosides had the highest average cost of $234 200.60, although these differences were not statistically significant (Table S3).

Generalized linear models including a propensity score were estimated to examine the association between the type of antibiotic used and costs. After adjusting for the propensity score, use of aminoglycosides was associated with higher outpatient costs at 30 days compared with use of other antibiotics. Polymyxin pharmacy costs were higher at 1 year when compared with other antibiotics. No other treatment regimen comparisons were significant in the adjusted models (Tables S4 and S5).

Discussion

Prior to the 2015 approval and widespread uptake of newer anti-CRE antibiotics, carbapenems and polymyxins were the most frequently used and recommended treatments for CRE-BSI.^7,15–17 Consistent with these data, our study found that hospitalized patients with CRE-BSI in the VA between 2013 and 2018 were frequently treated with carbapenems and polymyxins. However, between 2015 and 2018, carbapenem use decreased and ceftazidime/avibactam use increased, with ceftazidime/avibactam use being associated with 46% of the CRE-BSI cultures by 2018. This trend has also been observed in other national cohorts. A prior study analysing antibiotic prescription data for patients with CRE infections from a cohort of US hospitals showed declines in polymyxin and carbapenem use and increases in new anti-CRE agent use (including ceftazidime/avibactam) between 2015 and 2019.^15,18

Use of polymyxins in our study was associated with higher odds of all-cause mortality, while ceftazidime/avibactam had lower odds of all-cause mortality; however, these associations did not remain significant after adjustment. Ceftazidime/avibactam was associated with a reduction in 30 day and 1 year all-cause mortality. However, these associations did not reach significance, possibly due to the low percentage of cultures associated with that treatment (9.9%). These results agree with work published by others demonstrating ceftazidime/avibactam to be highly effective against CRE, with clinical cure rates reported as high as 85%.^13,19 Additionally, no patients treated with ceftazidime/avibactam in our study had a positive CDI culture within 6 weeks of CRE and, in general, these patients had the lowest total costs within 30 days of CRE. With these positive clinical outcomes combined with a lack of increase in cost, these results support the inclusion of ceftazidime/avibactam in the 2020 IDSA guidelines as a preferred treatment and it should be considered by clinicians as a treatment of choice in patients with CRE-BSI.^20,21

For cause-specific mortality, polymyxins were the only treatment associated with all three mortality outcomes. The reason for this is unclear as those who received polymyxin were not significantly sicker than those who did not receive polymyxins according to the co-morbidity and severity of illness scores included. Additionally, the treatment adequacy of polymyxins (based on susceptibility data) was equally as likely as the other regimens. However, polymyxin use has been associated with more adverse drug events, specifically acute kidney injury, than other antibiotics and this could have indirectly contributed to the increased mortality observed with the use of this antibiotic.²²

Fluoroquinolones were the only antibiotic found to be associated with CDI in this population. Considering that the use of this antibiotic was not strongly associated with a reduction in mortality, use of fluoroquinolones for persons with CRE-BSI should be limited. Additionally, fluoroquinolones are not recommended by the IDSA guidelines for the treatment of CRE infections.

There are several limitations to this study. We were not able to evaluate the use of newer β-lactam/β-lactamase combination antibiotics such as ceftolozane/tazobactam, imipenem/relebactam or meropenem/vaborbactam due to FDA approval occurring late in our study period. Antibiotic regimens were dynamic; patients often had daily antibiotic changes that occurred during both the empirical and definitive treatment time windows. Consequently, distinguishing the impact of a single antibiotic regimen was not possible. The specific genetic mechanism for carbapenem resistance was not identified and this could be a factor in treatment outcome. We did not capture treatments received outside our review window of 2 days before to 5 days after the culture date. Therefore, full duration of treatment was not a factor in our analysis. We were also unable to measure differences in dosage between the different treatments. We also may not have been able to properly distinguish between primary and secondary BSI because of the definition used. In addition, we were unable to determine the primary site of infection in cases where bacteraemia was determined to be secondary. Lastly, the number of patients that received certain antibiotics, namely ceftazidime/avibactam, was limited, which reduced our ability to detect a true difference.

However, there are also several strengths to this study. First, this was an in-depth national analysis of a large cohort of patients with this rare infection. Second, we were able to evaluate the use of the most frequent treatments for CRE-BSI to determine antibiotics associated with mortality. Third, the demonstrated agreement between electronic health record reviews and CDW data abstraction meant that our algorithm developed to define these complex infections was successful. Last, our study evaluated rarely reported secondary outcomes of CDI and cost.

Conclusions

In this study of 393 hospitalized VA patients with CRE-BSI, carbapenems and polymyxins were associated with higher mortality in the unadjusted analysis. However, after propensity score adjustment no associations between treatment and mortality were observed. There were no major differences in total costs with any of the antibiotics used, suggesting that treatment regimens can be selected based on clinical success rather than cost concerns.

Funding

This work was supported by The Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Health Services Research and Development, IIR 16-028 (PI: Charlesnika Evans) and RCS 20-192 (Charlesnika Evans) and Rehabilitation Research and Development Career Development Award B2826-W (Margaret Fitzpatrick).

Transparency declarations

The authors report that there are no conflicts of interest to declare.

Disclaimer

The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the US government.

Supplementary data

Tables S1 to S5 are available as Supplementary data at JAC Online.