Complicated carbapenem-resistant infections: a treatment pathway analysis in Italian sites

SUMMARY

Introduction

Efforts to curb a growing prevalence of carbapenem resistance are prominent worldwide and especially in countries where high levels of carbapenem resistance are reported, such as Italy. Complicated infections, including complicated urinary tract infections (cUTI), complicated intra-abdominal infections (cIAI), and hospital-acquired/ventilator-associated bacterial pneumonia (HABP/VABP), are often caused by carbapenem-resistant Gram-negative (CRGN) bacteria and as such, these infection sites and their causative bacteria are important areas of focus for healthcare practitioners seeking to follow good antimicrobial stewardship practices. The aim of this study was to assess the clinical management and associated clinical and economic outcomes of patients with cUTI, cIAI, and HABP/VABP resulting from CRGN bacteria in Italy.

Methods

We first conducted a hospital survey focusing on Gram-negative infections and their antibacterial susceptibility profile in four participating Italian hospitals. The second part of the study involved a non-interventional, retrospective single cohort chart review of 100 patients with cUTI, cIAI, or HABP/VABP caused by CRGN bacteria, in which patient characteristics, index hospitalization characteristics, infection characteristics, patient outcomes, treatment pathways, and healthcare resource use were assessed.

Results

The hospital survey demonstrated carbapenem resistance in approximately 17% of complicated infections, mostly associated with Acinetobacter baumannii. The non-interventional, retrospective cohort component showed that complicated CRGN infections were hospital- or healthcare-acquired in 99.0% of cases and were most often caused by Klebsiella pneumoniae (66.0%). Despite the carbapenem-resistant nature of the included infections, carbapenems were used in 19.0% of patients as empirical therapy, in 43.0% as late empirical (i.e. immediately before receipt of susceptibility test results), and in 64.0% as targeted therapy (post-susceptibility test result receipt). Colistin was used in 61.0% of patients after susceptibility results were available. High clinical and economic burden was evident, with the average length of hospital stay being greater than 50 days, clinical cure achievement in only 43.0% of patients, and an overall mortality rate of 65.0% by the end of the follow-up period.

Conclusion

Our results reflect the considerable burden associated with complicated CRGN infections in Italy and the limitations in current treatment strategies. Our study pinpoints potential areas for improvement. For example, regular and detailed local surveillance and state of the art microbial diagnostic capabilities might aid and hasten clinical decision-making and facilitate improved antimicrobial stewardship when treating complex CRGN infections. New therapeutic options which more appropriately address CRGN infections may assist in improving outcomes which are important to both patients and healthcare providers.

INTRODUCTION

Antibacterial resistance is recognized as a worldwide issue, associated with increased length of hospital stay, mortality, and overall healthcare costs [1]. Efforts to treat antibacterial-resistant infections properly and to prevent further development of antibacterial resistance have been the focus of many governments and organizations in recent years [2]. This is of particular importance when considering resistance to hitherto frequently used classes of antibacterial agents such as carbapenems. Several carbapenem-resistant pathogens are categorized as “critical” in the World Health Organization’s global priority list because of their alarming spread, especially in southern European countries [3, 4] [2]. Carbapenem resistance has been reported to be high in Italy: in 2017 19.9% of P. aeruginosa, 29.7% of K. pneumoniae and 78.7% of A. baumannii were resistant [3]. Furthermore, KPC-producing K. pneumoniae are considered to be endemic in the country and account for nearly all (95.1%) of the carbapenemase-producing Enterobacterales bloodstream infections [5, 6]. There has also been a recent emergence of bacterial strains with resistance to colistin, an antibacterial agent that can be used as a last-resort option for carbapenem-resistant (CR) infections but with considerable limitations, such as difficulty in achieving optimal dosing, high levels of renal toxicity, and problems in determining bacterial susceptibility [7, 8]. There is a difficult balance to strike when trying to treat patients with complicated infections both quickly and appropriately, and limitations in available treatments due to bacterial resistance may further complicate clinical decision-making [9]. A better understanding of current diagnostic and treatment practices is needed to help identify potential areas for improvement in terms of improved antimicrobial stewardship and clinical and economic outcomes for CR Gram-negative (CRGN) infections in Italy. The objective of this study was to characterize the clinical management of patients diagnosed with complicated urinary tract infections (cUTI), complicated intra-abdominal infections (cIAI), and hospital-acquired/ventilator-associated bacterial pneumonia (HABP/VABP) attributable to CRGN bacteria in Italy.

PATIENTS AND METHODS

Four hospitals participated in the study; the sites were all large (greater than 600 beds) teaching facilities. Carbapenem resistance was defined as per the European Committee on Antimicrobial Susceptibility Testing (EUCAST) susceptibility cutoff values in use at the time of culture collection [10]. Resistant or intermediate isolates were defined as CR and permitted inclusion into the study.

Preliminary survey

Hospital microbiological surveillance data from 2015–2017 were evaluated through a questionnaire. The number of cUTI, cIAI, and HABP/VABP infections per year, associated causative pathogens, and antibacterial resistance profiles were collected, by infection site.

Retrospective observational study

A non-interventional, retrospective, single cohort chart review was conducted to assess patients with cUTI, cIAI, or HABP/VABP attributable to CRGN bacteria in the same hospitals. Patients were enrolled by treating physicians in a retrospective manner from July 2017 backwards through a 30-month patient identification window period until a target number of patients (N=100) was reached.

Patients were eligible for inclusion if they were at least 18 years old, admitted to the hospital during the patient identification window period, had a confirmed diagnosis of cUTI, cIAI, or HABP/ VABP caused by a CRGN bacteria (based on a culture collected during their hospital stay), and had all medical data available from the date of culture collection to the end of follow-up (until date of death or up to 6 months after hospital discharge). A clinical diagnosis by the physician and a confirmatory culture for CRGN bacteria were mandatory to consider a potential case as an infection. Patients participating in a clinical trial during the observation period were excluded. The study was approved by the Ethics Committees of all participating sites. Italian regulations related to patient consent were followed, and signed patient consents were collected as appropriate prior to enrolment.

Study variables

Study variables included patient characteristics, index hospitalization characteristics, infection characteristics, details of treatment pathways, patient outcomes, and healthcare resource use during the index hospitalization. Patient characteristics collected were age, sex, body mass index (BMI), smoking status, alcohol consumption, comorbidities, hypersensitivity to antibacterial agents, reason for admission, ward of admission, and previous medical history, specifically prior nursing home admission, prior hospital stays, and invasive procedures and antibacterial agents received during the year prior to index hospitalization. Index hospitalization characteristics included length of stay, intensive care unit (ICU) need, ICU length of stay, and APACHE II (Acute Physiology And Chronic Health Evaluation II) score at ICU admission for those who were admitted to the ICU. Infection characteristics collected were infection site, presence of secondary bacteraemia, place of infection acquisition, pathogen isolated per patient, pathogen verification method, antibacterial susceptibility verification method and results, carbapenemase verification method and results, and isolate source.

Treatment pathway details included the antibacterial agents prescribed at culture collection, prior to culture susceptibility test results, and after culture susceptibility results. Healthcare resource use was described by the number of diagnostic tests and invasive procedures recorded during index hospitalization. Patient outcome variables collected included clinical cure and time to clinical cure (i.e. from culture collection to complete resolution of all signs and symptoms of infection). The EU-CARE study has provided novel evidence in the presentation of the flow treatments received by patients at 3 different cross-sectional time-points: at culture collection or next day (early empiric therapy), prior to culture susceptibility results (late empiric therapy), and after culture susceptibility results (first targeted therapy). Early empiric therapy corresponds to the first antibiotic treatment(s) that the patient received at date of culture collection or the day after but before susceptibility was known. Prior to culture susceptibility test results (late empiric therapy) includes those treatment(s) initiated between day 2 after culture collection and date of availability of antibiogram susceptibility results. And finally, first targeted therapy corresponds to treatment(s) prescribed after culture susceptibility results. Hereafter, first targeted therapy has been identified based on the following criteria:

1. treatment start date is after the date of antibiogram availability and no later than two days after the date of antibiogram availability;

2. treatment start date is prior to the date of antibiogram availability and treatment end date is at minimum two days after the date of antibiogram availability;

3. if no treatments are reported as per criteria defined above, the first treatment received between 3 and 7 days (inclusive) after the date of antibiogram availability is considered.

If other antibiotic treatments are initiated maximum 1 day after first antibiotic treatment start date, these are also considered as part of a combination therapy regimen. Additional variables collected were presence and duration of severe sepsis/ septic shock, mortality during index hospitalization and during the follow-up period, time to discharge, and the time to, and number of, readmissions for any diagnosis. Recurrence was defined as the development of new symptoms and signs of infection in association with a culture positive for CRGN following initial response to antibiotic therapy.

Statistical analysis

Descriptive statistics were analyzed with SAS Enterprise Guide version 7.13. Continuous variables are described as mean values and their standard deviations (SD), categorical variables are described by number and percentage. Furthermore, a subset analysis was performed on patients receiving carbapenems at any timepoint during index hospitalization (n=67) in order to assess their treatment pathways. Missing data were not included or imputed in the calculations, except for missing dates, where missing day values were substituted with mid-month dates in order to calculate length of time in days.

RESULTS

Preliminary survey

Hospitals reported an average of 737.5 (SD=430.8) cUTI, 547.5 (SD=452.29) cIAI, and 820.0 (SD=385.1) HABP/VABP cases (Figure 1A, Supplementary Table 1) in the three years prior to the survey. Altogether 16.9% of isolates were CR, with a similar frequency of CR levels across all three infection sites. KPC was the most common carbapenemase type observed in those isolates which were tested for carbapenemase production, with all participating hospitals reporting KPC in all three infection sites (Figure 1B, Supplementary Table 1). Oxacillinase type carbapenemase (OXA) was also common with three of the four hospital sites reporting OXA across all infection sites; New Delhi metallo-beta-lactamase (NDM) and Verona Integron-Mediated metallo-beta-lactamase (VIM) were less frequently identified.

Figure 1.

Hospital reported local Gram-negative infection characteristics.

Note: Data based on initial assessment performed by the sites (n=5) regarding the number and type of complicated Gram-negative infections over the last 3 years.

Figure 1A shows mean number of infections, colors represent ratio of antibiotic susceptibility/non-susceptibility.

Figure 1B shows percentage of hospital sites that reported different types of resistance with different infection types.

cIAI: Complicated intra-abdominal infection; cUTI: Complicated urinary tract infection; GN: Gram-negative; HABP/VABP: Hospital-acquired bacterial pneumonia/ventilator-associated bacterial pneumonia; KPC: Klebsiella pneumoniae carbapenemase; NDM: New Delhi metallo-beta-lactamase;

OXA: Oxacillinase type; VIM: Verona Integron-Mediated metallo-beta-lactamase.

Carbapenem resistance was highest for A. baumannii isolates (40%–70%) and K. pneumoniae isolates (20%–43%) (Figure 2, Supplementary Table 2). Escherichia coli, Pseudomonas aeruginosa, and other Enterobacterales isolates were less likely to be resistant to carbapenems (note: P. aeruginosa has endogenous resistance towards ertapenem). However, resistance was common towards other antibacterial groups: 38.3–72.5% against penicillins and 29.3–64.3% against fluoroquinolones, amongst isolates.

Figure 2.

Antibiotic resistance for different pathogens - Percentage of isolates that were resistant to each antibiotic class in pathogens observed in 2017.

Note: percent resistance refers to the percent non-susceptible (intermediate or resistant). Ertapenem resistance is not shown for Pseudomonas and Acinetobacter isolates, since these bacteria are intrinsically resistant to this antibiotic.

Retrospective observational study

Patient characteristics

Patients were mostly male, with an average age of 64.1 (SD=16.6) years (Table 1). Almost a third were overweight. Smoking and heavy drinking were uncommon. At least one comorbidity was noted for 94.0% of patients. Chronic lung disease, hypertension, and diabetes were each present in about a quarter of patients, and renal disease was also common. Nursing home stay was not reported frequently, but more than half of patients had a documented hospital stay in the year prior to index hospitalization. Altogether 38.0% of patients had received a prior course of antibacterial agents, most commonly penicillins (n=24). Prior carbapenem use was recorded for 11 patients. At admission, the most common diagnosis was infection or a respiratory disease; over half of patients were admitted to medical wards and a third to surgical wards.

Table 1.

Patient characteristics recorded at hospital admission.

Patient characteristics	No. (%)
At baseline
Age, years - mean (SD)	64.1 (16.6)
Sex
Female	36 (36.0%)
Male	64 (64.0%)
BMI categories
<18.5 kg/m2	10 (10.0%)
18.5–25 kg/m2	32 (32.0%)
>25 kg/m2	29 (29.0%)
Smoking status
Never smoked	23 (23.0%)
Previous smoking	18 (18.0%)
Current smoker	10 (10.0%)
Alcohol consumption
Never	20 (20.0%)
Occasional	12 (12.0%)
Heavy	1 (1.0%)
Comorbidities *
Chronic lung disease	25 (25.0%)
Diabetes	25 (25.0%)
Hypertension	24 (24.0%)
Renal disease	20 (20.0%)
Solid tumor	16 (16.0%)
None	6 (6.0%)
Hypersensitivity to antibiotics	3 (3.0%)
In the year prior to index hospitalization
Hospital admission	57 (57.0%)
Nursing home stay	5 (5.0%)
Nursing home resident at admission	4 (4.0%)
Invasive procedures	45 (45.0%)
Previous antibiotic administration*	38 (38.0%)
Penicillin	24 (63.2%)
Carbapenem	11 (29.0%)
Cephalosporins	8 (21.1%)
Other	13 (34.2%)
At admission
Reason for admission
Infection	29 (29.0%)
Respiratory	24 (24.0%)
Cardiovascular	12 (12.0%)
Gastrointestinal	11 (11.0%)
Neurological	6 (6.0%)
Metabolic/renal	5 (5.0%)
Trauma	2 (2.0%)
Other	11 (11.0%)
Ward of admission
Medical	53 (53.0%)
Surgical	33 (33.0%)
ICU	14 (14.0%)

^*Multi-response variable.

Index hospitalization characteristics

Mean length of stay was 52.9 (SD=0.5) days, with HABP/VABP patients having the longest hospitalization time (Figure 3, Supplementary Table 3). ICU need was high for all infection sites (38.2%–62.5%) but was highest for patients with HABP/ VABP, though patients with cUTI stayed in the ICU for the longest period (28.0 [SD=23.0] days). APACHE II scores at ICU admission, for those patients that required ICU, were also highest for cUTI (19.7 [SD=10.7]).

Figure 3.

Index hospitalization characteristics.

Note: A: Length of stay for different infection types (days). B: Intensive care unit need (percentage of all patients). C: Intensive care unit length of stay (days). D: APACHE II score. Worst score if more than one admission. Based on limited data (44% of patient data missing). Data are presented as mean values. APACHE II: Acute Physiology And Chronic Health Evaluation II; cIAI: Complicated intra-abdominal infection; cUTI: Complicated urinary tract infection; HABP/VABP: Hospital-acquired bacterial pneumonia/ventilator-associated bacterial pneumonia; ICU: Intensive care unit; LOS: Length of stay; SD: Standard Deviation.

Infection characteristics

The most common infection site included in the retrospective cohort was HABP/VABP, followed by cUTI and cIAI (Table 2). Secondary bacteraemia occurred in 44.0% of patients. The site of infection acquisition was either the hospital or another healthcare setting in nearly all cases, with only 1.0% of cases being community-acquired. The most common causative pathogen was K. pneumoniae, being implicated in two thirds of included patients. A. baumannii and P. aeruginosa were less common, and E. coli was only isolated from two patients. Bacterial verification was mostly performed via culture techniques, but MALDI-TOF was also used frequently (45.0% of cases). Antibacterial susceptibility was verified via an automated test or dilution methods in most cases, while E-test and disk diffusion methods were rarely used. The most common sources for isolates were urine and blood samples.

Table 2.

Infection characteristics.

Infection characteristics	No. (%)
Infection site
HABP/VABP	56 (56.0%)
cUTI	34 (34.0%)
cIAI	10 (10.0%)
Secondary bacteraemia	44 (44.0%)
Place of acquisition of infection *
Hospital-acquired	78 (78.0%)
Healthcare-acquired	21 (21.0%)
Community-acquired	1 (1.0%)
Pathogen isolated, per patient **
K. pneumoniae	66 (66.0%)
A. baumannii	23 (23.0%)
P. aeruginosa	12 (12.0%)
E. coli	2 (2.0%)
Pathogen verification **
Culture	80 (80.0%)
MALDI-TOF	45 (45.0%)
Antibiotic susceptibility verification **
Automated antimicrobial susceptibility test	42 (42.0%)
Dilution	26 (26.0%)
E-test	2 (2.0%)
Disk diffusion	2 (2.0%)
Sample type **
Urine	36 (36.0%)
Blood	36 (36.0%)
Other	27 (27.0%)
BAL	24 (24.0%)
Sputum	12 (12.0%)
Peritoneal fluid	4 (4.0%)
Pleural fluid	3 (3.0%)

^*Hospital-acquired: culture collected = 3 days after admission; Healthcare-acquired: culture collected <3 days after admission with previous healthcare exposure; Community-acquired: culture collected <3 days after admission with no previous healthcare exposure.

^**Multi-response variable.

BAL: Bronchoalveolar lavage; cIAI: Complicated intra-abdominal infection; cUTI: Complicated urinary tract infection; HABP/VABP: Hospital-acquired bacterial pneumonia/ventilator-associated bacterial pneumonia; MALDI-TOF: Matrix-assisted laser desorption ionization-time of flight mass spectrometry; PCR: Polymerase chain reaction.

Figure 4 (see also Supplementary Table 4) demonstrates that most isolates were tested for susceptibility to meropenem or imipenem, while ertapenem susceptibility was less frequently tested, and doripenem susceptibility was only tested against a few isolates. All tested A. baumannii and most P. aeruginosa isolates were classified as resistant, while K. pneumoniae and E. coli isolates were more likely to be susceptible towards the carbapenems tested. Carbapenemase tests were performed on 34 of the 103 isolates, with all tested isolates confirmed to be producers of KPC. The majority of carbapenemase testing was performed via Xpert Carba-R (73.5%). Additionally, the boronic acid test (20.6%) and Cepheid test (2.9%) were also used.

Figure 4.

Susceptibility test results.

Note: Susceptibility information for confirmatory cultures. Colors represent the proportion of isolates which are susceptible/non-susceptible; the height of columns indicates the percentage of isolates tested. Ertapenem resistance is not shown for Pseudomonas and Acinetobacter isolates, since these bacteria are intrinsically resistant to this antibiotic.

Treatment pathways

At culture collection, the majority of patients (n=52, 52%) did not receive an antibacterial agent. Among those who did, monotherapy (n=19, 19%) and combination (with two agents) (n=21, 21%) was common, while administration of three or more antibacterial agents was rare (n=7, 7%). Prior to the availability of susceptibility results, the most common treatment type was combination therapy (n=34, 34%), no treatment (n=27, 27%), monotherapy (n=20, 20%), or three or more antibacterial agents (n=19, 19%). After antibiogram results were available, the majority of patients received combination therapy (n=44, 44%), followed by combination therapy with three or more antibiotics (n=26, 26%), while monotherapy (n=21, 21%) and no treatment (n=8, 8%) were less common. The most common antibacterial agents prescribed after culture collection were meropenem (18.0%), colistin (14.0%), and piperacillin (14.0%). Immediately prior to the availability of susceptibility results, 41.0% of patients received meropenem, 25.0% colistin, and 18.0% piperacillin. After antibiogram results were available, meropenem and colistin were each administered to 61.0% of patients, and gentamicin and tigecycline to 14.0% and 13.0% of patients, respectively. Other antibiotics were prescribed less frequently at each timepoint (see Supplementary Table 5a).

Figure 5 shows treatment pathways for patients who received carbapenems at any timepoint (n=67) (in addition, see Supplementary Table 5b). As can be seen from the figure, immediately after culture collection more than half of patients (n=35) did not receive any antibiotics and 37.3% of patients (n=25) received either one or two antibiotics, with the most common regimen being a carbapenem with an additional antibiotic (n=8). By the second timepoint (immediately prior to receipt of susceptibility results), approximately half of patients (n=33) received a carbapenem plus one or two other antibiotics. After receiving the susceptibility results, this proportion increased to 82.1% of patients (n=55), while treatment combinations not involving any carbapenems became rare (n=3). At that point, the most common agent in combination with carbapenems was colistin (n=43). In addition, colistin (either alone or in combination) was also prescribed after receiving the susceptibility results to a further 18 patients not receiving carbapenems (data not shown in table or figure).

Figure 5.

Sankey figure of treatment pathways for patients who received a carbapenem during observation period.

Healthcare resource use

Table 3 shows data relating to healthcare resource use during index hospitalization. Blood cell count and biochemistry tests were the most common tests performed, but blood cultures, radiographic investigations, and urine cultures were also performed in at least 40.0% of patients. Endoscopy was the most common invasive procedure, followed by vascular catheterization, and percutaneous drainage.

Table 3.

Healthcare resource use: percentage of patients who underwent diagnostic tests and/or procedures during index hospitalization.

Test or procedure	No. (%)
Diagnostic test and/or procedure during index hospitalization
Blood cell count	99 (99.0%)
Biochemistry test	78 (78.0%)
Blood culture	67 (67.0%)
Radiograph	46 (46.0%)
Urine culture	44 (44.0%)
Bronchoscopy	36 (36.0%)
CT	29 (29.0%)
ECG	22 (22.0%)
Sputum analysis	16 (16.0%)
Other diagnostic test	3 (3.0%)
Invasive procedure during index hospitalization
Endoscopy	19 (19.0%)
Vascular catheter	15 (15.0%)
Percutaneous drainage	9 (9.0%)
Surgery (incl. surgical drainage)	5 (5.0%)
Urinary catheter	3 (3.0%)
Mechanical ventilation	2 (2.0%)
Cardiac catheterization	2 (2.0%)

Note: Table shows diagnostic tests and invasive procedures performed during index hospitalization and diagnostic tests in the follow up period for patients who have reported at minimum one diagnostic test related to infection treatment from 6 months after discharge (data regarding invasive procedures in this period was not available). CT: computed tomography, ECG: electrocardiogram.

Patient outcomes

Outcome data are summarized in Table 4. Clinical cure was achieved in 43.0% of patients in an average time of approximately three weeks. Severe sepsis or septic shock was present in 28.0% of patients during index hospitalization. Of the 61 patients who survived their index hospitalization, 26.3% (16/61) were readmitted within 60 days, with a mean time from discharge to readmission of 56.9 days (SD=51.9). Of these patients, recurrence of infection was the cause of readmission in 6.25% of patients. Half of patients readmitted had “other” (including cardiovascular, gastrointestinal or metabolic reason) or unknown reason recorded in their notes. Mortality was observed in 39.0% of included patients during their index hospitalization, with a mortality rate of 65.0% by the end of the follow-up period. The most common cause of death was deemed to be infection related.

Table 4.

Outcomes.

Characteristic	No. (%)
Clinical cure achieved*	43 (43.0%)
Time to clinical cure from culture collection – days - mean (SD)	20.6 (11.6)
Severe sepsis/septic shock	28 (28.0%)
Duration of severe sepsis/septic shock – days - mean (SD)	4.5 (3.8)
Mortality during index hospitalization	39 (39.0%)
Time to discharge from culture collection – days - mean (SD)	33.8 (52.5)
Time to ICU discharge since culture collection – days - mean (SD)	8.3 (14.6)
Time to death since culture collection – days - mean (SD)	134.9 (205.2)
Time to readmission from discharge following index hospitalization – days - mean (SD)***	56.9 (51.9)
Readmission rate at 30 days**	5 (8.8%)
Readmission rate at 60 days**	16 (26.3%)
Overall mortality (at chart review)	65 (65.0%)
Cause of death
Infection-related	28 (43.1%)
Other	16 (24.6%)
Unknown	21 (32.3%)

^*As defined by complete resolution of all signs and symptoms of infection.

^**Readmission rate was calculated using the number of alive patients at discharge as denominator.

^***Time to readmission since discharge was calculated for all patients who reported a hospitalization up to 6 months after hospital discharge. Those patients with missing information about status (alive or deceased) at 6 months following hospital discharge were removed from the population used to assess readmission, in order to avoid a potential bias to higher hospitalization rates.

DISCUSSION

The aim of this study was to assess the clinical management of patients diagnosed with cUTI, cIAI, and HABP/VABP resulting from CRGN bacteria in Italian hospitals. In the hospital survey we found that CR occurred in about 17% of the complicated infections in focus and that A. baumannii had the highest rates of CR. Additionally, in the retrospective study we found that complicated CRGN infections were hospital- or healthcare-acquired in 99% of cases and were most often caused by K. pneumoniae. Despite some use of rapid diagnostic procedures and frequent utilization of antibiotic combination treatments, outcomes were relatively poor – with an average length of hospital stay of over 50 days, achievement of clinical cure in less than half of patients, and an overall mortality rate of 65.0%. Due to the low number of patients (n=14) it was not possible to analyze mortality according to ICU and medical/surgical wards. These outcomes are in line with previously published international data on CRGN infections, such as patients with K. pneumoniae infections and patients with CR Pseudomonas bacteraemia [1, 11]. Appropriate (both effective and timely) antimicrobial therapy is known to improve the outcome of infections while being consistent with the principles of antimicrobial stewardship [12, 13]. Important attributes of appropriate treatment are the selection of effective empirical therapy, identification (ideally rapid) of the causative bacteria and antibiotic susceptibility, and subsequent timely modifications to prescribed treatments to ensure appropriate definitive therapy is being given. CRGN infections pose a particularly difficult challenge for healthcare providers trying to comply with antimicrobial stewardship principles, due to the higher chance of selecting an initial therapy which is inappropriate and the associated delay to appropriate treatment [13].

Considering the need to identify bacteria and antibiotic susceptibility properly, our study showed that bacterial identification in Italian hospitals was often achieved through MALDI-TOF, a rapid procedure previously reported to significantly reduce time to optimization of antibiotic therapy as well as reducing hospital costs [14]. Additionally, susceptibility verification was predominantly performed by automated susceptibility testing, a technique also associated with shorter diagnostic times [15]. On the other hand, according to the results of our cohort, PCR-based identification methods were not used extensively, and carbapenemase testing was limited to endemic KPC identification. Since the identification of carbapenemase type can help to optimize therapy, increased availability of such diagnostic methods might lead to improved prescribing choices and infection outcomes.

In this study we observed that 19 patients received a carbapenem as an early empirical therapy, 43 as a late empirical therapy, and 64 as part of the treatment regimen even after receiving the susceptibility data confirming that the infection was due to CRGN bacteria. This observation may explain some of the poor outcomes reported in our study cohort and indicates a need for additional therapeutic options for the management of CRGN infections in the Italian setting. Treatment pattern analysis performed in our study for patients receiving carbapenems at any timepoint describes trends in early and late empirical and targeted antibacterial therapy for these difficult-to-treat infections. Initial carbapenem therapy was observed in 19.0% of this subset of patients either as a monotherapy or in combination, which is in line with the 16.6% figure previously reported for CR Enterobacterales bloodstream infections in Italy [16]. However, overall carbapenem use was substantial, since they were prescribed – predominantly in combination with other antibiotics – in 43.0% of patients as late empirical therapy and in 64.0% of patients as targeted therapy. When looking at all patients, the most frequent concomitant therapy was colistin, perhaps due to the need for a broad coverage agent that can address the prevalent KPC-producers in Italy (recent evidence suggests that KPC-producing K. pneumoniae infections can be treated adequately with colistin-carbapenem combinations [17]. Colistin is associated with various negative consequences, including renal toxicity and suboptimal dosing, but may be utilized for these complicated infections due to the historical dearth of alternative effective agents [18]. However, even combination regimens, including colistin-based combinations, may not lead to better outcomes [19, 20]. This is reflected in our cohort, where clinical cure was relatively low and overall mortality high, despite use of combination therapy and last resort agents.

Other possible explanations for the poor outcomes observed in the study cohort relate to known patient-associated risk factors for worse outcomes, namely high levels of previous exposure to healthcare environments where antibiotic-resistant bacteria are likely to be present and the selection pressures which may be exerted by previous antibiotic administration [21]. Indeed, nearly all (99.0%) CRGN complicated infections were found to be healthcare-associated in our cohort, 57 patients had a documented hospital admission in the previous year and 38 had received prior antibiotic treatment, including 11 who were prescribed a carbapenem treatment. The identification of these risk factors within this Italian cohort supports the importance of considering these factors when evaluating the possibility of CRGN infections and planning empirical therapy, bypassing carbapenems in favor of agents which will address CR.

The strengths of the EU-CARE study include an extensive analysis of the management in Italy of complicated CRGN infections as well as new information on the real-world treatment patterns of patients with these difficult to treat conditions. Our study, however, has certain limitations as well. The small number of sites prohibits the generalization of results to the whole of the country, although survey data from the study sites is aligned with epidemiological data previously published for Italy [3]. The retrospective chart review data collection method assumed equivalent definitions regarding clinical terms and microbiological methods (e.g. dilution, disk diffusion, automated antimicrobial susceptibility test). However, these definitions were neither confirmed nor harmonized, which might have contributed to an unmeasured variance in the results. Additionally, reasons for clinical decision-making in the administration of antibiotics were not assessed in our study, although these might be important aspects of optimizing antimicrobial stewardship while treating infections caused by resistant bacteria. Furthermore, the chart review method might have resulted in missing data, which might explain some unexpected findings in the healthcare resource variables, i.e. that the rates of invasive procedures during index hospitalizations were reportedly low and that mechanical ventilation was only reported in 2.0% of cases despite 56.0% of the cases being HABP/VABP and 44.0% of patients requiring time in the ICU).

The literature also reports changes in resistance mechanisms among Enterobacteriaceae within some Italian hospitals over time [22]. Therefore, the findings of this study should be considered in context of their institutional setting and time frame of the data reported. Furthermore, in response to changing epidemiology and the availability of novel agents, it is important to recognize changes in clinical management that have not been captured within this study. In recent years new non-colistin-based regimens (including ceftolozane-tazobactam, ceftazidime-avibactam and/or meropenem-vaborbactam) have also been introduced into clinical practice impacting clinical outcomes and also local ecologies [23]. Between 2015 and 2019, there were significantly increasing trends in the European population-weighted mean percentages for carbapenem resistance. Across Europe, Italy was one of the regions with largest percentage of invasive K. pneumoniae isolates resistant to carbapenems [24]. This resistance situation remains problematic. The time period used for the study and the changes seen during the last years may affect the generalizability of our results to the current epidemiology and clinical management of these infections in Italy. In terms of type of infections, the epidemiological profile of each region is of major importance in order to define the most appropriate treatment algorithm. A recently published cross-sectional survey showed that the treatment of CRGN infections is highly variable with clinicians using a wide range of antibacterial strategies and combinations informed by clinical severity, local availability, and clinical experience [25]. Given the availability of novel non-colistin-based treatment regimens it is important to recognize their optimal use within the treatment pathway, and when possible inform treatment choice based on pathogen susceptibility profiles [23].

Finally, due to recent changes in the EUCAST definitions (i.e. for testing categories S, I, and R), our study might have included cases that would not currently qualify as CR [26].

In conclusion, our results reflect the multi-faceted medical burden associated with complicated CRGN infections. Our study pinpoints potential areas for improvement in treatment practices. For example, regular and detailed local surveillance and use of accurate and rapid microbial diagnostic techniques might aid clinical decision-making and facilitate improvements in antimicrobial stewardship when treating CRGN infections. Further research is needed to identify new therapeutic options, and to use the novel agents which are targeted to the treatment of CRGN infections, in order to optimize patient treatment and improve overall patient outcomes.

Supplementary Information

Supplementary Table 1 (data for Figure 1).

Hospital reported local Gram-negative infection characteristics.

Infection type	Characteristic	No. (%)
cUTI	Total Gram-negative infections - mean (SD)	737.5 (430.8)
	Carbapenem-resistant (% of total isolates)	21 (18.7%)
	Resistance type *
	KPC	4 (100%)
	NDM	2 (50.0%)
	VIM	2 (50.0%)
	OXA	3 (75.0%)
	Other	1 (25.0%)
cIAI	Total Gram-negative infections	547.5 (452.2)
	Carbapenem resistant (% of total isolates)	25 (19.1%)
	Resistance type *
	KPC	4 (100%)
	NDM	1 (25%)
	VIM	1 (25%)
	OXA	3 (75%)
	Other	0 (0%)
HABP/VABP	Total Gram-negative infections	820.0 (385.1)
	Carbapenem resistant (% of total isolates)	30.5 (18.9%)
	Resistance type *
	KPC	4 (100%)
	NDM	2 (50%)
	VIM	1 (25%)
	OXA	3 (75%)
	Other	1 (25%)
cUTI + cIAI + HABP/VABP	Carbapenem resistant (% of total isolates)	27.5 (16.9%)

Note: Data based on initial assessment performed by the sites (n=4) regarding the number and type of complicated Gram-negative infections over the last 3 years. Data are presented as mean (SD) or n (%).

^*Multi-response variable. Data shows number (%) of sites that reported type of resistance.

cIAI: Complicated intra-abdominal infection; cUTI: Complicated urinary tract infection; GN: Gram-negative; HABP/VABP: Hospitalacquired bacterial pneumonia/ventilator-associated bacterial pneumonia; KPC: Klebsiella pneumoniae carbapenemase; NDM: New Delhi metallo-beta-lactamase; OXA: Oxacillinase type; VIM: Verona Integron-Mediated metallo-beta-lactamase.

Supplementary Table 2 (data for Figure 2).

Antibacterial resistance for different pathogens.

Isolate	Antibacterial agent	Mean (SD) percentage of non-susceptible isolates in five hospitals
E. coli	Aminoglycosides	13.8 (8.9)
	Penicillins	38.3 (22.2)
	Doripenem	0.5 (0.7)
	Ertapenem	1.8 (2.9)
	Imipenem	0.3 (0.6)
	Meropenem	1.5 (2.4)
	Fluoroquinolones	44.8 (12.5)
	3rd gen cephalosporins	31.8 (12.2)
K. pneumoniae	Aminoglycosides	32.3 (22.2)
	Penicillins	55.0 (23.9)
	Doripenem	20.0 (21.2)
	Ertapenem	43.0 (29.4)
	Imipenem	33.0 (27.1)
	Meropenem	38.0 (24.6)
	Fluoroquinolones	55.5 (18.4)
	3rd gen cephalosporins	53.3 (23.4)
Other Enterobacterales spp.	Aminoglycosides	14.8 (11.5)
	Penicillins	55.3 (26.8)
	Doripenem	15.0 (14.1)
	Ertapenem	9.0 (10.7)
	Imipenem	10.7 (12.5)
	Meropenem	8.5 (11.1)
	Fluoroquinolones	36.8 (19.2)
	3rd gen cephalosporins	31.3 (11.1)
P. aeruginosa	Aminoglycosides	11.3 (12.7)
	Penicillins	51.3 (32.8)
	Doripenem	20.0 (7.1)
	Ertapenem	100.0 (0.0)
	Imipenem	17.0 (10.6)
	Meropenem	18.8 (9.7)
	Fluoroquinolones	29.3 (4.6)
	3rd gen cephalosporins	21.5 (6.2)
A. baumannii	Aminoglycosides	60.0 (39.5)
	Penicillins	72.5 (48.6)
	Doripenem	40.0 (56.6)
	Ertapenem	70.0 (47.6)
	Imipenem	65.8 (44.5)
	Meropenem	65.3 (44.2)
	Fluoroquinolones	64.3 (43.6)
	3rd gen cephalosporins	72.5 (48.6)

Note: Percentage of isolates that were non-susceptible (intermediate or resistant) to each antibacterial agent class in pathogens observed in 2017.

Supplementary Table 3 (data for Figure 3).

Index hospitalization characteristics.

	All patients (N=100)	HABP/VABP (N=56)	cUTI (N=34)	cIAI (N=10)
LOS (days)	52.9 (68.5)	58.7 (83.2)	46.7 (45.9)	42.0 (33.0)
ICU need	44 (44.0%)	35 (62.5%)	13 (38.2%)	5 (50.0%)
ICU LOS (days)	24.3 (20.4)	23.9 (20.2)	28.0 (23.0)	16.8 (16.3)
APACHE II score*	14.6 (8.6)	12.2 (6.9)	19.7 (10.7)	11.4 (3.1)

Note: Data are presented as mean (SD) or n (%).

^*Worse score if more than one admission. Based on limited data (43% of patient data missing).

APACHE II: Acute Physiology And Chronic Health Evaluation II; cIAI: Complicated intra-abdominal infection; cUTI: Complicated urinary tract infection; HABP/VABP: Hospital-acquired bacterial pneumonia/ventilator-associated bacterial pneumonia; ICU: Intensive care unit; LOS: Length of stay; SD: Standard Deviation.

Supplementary Table 4 (data for Figure 4).

Susceptibility test results.

	Doripenem		Ertapenem		Imipenem		Meropenem
	Tested	Resistant	Tested	Resistant	Tested	Resistant	Tested	Resistant
K. pneumoniae	18 (27.3%)	14 (77.8%)	46 (69.7%)	45 (97.8%)	34 (51.5%)	30 (88.3%)	65 (98.5%)	64 (98.5%)
P. aeruginosa	1 (8.3%)	0 (0%)	8 (66.7%)	8 (100%)	12 (100%)	12 (100%)	11 (91.7%)	10 (89.9%)
A. baumannii	0 (0%)	N/A	15 (65.2%)	15 (100%)	23 (100%)	23 (100%)	11 (47.8%)	11 (100%)
E. coli	2 (100%)	1 (50.0%)	0 (0%)	N/A	2 (100%)	0 (0%)	2 (100%)	2 (100%)

Note: Susceptibility information of confirmatory cultures related to carbapenem susceptibility. Data are presented as n (%).

Supplementary Table 5a.

Antibacterial treatments received during the study period.

	Immediately after culture collection *	Immediately prior to receipt of susceptibility results **	After receipt of susceptibility results ***
	(N=100)	(N=100)	(N=100)
Aminoglycosides - amikacin	0 (0.0%)	0 (0.0%)	1 (1.0%)
Aminoglycosides - gentamicin	1 (1.0%)	8 (8.0%)	14 (14.0%)
Aminoglycosides - tobramycin	1 (1.0%)	1 (1.0%)	1 (1.0%)
Carbapenems - ertapenem	0 (0.0%)	1 (1.0%)	4 (4.0%)
Carbapenems - imipenem/cilastatin	1 (1.0%)	2 (2.0%)	3 (3.0%)
Carbapenems - meropenem	18 (18.0%)	41 (41.0%)	61 (61.0%)
Cephalosporins - cefepime	1 (1.0%)	1 (1.0%)	0 (0.0%)
Cephalosporins - ceftazidime	0 (0.0%)	1 (1.0%)	1 (1.0%)
Cephalosporins - ceftolozane/tazobactam	7 (7.0%)	7 (7.0%)	2 (2.0%)
Cephalosporins - ceftriaxone	0 (0.0%)	1 (1.0%)	0 (0.0%)
Cephalosporins - cefuroxime	1 (1.0%)	0 (0.0%)	0 (0.0%)
Colistin	14 (14.0%)	25 (25.0%)	61 (61.0%)
Daptomycin	2 (2.0%)	2 (2.0%)	2 (2.0%)
Fluoroquinolones - ciprofloxacin	2 (2.0%)	3 (3.0%)	2 (2.0%)
Fluoroquinolones - levofloxacin	6 (6.0%)	8 (8.0%)	2 (2.0%)
Fluoroquinolones - lomefloxacin	1 (1.0%)	1 (1.0%)	0 (0.0%)
Fluoroquinolones - moxifloxacin	1 (1.0%)	1 (1.0%)	0 (0.0%)
Fosfomycin	0 (0.0%)	3 (3.0%)	8 (8.0%)
Glycopeptides - vancomycin	2 (2.0%)	4 (4.0%)	2 (2.0%)
Glycopeptides- teicoplanin	5 (5.0%)	7 (7.0%)	4 (4.0%)
Glycylcycline - tigecycline	2 (2.0%)	8 (8.0%)	13 (13.0%)
Linezolid	6 (6.0%)	6 (6.0%)	4 (4.0%)
Metronidazole	1 (1.0%)	1 (1.0%)	1 (1.0%)
Nitrofurantoin	0 (0.0%)	0 (0.0%)	1 (1.0%)
Penicillins - amoxicillin	0 (0.0%)	0 (0.0%)	1 (1.0%)
Penicillins - ampicillin	0 (0.0%)	1 (1.0%)	0 (0.0%)
Penicillins - piperacillin	14 (14.0%)	16 (16.0%)	5 (5.0%)
Tetracyclines - minocycline	1 (1.0%)	1 (1.0%)	1 (1.0%)
Tetracyclines - tetracycline	0 (0.0%)	1 (1.0%)	0 (0.0%)
Trimethoprim - sulfamethoxazole	0 (0.0%)	1 (1.0%)	3 (3.0%)

^*First antibacterial treatment(s) the patient received the next day after culture collection. If date of culture collection was the same as the date of susceptibility test result availability, the following definition was applied: treatment for which start date was before date of susceptibility result availability and date of finalization was after or up to two days prior to the date of susceptibility result availability.

^**Treatment for which start date was before date of susceptibility result availability and date of finalization was after or up to two days prior to the date of susceptibility result availability.

^***Treatments for which start date was after the date of susceptibility result availability and no later than two days after the date of susceptibility result availability. Treatment start date was prior to the date of susceptibility result availability and treatment end date was at minimum two days after the date of susceptibility result availability.

Supplementary Table 5b.

Antibacterial regimens received during the study period for patients who received carbapenem at any time during the study period.

	Immediately after culture collection *	Immediately prior to receipt of susceptibility results **	After receipt of susceptibility results ***
	(N=67)	(N=67)	(N=67)
Carbapenem monotherapy	5 (7.5%)	6 (9.0%)	6 (9.0%)
Carbapenem + 1 other	8 (11.9%)	21 (31.3%)	35 (52.2%)
Carbapenem + 2 other	5 (5.3%)	12 (17.9%)	20 (29.9%)
Carbapenem + 3 + other	1 (1.5%)	4 (6.0%)	3 (4.5%)
Monotherapy	5 (7.5%)	5 (7.5%)	0 (0%)
Combination involving 2 classes	7 (10.5%)	6 (9.0%)	2 (3.0%)
Combination involving 3 classes	1 (1.6%)	1 (1.6%)	0 (0.0%)
Combination involving 4+ classes	0 (0.0%)	0 (0.0%)	0 (0.0%)
No treatment	35 (52.2%)	12 (17.9%)	1 (1.6%)

^*First antibacterial treatment(s) the patient received the next day after culture collection. If date of culture collection was the same as the date of susceptibility result availability, the following definition was applied: treatment for which start date was before date of susceptibility result availability and date of finalization was after or up to two days prior to the date of susceptibility result availability.

^**Treatment for which start date was before date of susceptibility result availability and date of finalization was after or up to two days prior to the date of susceptibility result availability.

^***Treatments for which start date was after the date of susceptibility result availability and no later than two days after the date of susceptibility result availability. Treatment start date was prior to the date of susceptibility result availability and treatment end date was at minimum two days after the date of susceptibility result availability.