Predictors of Mortality in Patients with Infections Due to Carbapenem-Resistant Gram-Negative Bacteria

Introduction: Infections caused by carbapenem-resistant Gram-negative bacteria (CR-GNB) are a significant cause of mortality and represent a serious challenge to health systems. The early identification of mortality predictors could guide appropriate treatment and follow-up. We aimed to identify the factors associated with 90-day all-cause mortality in patients with CR-GNB infections. Methods: We conducted a cohort study from 1 January 2019 to 30 April 2022. The primary outcome was death from any cause during the first 90 days after the date of the first CR-GNB-positive culture. Secondary outcomes included infection relapse, invasive mechanical ventilation during follow-up, need for additional source control, acute kidney injury, Clostridioides difficile infection, and all-cause hospital admission after initial discharge. Bivariate and multivariate Cox-proportional hazards models were constructed to identify the factors independently associated with 90-day all-cause mortality. Results: A total of 225 patients with CR-GNB infections were included. Death occurred in 76 (34%) cases. The most-reported comorbidities were immunosuppression (43%), arterial hypertension (35%), and COVID-19 (25%). The median length of stay in survivors was 18 days (IQR 10–34). Mechanical ventilation and ICU admission after diagnosis occurred in 8% and 11% of cases, respectively. Both infection relapse and rehospitalisation occurred in 18% of cases. C. difficile infection was diagnosed in 4% of cases. Acute kidney injury was documented in 22% of patients. Mechanical ventilation after diagnosis, ICU admission after diagnosis, and acute kidney injury in the first ten days of appropriate treatment were more frequently reported among non-survivors. In the multivariate analysis, age (HR 1.19 (95%CI 1.00–1.83)), immunosuppression (HR 1.84 (95%CI 1.06–3.18)), and septic shock at diagnosis (HR 2.40 (95% 1.41–4.08)) had an independent association with death during the first 90 days after the CR-GNB infection diagnosis. Receiving antibiogram-guided appropriate treatment was independently associated with a lower risk of death (HR 0.25 (95%CI 0.14–0.46)). Conclusions: The presence of advanced age, immunosuppression, septic shock at diagnosis, and inappropriate treatment are associated with higher 90-day all-cause mortality in hospitalised patients with infections due to CR-GNB. Recognition of the risk factors for adverse outcomes could further assist in patient care and the design of interventional studies that address the severe and widespread problem that is carbapenem resistance.


Introduction
Infections caused by carbapenem-resistant Gram-negative bacteria (CR-GNB) are a significant cause of morbidity and mortality. The World Health Organization listed broadspectrum antimicrobial resistance (AMR) as one of the top-ten threats to global health [1]. Because carbapenems are an effective therapeutic alternative for multidrug-resistant (MDR) organisms, widespread carbapenem resistance represents a serious challenge to health systems. A progressive global increase in the incidence of CR-GNB in the last two decades has been identified, with reports of hospital outbreaks worldwide [2][3][4]. The latter has been associated with increased mortality and burden of disease [5]. A mortality of up to 50% has been reported among patients with bacteraemia due to carbapenem-resistant Enterobacterales (CRE) [6]. Additionally, the presence of carbapenemases has been independently associated with increased mortality [7]. In retrospective studies, the use of inappropriate antibiotic therapy, a higher APACHE score, a Charlson comorbidity index >3, advanced age, chemotherapy-induced neutropenia, and septic shock have been associated with increased mortality among patients with infections due to CR-GNB [8,9]. Based on estimates from a model conducted in the US in 2017, the average hospital cost of a single episode of an infection due to CR-GNB ranges between USD 22,484 and 66,031 [8].
It is essential to improve diagnostic methods and treatment strategies for patients with infections due to carbapenem-resistant organisms (CROs) [9]. As CROs disseminate, the factors associated with mortality could vary between regions [10]. In our region, scarce evidence exists regarding prognostic factors in patients with infections due to CROs. Unidentified factors could contribute to an excess in mortality, especially in regions with limited availability of first-line antibiotics. The early identification of mortality predictors could guide treatment and follow-up. Additionally, the recognition of factors associated with a worse prognosis may contribute to the design of research protocols focusing on high-risk patients.
We conducted a retrospective cohort study to identify the risk factors associated with 90-day all-cause mortality in patients with infections due to CR-GNB.

Methods
We conducted a retrospective cohort study in a tertiary care centre in Mexico City, which included all hospitalised patients aged 18 or older and diagnosed with any cultureproven infection due to carbapenem-resistant Enterobacterales or non-fermenting GNB during the period from 1 January 2019 to 30 April 2022. A CRO was considered when nonsusceptibility to at least one carbapenem was documented. Patients with community-or hospital-acquired infections were included. Community-and hospital-acquired infections were defined as previously recommended [11]. All samples sent to the clinical microbiology laboratory were screened, regardless of their origin. We excluded patients in whom the CR-GNB isolate was considered not clinically significant according to the attending team's criterion.
Demographics, clinical, treatment, laboratory, and microbiology data were collected from the electronic medical records. Immunosuppression was considered when immunosuppressive medication (e.g., chemotherapeutic agents, drugs to prevent graft rejection, methotrexate, ≥10 mg of prednisone or equivalent for the last 14 days, monoclonal antibodies) or comorbidities (e.g., solid malignant tumours, hematologic malignancy, solid organ transplant, human immunodeficiency virus infection, and connective tissue disorders) were present. Screening for CRO colonisation is not a routine practice in our centre. The treatment data included antimicrobial prescription and source control procedures. Antibiogram-guided appropriate treatment was considered when a combination therapy with at least one active antimicrobial, as reported by antibiogram, was used. For carbapenemase-harbouring isolates, carbapenem monotherapy was considered inappropriate. Data regarding the outcomes were also retrieved from the medical records. The participants were followed up for 90 days after the date of the first CR-GNB-positive culture.
The primary outcome was death from any cause during the first 90 days after the date of the first CR-GNB-positive culture. An infection-related death was considered when patients died because of direct infectious complications (e.g., septic shock, pneumonia, source control surgical complications) or in cases of persistent signs of infection such as fever, persistently elevated leucocyte counts, C-reactive protein or procalcitonin, or persistent positive blood cultures. The secondary outcomes included death from any cause during the first 30 days after the date of the first CR-GNB-positive culture, length of stay, infection relapse, defined by the presence of clinical deterioration as assessed by the attending clinical team in addition to microbiologic confirmation of the index isolate, invasive mechanical ventilation (IMV) after the diagnosis of infection, need for additional source control, acute kidney injury (e.g., creatinine increase of >0.3 mg/dL of serum creatinine when compared to baseline) during the first ten days after antimicrobial treatment initiation, Clostridioides difficile infection, and all-cause hospital admission after initial discharge within the first 90 days after the date of the first CR-GNB-positive culture.
Considering a primary outcome probability of 50% [2,3], a mean absolute percentage error of 10%, and the identification of up to 10 potential predictors, we calculated a sample size of at least 165 patients [14]. Descriptive statistics were reported using mean and standard deviation or median and interquartile range (IQR) according to the variables' distribution determined by the Shapiro-Wilk test. Chi-square, Fisher's exact test, T-test for independent samples, and rank sum tests were used for comparisons between groups. To identify the factors associated with mortality, a bivariate analysis was performed to calculate the hazard ratio (HR) and 95% confidence intervals (95%CI). A multivariate Cox-proportional hazards model was constructed to identify the factors independently associated with 90-day all-cause mortality. The model was constructed using variables of biological importance according to previous reports. Variables with interactions as assessed by Mantel-Haenszel Chi were not included in the model (see Supplementary Materials for detailed information). Missing data were not replaced. A p-value of 0.05 was considered statistically significant. The analysis was carried out using STATA V15 (Houston, TX, USA). Because of the study's retrospective nature, the informed consent requirement was waived. The study, including the waived informed consent, was approved by the Institutional Review Board (ref. 4022). All personal data were protected according to national and international standards.

Results
A total of 288 patients with CRO isolates were screened during the study period. A total of 225 patients were included. All-cause 90-day mortality occurred in 76/225 (34%) cases of which 51/76 (67%) died from infectious causes, and 25/76 (33%) died from non-infectious causes ( Figure 1).

Results
A total of 288 patients with CRO isolates were screened during the study period. A total of 225 patients were included. All-cause 90-day mortality occurred in 76/225 (34%) cases of which 51/76 (67%) died from infectious causes, and 25/76 (33%) died from noninfectious causes ( Figure 1).
Death from any cause during the first 30 days after the date of the first CR-GNBpositive culture occurred in 54/225 (24%) patients. The median length of stay in survivors was 18 days (IQR 10-34). The median time from diagnosis to death was 13 days (IQR 4-38). Mechanical ventilation and ICU admission after diagnosis occurred in 19/225 (8%) and 25/225 (11%) patients, respectively. Both infection relapse and rehospitalisation for any cause occurred in 41/225 (18%) cases. A C. difficile infection was diagnosed in 9/225 (4%) patients. Acute kidney injury was documented in 40 of 181 (22%) patients with available data. Mechanical ventilation after diagnosis, ICU admission after diagnosis, and acute kidney injury in the first ten days of appropriate treatment were more frequently reported among non-survivors (Table 2). In the bivariate analysis, increasing age, respiratory tract infections, bone and soft tissue infections, Charlson comorbidity index >3, heart disease, immunosuppression, renal replacement therapy, IMV at diagnosis, septic shock at diagnosis, chemotherapy prescription 180 days prior to diagnosis, ICU at diagnosis, lower levels of haemoglobin, leukocytes, platelets, and albumin at diagnosis, higher levels of glucose, creatinine and C-reactive protein at diagnosis, acute kidney injury during treatment, IMV after diagnosis, and ICU stay after diagnosis were associated with higher HR for death during the first 90 days after the CR-GNB infection diagnosis. Intraabdominal infections and antibiogram-guided appropriate antibiotic treatment were associated with lower HR for death during the first 90 days after the CR-GNB infection diagnosis. Table S10 summarises the non-adjusted bivariate analysis.

Discussion
In our study, older age, immunosuppression, and septic shock at diagnosis of the infection were independently associated with a higher risk of death in patients with infections due to CR-GNB. In contrast, receiving antibiogram-guided appropriate treatment was independently associated with a lower risk of death. Our results could be explained by the fact that advanced age and immunosuppression pose a higher risk for increasingly severe forms of infection; septic shock remains the most clinically severe form of an infectious disease. Our results are in accordance with previous reports. In Scotland, Zhao et al.

Discussion
In our study, older age, immunosuppression, and septic shock at diagnosis of the infection were independently associated with a higher risk of death in patients with infections due to CR-GNB. In contrast, receiving antibiogram-guided appropriate treatment was independently associated with a lower risk of death. Our results could be explained by the fact that advanced age and immunosuppression pose a higher risk for increasingly severe forms of infection; septic shock remains the most clinically severe form of an infectious disease. Our results are in accordance with previous reports. In Scotland, Zhao et al. [15] reported that age >60 years and organ failure were associated with higher 30-day mortality in patients with infections due to carbapenemase-producing organisms. In the study carried out by Gualtero et al. [16], the presence of septic shock was associated with increased 30-day mortality in patients with infections due to CRE. Similarly, immunosuppression was associated with higher mortality in patients with CRE infections in a study conducted in Israel [17]. A higher mortality has been associated with inappropriate treatment in the context of CR organisms [18][19][20][21]. Additionally, clinical trials have reported an impact on mortality based on the type of treatment administered [22][23][24]. Inappropriate and second-line treatment options have been associated with unfavourable clinical response and toxicity [25]. Our mortality rate remains within the range previously reported [26].
Respiratory tract and intraabdominal infections contributed to most of the CR-GNB infections. According to our results, the site of infection was not associated with a higher risk of death. To our knowledge, there are no studies comparing mortality between different sites of infection in the setting of CR. A high prevalence of chronic degenerative diseases and immunosuppressive conditions was found in the studied population, which is explained by the characteristics inherent to the population treated in our institution. Unlike previous reports [15,17], we found no isolate-driven differences in mortality. The latter could be because the low number of Acinetobacter sp. isolates. Of note, isolate frequency varied according to the type of infection. Higher frequencies of P. aeruginosa and Enterobacteriaceae in respiratory tract and intraabdominal infections, respectively, were expected. Although some studies correlate the presence of acute kidney injury with higher mortality [27,28], evidence in patients with infections due to CR-GNB infections is scarce.
Among the limitations of our study, its retrospective nature must be considered. Regardless of the retrospective design, systematic data gathering was undertaken, and missing information was scarce and accounted for during the statistical analysis. The singlecentre nature of our study could limit the applicability of the results. In our study, the effect of previous CRE colonisation was not measured. Also, we were unable to report complete genotypic profiles; therefore, specific carbapenem resistance mechanisms' contributions to mortality could not be analysed. In most of the reported studies, complete molecular data on resistance mechanisms are unavailable, so their absence does not necessarily imply a major bias. Although including different infectious syndromes and isolates in the same model could limit our results, the fact that none of them were independently associated with a higher risk of death may provide further strength to our results, being that we describe factors that are associated with worse outcomes regardless of specific clinical scenarios. Nevertheless, sample heterogeneity regarding isolates and clinical scenarios must be considered when interpreting our results. Previous studies have suggested that, rather than carbapenem resistance, difficult-to-treat resistance is associated with mortality [29]. Of note, our study focuses predominantly on CR isolates. Additionally, although antibiogramguided appropriate treatment was independently associated with lower mortality, the fact that some patients not receiving appropriate treatment died within the first days after diagnosis must be considered. Given that the study was not initially designed accordingly, the effect of different antibiotics and antibiotic combinations could not be measured. The significant sample size and extensive data gathering represent strengths in our study. Of note, we chose to study outcomes at 90 days after diagnosis to ensure a better understanding of the long-term impact of infections due to CRO. The latter enabled us to document outcomes that could otherwise be unnoticed. Additionally, patients with infections due to CRO tend to have a higher frequency of non-infectious comorbidities that could impact the outcomes after a 30-day follow-up.

Conclusions
Prospective studies are needed to support our findings. Also, more information regarding resistance mechanisms and mortality is warranted. Because antibiogram-guided appropriate treatment was associated with a lower risk of death, we believe that our study highlights the fact that implementing fast and reliable diagnostic methods is necessary to reduce the time to diagnosis and, hence, appropriate treatment. An accurate diagnosis may contribute to reducing mortality attributed to suboptimal treatment. The recognition of prognostic factors is essential to implement adequate preventive and therapeutic care to improve patient outcomes. Additionally, the recognition of risk factors for adverse outcomes could further assist in patient care and the design of interventional studies that address the severe and widespread problem that is carbapenem resistance.
Advanced age, immunosuppression, septic shock at diagnosis, and inappropriate treatment are associated with higher 90-day all-cause mortality in hospitalised patients with infections due to CR-GNB.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/antibiotics12071130/s1. Tables S1-S10: Table S1. Risk factors for the acquisition of CR-GNB and mortality. Table S2. Laboratory values at the time of diagnosis of CR-GNB infection.

Funding:
The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Informed Consent Statement:
We confirm that all experiments and data management were performed in accordance with the Declaration of Helsinki. Because of the study's retrospective nature, the informed consent requirement was waived by the Institutional Review Board (Comité de Investigación & Comité de Ética en Investigación del Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán). All personal data were protected according to national and international standards.
Data Availability Statement: The datasets used and/or analysed during the current study are available from the corresponding author (B.A.M.-G.) on reasonable request.