Antibiotic Consumption During the Coronavirus Disease 2019 Pandemic and Emergence of Carbapenemase-Producing Klebsiella pneumoniae Lineages Among Inpatients in a Chilean Hospital: A Time-Series Study and Phylogenomic Analysis

Abstract Background The impact of coronavirus disease 2019 (COVID-19) on antimicrobial use (AU) and resistance has not been well evaluated in South America. These data are critical to inform national policies and clinical care. Methods At a tertiary hospital in Santiago, Chile, between 2018 and 2022, subdivided into pre- (3/2018–2/2020) and post–COVID-19 onset (3/2020–2/2022), we evaluated intravenous AU and frequency of carbapenem-resistant Enterobacterales (CRE). We grouped monthly AU (defined daily doses [DDD]/1000 patient-days) into broad-spectrum β-lactams, carbapenems, and colistin and used interrupted time-series analysis to compare AU during pre- and post-pandemic onset. We studied the frequency of carbapenemase-producing (CP) CRE and performed whole-genome sequencing analyses of all carbapenem-resistant (CR) Klebsiella pneumoniae (CRKpn) isolates collected during the study period. Results Compared with pre-pandemic, AU (DDD/1000 patient-days) significantly increased after the pandemic onset, from 78.1 to 142.5 (P < .001), 50.9 to 110.1 (P < .001), and 4.1 to 13.3 (P < .001) for broad-spectrum β-lactams, carbapenems, and colistin, respectively. The frequency of CP-CRE increased from 12.8% pre–COVID-19 to 51.9% after pandemic onset (P < .001). The most frequent CRE species in both periods was CRKpn (79.5% and 76.5%, respectively). The expansion of CP-CRE harboring blaNDM was particularly noticeable, increasing from 40% (n = 4/10) before to 73.6% (n = 39/53) after pandemic onset (P < .001). Our phylogenomic analyses revealed the emergence of two distinct genomic lineages of CP-CRKpn: ST45, harboring blaNDM, and ST1161, which carried blaKPC. Conclusions AU and the frequency of CP-CRE increased after COVID-19 onset. The increase in CP-CRKpn was driven by the emergence of novel genomic lineages. Our observations highlight the need to strengthen infection prevention and control and antimicrobial stewardship efforts.

Antimicrobial resistance (AMR) constitutes a major health crisis causing substantial global disease and economic burden worldwide [1][2][3]. A recent report estimated 1.2 million deaths directly attributable to AMR in the year immediately prior to the emergence of coronavirus disease 2019 (COVID-19) [1]. Further, the impact of AMR is expected to increase, with estimates of approximately 10 million global annual AMR-related deaths by 2050 [4]. The World Health Organization (WHO) declared AMR as one of the most critical public health threats of the century [5].
COVID-19 led to a sharp increase in hospitalizations, a large proportion of which corresponded to high-complexity patients requiring admission to intensive care units (ICUs), invasive procedures, and prolonged hospital stays, in addition S20 • CID 2023:77 (Suppl 1) • Allel et al Clinical Infectious Diseases S U P P L E M E N T A R T I C L E to shortages of healthcare personnel and protective equipment, especially early in the pandemic [6][7][8]. There is growing concern that COVID-19 might have resulted in higher antimicrobial use (AU) and in lapses in infection prevention and control (IPC) practices, both of which could have accelerated the spread of AMR [9][10][11][12]. Recent studies showed an escalation in AU during the pandemic, with up to 74.6% of patients with COVID-19 receiving one or more antibiotics [13,14], despite the relatively low occurrence of secondary bacterial coinfections [15,16]. The most frequently prescribed antibiotics were β-lactams (30%), fluoroquinolones (20%), and macrolides (18.9%) [13]. One study reported a significant increase in the use of broad-spectrum β-lactams (eg, cefepime, piperacillin/tazobactam, and carbapenems) and other last-resort antibiotics (eg, colistin and ceftazidime/avibactam) during the first pandemic peak [17].
Carbapenem-resistant Enterobacterales (CRE) are listed as critical-priority pathogens by the WHO [18]. A report from the US Centers for Disease Control and Prevention highlighted increases in both hospital-onset infections due to CRE and AU in inpatient settings during the first year of the pandemic [19]. Carbapenemase-producing (CP) CRE (CP-CRE) are particularly concerning as they harbor highly efficient enzymes often contained on mobile genetic elements that facilitate their spread, posing a daunting challenge for clinicians and IPC teams. A recent report alerted about an increased detection of CP-CRE after the COVID-19 pandemic in Latin America [20]. However, the magnitude of the impact of COVID-19 in the emergence of AMR remains unknown.
In Chile, official reports have shown that the most important CRE is carbapenem-resistant (CR) Klebsiella pneumoniae (CRKpn), with a prevalence of approximately 35-40%. However, in contrast to other Latin American countries, the prevalence of CP-CRE prior to the pandemic was conspicuously low in Chile [21]. In this study, we evaluated the potential impact of the COVID-19 pandemic on AU and CRE. Moreover, we assessed the emergence of CP-CRE following the COVID-19 pandemic onset.

Study Design and Sample Analysis
We collected hospital-wide data on AU and the frequency of CRE isolation in a public tertiary-care hospital in Santiago, Chile, with 391 beds and a catchment area of approximately 423 000 population (annual hospital discharges: ∼24 300) from March 2018 until March 2022. For context, the first patient with COVID-19 in Chile was diagnosed on 3 March 2020, and antimicrobial stewardship and IPC practices remained unchanged during the pandemic. We compared two years before the pandemic (pre-COVID-19, March 2018-February 2020) with two years after the onset of COVID-19 in Chile (COVID-19, March 2020-February 2022), combining various datasets and analytical strategies.

Data Collection and Processing
Data were abstracted from the hospital's epidemiological and pharmacy records and included total number of beds, patient discharges, patient-days, and intravenous AU for all adult patients admitted to acute care wards during the study period. Acute care wards refer to any patient admitted from the emergency department or by a general practitioner, along with those electively admitted for a surgical procedure. Additionally, we obtained data on monthly ICU admissions and laboratory-confirmed COVID-19 discharges of adult subjects. Antimicrobial use was expressed in defined daily doses (DDDs) per 1000 patient-days and calculated for each intravenous compound as per WHO recommendations [22]. Antibiotics were classified into three groups: (1) broadspectrum β-lactams (ie, ceftazidime, cefepime, piperacillin/ tazobactam, ertapenem, meropenem, imipenem), (2) carbapenems (ie, imipenem, meropenem, and ertapenem), and (3) colistin, a drug frequently used against CP-CRE. Antibiotics evaluated in the study are presented individually in Supplementary Figure 1.
Throughout the study period, we prospectively collected all clinical CRE isolates (ie, nonsusceptible to ≥1 carbapenem as per Clinical and Laboratory Standards Institute [CLSI] 2022) recovered from invasive infections (ie, bloodstream, sterile fluids, or tissues). Isolates were sent to a central laboratory where species identification was reconfirmed by MALDI-TOF (matrix-assisted laser desorption/ionization-time of flight) mass spectrometry. The antibiotic susceptibility profile was reconfirmed using the disk diffusion method following CLSI 2022 [23]. Testing was performed using a multiplex polymerase chain reaction (PCR) designed to detect the three carbapenemases most frequently reported in the country (ie, Klebsiella pneumoniae carbapenemase [bla KPC ], New Delhi metallo-β-lactamase [bla NDM ], and Verona integron-encoded metallo-β-lactamase [bla VIM ]) and was performed in all CRE isolates. Finally, given their high frequency and clinical relevance, we performed wholegenome sequencing (WGS) on all CRKpn isolates recovered during the study period.

Statistical Analyses
Descriptive statistics were used to visualize monthly AU, ICU admissions, and COVID-19 patient discharges. A second-order polynomial fit was adjusted to the data as it presented the best goodness-of-fit (according to the Akaike information criterion [AIC]). The AU rate for each antibiotic group expressed by DDDs per 1000 patient-days was compared between pre-and post-pandemic onset. To further understand AU over time, we calculated a baseline average monthly AU between March 2018 and February 2019. Using this information, we estimated the monthly percentage change for March 2019-February 2020 (prepandemic) and for the two years post-pandemic onset (March 2020-February 2022).
We used interrupted time-series analyses for each antibiotic group [24,25] to evaluate the impact of COVID-19 on AU, adjusting for seasonality and autocorrelation. First, we logarithmically transformed AU rates to adjust their variance over time and computed a first-order differentiation between consecutive time points to correct stationarity. Subsequently, we tested autocorrelation and seasonality among AU group variables [25]. We used an autoregressive integrated moving average (ARIMA) approach through an automated algorithm, based on the best goodness-of-fit reported (eg, lowest AIC/Bayesian information criterion [BIC]), resulting in a seasonal ARIMA (1,0,0) (0,1,1) model [24]. A seasonal ARIMA model is classified as an ARIMA(p,d,q) x (P,D,Q), where (p,d,q) refers to the seasonal and (P,D,Q) to the non-seasonal component. P or p = number of seasonal autoregressive terms, D or d = number of seasonal differences, Q or q = number of seasonal moving average terms. The interrupted side of the model comprised step change and ramp components, derived from any random shift and slope changes in AU over time after the pandemic onset [25]. Finally, we generated a counterfactual scenario related to a hypothetical absence of the COVID-19 pandemic to contrast observed and estimated AU through the backward prediction of the time series as if no random shift and slope changes would have existed. Analyses were conducted using R version 3.2.1 (R Foundation for Statistical Computing).

Whole-Genome Sequencing and Phylogenomic Analyses
We performed WGS using Illumina MiSeq with the Illumina DNA library prep kit (Illumina, Inc). We used FASTQC and MultiQC to determine the read's quality, and Trimmomatic to pair the reads [26,27]. The genomes were assembled de novo with SPAdes, and the quality of the assemblies was assessed with QUAST [28,29]. We used MLST 2.19.0 [30] and ABRicate v1.0.1 15 to determine the sequence type (ST) and the presence of carbapenemases. We annotated genome assemblies with Bakta [31] and evaluated the pangenome using Roary v3.13.0 [32]. A maximum likelihood phylogenomic tree was performed using a core genome definition of 99% with RAxML 8.2.12 [33]. Finally, a recombination-free phylogenomic tree was generated with Clonal Frame ML v1.12 [34] and visualized with the interactive Tree Of Life (iTOL) tool [35].

Ethics
Our study was approved by the Research Ethics Committee of the Clinica Alemana, Universidad del Desarrollo Faculty of Medicine (Institutional Review Board [IRB] 2021-24, Protocol number #UIEC1047).

Hospital Characteristics and Epidemiological Analyses
The first patient with COVID-19 in Chile was diagnosed in early March 2020 and the first pandemic wave peaked in June 2020 [36]. During this peak, our hospital discharged 530 patients with COVID-19 (Supplementary Figure 2A). The total number of beds and average monthly hospital discharges did not significantly vary during the study period. ICU admissions substantially increased after the pandemic onset, with an average of 11 and 25 ICU admissions in the pre-and post-pandemic period, respectively (P < .001). Most ICU admissions (80%) during the pandemic period were patients aged older than 60 years (Supplementary Figure 2B).

Antibiotic Use Over Time and Impact of COVID-19
Compared with pre-COVID-19, we observed a significant increase in mean DDDs per 1000 patient-days during COVID-19, with an overall higher AU of broad-spectrum β-lactams (78.1 vs 142.5; P < .001), carbapenems (50.9 vs 110.1; P < .001), and colistin (4.1 vs 13.3; P < .001) ( Figure 1 and Table 1). Noticeably, the highest surge in AU of broadspectrum β-lactams, carbapenems, and colistin was observed approximately 12 months after the pandemic onset, peaking at 137%, 246%, and 705%, respectively ( Figure 1B). The monthly variation in AU for individual antibiotics is provided in Supplementary Figures 1 and 3. Cefepime, ertapenem, imipenem, meropenem, and colistin drove the increasing trend in consumption among the different antibiotic groups (Supplementary Figure 3). The AU of colistin, imipenem, and meropenem increased after COVID-19's onset in the ICU and in general wards, but remained stable in the emergency department (Supplementary Figure 4).

DISCUSSION
Understanding the drivers of AMR is critical to prevent the spread of multidrug-resistant organisms. Our data from a large public hospital in Chile show an association of the COVID-19 pandemic with increases in broad-spectrum antibiotic use and CRE infections. Notably, during the pandemic period we observed a significant increase in the proportion of CP-CRE, which was particularly relevant for CP-CRKpn, with an approximately 7-fold increase in isolates encoding bla KPC or bla NDM . This increase was driven by the appearance of two distinct genomic lineages of CP-CRKpn: ST1161 (harboring bla KPC-2 ) and ST45 (harboring bla NDM-7 ).
The increase observed in CP-CRE, and especially in bla NDM -harboring organisms, which was previously uncommon in Chile, has been reported in other Latin American countries during the pandemic [20]. In October 2021, the Pan American Health Organization issued an alert on the emergence of and increase in new combinations of carbapenemases in Enterobacterales in the region [37]. Although we did not find CRE harboring more than one carbapenemase, several countries in Latin America have reported the detection of dualproducers after the pandemic [20]. The rapid dissemination of CP-CRKpn ST45 harboring bla NDM-7 observed in 2021 may suggest in-hospital transmission rather than multiple introductions. Hospitals, from different regions, reported challenges maintaining IPC practices, contributing to increases in healthcare-associated infections [38,39]. Importantly, as shown by our data and official reports, our study was performed in a setting of low CP-CRE prevalence pre-COVID- 19 [21], which provides a perfect setting to  ARIMA (1,0,0)(0,1,1 The AR term refers to autoregressive order; the Ramp coefficient indicates the increment at each time point of the time series after the COVID-19 pandemic. The Step change coefficient indicates the augment rate immediately following the intervention; SM is for seasonal moving average. The model used the logged form of the difference in antibiotic consumption over time (by group); hence, coefficients should be transformed for interpretation. The logged time series and autocorrelation functions were computed to indicate if the time series was stationary. Our analysis of the model's residuals indicated they were uncorrelated and had a zero mean. Significance level, α = 5.
assess the COVID-19 impact on the emergence of CP organisms. Our phylogenomic analyses of CRKpn revealed that the increase in CP-CRKpn during 2021 was primarily driven by the emergence of two genomic lineages. ST1161 carried bla KPC-2 , a class A enzyme frequently observed in CRKpn in different parts of the world. In contrast, strains of ST45 harbored bla NDM-7 , a class B metallo-enzyme against which there are very few, if any, reliable therapeutic options. While bla NDM-7 was also found in CP-CRKpn from other genomic lineages (ie, ST25 and ST528), bla KPC-2 was only observed in ST1161, suggesting that bla NDM-7 could be located in a mobile genetic element that facilitates its horizontal transmission into different genomic lineages and perhaps species. Moreover, the fact that bla NDM was observed in non-K. pneumoniae CRE prior to the pandemic and increased during the pandemic, mainly driven by E. cloacae complex, may hint towards horizontal transmission of this genetic trait. The study of genomic platforms with long-read sequencing analyses and transmission dynamics is part of our future research endeavors.
In addition to an increase in CP organisms, we observed an increase in AU after the pandemic onset. Our findings are consistent with previous reports from China suggesting that approximately 70% of patients with COVID-19 received antibiotic treatment during the early stages of the pandemic [40,41]. We observed a prolonged and consistent increase in broad-spectrum β-lactams, carbapenems, and colistin after the first pandemic wave. Antimicrobial use peaked soon after the first year since the pandemic onset and coincided with the increase in CP-CRE. While there is a temporal correlation, our data do not allow us to establish causality. Therefore, the role of the increases in AU in selecting for CRE in general, and CP-CRE in our hospital, remains unclear. Several studies have demonstrated AU to be an independent risk factor for CRE colonization, including a meta-analysis focused on CRKpn [42,43]. Further studies are needed to evaluate the appropriateness and drivers of AU in the hospital and its role in the emergence of CP-CRE.
Our study has several limitations. First, we only performed PCR detection for bla KPC , bla NDM, and bla VIM ; therefore, it is possible that we missed other relevant carbapenemases, leading to an underestimation of the number of CP-CRE isolates. Indeed, a recent communication reported the first detection of bla OXA-48 in CRKpn and Escherichia coli in Chile during the pandemic [36]. However, we did perform WGS in all CRKpn and no other carbapenemases were observed in these analyses. Second, while we analyzed the genomes of all CRKpn (which were by far the most frequent bacterial species), our WGS data did not include other organisms (eg, CR-E. cloacae complex), limiting our ability to draw conclusions about relevant observations such as the expansion of bla NDM -harboring organisms. Third, our analyses are ecological by nature and preclude conclusions regarding any causal effect. Although AU is one of the main drivers of AMR [44], our data do not allow us to rule out the influence of confounding factors, therefore hampering our ability to establish direct causality between AU increase and the emergence of CP-CRE.
Despite these limitations, this is the first report examining the temporal association between COVID-19 and its impact on AU and AMR in Chile. It draws attention to the emergence of genomic lineages of CP-CRE that pose treatment challenges and emphasizes the need for improved antibiotic stewardship and enhanced IPC measures to prevent their spread within healthcare facilities. The use of genomic surveillance provides data to help understand whether there were multiple introductions of new strains or if there is an expansion of a single strain, which hints towards healthcare transmission. It is not known whether bla KPC-2 ST1161 or bla NDM-7 ST45 CRKpn will spread rapidly within Chilean or South American hospitals, but increased vigilance will be warranted.
In summary, our analyses show that AU rate and AMR increased during COVID-19 surges in Chile. Additional studies are necessary to understand the specific ways in which the burden of the pandemic affected AU and AMR rates and whether the increases in AU observed in our data directly increased the risk of AMR among our population. Our findings also highlight the need to build capacity for IPC and antimicrobial stewardship programs. As we move into next phase of the COVID-19 pandemic and recovery, it will be critical to emphasize the need for strong IPC programs, one of the cornerstones of a resilient healthcare system.

Lesson Learned
Strengthening our capabilities to ensure appropriate AU, rapid genome-based surveillance of emerging multidrug-resistant pathogens, and efficient IPC programs is crucial to tackle AMR in the future.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.