Antibiotic resistance associated with the COVID-19 pandemic: a systematic review and meta-analysis


                  Background
                  COVID-19 and antimicrobial resistance (AMR) are two intersecting global public health crises.
               
                  Objective
                  We aim to describe the impact of the COVID-19 pandemic on AMR across healthcare settings.
               
                  Data Source
                  A search was conducted in December 2021 in World Health Organization’s COVID-19 Research Database with forward citation searching up to June 2022.
               
                  Study Eligibility
                  Studies evaluating the impact of COVID-19 on AMR in any population were included and influencing factors were extracted. Reporting of enhanced infection prevention and control (IPAC) and/or antimicrobial stewardship programs (ASPs) were noted.
               
                  Methods
                  Pooling was done separately for Gram-negative and Gram-positive organisms. Random effects meta-analysis was performed.
               
                  Results
                  Of 6036 studies screened, 28 were included and 23 provided sufficient data for meta-analysis. The majority of studies focused on hospital settings (n=25, 89%). The COVID-19 pandemic was not associated with a change in the incidence density (IRR 0.99, 95% CI: 0.67 to 1.47) or proportion (RR 0.91, 95% CI: 0.55 to 1.49) of MRSA or VRE cases. A non-statistically significant increase was noted for resistant Gram-negatives (i.e., ESBL, CRE, MDR or carbapenem-resistant Pseudomonas aeruginosa or Acinetobacter baumannii, IRR 1.64, 95% CI: 0.92 to 2.92; RR 1.08, 95% CI: 0.91 to 1.29). The absence of reported enhanced IPAC and/or ASP initiatives was associated with an increase in Gram-negative AMR (RR 1.11, 95%CI: 1.03 to 1.20). However, a test for subgroup differences showed no statistically significant difference between the presence and absence of these initiatives (P=0.40).
               
                  Conclusion
                  The COVID-19 pandemic could have hastened the emergence and transmission of AMR, particularly for Gram-negative organisms in hospital settings. But there is considerable heterogeneity in both the AMR metrics utilized and the rate of resistance reported across studies. These findings reinforce the need for strengthened infection prevention, antimicrobial stewardship, and AMR surveillance in the context of the COVID-19 pandemic.
               
                  PROSPERO registration
                  
                     CRD42022325831.
               


Background
High antibiotic use in patients with COVID-19 threatens to contribute to the antimicrobial resistance (AMR) crisis. Although antibiotics do not treat COVID-19, they are commonly used because of initial diagnostic uncertainty in patients presenting with respiratory illness, and of concern for bacterial co-infection or secondary infection in those with confirmed COVID-19. In previous rapid reviews, we found high antibiotic prescribing (approximately 75%) to patients with COVID-19 despite the relatively low bacterial infection rates, particularly in patients outside of the ICU setting (<10%). [1][2][3] Our most recent systematic review identified COVID-19 patients as a potential important reservoir for antimicrobial resistance. Over 60% of patients with COVID-19 who had a bacterial infection carried a highly resistant organism. 4 Due to person-to-person transmission of organisms, particularly in healthcare settings, this presents a threat to the broader population beyond those with COVID-19.
While substantial inappropriate antibiotic prescribing has occurred in patients with COVID-19, antibiotic use for other infectious syndromes has declined early in the pandemic, particularly in community settings. 5,6 This could potentially be due to the attenuation of transmission of other viral and bacterial pathogens due to public health measures to contain COVID-19, including physical distancing and masking. Enhanced infection prevention and control activities in healthcare settings could further mitigate the impact on AMR. 7 Given potentially opposing effects, it is unclear how selection of AMR in bacteria has occurred across populations during the pandemic. Emerging data from the United States Centers for Disease Control and Prevention suggests the pandemic has resulted in rising rates of AMR, including carbapenem-resistant Acinetobacter and extended spectrum beta-lactamase producing Enterobacterales. 8 While we have reported that antimicrobial resistance is high in individual patients with COVID-19 and bacterial infection, the ecological impact of the pandemic on AMR at the population level is not yet well-described. In this analysis, we present the findings of a systematic review and meta-analysis describing the impact of the COVID-19 pandemic on AMR across healthcare settings.

Searches
We performed a comprehensive search of the World Health Organization (WHO) COVID- 19 Research Database for published literature in any language from January 1, 2019 to December 1, 2021. The WHO COVID-19 Research Database is a comprehensive multilingual source of COVID-19 literature updated weekly that includes citations from Medline, Scopus, CINAHL, ProQuest Central, Embase, and Global Index Medicus. 9 The J o u r n a l P r e -p r o o f search strategy was structured to include co-infection or secondary infection terms and bacterial infection terms which were applied to the COVID-19 literature in the database. The full search strategy is available in the supplement. Forward citation searching was performed in Google Scholar to capture more recent publications up to June 2022. 10

Study Eligibility
All studies in inpatient and outpatient settings were eligible for inclusion. The following inclusion and exclusion criteria were applied:

Inclusion Criteria
1. Study provides data on AMR before (before January 2020, or as identified by authors) vs. during the COVID-19 pandemic (January 2020 or later, or as identified by authors) in a specific healthcare setting. 2. AMR is reported as 1) incidence density rate (e.g., rate per 1000 patient days or per patient population), and/or 2) effect measure (e.g., risk, odds, rate ratio) of antimicrobial resistance, and/or 3) prevalence of antimicrobial resistant organisms (e.g., methicillin-resistant Staphylococcus aureus (MRSA) out of all Staphylococcus aureus).
Exclusion Criteria 1. Reviews, editorials, case studies, case series, letters, pre-print publications, dissertations, poster presentations. 2. Studies including <100 patients. 3. Studies combining bacterial and non-bacterial co-infection as a single metric.

Population
Individuals receiving care in any healthcare setting and in any age group.

Main Outcomes
The main outcome is the incidence of AMR in the population associated with COVID-19, either expressed as an incidence density rate (antibiotic resistant infections per 1000 patient days) or proportion (e.g. proportion of S. aureus that were MRSA, proportion of patient admissions with resistant infection).

Data Screening and Extraction
Records were managed using Covidence bibliographic software. All titles and abstracts were screened by a single author (in our previous review, 4 there was substantial reviewer agreement, kappa: 0.66). Full text screening was performed by at least a single author (in the previous review, we determined kappa to be substantial at 0.62 to 0.68). A single review author extracted study characteristics and data according to a pre-defined list of study elements, with a second check by another review author. Study characteristics including design, patient population, and AMR metrics were extracted. We also extracted whether the authors indicated infection prevention and control (IPAC) measures were strengthened during the pandemic and/or whether there was an antimicrobial stewardship program (ASP) in place. This was categorized into two groups: 1) reporting of enhanced IPAC or ASP or 2) reported no enhanced IPAC/ASP OR did not report enhanced IPAC/ASP. These variables were extracted in order to stratify changes in AMR based on potential AMR-mitigating factors.

Risk of Bias Assessment
We used a 10-item validated risk of bias in prevalence studies tool incorporated into data extraction. 11

Data Analysis
Findings were summarized descriptively. In studies providing complete numerator and denominator data, incidence rate ratios (IRR) were pooled using a GLMM random-effects meta-analysis and risk ratios (RR) were pooled using Mantel-Haenszel random effects metaanalysis with between-study variance estimated using the Paule-Mandel estimator. Results were presented in forest plots and pooled across Gram-positive and Gram-negative organisms, stratified by the reporting of enhanced IPAC measures and/or ASP. All analyses were carried out using R version 4.1.2 with the packages metafor and meta.

Results
Of 6036 studies identified via literature search, 28 were eligible for inclusion (18 via full-text screening, 9 via forward citation screening, and 1 expert-identified; Figure 1). 13-40 The most common countries of origin were the United States (n=4), Italy (n=4), and Brazil (n=3). Patient populations studied included all hospitalized patients (n=17), those hospitalized in intensive care units only (n=5), special populations (e.g., oncology, surgery) (n=3), mixed hospitalized and community-dwelling patients (n=2), and community-dwelling patients only (n=1). Studies evaluated a range of both community-and healthcare-acquired infection.
Combined healthcare and community-acquired infection or setting of acquisition commonly undistinguished (n=15), followed by only healthcare-associated (n=11), and only communityacquired (n=2). The majority of studies derived resistance data from clinical specimens (n=20), six included both clinical and screening specimens or did not specify the type of specimen, followed by two studies using screening specimens only. Most studies had moderate risk of bias (n=18), followed by low (n=5), and high (n=5) risk of bias ( Table 1). The most common reasons for downgraded risk of bias included inconsistent or lack of reporting on mode of data collection across study subjects, lack of reporting of case definitions, and lack of reporting of complete numerator and denominator data (Supplementary Table 1).

Measures of Antimicrobial Resistance
Incidence density (e.g., cases of resistant infections per 1000 patient days) was most commonly used to measure a change in AMR associated with COVID-19 (n=11) or proportion of isolates or infections (e.g., percentage of S. aureus cases that were MRSA, n=11), followed by incidence (e.g., cases per admission or discharges, n=5), and other (standardized infection ratio, point prevalence n=2). Study details and AMR metric directionality are provided in Table 1. Of the 28 eligible studies, 23 (82%) provided raw numerator and denominator data to facilitate meta-analysis.

Resistance in Gram Positive
Organisms MRSA Over 6,848,357 patient days of follow-up, our meta-analysis found that the COVID-19 pandemic was not associated with a change in incidence rate of MRSA (IRR 1.03, 95%CI: 0.65 to 1.62, I 2 =95%, n=5). Similarly the COVID-19 pandemic was not associated with a change in the proportion of cases that were MRSA (RR 0.91, 95%CI: 0.60 to 1.36, I 2 =93%, n=7).

VRE
Over 356,056 patient days, meta-analysis shows that the COVID-19 pandemic was not associated with a change in the incidence of VRE ( (ESBL)-producing (or third generation cephalosporin resistant) Enterobacterales One study with 87,204 patient days of follow-up found an increased IRR associated with the COVID-19 pandemic (IRR 15.20, 95%CI: 4.90 to 47.14). However the proportion of cases with an ESBL-producing organism was not significantly altered with COVID-19 (RR: 1.10, 95%CI: 0.91 to 1.33, I 2 =94%, n=8).

Overall Gram-Negative Resistance and Association with Infection Prevention and Antimicrobial Stewardship
When pooling all resistant Gram-negative organisms, there was a non-statistically significant association between COVID-19 pandemic and the incidence rate (IRR 1.64, 95%CI: 0.92 to 2.92, I 2 =93%, n=14) as well as the proportion of cases that were resistant (RR 1.08, 95%CI: 0.91 to 1.29, I 2 =92%, n=22). The lack of reporting of enhanced IPAC and/or ASP was significantly associated with an increase in Gram-negative AMR (RR 1.11, 95%CI: 1.03 to 1.20, I 2 =88%, n=5), whereas no significant association with AMR was seen in studies that did report such initiatives (RR 0.80, 95%CI: 0.38 to 1.70, I 2 =90%, n=17). A test for subgroup differences showed no statistically significant difference between the presence and absence J o u r n a l P r e -p r o o f of reported enhanced IPAC/ASP interventions when evaluating changes in AMR (P=0.40). Figure 4 and 5.

Discussion
AMR definitions and reporting in the context of the COVID-19 pandemic is variable with substantial heterogeneity in reported outcomes among studies. We found that although the COVID-19 pandemic was not associated with a change in Gram-positive AMR, Gramnegative resistance appears to have increased (MDR or carbapenem-resistant P. aeruginosa or Acinetobacter spp., ESBL and CRE), particularly in settings where enhanced IPAC and ASP initiatives were not reported.
A recent special report from the US Centers for Disease Control and Prevention (CDC) found a 15% increase in the rate (per discharge or admission) of resistant organisms including carbapenem-resistant Acinetobacter, MRSA, CRE, and ESBL. 8 Study location, burden of COVID-19, burden of non-COVID-19 respiratory infections, background epidemiology, and antimicrobial prescribing practices may partially explain the difference between the CDC data and our findings. The apparent increase in incidence of Gram-negative AMR but not Grampositive AMR suggests antibiotic prescribing may play an important role, given the high use of beta-lactam/beta-lactamase inhibitors and third generation cephalosporins in patients with COVID-19. 3 Nevertheless, both the CDC report and this systematic review present a concern that the COVID-19 pandemic may play a role in increasing rates of AMR in the population.
The concern for increasing AMR in the context of COVID-19 has been previously highlighted. 41,42 Inappropriately high usage of antibiotics in patients with COVID-19 selects for resistant organisms which can potentially be transmitted to the broader population. 42 We have previously shown that in the context of COVID-19 co-infection and secondary infections, 38% of organisms and 61% of patients harbour AMR. 4 This analysis extends the concern for drug resistance beyond COVID-19 infected patients themselves to document an ecologic impact of the pandemic on AMR.
The relationship between COVID-19 and AMR is complex, as several factors such as improved hand hygiene, personal protective equipment use, and physical distancing may help to reduce the transmission of AMR organisms, at least temporarily while such enhanced measures are in place. 7 On the other hand, shortage of medical personnel and personal protective equipment during the pandemic could thwart this effect. Our findings reinforce that infection prevention and control activities are important mitigating factors limiting the growth of AMR associated with COVID-19.
While this study provides a broad global view of AMR in the context of COVID-19 from an ecological perspective, there are several important limitations. There is significant heterogeneity in methodology and AMR outcome measures reported across studies. At least five different AMR metrics were provided (incidence density, incidence per admission/discharge, proportion of infections, standardized infection ratio, point prevalence), which prevents direct comparison and makes meta-analysis challenging. A lack of adjustment for confounding factors raises the possibility that changes in AMR over time may be due to changes in patient populations or other underlying factors. And lack of longitudinal data also limits the ability to account for existing temporal trends in AMR incidence and prevalence, as well as more distal changes that continue to evolve after the pandemic. Differences in regional baseline rates of AMR or epidemiology may also account for the heterogeneity seen, and as such individualized assessment of regional AMR surveillance data is needed. Many studies did not comment on other confounding factors such as the presence or intensity of their infection prevention and control or antimicrobial stewardship program, so in studies reporting these interventions, the association with reduced AMR may represent correlation rather than causation. Several studies only reported a small number of pathogens with AMR, hence there is a risk of selective outcome reporting. It is also important to note that less than 20% of studies had low risk of bias, suggesting that higher quality studies are needed to better understand the impact of COVID-19 on AMR.

Conclusion
The COVID-19 pandemic could play an important role in the emergence and transmission of resistant pathogens, particularly for Gram-negative organisms in hospital settings. There is considerable heterogeneity in both the AMR metrics utilized and the rate of resistance reported across studies. Our findings reinforce not only the need for strengthened infection prevention and antimicrobial stewardship, but also robust and consistent AMR surveillance as part of the pandemic response and recovery.

Transparency Declaration
This study was supported by funding from the World Health Organization

Conflict of Interest Disclosure
The authors have no conflicts of interest to declare SB is staff at WHO. This paper solely reflects the view of the authors and do not necessarily reflect the view of the Organization.

Author Contributions
Concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Langford
Critical revision of the manuscript for important intellectual content: All authors.