Antimicrobial resistance in patients with COVID-19: a systematic review and meta-analysis

The study addresses whether and to what extent bacterial co-infections and secondary infections occur in COVID-19 patients and how prevalent antimicrobial resistance is in these infections. Despite low reported rates of bacterial co-infection (around 3–8%) in COVID-19, antibiotic prescribing remains high (50–75%), raising concerns of accelerating antimicrobial resistance. Surges in COVID-19 hospitalizations have coincided with increases in resistant infections (e.g., MRSA, VRE). The purpose is to quantify prevalence and predictors of bacterial infections and AMR in COVID-19 patients to inform clinical management and stewardship.

Prior reports indicated low prevalence of bacterial co-infection at presentation but higher rates of secondary infections, particularly in critically ill patients, alongside high antibiotic use. Concerns have been raised about pandemic-related increases in AMR and healthcare-associated infections. Existing reviews highlighted coinfection and antibiotic prescribing patterns but lacked comprehensive quantification of AMR prevalence. This study adds large-scale, global estimates focusing on AMR in COVID-19-associated bacterial infections and identifies predictors such as ICU setting and comorbidities.

Design: Systematic review and meta-analysis per Cochrane Rapid Reviews guidance. Registration: PROSPERO CRD42021297344. Data sources: WHO COVID-19 Research Database (MEDLINE, Scopus, CINAHL, ProQuest Central, Embase, Global Index Medicus), any language, Jan 1, 2019–Dec 1, 2021. Inclusion: Studies (cohort, case series, case-control, RCTs) with ≥50 patients, hospitalized or mixed inpatient/outpatient populations, with microbiologically confirmed bacterial infections (excluding nasopharyngeal-only testing) in lab-confirmed COVID-19. Exclusions: Reviews, editorials, letters, preprints, conference abstracts; non-microbiologically confirmed infections; studies not differentiating bacterial from fungal/other viral infections. Outcomes: Primary—prevalence of bacterial co-infection (≤48 h), secondary infection (>48 h), and prevalence of AMR among bacterial infections. Data extraction: Study setting, region, period; demographics (age, sex, comorbidities); COVID-19 therapeutics (e.g., IL-6 inhibitors, corticosteroids); microbiology methods; AMR data per patient and per organism. Bacteria considered per WHO priority list; multidrug resistance defined as non-susceptibility to ≥1 agent in ≥3 classes (or per authors). AMR inclusion threshold required susceptibility/resistance data for ≥4 species (or all if <4 reported) to minimize underestimation. Analysis: Random-effects meta-analysis using generalized linear mixed models (random intercept logistic regression, logit link); between-study variance by maximum-likelihood estimator; heterogeneity via I². Sensitivity analyses stratified by study quality (risk of bias tool for prevalence studies). Meta-regression (univariable and multivariable) to assess effects of patient characteristics, setting (ICU vs general), geography, income level, temporal trends; adjusted for age and COVID-19 severity index (ARDS proportion, ICU proportion, mechanical ventilation proportion). Software: R (metafor, meta). Role of funder: WHO staff involved in design, interpretation, and writing; no role in data collection or analysis.

• Studies included: 148 observational studies (mostly retrospective cohorts), 362,976 patients (Dec 2019–May 2021) across >40 countries; 47% Europe, 27% Americas, others across Western Pacific, Eastern Mediterranean, South-East Asia, and Africa. Most conducted in high-income countries (≈76%). - Settings: 59% all hospitalized; 37% ICU-only; 4% mixed inpatient/outpatient. Median ICU proportion in reporting studies 50%; mechanical ventilation 46%. - Prevalence: • Bacterial co-infection overall 5.3% (95% CI 3.8–7.4); hospitalized 4.9% (3.2–7.6); ICU 8.4% (6.0–11.7). • Secondary bacterial infection overall 18.4% (14.0–23.7); hospitalized 8.4% (6.7–10.3); ICU 39.9% (31.1–49.5). • ICU admission associated with higher odds of secondary infections vs general inpatients (adjusted OR 7.52, 95% CI 4.69–12.06). - Microbiology: 17,423 bacterial isolates reported in 130 studies. Common organisms: Staphylococcus aureus (15%), Klebsiella spp (15%), Pseudomonas spp (10%). In co-infections: S. aureus (21%), coagulase-negative staphylococci (14%), E. coli (12%). In secondary infections: Klebsiella spp (16%), Pseudomonas spp (12%), S. aureus (12%). - Antimicrobial resistance prevalence (comprehensive AMR reporting): • Per-patient: 60.8% (95% CI 38.6–79.3) of bacterial infections resistant (17 studies). • Per-organism: 37.5% (26.9–49.5) isolates resistant (42 studies). • Heterogeneity substantial: I² ≈86% per-patient; ≈96% per-organism. - Pathogen-specific high resistance: • Stenotrophomonas spp: 100% MDR (95% CI 82.6–100). • Acinetobacter spp: 96.5% MDR (86.9–99.1); 95.9% carbapenem-resistant (84.1–99.0). • Klebsiella spp: 88.3% colistin-resistant (6.7–99.9); 69.2% carbapenem-resistant (49.6–83.6). • Carbapenem-resistant Acinetobacter baumannii: 96% of isolates carbapenem resistant; carbapenem-resistant Enterobacterales prevalent (e.g., Klebsiella). - Predictors of AMR: • Higher AMR odds in ICU populations vs general inpatients (adjusted OR 3.69, 95% CI 1.27–10.68). • Higher AMR odds in studies from low- and middle-income countries compared with high-income countries (per-patient and per-organism analyses). • Per-organism adjusted predictors: IL-6 inhibitor use (adjusted OR 1.44, 95% CI 1.03–2.01); diabetes (1.95, 1.19–3.21). • Per-patient adjusted predictors: cardiovascular disease (OR 118.78, 95% CI 7.78–1814.67), diabetes (3.14, 2.19–4.50), IL-6 inhibitor use (1.70, 1.12–2.56), antibiotic use (2.70, 1.28–5.70). - Diagnostic methods: Predominantly culture-based; many studies applied criteria to distinguish infection from colonization. - Risk of bias: Mostly low to moderate; sensitivity analyses by quality showed similar AMR prevalence estimates.

The review demonstrates that bacterial co-infections at presentation among hospitalized COVID-19 patients are uncommon, whereas secondary infections—particularly in ICU—are frequent. Despite this, antibiotic use has been high, underscoring the need for stewardship to avoid exacerbating AMR. A substantial proportion of bacterial infections and isolates in COVID-19 patients are antimicrobial-resistant, with critically ill ICU populations at greatest risk. Elevated AMR in low- and middle-income settings highlights global inequities in diagnostics, surveillance, and stewardship capacity. Identified predictors (ICU care, antibiotic exposure, IL-6 inhibitor use, diabetes, cardiovascular disease) align with mechanisms of nosocomial infection risk and selection pressure from antibiotics and immunomodulation. These insights support targeted stewardship, infection prevention and control, and global AMR surveillance, particularly amid ongoing pandemic pressures.

This study provides global estimates of the prevalence of bacterial co-infections and secondary infections in COVID-19 and quantifies the high burden of antimicrobial resistance among these infections, especially in ICU settings and low- and middle-income countries. Findings support judicious empirical antibiotic use, enhanced AMR surveillance, and strengthened stewardship and infection prevention and control. Future research should include comprehensive, patient-level, multicountry studies to clarify predictors of AMR, expand representation of under-studied regions, and assess the broader population-level impact of the COVID-19 pandemic on AMR transmission and outcomes.

• Incomplete AMR reporting: Many studies reported susceptibility for only one or two organisms, potentially biasing estimates. Inclusion threshold (≥4 species) may still underestimate or, conversely, overestimate AMR if reporting favored high-resistance settings. - Sampling bias: Microbiological testing may have been more frequent in patients failing therapy, inflating resistance estimates. - Geographic imbalance: Under-representation of regions outside Europe and North America, particularly LMICs; high heterogeneity and wide confidence intervals in regional comparisons. - Meta-regression limitations: Likely underpowered for smaller associations; study-level (not patient-level) covariates limit inference. - Societal impact not assessed: The study focuses on hospitalized patients, not community transmission or broader AMR dynamics during the pandemic.