According to institutional infection control policy, perirectal and nasal surveillance swab samples were collected on all patients from Afghanistan at admission to LRMC and WRNMMC. Swab samples were used to screen for methicillin-resistant Staphylococcus aureus (MRSA), extended spectrum β-lactamase (ESBL)–producing bacteria, carbapenemase-producing Enterobacterales (CPE), and vancomycin-resistant Enterococcus (VRE). We defined an MDR organism (MDRO) as MRSA, VRE, CPE, or isolates recovered from clinical or surveillance cultures that were resistant to >1 agent in 3 different antimicrobial classes. We considered recovered isolates that met those criteria surveillance culture to be positive. Patients with negative surveillance cultures were immediately released from isolation; patients with positive surveillance or clinical cultures remained in isolation and on contact precautions to prevent nosocomial transmission. Isolates from all MDROs were sent to the MRSN for whole-genome sequencing. During September–October 2021, WRNMMC and LRMC submitted 171 MDR bacterial isolates from 80 inpatients and outpatients from Afghanistan. We used Illumina Miseq and MiSeq Reagent Kit version 3 (Illumina, https://www.illumina.com) for sequencing (600 cycles, 2 × 300 bp).

Figure 1

Figure 1. Cluster network of all isolates from molecular surveillance of multidrug-resistant bacteria among refugees from Afghanistan in 2 US military hospitals during Operation Allies Refuge, 2021. Isolates are grouped and colored...

Among 80 patients with MDR isolates, 42 were male and 38 were female, and their median age was 26 years (range 1 week to 83 years). The top 5 clinical diagnoses were blast trauma (28%; n = 22), gunshot wound (15%; n = 12), urinary tract infection (11%; n = 9), abscess (10%; n = 8), and pregnancy (6%; n = 5). Among the isolates analyzed, E. coli was the most prevalent (58%; n = 99) bacterial species, followed by K. pneumoniae (13%; n = 22), S. aureus (12%; n = 20), and A. baumannii complex (5%; n = 9) (Figure 1).

After removal of serial isolates from the same patient (same sequence-type determined by in silico multilocus sequence typing), we used the deduplicated dataset of 132 isolates from 80 patients to investigate the prevalence of species and antimicrobial resistance genes. Of the 132 deduplicated isolates, E. coli was the most prevalent (64%; n = 85), followed by K. pneumoniae (12%; n = 16), S. aureus (11%; n = 14), and A. baumannii (5%; n = 6) (Table 1). Of the 80 distinct patients, 59 were identified with solitary MDRO-positive surveillance cultures of E. coli (76%; n = 45), Klebsiella (7%; n = 4), and Enterobacter spp. (2%; n = 1). Nine patients were colonized with multiple species in addition to >1 strain of E. coli: 5 (9%) with E. coli and K. pneumoniae, 2 (3%) with E. coli and VRE, and 2 (3%) with E. coli and 3 additional species. Overall, we found 92% (n = 54) of colonized patients carried an ESBL-positive E. coli. At LRMC, 85% (n = 30; data not shown) of patients who received surveillance cultures were colonized with >1 MDRO.

Among deduplicated E. coli isolates, 84% (n = 71) were ESBL-positive, mostly due to bla_CTX-M-15; 16% (n = 14) were carbapenemase-positive, including 9 isolates carrying bla_NDM-5 metallo-β-lactamase. We noted high diversity among the 85 E. coli isolates and identified 45 distinct sequence types (STs). Lineages ST10 (8 patients) and ST69 (6 patients) were most common. We identified only 2 patients with extraintestinal pathogenic E. coli lineage ST131. A total of 15 patients carried 2 (n = 11) or 3 (n = 4) distinct E. coli lineages. A single isolate of ST38 was shown to coproduce New Delhi metallo-β-lactamase (NDM) 5 and oxacillinase (OXA) 181 carbapenemases (Table 2).

We identified 16 distinct lineages among the 15 patients carrying K. pneumoniae. Overall, 81% (n = 13) of K. pneumoniae isolates carried bla_CTX-M-15, and 19% (n = 3) were carbapenemase-positive due to bla_NDM-1, bla_OXA-181, and bla_OXA-232.

Among other identified MDRO isolates, we identified various antimicrobial resistance genes. Of the 6 patients with A. baumannii–positive cultures, 4 carried the epidemic clone ST2 that had bla_OXA-23 carbapenemase. Two of the E. cloacae isolates carried bla_CTX-M-15. All 14 S. aureus isolates carried the mecA gene, and all 3 E. faecium isolates carried the vanA gene. None of the 4 P. aeruginosa isolates carried high-risk AMR genes.

Figure 2

Figure 2. Phylogenetic analysis of highly related isolates from molecular surveillance of multidrug-resistant bacteria among refugees from Afghanistan in 2 US military hospitals during OAR, 2021. A) Escherichia coliST44, ST69,...

To identify clusters of high genetic relatedness, we compared genomes of all isolates from this study to historical isolates unrelated to OAR that were collected from LRMC and Bagram Air Force Base (AFB). Bagram AFB was located in the northeastern province of Parwan in Afghanistan, 40 km north of Kabul, and was evacuated by US military personnel in July 2021. Bagram AFB housed the Craig Joint Theater Hospital that provided healthcare to US and coalition forces. We identified a total of 6 genetic clusters: 3 E. coli, and 1 cluster each of K. pneumoniae, A. baumannii, and E. faecium (Figure 2). We detected 3 clusters of epidemic E. coli, ST44, ST69, and ST648 (Figure 2, panel A). The ST44 cluster contained isolates from 2 OAR patients whose admissions overlapped at WRNMMC; those isolates differed by 14­–15 single-nucleotide polymorphisms (SNPs). The ST69 and ST648 clusters comprised 1 OAR patient from 2021 and 1 Bagram AFB patient whose isolates were collected in 2018; isolates in those clusters differed by 30 SNPs.

In the K. pneumoniae cluster, 3 serial ST15 isolates that carried bla_OXA-232, bla_CTX-M-15, and rmtF (16S rRNA methyltransferase) were isolated from a single OAR patient at WRNMMC. Those isolates were highly related (17–45 SNPs difference) to isolates previously collected from 5 patients at Bagram AFB in 2018 (Figure 2, panel B). That OAR patient was also colonized with VRE and E. coli ST38 and ST648. In the A. baumannii cluster, 3 OAR patients from LRMC carried ST2 isolates that were highly related to a protracted nosocomial outbreak at Bagram AFB during 2019–2020 (Figure 2, panel C), and differed only by 11–38 SNPs. Last, 3 patients in the E. faecium cluster each had an isolate belonging to the ST80 epidemic clone (Figure 2, panel D). Two of those E. faecium isolates only differed by 7 SNPs from 2 patients in isolation at the same time at LRMC. Those patients were wounded in Afghanistan 2–3 weeks before evacuation to LRMC and were admitted 1 day apart from each other.

Among the S. aureus isolates, we identified ST1482 in 43% (n = 6) of patients; the other patients carried clones ST30 (n = 4), ST22 (n = 3), and ST772 (n = 1). Although many patients shared the same ST, we did not identify any clusters of isolates sharing high levels of genetic relatedness.

In this study, MDRO surveillance was conducted at 2 US military hospitals that provided medical care to refugees from Afghanistan during OAR. We observed high rates of MDRO colonization with E. coli and found that many isolates carried ESBL, carbapenemase, and 16S methyltransferase genes. Furthermore, we observed many AMR K. pneumoniae, A. baumannii, E. faecium, and S. aureus clinical isolates. We found that >75% of ESBL-producing E. coli isolates produced CTX-M-15. Among the 14 carbapenemase-producing E. coli isolates, we isolated E. coli ST648 coproducing NDM-5 and RmtB in samples from 5 patients. ST648 has been characterized by a combination of multidrug resistance, high-level virulence, and biofilm formation, similar to the global high-risk ST131 clonal lineage (7). One patient was colonized with K. pneumoniae ST11 that coproduced NDM-1 and RmtC, a lineage that is commonly known for high virulence and resistance to all β-lactams, including carbapenems, and resistance to aminoglycosides (8).

One patient from WRNMMC had 3 serial K. pneumoniae ST15 isolates from an abdominal wound and surveillance cultures. Surveillance cultures from that patient were also positive for VRE, P. aeruginosa, and 2 ESBL-producing strains of E. coli, ST38 and ST648. In addition to bla_OXA-232, bla_CTX-M-15, and rmtF, the K. pneumoniae ST15 isolates carried markers of hypervirulence and hypermucoviscosity (data not shown). The siderophores, yersiniabactin and aerobactin, rmpA2 (regulator of mucoid phenotype), and peg344 (a drug and metabolite transporter) were detected in those isolates, and all of those have been associated with invasive disease in immunocompetent patients (9). OXA-232–producing K. pneumoniae ST15 predominantly is found in China, where multiple nosocomial outbreaks have been reported, but the lineage also has been described globally (10).

Nosocomial outbreaks causing severe illness and death have been reported in medical treatment facilities in Iraq and Afghanistan and are a great concern for the US military healthcare system (11). Although phylogenetic analysis in this study revealed multiple clusters of highly related isolates, patients in those clusters were likely colonized or infected before entering US military hospitals. Prevalence of high-risk AMR genes is high in Afghanistan, and patients could have acquired environmental MDRO; however, those possibilities would be difficult to determine because microbiologic sampling was conducted after evacuation to LRMC. Alternatively, patients could have acquired MDR bacteria during hospitalization in Afghanistan before evacuation. However, we do not know the extent of medical treatment patients received in Kabul.

The A. baumannii ST2 isolates collected during this study were genetically related (11–38 SNPs) to nosocomial ventilator-associated pneumonia and bloodstream infections collected at Bagram AFB during 2019–2020 (C. Anderson et al., unpub. data). Two clusters were identified in that outbreak, comprising 10 patients from Afghanistan with A. baumannii ST2. The 2 clusters containing ST2 were separated by 80 allelic differences, which suggests 2 distinct strains of ST2 were involved in nosocomial transmission. Isolates within those clusters were found to differ by only 3–13 SNPs in 1 cluster and 1–4 SNPs in the other. Patients in that outbreak were transferred to Bagram AFB after initial treatment in hospitals in Afghanistan, but an outbreak investigation at Bagram AFB did identify lapses in infection control protocols that could have led to nosocomial transmission within those clusters after independent introductions of the strains from hospitals in Afghanistan. Despite the high prevalence of MDROs observed during OAR and the genetic similarities between some of the isolates, we found no evidence to suggest nosocomial spread occurred at LRMC or WRNMMC. That point highlights how crucial rapid screening, patient cohorting, and preemptive isolation upon hospital admission are to preventing nosocomial outbreaks within the military healthcare system.

We observed high rates of MDRO colonization in this study. Risk factors for MDRO acquisition include international travel, travel-related diarrhea, antibiotic use, and prolonged hospitalization (12). Among military personnel, Afghanistan-based personnel have been found to have a 5.5-fold higher prevalence of MDR E. coli colonization than US-based personnel (13). At LRMC, >80% of patients from Afghanistan had a positive MDRO surveillance culture, much higher than the 30% rate among US patients during the same timeframe (data not shown). Those rates are consistent with studies that describe high MDRO colonization rates or MDRO outbreaks among refugees from regions with high AMR prevalence. A 2016 study in Germany observed an MDRO prevalence in hospital-admitted refugees of up to 60.8%, and a preponderance of ESBL-producing Enterobacterales (14). Of note, >50% of the refugees in this study were from Syria or Afghanistan, and only 16.7% of nonrefugee patients were found to be positive for MRDOs. Similar observations in Germany, such as increased reports of NDM-1– producing K. pneumoniae in refugees and injured soldiers evacuated from Ukraine, have also been reported (15). Of note, prior hospitalization in Ukraine is now considered a major risk factor for MDRO colonization, because infection control measures have become inadequate due to limited resources and personnel since the 2022 invasion by Russia.

Although this study provides a unique snapshot of the AMR burden in Afghanistan, our results are limited by the small sample size and sampling bias because data were only available for refugees who were evacuated from Kabul and received care at LRMC or WRNMMC. This study also lacked detailed medical history before medical evacuation, such as the extent of hospitalization in Kabul. Further, microbiologic sampling was conducted after patients were evacuated from Afghanistan. Therefore, whether MDRO acquisition occurred during hospitalization or within the community environment because of high AMR prevalence in this region cannot be determined.

In summary, this study describes a high prevalence of MDRO among refugees from Afghanistan evacuated to US military hospitals during OAR. Molecular surveillance identified multiple high-risk clonal lineages that were characterized by extensive AMR and hypervirulence profiles and were genetically related to historical isolates collected during the war in Afghanistan. Unlike other reports that focused primarily on war wounds among military personnel injured in Afghanistan, this study uniquely provides detailed genomic AMR data, from both clinical and surveillance isolates, on refugees from Afghanistan who received care within the US military healthcare system. When viewed in the context of rising global AMR transmission and unprecedented population displacement, these data are a reminder of how crucial robust infection control measures and surveillance are to protecting public health.