Environmental reservoirs of carbapenem-resistant organisms in the intensive care unit: a multicenter longitudinal study in two middle-income country hospitals

Abstract

Background

There is limited data regarding environmental reservoirs of carbapenem-resistant organisms (CRO) during non-outbreak settings in resource-limited hospitals, or the role of these reservoirs in healthcare transmission.

Methods

Prospective longitudinal study in which sinks and high-touch surfaces (HTS) were sampled prior to room cleaning in intensive care units (ICUs) in two hospitals (hospital A, Argentina, and hospital B, India), July 2023-February 2024. Selective media was used to recover CROs. Whole genome sequencing (WGS) and single nucleotide polymorphism (SNP) pairwise analysis were performed on environmental and clinical isolates to evaluate bacterial transmission dynamics. Metagenomic sequencing was performed to evaluate bacterial diversity of environmental samples.

Results

Of 541 environmental samples collected, 47.9% in hospital A and 97.5% in hospital B grew at least one CRO. Most CROs tested for the presence of a carbapenemase were positive (63.9–91.0% for hospital A and B isolates, respectively). Carbapenemase producer (CP)-Acinetobacter baumannii and CP-Pseudomonas spp. predominated in HTS and sinks samples, respectively, in hospital A; while CP-Klebsiella pneumoniae predominated in hospital B samples. WGS of 113 CRO isolates and SNP analysis demonstrated certain lineages established enduring reservoirs in the ICUs environment (e.g., blaVIM-36 P. aeruginosa ST395 isolates with 2–9 SNP difference were detected in sinks over 7 months). Several clusters involving environmental and clinical isolates that shared an epidemiological link and displayed ≤ 10 SNP difference were identified (e.g., blaOXA-23 A. baumannii ST195 isolated from three unique patients who stayed in the same private room on sampling months 4, 5, 6 and 7, and from HTS of that room on sampling month 5 displayed 0–3 SNP difference). Metagenomic analysis identified additional AMR genes of clinical importance.

Conclusions

CROs were abundant and persisted in the ICU environment in countries with high prevalence of MDROs. Our data suggests movement of clones between the environment and patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13756-026-01768-x.

Background

Healthcare-associated infections (HAIs) and antimicrobial resistance (AMR) are more prevalent in low- and middle-income countries (LMIC) than high-income countries [1–4]. Antibiotic use and healthcare transmission have been identified as key factors contributing to AMR in humans (5). The hospital-built environment has been implicated as a source of healthcare transmission of multidrug-resistant organisms (MDROs) [6–9]. A recent study showed recent hospitalization, hospitalization in more crowded units, and hospital stay > 4 days were strongly associated with carbapenem-resistant organisms (CROs) colonization [10]. However, there is limited data on environmental reservoirs of MDROs in non-outbreak settings, especially in LMICs. In this study, we aimed to characterize environmental reservoirs of CROs in the intensive care unit (ICU) in two middle-income countries, and to assess healthcare CRO transmission.

Methods

Study aim, design and setting

The aim of this study was to characterize environmental reservoirs of CROs in the ICU of two middle income country hospitals, and to investigate their role in healthcare transmission. We prospectively collected environmental samples from high-touch surfaces (HTS) and sinks from ICUs of two hospitals over 8 consecutive months, July 2023 to February 2024. Specifically, we included two ICUs from a 500-bed acute care non-profit teaching hospital in Argentina (hospital A), and two ICUs from a 1926-bed acute care public teaching hospital in India (hospital B). The ICUs in Hospital A have a mix of private rooms and open bay design, while both ICUs in Hospital B are open bay units. Characteristics of participating hospitals’ infection prevention and environmental cleaning programs are summarized in Suppl. Table 1. This study was supported by the Centers for Disease Prevention and Control through a Broad Agency Announcement. Study sites received financial support through JHU (award recipient) to implement study protocols. JHU provided technical support for all study activities.

Environmental sampling and processing protocols

As practices, room design, and room equipment are different between sites, we developed a list of HTS based on observations of healthcare worker-patient care interactions at each participating site using a standardized tool (Supplementary material). All dry and wet surfaces were collected by trained personnel prior to routine room cleaning using a composite sponge technique capable of sampling up to 2,258.06 cm [11]. We constructed composite samples to include one large surface area (e.g., bed rails) and 2–3 smaller sites (see Supplementary material for full description of composite samples and processing protocol). Each room/patient area was sampled once a month (detailed description of the sampling schedule is summarized in Suppl. Table 3). All sink drain samples were collected with the ESwab® 480C Regular Flocked Swab with Liquid Amies medium [12].

Assessment of microbial burden

We assessed the microbial bioburden of the environmental samples by plating samples on tryptic soy agar (TSA) with 5% Sheep Blood (TSA-SB) undiluted, and in tenfold and 100-fold dilutions (more details in Supplementary material). Colonies were manually counted within 48h following a clockwise quadrant approach. Any count > 100 colonies/plate was registered as too numerous to count (TNTC).

Bacterial identification, antimicrobial susceptibility testing (AST), and carbapenemase detection

For organism identification, samples were either plated directly on a MacConkey agar plate with an ertapenem and a meropenem disk or after sample enrichment with tryptic soy broth (TSB) (hospital A, plated enriched samples onto CHROMagar™ KPC). For any growth ≤ 25 mm of disk(s) or any growth on the CHROMAgar™ KPC plate, bacteria were subcultured for further isolation on a TSA-SB plate (each morphology on a new plate) [13]. Hospital A used matrix assisted laser desorption ionization (MALDI)-TOF (Bruker) for organism identification and Phoenix-BD for AST. Hospital B used Vitek Compact 2 (bioMérieux) for both organism identification and AST. Carbapenemase (CP) detection was performed with the BD MAX™ CPO assay (Beckton Dickinson) (hospital A), or the modified carbapenem inactivation method (mCIM) (hospital B), modified to use CAMHB broth and 10 µL inoculum [14] on each of the pure morphology growing on the TSA-SB agar purity plates, and lateral flow assay.

Clinical samples

A subset of CRO isolates from rectal samples obtained through weekly surveillance protocols in participating ICUs or from clinical cultures obtained during routine care between September 2023-January 2024 were included in whole genome sequencing (WGS). CRO identification in clinical specimens followed the same approach previously described (CHROMagar™ KPC media in hospital A and MacConkey culture with ertapenem and meropenem disks in hospital B).

Next-generation sequencing (NGS) of environmental and clinical samples

Whole-genome sequencing (WGS) was performed on all environmental CRO enterobacterales, Pseudomonas aeruginosa and A. baumannii isolates recovered from hospital A and in a subset of these isolates from hospital B. Fewer isolates were sequenced in hospital B due to limited resources for processing. Both WGS and shotgun metagenomic sequencing (mNGS) was performed locally at the National Institute of Microbiology “Dr. Carlos G. Malbrán” for samples collected from hospital A and on premises for samples from hospital B following standardized protocols for the respective site. Human reads were removed prior to downstream analysis using standard human genome filtering. Details on WGS and metagenomic sequencing procedures are described in Supplementary material.

Cluster identification

Genetic relatedness was assessed by pairwise single nucleotide polymorphisms (SNPs) analysis between isolate pairs of same species and multilocus sequence typing [MLST] using Snippy [15]. A cluster was defined as isolates with ≤ 10 SNP difference from samples with a spatiotemporal connection (i.e., collected from the same unit at different time points or collected on the same date from different rooms or units) as determined in previous studies [16, 17].

Statistics

Descriptive statistics were used to report the proportion of samples growing CROs or CP-CROs. Analysis was performed in R (version 4.3.1). To estimate bacterial quantity, we calculated median and range per cc/surface area (CFU/mL/cm²).

Results

Environmental bacterial burden by quantitative and qualitative culture-based methods

A total of 541 environmental samples were collected over 8 months (Suppl. Table 3). The microbial burden by area sampled is shown in Table 1. The highest burden was observed in sink splash zones (median growth 1,105.80–1,439.03 CFU/cm² in hospital A and B, respectively) and sink drains (median 1 × 10⁶ CFU/mL). Overall, 47.9% (146/305) of environmental samples from hospital A and 97.5% (232/238) samples from hospital B, grew at least one CRO on culture (median CRO isolates per sample was 1 [IQR 1–3]), Table 2. Most CROs tested were positive for a carbapenemase (74.5%, hospital A and 93.2%, hospital B. The distribution of carbapenemase varied by bacterial species and area sampled (Figs. 1 and 2). Predominant CP-CROs isolated from sinks included bla_VIM-Pseudomonas spp. (hospital A) and bla_NDM- K. pneumoniae (hospital B). HTS nearest to patient beds showed predominance of bla_OXA-23-Acinetobacter spp. and NDM-K. pneumoniae (hospital A), and bla_NDM- K. pneumoniae and blaNDM + blaOXA-48 K. pneumoniae (hospital B). CP-CROs were absent from nursing station samples in hospital A. Among CP-CRO isolates from hospital A, 5% (5/99) were double producers including bla_KPC + bla_{NDM (}C. freundii [n = 2], bla_NDM + bla_OXA-23 (A. baumannii [n = 2]), and bla_NDM + bla_OXA-58 (A. baumannii [n = 1]). In hospital B, 32% (68/212) were double producers, including bla_OXA-48-like + bla_NDM (K. pneumoniae [n = 51], E. coli [n = 13], E. cloacae [n = 2]) and bla_VIM + bla_NDM (K. pneumoniae [n = 1], and Chromobacterium violaceum [n = 1]). Additional mechanisms of resistance to carbapenems were present in isolates from both hospitals including point mutations associated with efflux pumps (mexR Q94STOP, adeS H189Y), porins (oprD W339STOP, oprD V359L, OmpK36_T136TDT, OmpK36_D135DGD) and binding to penicillin binding proteins (FtsI_N337NYRIN, Suppl. Tables 4, 5, and 6).

Table 1.

Aggregate microbial growth by sampling area at each site (rounded to the nearest whole number). Samples were plated using non-selective media (tryptic soy agar with sheep blood [TSA-SB]), undiluted, tenfold, and 100-fold dilution. Samples collected in two middle-income country hospitals between July 2023 and February 2024

Sample area	Hospital A			Hospital B
	Median (Min–Max) colony forming units/surface area (CFU/cm2)			Median (Min–Max) colony forming units/surface area (CFU/cm2)
	ICU A	ICU B	Overall	ICU A	ICU B	Overall
Patient room	11 (< 1–1743)	2 (< 1–1735)	4 (< 1–1743)	39 (1–1608)	31 (< 1–1542)	35 (< 1–1608)
	n = 63	n = 38	n = 101	n = 78	n = 104	n = 182
Nursing station	1 (< 1–1845)	1 (< 1–28)	1 (< 1–1845)	19 (3–1984)	5 (< 1–130)	8 (< 1–1984)
	n = 8	n = 8	n = 16	n = 14	n = 14	n = 28
Sink splash zones	1106 (1–2542)	1106 (2–1929)	1106 (1–2542)	1,610 (1610–1610)	1268 (1268–1268)	1439 (1268–1610)
	n = 19	n = 27	n = 46	n = 7	n = 7	n = 14

Table 2.

Carbapenem-resistant organism (CRO) and carbapenemase-producing CRO (CP-CRO) detection among environmental samples (aggregated data). Samples were collected from ICUs in two middle-income country hospitals between July 2023 and February 2024. Not all CROs were tested for CP production (see Methods for criteria for CP testing)

Specimen type collected in intensive care units	Sample collections, n	CRO positive samples, n (%)	CP-CRO among CRO positive samples tested for CP, n (%)	Median CRO isolates per CRO positive specimen, n (IQR)	Unique CP-CRO isolates/CRO isolates tested for CP, n (%)
Hospital A	305	146 (47.9)	79 (74.5)	2 (1–3)	99/155 (63.9)
Nursing station	16	1 (6.2)	0	1 (1–1)	0/0
Common area sink drain	10	9 (90)	9 (100)	3 (2–3)	15/20 (75)
Common area sink splash zone	16	16 (100)	14 (93.3)	3 (2–3)	23/30 (76.7)
Patient bathroom sink drain	24	24 (100)	13 (81.2)	2 (2–3)	16/26 (61.5)
Patient bathroom sink splash zone	56	53 (94.6)	17 (58.6)	2 (1–3)	18/37 (48.6)
Patient room/area	183	43 (23.5)	26 (70.3)	1 (1–1)	27/42 (64.3)
Hospital B	238	232 (97.5)	178 (93.2)	2 (1–2)	212/233 (91)
Nursing station	14	13 (92.9)	10 (90.9)	2 (1–2)	11/12 (91.7)
Common area -trolley	14	13 (92.9)	7 (87.5)	1 (1–2)	8/9 (88.9)
Common area sink drain	14	14 (100)	13 (100)	2 (2–3)	19/20 (95)
Common area sink splash zone	14	13 (92.9)	11 (91.7)	2 (2–2)	15/18 (83.3)
Patient room/area	182	179 (98.4)	137 (93.2)	2 (1–2)	159/174 (91.4)

HTS: high-touch surface (a full list of what HTS were included for sampling is shown in the Supplementary material

Fig. 1.

Carbapenemase results by species and by sample area among carbapenem-resistant organisms (CROs) recovered from the ICU environment in Hospital A, n=155 (A) and Hospital B, n=233 (B), July 2023-February 2024. Per protocol, carbapenemase detection was investigated in carbapenem-resistant (CR)-Enterobacterales, CR-Pseudomonas spp. isolates (both Hospitals), and CR-A. baumannii (Hospital A)

Fig. 2.

Clusters of carbapenem resistant organisms (CROs) detected in participating intensive care units (ICUs) in Hospital A during the study period, July 2023 – January 2024. Each node represents a clinical (asterisk) or an environmental isolate (circles with border represent sink samples and those without a border represent samples from dry surfaces). Isolates were considered of the same cluster if they belonged to the same sequence typing (ST) and had ≤10 single nucleotide polymorphism (SNP) difference from other isolates recovered from samples collected at different time points during the study period or on different rooms/units. There was a total of 15 clusters identified. Dotted and solid lines were used to help track different clusters

Environmental bacterial diversity by metagenomic analysis

The environmental community composition was analyzed in a subset of 11 metagenomic samples (Suppl. Figure 2). A total of 3056–2771 genera were detected in samples from hospitals A and B, respectively (or 217–62158 respectively, when restricting the analysis to bacteria genera only). Overall, the most abundant genera detected in samples from hospital A were Methylorubrum, Fusarium, Sphingobium/Sphingomonas, and Pseudomonas. Clinically relevant genera such as Stenotrophomonas and Acinetobacter were also detected. Acinetobacter was the most abundant genus across all the samples from hospital B, followed by Escherichia, Pseudomonas, Stutzerimonas, Staphylococcus, and Streptococcus. Candida spp. constituted a median of 0.03–0.06% of reads in hospitals A and B, respectively. Clinically relevant carbapenemase genes were detected, including genes not detected through WGS of isolates (e.g., in Argentina blaIMP and blaBKC genes, and in India blaBKC, Suppl. Tables 7 and 8).

Persistence of CROs in the hospital-built environment and potential role of environmental reservoirs in healthcare transmission

WGS was performed on 113 CROs isolated from environmental samples (97 from hospital A and 16 from hospital B) and 19 CROs isolated from unique patients (13 from hospital A and six from hospital B). Detailed description of the AMR profile for sequenced isolates is shown in Suppl. Tables 6, 9–12. SNP analysis of isolates from hospital B showed a wide SNP range (89–5888 SNPs), Suppl. Table 6. In contrast, we identified several clusters in hospital A, with 60% (58/97) of environmental and 69% (9/13) of patient isolates linked to a cluster, Fig. 2 and Suppl. Table 13. We herein describe a few examples. include two distinct clusters of bla_OXA-23 producing A. baumannii ST195 involving environmental and clinical isolates. These two clusters had a SNP difference of 11–15 SNPs, and isolates within each cluster had < 5SNP difference. One cluster (0–3 SNP difference between isolates) involved rectal samples of four different patients staying in the same room (individual occupancy) over a 5 month-period as well as HTS samples collected in this room. The other bla_OXA-23 A. baumannii ST195 cluster involved isolates recovered from a sink in a common area in the ICU as well as the sink in a patient’s room sink and HTS (0–4 SNP difference between implicated isolates). Sinks were also reservoirs of CP-CRE. We identified a cluster of bla_NDM-1 producing C. freundii ST257 recovered from sinks from different patient rooms across both ICUs over a 7 month-period (0–4 SNP difference between isolates). We found co-existence of CP-CRE in sinks such as bla_NDM-1 producing C. freundii ST8 and bla_NDM1 producing E. cloacae ST90, which also carried the same plasmid marker (IncC) (Suppl. Table 14).

Discussion

Antimicrobial resistant infections remain a threat to public health, especially in LMICs [18]. We performed a longitudinal evaluation of the ICU environment in two middle-income country hospitals (Argentina and India) with a focus on CRO reservoirs. We recovered CP-CROs in a quarter of all samples in hospital A and in almost all samples in hospital B. We found evidence of cross-contamination and of CRO persistence in the hospital environment (CROs with similar MLST and AMR profile and few SNP differences between isolates recovered from different rooms on same date or recovered from HTS or sinks in different time points during the study). Furthermore, genetic analysis supported transfer between patients and environmental reservoirs as we found closely related isolates from patients and the environment over several months which would be less likely to be caused by contamination of HCWs hands. These results highlight the role of the hospital environment in healthcare transmission of CROs and inform future efforts to minimize this risk.

Similar to previous studies [9], we found greater bacterial burden in sinks (sink drains and sink surfaces) than HTS near patients (i.e., HTS of beds and equipment around beds). The predominant pathogens recovered from sinks in our study differed by country, with bla_VIM P. aeruginosa and bla_NDM C. freundii predominating in hospital A, and bla_NDM K. pneuomoniae in hospital B. We found well-established CP-CRO reservoirs in sinks as illustrated by closely related C. freundii isolates (< 5 SNP difference among bla_NDM-1 C. freundii ST257 and bla_NDM-1 C. freundii ST8, respectively) recovered from same the same sink over several months. Furthermore, we found the same plasmid marker (IncC plasmid) among these bla_NDM-1 producing enterobacterales providing a possible explanation for transmission of AMR genes. IncC plasmid mediated transmission of bla_NDM among enterobacterales has been previously reported in Argentina [19]. This work has relevance for infection prevention and control (IPC) programs. There is no consensus on optimal disinfectants to decontaminate sinks remaining an important research gap [20]. Meanwhile, IPC programs can implement comprehensive sink infection prevention programs including water infection risk assessments, strict sink etiquette, assessment of quality of sink cleaning, and assessment of disinfectant solutions. Simple changes such as avoiding disposal of patient wastewater into sinks, storing medical supplies and preparing medications in dedicated areas away from sinks can help reduce risks, although sustaining changes and securing resources to monitor compliance are limiting factors in resource limited settings [20, 21]. Attention to sink design when installing or modifying sinks is important as sink depth was a strong predictor of pathogen dispersion outside of the sink [22].

CROs were also found on HTS near patients (24% of samples in hospital A and 98% of samples in hospital B), including CR-Acinetobacter spp. and CRE. SNP analysis and epidemiologic links support bacterial transfer between patients and the environment in participating units. Our study was not designed to assess the directionality of such transfer; however, several other studies have evaluated this and have shown complex transmission dynamics. One study assessed bacterial colonization of patient rooms over a 30-week period after hospital opening and found that in a relatively short period (5–7 weeks), site-specific microbiota was established, and that patients had greater impact on establishing the environmental microbiome than other factors such as temperature or humidity [23]. While we did not assess the quality of cleaning and disinfection practices in our study, challenges associated with disinfection of HTS even with more exhaustive cleaning and in well-resourced settings have been reported [24, 25]. These data highlight the need for more research to identify effective approaches to mitigate CRO reservoirs in the hospital environment.

Genomic surveillance is an important tool to monitor emergent mechanisms of AMR. We recently reported on a new Guiana Extended Spectrum (GES) enzyme with carbapenemases activity (GES-66) in an environmental K. pneumoniae isolates recovered from a sink in hospital [26]. We found several P. aeruginosa isolates belonging to high-risk clones including ST235 (isolated from HTS near patients), and ST233 and ST244 (both from sink samples).

While we included a limited number of samples for metagenomic analysis, it provided insights into the wider scope of AMR genes present in the environment that were not identified on selected isolates [27]. For example, bla_IMP was detected in three metagenomic samples from Argentina. Outbreaks of bla_IMP-1 have been previously reported in Argentina involving A. ursingii [28].

There are several limitations to this study. We collected samples before cleaning as we wanted to maximize bacterial recovery which may have overestimated the bacterial burden of environmental surfaces. However, we found evidence of persistent contamination over weeks and months. We could not identify clusters in hospital B despite a very high burden of CROs in the environment and our robust approach to assessing variant detection and genetic relatedness. This is likely related to the limited number of isolates sequenced in hospital B, which cares for a much larger number of patients than hospital A, in a region with a very high prevalence of CROs. A previous study by Robinson and colleagues showed that nosocomial transmission was the likely mechanism of CRO acquisition in neonates in hospital B [29]. Characterizing the mode of CRO transmission within the hospital was not within the scope of this project. Therefore, we did not collect detailed epidemiological data (e.g., shared staff or equipment between patients, etc.). The carbapenemase identification technique used in hospital B was not validated for Acinetobacter spp. (which were excluded from CP analysis) resulting in underestimation of the burden of CP-Acinetobacter spp. in hospital B. However, WGS of A. baumannii isolates from hospital B showed the presence of various carbapenemases (bla_NDM-1, bla_OXA-23, bla_OXA-232).

Conclusions

In this multicenter study of environmental contamination of the ICU in two middle-income country hospitals, we found a large burden of CP-CROs on surfaces and sinks. Using pairwise SNP analysis, we found evidence of persistence of CROs and of cross-contamination within the hospital-built environment. Our study illustrates the urgent need to develop effective and sustainable programs to mitigate environmental reservoirs of CROs in healthcare.

Author contributions

VF, MR, CR, KS, PS and AS designed the study, analyzed and interpreted data and wrote the manuscript. RM, AF, RK, BR, SS, MN, HP and YB collected data, interpreted data and reviewed the manuscript. FM, AC, FP, WF, AS, JL analyzed data, interpreted results, and wrote the manuscript. AG analyzed data and critically reviewed the manuscript. AMZ, JNW, KAM, APD, and MP assisted with study design, data interpretation and review of the manuscript.

Funding

This study was funded by the United States Centers for Disease Control and Prevention (CDC) (contract # 75D30121D12750).

Data availability

The nucleotide sequences are available in GenBank under Bioproject PRJNA1140135 and PR-India.

Declarations

Ethics approval

This study was approved by the Johns Hopkins Medicine Institutional Review Board as Non-Human Subjects Research /Quality Improvement (QI) and reviewed by local ethic committees at participating sites. This activity was reviewed by CDC, deemed not research, and was conducted consistent with applicable federal law and CDC policy (See e.g., 45 C.F.R. part 46.102(l)(2), 21 C.F.R. part 56; 42 U.S.C. §241(d); 5 U.S.C. §552a; 44 U.S.C. §3501 et seq).

Competing interests

The authors declare no competing interests.