Introduction
Antimicrobial resistance is a global public health challenge. Multidrug-resistant organisms (MDROs) readily transmitted in healthcare settings can cause infections with limited therapeutic options. In 2022, six key healthcare-associated MDROs, such as vancomycin-resistant Enterococcus species and carbapenem-resistant Enterobacterales (CRE), caused an estimated 569,749 infections in the United States. Reference Wolford, McCarthy, Baggs, Hatfield, Maillis and Olubajo1 Despite intensive public health efforts to reduce their spread, rates of hospital-onset MDRO infections increased from 2019 to 2022. 2
A major challenge to controlling these pathogens is their ability to colonize body sites without causing signs or symptoms of infection. People who are colonized with MDROs are at higher risk of developing infections caused by the colonizing pathogen; increased pathogen relative abundance is associated with higher infection risk. Reference Shimasaki, Seekatz, Bassis, Rhee, Yelin and Fogg3 Patients (Patients in this manuscript may refer to hospital patients and/or residents of long-term care facilities) colonized with MDROs can serve as reservoirs of transmission to other patients in healthcare settings. Increased pathogen relative abundance is also associated with greater transmission risk. Reference Alhmidi, Cadnum, Koganti, Jencson, Bonomo and Wilson4 While the duration of colonization of these pathogens is poorly defined, studies suggest that it can be prolonged. Reference Zimmerman, Assous, Bdolah-Abram, Lachish, Yinnon and Wiener-Well5 Current strategies to reduce the spread of MDROs in healthcare settings are based on hand hygiene, environmental cleaning and disinfection, and the use of barriers (e.g., gowns and gloves) to reduce contamination of healthcare personnel. Reference Siegel, Rhinehart, Jackson and Chiarello6 These interventions are challenging to implement, costly, and resource intensive.
Although new antimicrobials have been developed to treat MDRO infections, few new agents have been developed specifically to reduce or eliminate MDRO colonization load (i.e., pathogen reduction agents [PRAs]). While nasal decolonization for S. aureus and antiseptic bathing has been used successfully to reduce MDRO colonization and infection across multiple settings, these interventions have limitations, especially for gram-negative bacteria whose primary colonization site is the gastrointestinal tract. Reference Mangalea, Halpin, Haile, Elkins and McDonald7 Novel PRAs, including those that reduce MDRO burden in the gastrointestinal tract, could reduce the risk of MDRO infection among colonized patients and reduce transmission, preventing serious disease and death. Using CRE as an example pathogen, we modeled the effect of a targeted novel hypothetical pathogen reduction intervention in acute care hospitals (ACHs) and long-term care facilities on the burden of CRE bloodstream infections (BSIs) and deaths in the United States.
Methods
Transmission models
We used a two-stage modeling approach with single facility and national healthcare facility network models. For the single-facility model, we adapted a compartmental model of CRE transmission in a healthcare facility from Toth et al. Reference Toth, Samore and Nelson8 This model utilized seven compartments for patients within the healthcare facility. Figure 1 shows the flow of patients through disease states within a healthcare facility, as described below. Parameters defining the probability of transitioning between disease states were obtained from the literature or via model calibration to CDC’s Antimicrobial Resistance Threats Report, 2,9,10 2018–2022 data from the Emerging Infections Program, 11 and published Reference Shimasaki, Seekatz, Bassis, Rhee, Yelin and Fogg3 and unpublished data on BSIs in LTACHs. A comprehensive description of the model, including a complete list of parameters and literature references used for parameter estimation, is provided in the Supplementary Appendix.
Patient flow in carbapenem-resistant Enterobacterales transmission model.
Patient flow occurred as follows:
-
Patients were admitted to the facility as either susceptible or colonized (but undetected by the facility) based on the probability of colonization at admission.
-
Patients progressed from susceptible to colonized based on a transmission parameter (β), the prevalence of colonization, and the effect of Contact Precautions in reducing CRE contamination of healthcare personnel and the environment for those known to be colonized.
-
Once colonized, patients could (i) become detected as colonized through clinical testing or screening, (ii) progress to BSI, or (iii) clear colonization and become susceptible to recolonization.
-
Patients with BSI could recover from the infection while remaining colonized or die due to infection; we assumed these patients would be treated for infection and would not clear colonization until after they recovered from the infection. Patients who recovered from a BSI but were still colonized were assumed to be detected as colonized.
Patients who were detected as colonized, either through clinical testing or screening or had a BSI, were assumed to be on Contact Precautions and remained on Contact Precautions in all subsequent states, reducing their transmission and providing a protective effect once they became susceptible. Patients with BSIs were assumed to have higher mortality rates than other patients (see Supplemental Table 1); they were discharged at varying rates based on the assumed capacity of the facility type to care for a BSI. Reference Toth, Samore and Nelson8,Reference Falagas, Tansarli, Karageorgopoulos and Vardakas12 All other patients were subject to a discharge rate based on the average annual facility length of stay (LOS) and an underlying death rate held constant across all facility types. The model’s time horizon was one year.
To assess the impact of intervention strategies across the national healthcare facility network, we included non-federal ACHs, critical access hospitals (CAHs), skilled nursing facilities (SNFs), ventilator-capable skilled nursing facilities (vSNFs), long-term acute care hospitals (LTACHs), and other facilities (psychiatric hospitals, inpatient rehabilitation). CRE transmission could occur in all facility types except the other facilities category. We used the 2019 Medicare Cost Reports to develop a list of all healthcare facilities in the United States and obtain annual admissions and patient days for each facility to estimate the average facility LOS. 13 CAHs were split into two parameterization categories (average facility LOS > = 4 days, average facility LOS <4 days) to account for their provision of both acute inpatient care and skilled nursing facility level care (i.e., swing-bed services).
The single-facility model was run to equilibrium for each intervention strategy for each facility. We used the facility-specific discharge rate, admission colonization prevalence, natural colonization clearance rate, and facility colonization prevalence from the single-facility model to calculate an effective transmission parameter for each intervention strategy for each facility.
We used the national network of connected facilities for the national facility model. Each facility was run through a simple susceptible-infectious-susceptible compartmental model using the effective transmission parameters obtained from the single-facility model above. Using 2018–2019 Medicare claims and Minimum Dataset to estimate total direct patient transfers between facilities within a year, facility connections through patient transfers were defined. Non-transfer facility admissions used the 2018 U.S. Census Bureau state population estimates to represent the community. 14 The network model estimated the total number of incident colonizations occurring within the network by facility type annually. The incident colonizations and equations for progression to BSI and death derived from the single-facility model were used to estimate BSIs and deaths due to BSIs (see Supplemental Methods pg. 6).
Intervention strategies
Modeled interventions included screening to detect unrecognized CRE colonization and treating colonized patients with a hypothetical PRA. Screening type and frequency were based on CDC’s novel and targeted MDRO prevention strategies. 15 Patients could be screened as part of a facility-wide screening (i.e., point prevalence survey (PPS) conducted at regular intervals), and screening tests were assumed to have 80% test sensitivity. Reference Gorrie, Mirceta, Wick, Judd, Wyres and Thomson16 Administration of a hypothetical PRA was assumed to reduce the mean time to clearance from 387 days to 30 days. We assumed the agent was 100% effective but also modeled the effect of a partially effective (25%, 75%) agent. Previous studies have suggested that vSNFs and LTACHs play an influential role in regional MDRO transmission. Reference Cincotta, Walters, Ham, Octaria, Healy and Slayton17 Thus, we prioritized these facility types for intervention in our models, simulating the following strategies for the use of the novel agent:
-
In vSNFs and LTACHs only, PRA administration to only clinically detected colonized patients (i.e., no screening)
-
In vSNFs and LTACHs only, PRA administration to clinically detected colonized patients and those detected through patient screening (PPS every 90 days).
-
In all eligible facility types (i.e., all facility types except the other facilities category), PRA administration to only clinically detected colonized patients (i.e., no screening)
-
In eligible facilities other than vSNFs and LTACHs, PRA administration to only clinically detected colonized patients, and in vSNFs and LTACHs, PRA administration to colonized patients identified through clinical detection or screening.
To estimate the impact of each intervention strategy, we subtracted the number of CRE BSIs occurring in the intervention scenario from the number of CRE BSIs occurring in the baseline scenario (no PRA or screening). We defined the direct effect as the number of CRE BSIs and deaths averted among detected patients who received the agent and the indirect effect as the number of CRE BSIs and deaths averted among patients who did not receive the agent. The proportion of the total BSIs averted by the intervention due to patients directly treated with a PRA was calculated for each facility in the single-facility model and then used as a multiplier on the total impact of the intervention in the national model. Direct and indirect effects were estimated for the PRA administration to only clinically detected colonized patients in the all eligible facility types intervention strategy. Analyses were conducted using R 4.4.0. 18
Sensitivity analyses
We conducted several sensitivity analyses to evaluate the influence of key parameters and assumptions. We assessed the impact of varying admission prevalence estimates and varying the rate of natural clearance of colonization. We also examined the effect of varying progression from colonization to BSI by facility type.
This activity was reviewed by the CDC, deemed research not involving human subjects, and was conducted consistent with applicable federal law and CDC policy (See e.g., 45 C.F.R. part 46; 21 C.F.R. part 56; 42 U.S.C. §241(d), 5 U.S.C. §552a, 44 U.S.C. §3501 et seq).
Results
An estimated 2,086 CRE BSIs and 622 associated deaths occurred nationwide in the baseline scenario. Of these BSIs, 893 (42.8%) occurred in ACHs [n = 3,229 facilities], 638 (30.6%) in LTACHs [n = 356], 495 (23.7%) in vSNFs [n = 321], 60 (2.9%) in SNFs [n = 13,667], and 1 (0.03%) in CAHs [1,333]. LTACHs had the highest rate of BSIs (1.2 infections per 10,000 patient-days), followed by vSNFs (0.3), ACHs (0.06), CAHs (0.002), and SNFs (0.001).
Figure 2 includes annual modeled national incidence of CRE BSIs by intervention strategy. When a PRA was administered to clinically detected patients in LTACHs and vSNFs only, 594 CRE BSIs and 177 associated deaths occurred, a decrease of 71.5% relative to the baseline scenario. Most (90.4%) of the BSIs occurred in ACHs, followed by vSNFs (4.5%), SNFs (4.5%), and LTACHs (0.5%). When PRA was administered to colonized LTACH patients and vSNF residents identified through a combination of quarterly PPSs and clinical detection, the numbers of CRE BSIs (n = 554) and deaths (n = 165) were lower than the clinically detected-only scenarios. Incidence was reduced by 73.5% relative to the baseline scenario and 6.9% relative to the clinically detected-only in LTACH and vSNFs strategy. When a PRA was administered to clinically detected patients in all facility types, CRE BSIs decreased by 91.2% from baseline. Stratifying by facility type, 153 (83.2%) BSIs occurred in ACHs, 25 (13.4%) in vSNFs, 2 (1.2%) in LTACHs, and 4 (2.2%) in SNFs for this intervention strategy. When a PRA was administered to clinically detected patients in all facility types and combined with 90-day PPSs in LTACHs and vSNFs, CRE BSIs decreased by 92.8% relative to baseline. In this strategy, 97.3% (n = 147) of BSIs occurred in ACHs, 2.6% (n = 4) in SNFs, and 0.1% (n = 0.1) in LTACHs.
Annual modeled incidence of carbapenem-resistant Enterobacterales bloodstream infections by facility type and intervention strategy. Intervention strategies – baseline (no intervention); pathogen reduction agent [PRA] administration to clinically detected patients in long-term acute care hospitals [LTACH] and ventilator-capable skilled nursing facilities [vSNFs] (PRA LTACH/VSNF); PRA administration to clinically detected patients and patients detected via screening every 90 days in LTACH and vSNFs (Screen/PRA LTACH/VSNF); PRA administration to clinically detected patients in all facility types (PRA all facilities); PRA administration to clinically detected patients in all facility types and to patients detected via screening every 90 days in LTACH and vSNFs (Screen LTACH/VSNF PRA all).
Direct effects of pathogen reduction were substantially lower than indirect effects (i.e., benefit to patients who did not receive the agent through transmission prevention) (Figure 3). When a PRA was implemented in all facility types nationwide without screening in LTACHs and vSNFs, 1,902 BSIs were averted, but only 190 (10.0%) BSIs averted were in the patients who received the PRA. The proportion of BSIs averted in patients who received the agent varied by facility type: 6.9% in ACHs, 6.7% in LTACHs, 14.4% in vSNFs, 39.9% in CAHs, and 52.3% in SNFs.
Annual modeled carbapenem-resistant Enterobacterales bloodstream infections (BSIs) averted, direct and indirect effects. Direct and indirect effects estimated for the PRA administration to only clinically detected colonized patients in all eligible facility types intervention strategy.
Figure 4 includes annual modeled national incidence of CRE BSIs by intervention strategy when the PRA was 75% or 25% effective. When clinically detected patients in LTACHs and vSNFs received a 75% effective agent or 25% effective agent, there were 67.2% and 57.0% fewer CRE BSIs relative to the baseline scenario, respectively. When clinically detected patients in all eligible facility types were included, 86.0% and 74.8% reductions were observed using a 75% or 25% effective agent, respectively.
Annual modeled incidence of carbapenem-resistant Enterobacterales bloodstream infections by facility type and pathogen reduction agent effectiveness. Intervention strategies – pathogen reduction agent [PRA] administration to clinically detected patients in long-term acute care hospitals [LTACH] and ventilator-capable skilled nursing facilities [vSNFs] (PRA LTACH/VSNF); PRA administration to clinically detected patients in all facility types (PRA all). An estimated 2,086 CRE BSIs occurred in the baseline scenario (not shown).
Findings from sensitivity analyses were similar to primary analyses. These results are presented in the Supplementary Appendix.
Discussion
Using CRE as an example pathogen, our analysis suggests that a pathogen reduction intervention with a hypothetical agent could substantially reduce the burden of MDRO infections and deaths among patients in healthcare facilities in the United States. Implementing a PRA in LTACHs and vSNFs with or without screening averted 69–75% of CRE BSIs. When expanded to all facility types, greater reductions were achieved. BSI prevention effects extended beyond patients who received the agent, with substantial indirect effects. Indirect effects were largest in ACHs, LTACHs, and vSNFs. A previous modeling-based study of a hypothetical decolonizing antibiotic similarly found large indirect effects. Reference Toth, Samore and Nelson8 Notably, pathogen reduction-based strategies were also highly impactful with a partially effective agent. Our results could be extended to other pathogens, including endemic causes of HAIs, with similar natural history and epidemiology. The number of patients who could benefit from a PRA is substantial, considering the number of patients colonized or infected with these pathogens and the large indirect effects that we demonstrated. In the United States, high MDRO colonization prevalence in long-term care settings has been reported in several studies. Reference McKinnell, Miller, Singh, Gussin, Kleinman and Mendez19–Reference Lin, Lyles-Banks, Lolans, Hines, Spear and Petrak21 Additionally, CRE prevalence among hospitalized patients is high in many low- and middle-income countries, ranging from 15% in Chile to 37% in Bangladesh. Reference Styczynski, Herzig, Luvsansharav, McDonald and Smith22 This suggests a significant global need for these agents.
Pathogen reduction offers several advantages over current approaches to control MDROs, which focus on adherence to core infection control measures, such as hand hygiene, use of gowns and gloves, and environmental cleaning, and on targeted approaches to identify patients with MDRO colonization or infection for placement on Transmission-based Precautions. 15 Reducing MDRO burden in patients who are colonized directly benefits the patient by reducing their risk of progressing from colonization to infection, preventing severe outcomes such as BSI and death. Pathogen reduction may also reduce patient shedding, reducing potential for transmission through the environment, medical equipment, and transient colonization of healthcare workers. Finally, pathogen reduction could reduce the need for Transmission-based Precautions for MDROs. This would potentially benefit healthcare workers, healthcare facilities, and patients through reduced staff time for donning and doffing personal protective equipment, reduced environmental impact of disposable barrier precautions, improved patient flow due to decreased requirement for single patient rooms, and reduced patient isolation.
The optimal characteristics of a novel MDRO PRA include a low potential for developing resistance to the agent and ease of administration. 23 PRAs that could increase colonization resistance in at-risk patients who are not colonized would offer additional benefits. While their safety and efficacy for pathogen reduction have yet to be established, approaches such as bacteriophages and live microbial consortia have shown promise in reducing pathogen load. Reference Mangalea, Halpin, Haile, Elkins and McDonald7 There are multiple barriers to the development and regulatory approval of these agents. Current standards for regulatory approval require demonstration of an effect on clinical outcomes such as infection or death, which are relatively infrequent among colonized patients, necessitating large and expensive randomized controlled trials. While surrogate trial endpoints (e.g., pathogen load measures and other factors correlated with patient outcomes) for MDRO infection risk could make these trials more feasible, they have not been established and would require validation studies using new diagnostics that have yet to be standardized. Unlike new antibiotics developed to treat highly resistant infections that are reserved for limited early use, a pathogen reducing agent with the above-noted attributes would optimally be used widely to achieve maximal benefit and a potential early return on investment. Articulating these attributes through development of a target product profile could foster product development by informing preclinical development and clinical trial design. 24 Ongoing and expanded partnerships between government, academia, and industry will be critical to overcoming these challenges and paving the way for developing and using PRAs.
The findings in this report are subject to at least four limitations. First, the natural history of CRE colonization and infection is not well-defined and may vary by setting and pathogen; model estimates are sensitive to a shorter estimated duration of colonization. However, our sensitivity analysis shows that results remain consistent within the range of CRE colonization duration estimates in the literature. Second, key parameter estimates might not be generalizable to all facilities, including CRE prevalence upon admission, risk of BSI, and effect of Contact Precautions. Third, patient transfer patterns were calculated using 2018–2019 Medicare data, which is limited to older adults and might not reflect current transfer patterns of all patients. Fourth, the model was calibrated to national BSI estimates in ACHs and proportional estimates in other facility types; reporting and testing might vary in other facility types and might not be nationally representative.
MDROs are a significant threat to patients in healthcare facilities. Patients who are colonized with these pathogens are at risk of developing difficult-to-treat infections and can serve as a source of spread to other patients in healthcare settings. New strategies are needed to reduce the impact of MDROs. Development of novel products that reduce MDRO colonization, even if only partially effective, could avert a substantial number of infections and deaths caused by these pathogens.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2025.10335.
Acknowledgments
None.
Financial support
This work did not receive any non-CDC funding support.
Competing interests
All authors have completed and submitted the International Committee of Medical Journal Editors form for disclosure of potential conflicts of interest. No potential conflicts of interest were disclosed. The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
