Introduction
The prevalence of multidrug-resistant gram-negative bacteria (MDRGNB) is increasing worldwide and is of particular concern in vulnerable populations in acute care settings. Reference Abayneh and Worku1–Reference Santoro, Franceschini and Meschiari5 Although multiple mechanisms confer β-lactam resistance, β-lactamase production remains the most common mechanism of resistance. Reference De Angelis, Giacomo and Posteraro6 Enterobacterales, Pseudomonas aeruginosa, and Acinetobacter baumannii are identified as some of the most relevant pathogens with high rates of resistance to several classes of antibiotics, limiting treatment options for infections caused by these organisms. Reference Morris and Cerceo4,Reference Bassetti, Peghin and Vena7–Reference Oliveira and Reygaert10
The rise in MDRGNB poses an increasing problem in acute care, surgery, hospital wards, intensive care units (ICUs), and within the community. Reference Exner, Bhattacharya and Christiansen3,Reference Cerceo, Deitelzweig and Sherman8,Reference Gandra, Tseng and Arora11–Reference Souli, Galani and Giamarellou13 It is important to quantify the burden to have a clear understanding of the epidemiology of MDRGNB, including its regional variability. Reference Morris and Cerceo4,Reference Nordmann and Poirel14–Reference Tang, Apisarnthanarak and Hsu16 Likewise, it is equally important to understand risk factors for MDRGNB infection to ensure appropriate steps are taken to mitigate their spread and to ensure that appropriate empiric antibiotic therapy is started as early as possible. Reference Bassetti, Carnelutti and Peghin17–Reference Patolia, Abate and Patel20 Not only does delayed appropriate therapy increase the risk of death, Reference Lodise, Patel and Kwa21 but studies have also shown that it also prolongs hospitalization. Reference Raman, Avendano and Berger22 Moreover, prolonged hospitalization places patients at risk for developing subsequent antibiotic-resistant infections, Reference Friedman, Temkin and Carmeli23 which can ultimately lead to further antibiotic usage Reference Tellado, Sen and Caloto24 and exacerbate institutional antimicrobial resistance (AMR) patterns. Reference Dellit, Owens and McGowan25 Therefore, it is important to investigate whether published current risk factor associations and their related scoring systems accurately identify patients at risk particularly for MDRGNB infections. Reference Aliberti, Di Pasquale and Zanaboni26,Reference Cano, Gutiérrez-Gutiérrez and Machuca27
This review aims to provide an overview of relevant studies on risk factors for colonization and/or infection with MDRGNB, as well as evaluate scoring systems, focusing on those aiming to determine the risk of MDRGNB infection, colonization and/or mortality. We conducted a search for data published in international public databases (PubMed, Scopus, Google Scholar, SciELO—for Latin American and the Caribbean (LAC)—studies and LILACS) dating from January 2014 to June 2023, using as search criteria within the title section and within the abstract section [Title/Abstract] the following terms: “antimicrobial resistance,” “Gram-negative bacteria,” “extended spectrum β-lactamase,” “carbapenem-resistant,” together with [AND/OR] “Enterobacterales,” “Klebsiella pneumoniae,” “Escherichia coli,” “Pseudomonas aeruginosa,” “Acinetobacter baumannii,” together with [AND/OR] “risk factors,” “risk assessment,” “scoring system,” and “score.” The quality of the studies was not formally evaluated but we have excluded publications not considered as peer reviewed. Duplicates were deleted before screening. Selected documents were reviewed for full-text eligibility and furthermore, we manually selected all those publications reporting a risk score and/or risk assessment to determine the risk of MDRGNB infection and/or mortality. Discrepancies were solved among all authors. During this search, a final number of 81 published studies were finally included in this review and, 23 were used to build the comparative assessment analysis presented particularly in Table 2, as well as in Table 3.
Table 1. Single and MDR percentage in isolates of E. coli, K. pneumoniae, P. aeruginosa and Acinetobacter spp. reported by EARS-NET in EU/EEA countries during 2015–2019
Data source: EARS-Net, European Antimicrobial Resistance Surveillance Network; EU/EEA, European Union/European Economic Area. 45
Table 2. Comparative analysis of risks factors for MDR-GNB infections or colonization on evaluation studies
1 Risk factors are expressed in the association measure Odds Ratio (OR) with 95% confidence intervals, 2 BRE: resistant to carbapenems, ceftazidime and piperacillin/tazobactam, 3 within the last 3 months, 4 within the last 30 days, 5 P. aeruginosa carbapenem-resistant infection, 6 P. aeruginosa BRE infection, 7 gram-negative bacilli resistant to 3 different antibiotic families, 8 immunosuppressive therapy with steroids, tacrolimus, sirolimus, cyclosporine, mycophenolate, antithymocyte globulin, or chemotherapy (with alkylating agents) within the past 3 months, 9 Steroids within the past 3 months, 10 urinary tract infection, 11 carbapenem-resistant Enterobacteriaceae, 12 within the last year, 13 within the last 6 months, 14 Esophagus-gastro-duodenoscopy or colonoscopy within the past 6 months, 15 previous colonization by CRE, 16 abdominal surgery a year or less ago, 17 hospital stay longer than 5 days, 18 carbapenem-resistant Klebsiella pneumoniae, 19 ESBL E.coli, K. pneumoniae and P.mirabilis, 20 age ≥ 70 years, 21 emergency room consult within the last 4 weeks, 22 nursing home resident, 24 β-lactams or quinolones within the past 3 months, 25 aminoglycosides in the last 3 months, 26 carbapenems,27 age ≥ 55 years, 28 hospital-acquired infection, 29 gospitalization in ICU,30 prolonged hospitalization, 31 previous infection with E.coli ESBL,32 3rd or 4th generation cephalosporins, 33 liver cirrhosis, 34 age > 60 years, 35 chronic obstructive pulmonary disease, 36 intra-abdominal infection.
Table 3. Performance of risk scores predicting infection, colonization, or mortality by MDR bacteria
Global burden of bacterial AMR
One of the most comprehensive studies evaluating AMR burden was published in 2022. Reference Murray, Ikuta and Sharara28 The study assessed estimated deaths and disability-adjusted life-years (DALYs) directly attributed to bacterial AMR. It examined the deaths and DALYS associated with 23 different pathogens and multiple pathogen–drug combinations in 204 countries in 2019. Reference Murray, Ikuta and Sharara28 The researchers used systematic literature reviews, hospital systems, surveillance systems, and other sources to obtain a wide range of data. Reference Murray, Ikuta and Sharara28 Furthermore, they employed predictive statistical modeling to estimate AMR burden, even in locations lacking specific data. Reference Murray, Ikuta and Sharara28 In summary, three infectious syndromes were associated with the highest AMR burden in 2019: (i) lower respiratory infections, (ii) bloodstream infections, and (iii) intra-abdominal infections. Reference Murray, Ikuta and Sharara28 These three syndromes accounted for 79% of deaths attributed to AMR in 2019. Notably, six pathogens were responsible for more than 250,000 deaths associated with AMR; those pathogens were: E. coli, S. aureus, K. pneumoniae, Streptococcus pneumoniae, A. baumannii, and P. aeruginosa. Although the order of attributable mortality of the pathogens varied depending on the world’s region, together, these six pathogens were responsible for a calculated 929,000 deaths attributed to AMR, and 3.57 million deaths associated with AMR globally in 2019. Reference Murray, Ikuta and Sharara28
As the prevalence of MDR organisms is constantly increasing, along with its associated mortality, identifying risk factors for developing an MDRGNB infection could greatly affect patient care and infection management. Reference Patolia, Abate and Patel20,Reference Anaya-Baz, Maldonado and Palacios-Baena29
The increase in AMR is a global concern; the World Health Organization’s (WHO) in its most recent Global Antimicrobial Resistance and Use Surveillance System (GLASS) report from 2022, involving surveillance data from 109 countries collected during 2017–2020, revealed multiple concerning findings including, increased rates of MDR E. coli from blood isolates by more than 15% in 2020 when compared to the rates of 2017, particularly for meropenem (0.5% [in 2017]–0.9% [in 2020]) and third generation cephalosporins (20.2% [in 2017]–24.0 [in 2020]). Similarly, they reported high levels of resistance in K. pneumoniae to third-generation cephalosporins (59%–64.7%) and to fourth-generation cephalosporins (57.4%), with isolates often showing multidrug resistance. Also, a high percentage of resistance to carbapenems in Acinetobacter spp. (> 50%), 73.1% (in 2017) and 72.9% (in 2020). Reference Ali Fikri and Syamsul Arifin30
Overview of AMR in GNB: regional dissemination
Multiple antibiotic-resistant organisms have spread worldwide; particularly extended-spectrum β-lactamase (ESBL)-producing E. coli. in India and Latin America, this group represents up to 40% of the worldwide burden of clinical isolates of ESBL-producing E. coli. 31,32 CR-K. pneumoniae is common (6%–28%) in China, the US, and Latin America. New Delhi metallo-β-lactamase (NDM)-producing Enterobacterales have been increasingly observed over the past 15 years in South Asia, China, the USA, and Europe. 31,Reference Johnson and Woodford33–Reference Pillai, McGeer and Low35 Variants of several β-lactamase genes, particularly ESBLs, such as bla CTX-M, bla TEM, bla SHV, bla PER, bla VEB and bla TLA variants, are reported worldwide, among which bla CTX-M group is the most prevalent group. Reference Abrar, Ain and Liaqat36 Carbapenem-hydrolyzing enzymes such as K. pneumoniae carbapenemases (KPC), NDM-1, the imipenemases (IMP), Verona integron-encoded metallo-β-lactamases (VIM) and OXA (oxacillinase)-48 have become widely disseminated. However, there are important differences in the prevalence of these enzymes across countries and particular geographical regions. Reference Ageevets, Partina and Lisitsyna37,Reference van Duin and Doi38 The successful global dissemination of these carbapenemases is primarily attributed to the presence of their genes on mobile genetic elements, most commonly plasmids. These elements often co-harbor resistance to multiple classes of antibiotics, leading to the emergence of pan-drug-resistant organisms. Reference Hawkey39,Reference Weingarten, Johnson and Conlan40
The regional and national variations, whether in terms of ESBLs or carbapenemases, are mainly due to variations in population density, hygiene and integrity of the country’s infrastructure, use of antibiotics, and infection control practices. Reference Bush and Bradford41 As for its international spread, it was attributed largely to international travel, including globalized medical tourism. Reference Ahammad, Sreekrishnan and Hands42
Globalization, along with the increase in antibiotic use worldwide, has significantly affected the evolution and spread of antibiotic-resistance genes over the past 25 years. Studies have shown that 8% of E. coli isolates in the US produce ESBL, while in Latin America and South Asia the percentage may reach 32% and 33%, respectively. Reference Guzmán-Blanco, Labarca and Villegas43,Reference Islam, Heffernan and Naicker44 In 2019, more than half of E. coli isolates reported to EARS-Net and more than a third of K. pneumoniae isolates were resistant to at least one antimicrobial group of interest, and combined resistance to multiple antimicrobial groups was common. 45 While carbapenem resistance remained rare in E. coli, in contrast, several countries reported carbapenem resistance percentages above 10% for K. pneumoniae. In this report, carbapenem resistance was also common in P. aeruginosa and Acinetobacter spp. 45 A summary of this international report, covering more than 30 countries, is given in Table 1.
Risk factors and risk assessment scores for colonization and infection by MDRGNB
Infections caused by MDRGNB mainly involve carbapenem-resistant (CR) and ESBL-producing pathogens. Most of these infections are healthcare-acquired; however, some originate in the community. Reference Maechler, Pena Diaz and Schroder46,Reference Poulou, Spanakis and Pournaras47 Multiple publications have addressed risk assessment for ESBL-producing and CR- related infection and colonization. Table 2 describes and compares studies that identified risk factors associated with these conditions and their corresponding odds ratio.
It is currently uncertain whether it is necessary to conduct active fecal screening for ESBLs at hospital admission. Reference Tenover and Goff48 The prevalence of ESBL-producing Enterobacterales (ESBL-PE) infections among carriers at ICU admission was reported with a rate of 15% in a recent 7-year study. Reference Razazi, Rosman and Phan49 From this study, a clinical risk score designed to predict ESBL-PE infection was derived from five independent risk factors associated with ESBL-PE infection in carriers (ie, age >60 years, cirrhosis, broad-spectrum antibiotic treatment within the previous 3 months, urinary or intra-abdominal infection, and absence of chronic pulmonary disease). Based on these variables, an ESBL risk score was created to guide empirical carbapenem therapy against ESBL-producing Enterobacterales infections, scores of 4–5, and 6–7 had a prevalence of 26%, and 49%, respectively. Reference Razazi, Rosman and Phan49 Based on a systematic review and meta-analysis, ESBL colonizers were more likely to have a history of recent antibiotic treatments or healthcare facility utilization. Reference Flokas, Alevizakos and Shehadeh50 However, predicting colonization is not easy; many risk factors are nonspecific and are also common to all MDRGNB. Reference Burillo, Muñoz and Bouza51 Nevertheless, some scores of risk stratification reliably identify colonization by resistant bacteria. An example is a scoring system devised by Tumbarello et al., which is used to identify ESBL colonization in patients with GNB infections (E. coli, Klebsiella spp. or Proteus mirabilis) at hospital admission. Reference Tumbarello, Trecarichi and Bassetti52
Conversely, CRE infections have become an urgent public health threat with higher morbidity and mortality, as well as limited antibiotic treatments available. Reference Salomao, Guimaraes and Duailibi53 Although the CDC does recommend surveillance for CR gram-negative pathogens for infection control purposes, some studies also have shown that the detection of colonization by a CR organism (CRO) could guide empirical therapy in certain populations such as hematology patients or kidney transplant recipients. Reference Tenover and Goff48 The primary risk factors for developing CRE infections include longer hospital stays, ICU admission, invasive procedures, recent surgery, recent immunosuppression/immunodeficiency, severe underlying comorbidities, prior antibiotic exposure, and colonization/carriage with CRE. Reference Cano, Gutiérrez-Gutiérrez and Machuca27,Reference Giannella, Trecarichi and De Rosa54 Additional studies indicate that ICU admission, transfer between wards, prolonged hospital stay and sharing a room with known carriers are predominant factors related to CRE colonization. Reference Pano Pardo, Serrano Villar and Ramos Ramos55,Reference Torres-Gonzalez, Cervera-Hernandez and Niembro-Ortega56 There is also evidence showing high rates of CRE infections up to 45% in critically ill individuals among patients who were previously colonized but asymptomatic. Reference Kontopoulou, Iosifidis and Antoniadou57,Reference Zarakolu, Eser and Aladag58 A systematic review of 92 studies found that the most important risk factors for CR-GNB include previous antibiotic use (especially carbapenems), previous colonization, mechanical ventilation, previous ICU stay, dialysis, catheter use, length of hospital stay, comorbidities and, an increase in Acute Physiology and Chronic Health Evaluation (APACHE) II score. Reference Palacios-Baena, Giannella and Manissero19 To better predict the development of infections, risk factors have been evaluated and scored for carriers of these CR bacteria.
The Giannella risk score (GRS) has been used to identify the risk of KPC-producing K. pneumoniae bloodstream infection (BSI) among rectal carriers. Reference Giannella, Trecarichi and De Rosa54 Notably, multi-site colonization was considered the most reliable predictor. Reference Giannella, Trecarichi and De Rosa54 Further cohort studies have measured the utility of the GRS and determined a cutoff point (of 7) to discriminate between low and high risk of all KPC-producing K. pneumoniae infection. Reference Cano, Gutiérrez-Gutiérrez and Machuca27
Lastly, the acquisition of MDR P. aeruginosa is significantly related to ICU admission but is also related to other factors including prior use of broad-spectrum (cephalosporins, aminoglycosides, quinolones, carbapenems) high invasive-device scores and prior hospital stays, advanced age, and human immunodeficiency virus infection. Reference Kaye and Pogue9,Reference Tumbarello, Trecarichi and Bassetti52,Reference Aloush, Navon-Venezia and Seigman-Igra59–Reference Tacconelli, Tumbarello and Bertagnolio61 In the case of MDR A. baumannii infections, prior colonization plays an important role as a risk factor; as well as, previous antibiotic use, prolonged or prior hospitalization, previous ICU stay, or invasive procedures and repeated skin grafts. Reference Blanco, Harris and Rock62–Reference Sheng, Liao and Lauderdale65
Based on previous studies and the fact that previous colonization with a MDRGN can lead to an infection, rectal surveillance in patients hospitalized in critical care areas in hospitals with high prevalence of MDRGN is suggested, as this practice can help guide treatment for critically ill patients. Reference Jaiswal, Gupta and Kumar66–Reference McConville, Sullivan and Gomez-Simmonds68
Risk factors and risk scores for mortality in MDRGNB
Data from the WHO confirms that the risk of death attributable to infection by antibiotic-resistant microorganisms is twice that of non-resistant bacteria. Reference Falagas, Tansarli and Karageorgopoulos69 Falagas et al. published a meta-analysis of nine studies on mortality following CRE infections in 2012, reporting that 26%–44% of deaths in seven studies were attributable to carbapenem resistance. Reference Falagas, Tansarli and Karageorgopoulos69 A recent multinational prospective cohort study that looked at the effect of carbapenem resistance on outcomes of BSI caused by Enterobacterales in low- and middle-income countries (LMICs) showed that carbapenem resistance was associated with increased length of hospital stay and mortality. Reference Stewardson, Marimuthu and Sengupta70 In Latin American countries, BSI by CRE has been shown to increase in-hospital mortality four times, compared to non-CRE infection (OR = 4.0 CI 95% (1.7–9.5)). Reference Villegas, Pallares and Escandón-Vargas71 However, an analysis of 24 studies on infections caused by MDRGNB in ICUs published in 2016, did not confirm a direct association between infections due to MDRGNB and mortality in ICU patients. Reference Paramythiotou and Routsi72
Many factors confound the identification of main risk predictors for mortality in MDRGNB-associated infections, including variability in patient populations, late onset of appropriate initial antibiotic therapy, lack of appropriate antibiotic therapy on site in everyday practice, heterogenicity in the outcomes of each study, and the lack of availability of a quick and specific diagnostic test to identify MDR bacteria. Reference Exner, Bhattacharya and Christiansen3 Despite inconsistencies in studies defining risk factors, here we point out studies with common risk factors reported to be associated with higher risk for mortality in MDRGNB.
Patolia et al. conducted a retrospective observational cohort study and used data collected over a 13-month period from the electronic health records of patients with gram-negative bacteremia at a single university medical center; 177 patients were included in the analysis, 46 of which (26%) had MDRGNB bacteremia with a mortality rate of 34.8%, compared to 13.7% in the non-MDR-gram-negative bacteremia group (P = 0.002). In particular, inappropriate empiric antibiotics (OR: 7.59, 95% CI: 1.68–34.34), urinary catheter as a source of infection (OR: 5.68, 95% CI: 1.37–23.5), intra-abdominal source of infection (OR: 3.66, 95% CI: 1.14–11.73), end-stage liver disease (OR: 3.64, 95% CI: 1.07–12.3) and solid organ malignancy (OR: 3.64, 95% CI: 1.25–10.56) were significant independent risk factors for mortality in patients with MDRGNB. Additional risk factors were identified in the multivariate analysis as significant for the development of a MDRGNB infection: diabetes mellitus (OR: 2.8, 95% CI: 1.00–4.88), previous antibiotic use (OR: 2.93, 95% CI: 1.25–6.87), and urinary catheter as a source of infection (OR: 5.96, 95% CI: 1.78–19.94). Reference Patolia, Abate and Patel20
For A. baumannii, significant mortality has been reported with CR-Acinetobacter baumannii (CRAB) infection although limited data on contributing microbiological factors have been published. In a retrospective study of 164 patients, Hyo-Ju Son et al. found that 90 (55%) of the 164 patients died within 30 days. Reference Son, Cho and Bae73 In the multivariate analysis of study findings, independent risk factors for mortality were a non-eradicated focus, septic shock, and inappropriate antimicrobial therapy. Reference Son, Cho and Bae73 A recent meta-analysis evaluating the predictors of mortality in patients infected with CRAB identified that the most relevant risk factors for mortality were inappropriate empirical antimicrobial treatment (OR, 5.04; 95%), septic shock (OR, 5.65; 95), chronic liver (OR, 2.36; 95%), and chronic renal disease (OR, 2.02; 95%). Reference Du, Xu and Yao74
Infection with MDR P. aeruginosa is also associated with high mortality. In a retrospective study of patients diagnosed with P. aeruginosa BSIs in two Italian university hospitals, Tumbarello et al. Reference Tumbarello, Repetto and Trecarichi75 compared risk factors for isolation of MDR or non-MDR P. aeruginosa in blood cultures. Presentation with septic shock, infection due to MDR P. aeruginosa and inadequate initial antimicrobial therapy were the variables independently associated with 21-day mortality. Reference Tumbarello, Repetto and Trecarichi75
For CP-CRE, Tamma et al. conducted an observational study that compared 14-day mortality in patients with CP-CRE and non-CP-CRE bacteremia. After adjusting for severity of illness on day 1 of bacteremia, underlying medical conditions, and differences in the antibiotic treatment administered, the odds of dying within 14 days was more than four times greater for CP-CRE compared with non-CP-CRE bacteremia patients (aOR, 4.92; 95% CI: 1.01–24.81). Reference Tamma, Goodman and Harris76
The INCREMENT CPE score was developed, and subsequently validated, to assess mortality risk – and subsequently guide the initiation of empiric therapy – using objective clinical criteria in CPE-colonized patients, including KPC-producing K. pneumoniae. The underlying logistic regression model identified the following variables in BSIs and assigned a point-based weighting, to which a threshold was applied. Reference Cano, Gutiérrez-Gutiérrez and Machuca27,Reference Gutiérrez-Gutiérrez, Salamanca and de Cueto77 INCREMENT-CPE has been externally validated in further studies. Reference Papadimitriou-Olivgeris, Bartzavali and Georgakopoulou78,Reference Machuca, Gutiérrez-Gutiérrez and Rivera-Espinar79
Finally, a meta-analysis published by Vardakas et al. included 30 studies, 25 of which were retrospective, nine provided data on predictors of mortality for MDRGNB infections only, and 21 provided data for MDRGN versus non-MDRGNB infections. Reference Vardakas, Rafailidis and Konstantelias80 Within the studies, Acinetobacter spp., P. aeruginosa and Enterobacterales were the most studied bacteria. There was significant diversity among studies regarding evaluated predictors of mortality. Nevertheless, the most reported independent predictors of mortality were infection severity and underlying diseases, followed by multidrug resistance, inappropriate treatment and increasing age. In the studies that included only patients with MDRGNB infections, cancer (RR: 1.65, 95% CI: 1.13–2.39) and prior or current ICU stay (RR: 1.27, 95% CI: 1.02–1.56) were the risk factors associated with mortality. Reference Vardakas, Rafailidis and Konstantelias80
Author’s opinion: advantages and disadvantages of risk scores
There are multiple advantages of the scoring systems; they are simple to calculate and are based on readily available data at hospital admission. Reference Giannella, Trecarichi and De Rosa54,Reference Tumbarello, Repetto and Trecarichi75,Reference Boyd, Vasudevan and Moore81,Reference Lee, Chu and Hsieh82 Such scores can shorten the time to initiate appropriate empiric therapy, help optimize therapy and avoid inappropriate and potentially toxic therapies. Reference Cano, Gutiérrez-Gutiérrez and Machuca27,Reference Giannella, Trecarichi and De Rosa54,Reference Kengkla, Charoensuk and Chaichana83,Reference Tartof, Kuntz and Chen84 Such scoring systems also help de-escalate therapies, avoiding toxicity and costs associated with extended antibiotic use. Reference Tumbarello, Trecarichi and Bassetti52,Reference Giannella, Trecarichi and De Rosa54 In studies evaluating these scoring systems, those assessing MDRGNB risk performed well, providing significant and accurate predictions of infection or mortality with good powers of prediction. Reference Cano, Gutiérrez-Gutiérrez and Machuca27,Reference Tumbarello, Trecarichi and Bassetti52,Reference Giannella, Trecarichi and De Rosa54,Reference Kengkla, Charoensuk and Chaichana83 Table 3 summarizes the different studies evaluated for this review and compare the different risk scores for the prediction of infection or mortality by resistant bacteria. We found that the sensitivity of scores ranges between 74%–96%, the specificity from 42%–91%, and the negative predictive values from 68%–99%. Reference Cano, Gutiérrez-Gutiérrez and Machuca27,Reference Tumbarello, Trecarichi and Bassetti52,Reference Giannella, Trecarichi and De Rosa54,Reference Lee, Chu and Hsieh82,Reference Kengkla, Charoensuk and Chaichana83,Reference Johnson, Anderson and May85
Despite the advantages of using risk scores, we found that one of the major disadvantages with MDRGNB risk scoring systems is that multiple methods have not been externally validated in wider populations and may not be generalizable to patient cohorts in different regions. Reference Cano, Gutiérrez-Gutiérrez and Machuca27,Reference Giannella, Trecarichi and De Rosa54,Reference Boyd, Vasudevan and Moore81–Reference Kengkla, Charoensuk and Chaichana83,Reference Henderson, Luterbach and Cober86,Reference Teysseyre, Ferdynus and Miltgen87 In addition, it was not clear which sub-populations would benefit most from this risk assessment, as some studies evaluated specific populations Reference Lodise, Patel and Kwa21,Reference Guzmán-Blanco, Labarca and Villegas43,Reference Son, Cho and Bae73,Reference Tamma, Goodman and Harris76 or specific sites of infection. Reference Guzmán-Blanco, Labarca and Villegas43,Reference Stewardson, Marimuthu and Sengupta70,Reference Tamma, Goodman and Harris76 The utility of a scoring system depends on the intended application and its ultimate value. Moreover, many of the studies assessing these scores have been retrospective, and do not consider the different methods and patient populations involved. Reference Lee, Chu and Hsieh82,Reference Henderson, Luterbach and Cober86,Reference Palacios-Baena, Gutierrez-Gutierrez and De Cueto88 It is also worth noting that the quality of data and medical practice may not be consistent across all sites. Reference Lee, Chu and Hsieh82,Reference Henderson, Luterbach and Cober86,Reference Palacios-Baena, Gutierrez-Gutierrez and De Cueto88 Moreover, scoring systems are unable to determine the duration of carrier status prior to the development of an infection. Reference Giannella, Trecarichi and De Rosa54 In some scoring systems, the cutoff value for the score that represents a risk is critical and can change the indicated outcome, and more data is needed to determine the reliability of risk prediction. Due to the difference in rates of resistance at various locations, some scoring systems may also require local adaptation of the test parameters or changes in cutoff values. Reference Johnson, Anderson and May85 Another limitation is that most risk scoring systems are intended for use with specific pathogens/resistance mechanisms rather than multiple different mechanisms of resistance of GNB, and this may restrict their usefulness and widespread use, or dictate that multiple different scores may need to be used. Reference Boyd, Vasudevan and Moore81,Reference Johnson, Anderson and May85–Reference Teysseyre, Ferdynus and Miltgen87 Nonetheless, examples like the INCREMENT-ESBL scoring system, which has been evaluated at multiple different locations internationally, has more reliable evidence-based data to support its use. Despite these current drawbacks, risk scoring systems show promise and have the potential to expedite appropriate treatment and substantially improve outcomes in serious MDRGNB infections.
Conclusions
This review underscores the concerning prevalence of antibiotic resistance in MDRGNB, specifically focusing on the clinically relevant Enterobacterales, P. aeruginosa, and A. baumannii. The escalating global presence of these bacteria emphasizes the critical need to identify colonization and infection risk factors which are significantly linked to mortality. Selecting multiple publications which reported risk factors with association measures and compared the different OR values that might aid when assessing risk factors to predict colonization and/or infection with MDR bacteria may help the selection of empiric treatment in a more appropriate way. Although the accuracy of published risk factor scores and assessments varies, it is imperative to consider these factors and utilize available tools judiciously. We found that most risk scores have a high sensitivity, while most of them are not specific. Nonetheless, we believe that used in a judicious and individualized way in each clinical situation, these tools can aid healthcare workers in expediting appropriate empiric therapy, guiding the selection of optimal treatment, applying necessary infection control and prevention measures and thereby improve the overall clinical outcomes for patients. In the future, machine learning, as well as AI, could revolutionize the prediction of colonization and infection risk. By combining patient-specific risk factors with dynamic, machine-learning-driven approaches, these advanced models have the potential to significantly enhance predictive accuracy. Unlike traditional methods, machine learning models and AI can integrate vast amounts of data, including clinical, microbiological, and demographic information, in real time. This allows for the generation of personalized risk assessments tailored to each patient’s unique profile, enabling earlier and more targeted interventions.
However, it is important to note that while these risk assessment tools can be valuable in initiating empirical therapy for at-risk patients, they should not supersede clinical judgment or the availability of clinical data to prevent the overuse of broad-spectrum antibiotics.
Acknowledgments
The authors would like to thank Dr Pranita D. Tamma for her valuable suggestions for the final version of this manuscript. Also, to Clóvis Arns da Cunha for his contribution to the conception and design of this review. To Katy Holliday, from Lucid Group Communications Ltd., Marlow, Buckinghamshire, UK, for providing editorial and medical writing support, who were able to review and provide feedback on medical accuracy and fair balance.
Financial support
Editorial and medical writing support was funded by Pfizer.
Competing interests
The authors declare the following financial interests/personal relationships that may be considered as potential competing interests. M.V.V. and C.J.P have received honoraria and research grants from MSD, Pfizer, bioMérieux, West and GPC pharma. S.S.K has received honoraria for participating in advisory boards and as a speaker from MSD, Pfizer, GSK, and Menarini. P.K has received honoraria for participating in an advisory board from Pfizer and as a speaker from bioMérieux, MSD and Pfizer. The other authors declare no competing financial interest. All authors, except J.C.G.B, C.J.P, N.R.A, and L.W.A received an honorarium from Pfizer to attend a steering committee meeting to discuss their regional perspective on MDR gram-negative infections and AMS strategies. The output of these discussions was used to create to a framework for this manuscript prior to its development.
Ethical standard
None.
Open access
