Acinetobacter baumannii is a significant Gram-negative opportunistic pathogen in healthcare-associated infections, notorious for its robust environmental persistence and remarkable potential to acquire antimicrobial resistance (1–4). Unlike other Acinetobacter species that often act as environmental or skin commensals, clinically isolated A. baumannii is unequivocally pathogenic, predominantly infecting hospitalized critically ill patients and posing a severe threat, particularly within intensive care units (ICUs) (3, 5–7). Its ability to persist on inanimate surfaces and form biofilms not only facilitates enduring presence and spread in hospital settings but also frequently leads to diverse serious infections, including ventilator-associated pneumonia (VAP) (8–10), bloodstream infections (BSI) (3, 11, 12), wound infections, and meningitis (13–15). Notably, pneumonia represents the most common clinical manifestation of Acinetobacter infection. Globally, the majority of A. baumannii strains causing hospital-acquired pneumonia/ventilator-associated pneumonia (HAP/VAP) are multidrug-resistant, including carbapenem-resistant A. baumannii (CRAB), with a recent meta-analysis reporting a pooled prevalence of 79.9% (95% CI 73.9–85.4%) (16). Of greater concern is the propensity of A. baumannii to develop multidrug resistance. While carbapenems were historically considered last-line therapeutic agents for its infections (17), current global surveillance data indicate alarmingly high levels of carbapenem resistance. Consequently, the World Health Organization has designated CRAB a “critical priority” pathogen, leaving clinicians with severely limited therapeutic options (2, 3, 18).

This therapeutic impasse directly translates into dire clinical outcomes. Carbapenem-resistant A. baumannii bloodstream infection (CRAB-BSI) is associated with exceedingly high mortality. Surveillance data show that the global prevalence of CRAB approaches 45%, with infection rates exceeding 70% among clinical isolates in China (3, 19). Once CRAB-BSI develops, mortality in ICU patients can surpass 60%, a risk significantly higher than that associated with bloodstream infections caused by other multidrug-resistant Gram-negative bacteria (19–21). It is noteworthy that compared to other pathogens causing ICU-acquired BSIs, A. baumannii, although less frequent, is most often healthcare-associated and is the pathogen most likely to exhibit carbapenem resistance (22). This elevated mortality stems from the convergence of two principal factors. On one hand, the scarcity of effective antimicrobial agents makes initial empirical treatment failure or delay common (1, 23, 24). On the other hand, these infections almost exclusively occur in critically ill patients with severe underlying conditions, organ dysfunction, or immunosuppression, whose physiological reserve is already profoundly depleted (25–27). Furthermore, patient-to-patient transmission serves as a primary driver of colonization and infection, particularly during outbreaks (28).

The COVID-19 pandemic may have further exacerbated this crisis (29–31). A significant surge in CRAB infections within ICUs was observed globally during the pandemic. For instance, in Europe, bloodstream infections caused by Acinetobacter spp. increased by 57%, and their carbapenem resistance rate rose from 48.4% in 2018–2019 to 65.8% in 2020–2021 (18, 32). The pandemic led to a massive influx of patients with viral pneumonia into ICUs, where widespread glucocorticoid use and prolonged hospitalization, coupled with the strain on infection control measures due to system overload, have been associated with increased CRAB infection rates in multiple studies (33–35). Therefore, in managing CRAB-BSI—characterized by limited treatment options and complex host factors—a pivotal preliminary step in clinical decision-making is the accurate early identification of patients at the highest risk of mortality. Early and precise risk stratification is becoming increasingly crucial for guiding decisions on initiating intensive treatment regimens (e.g., combination therapy or novel agents) and for the rational allocation of critical care resources.

Despite the clear importance of accurate risk assessment, practical tools readily applicable at the ICU bedside remain scarce. While recent studies have explored complex algorithms, including machine learning, to predict outcomes in CRAB infections, the integration of such models into time-pressured clinical workflows is often hindered by their “black-box” nature, reliance on extensive specific variables, and computational complexity (36, 37). Consequently, within the realm of CRAB-BSI management, there is a pressing need for a tool that leverages routinely available clinical parameters, is intuitive, and can rapidly provide an individualized risk score to inform early clinical decisions. To address this gap, this study aims to identify independent risk factors for 30-day mortality following CRAB-BSI in ICU patients through a retrospective cohort analysis and to subsequently develop and validate a visual nomogram prediction model. The goal is to provide clinicians with a concise and practical tool for risk assessment.

This study substantiates that the prognosis of ICU-acquired CRAB-BSI is determined not merely by the antimicrobial resistance profile of the pathogen, but by a critical convergence of impaired host defense and the severity of the systemic inflammatory response. This paradigm shift is robustly supported by our microbiological findings, which revealed no significant difference in antimicrobial susceptibility profiles between survivors and non-survivors (Table 3). In an era of near-pan-resistance, where effective pathogen-targeted therapeutic options are severely constrained, patient survival hinges more decisively on the individual’s physiological reserve and the clinical response to the infective inflammatory insult. To operationalize this insight, we developed a nomogram prediction model that integrates key clinical markers reflecting these two dimensions—namely, indicators of host vulnerability and markers of inflammation/organ dysfunction—thereby translating a theoretical construct into a practical tool for rapid bedside assessment.

The independent risk factors identified in this study can be conceptualized within two interrelated dimensions: “baseline host vulnerability” and “the intensity of the infection-induced systemic inflammatory response.” The former defines the intrinsic risk reserve of a patient prior to encountering the infection, while the latter dynamically quantifies the pathophysiological burden imposed by the host’s defensive reaction post-infection.

Within the dimension of host vulnerability, advanced age (OR = 1.04) emerged as an independent risk factor, consistent with the established pathophysiology of immunosenescence and the decline in multi-organ physiological reserve associated with aging (38–40). Impaired consciousness was also significantly associated with mortality (OR = 5.10), a relationship underpinned by a dual mechanism. It serves both as a direct marker of severe systemic inflammation or septic encephalopathy (41) and as a clinical condition that elevates the risk of aspiration pneumonia and complicates airway management, potentially initiating a vicious cycle of clinical deterioration (42, 43). Furthermore, a history of glucocorticoid use—an iatrogenic vulnerability factor requiring consideration within its specific temporal context—demonstrated particularly strong predictive value.

A history of glucocorticoid use stood out as the strongest predictor (OR = 5.82), warranting nuanced clinical interpretation. This variable likely encapsulates two distinct yet complementary high-risk scenarios in critically ill patients: (1) a state of chronic immunosuppression due to maintenance therapy for an underlying condition, or (2) the emergent immunomodulatory treatment for an acute, severe inflammatory condition (e.g., septic shock, acute respiratory distress syndrome). In both scenarios, the common pathway to poor outcome is a compromised host defense at the time of bacteremic challenge—whether stemming from baseline immune dysfunction or from iatrogenic immunosuppression superimposed on critical illness. This finding strongly reinforces the paradigm that in CRAB-BSI, patient survival is often governed more by the functional reserve of the host than by the antimicrobial resistance profile of the pathogen (44–46).

The clinical relevance of these risk factors is further amplified when considered within the specific historical context of our study period. The interpretation of our findings, particularly the strong predictive value of prior corticosteroid use, must be carefully contextualized within the timeline of our study, which spanned the COVID-19 pandemic (2020–2023). This global health crisis constituted a defining event within our decade-long study period (2013–2023), profoundly altering ICU admission patterns, therapeutic protocols, and infection epidemiology worldwide (47, 48). While our inclusion criteria were based on a diagnosis of CRAB-BSI and did not stratify patients by COVID-19 status, our cohort objectively mirrors the evolution of the critically ill population before, during, and after this unique phase. Evidence-based guidelines during the pandemic led to the widespread use of glucocorticoids for severe pneumonia (49), inevitably increasing the proportion of ICU patients receiving such immunomodulatory therapy in the latter part of our study. Consequently, the variable “glucocorticoid use” likely captures the synergistic high-risk confluence of “extensive immunomodulatory therapy” and “a concentrated influx of critically ill hosts into the ICU” during the pandemic period. This highlights the paramount importance of iatrogenic immunosuppression as a marker of host vulnerability under specific temporal and therapeutic conditions. It further suggests that during future public health crises imposing similar strains on critical care systems, risk assessment for CRAB-BSI must remain highly vigilant for such acquired immunosuppressive factors.

Concurrently, our model incorporates direct measures of the severity of the systemic inflammatory response, offering a physiological perspective for early mortality risk assessment in CRAB-BSI. The Sequential Organ Failure Assessment (SOFA) score, a cornerstone for evaluating sepsis-related organ dysfunction (50, 51), was an independent predictor (OR = 1.26), underscoring that the degree of multi-organ failure is central to determining prognosis. The cutoff value (> 5) identified in our study provides an objective threshold for bedside risk recognition. C-reactive protein (CRP) level dynamically reflects the intensity of the systemic inflammatory response (41, 52, 53). In the context of near-pan-resistant CRAB infections and the increasingly complex clinical challenges exemplified by events like the COVID-19 pandemic, the prognostic value of parameters like SOFA and CRP—which reflect the host’s immediate physiological state—transcends that of purely microbiological data. This shift in focus is critical: it implies that within the therapeutic constraint of limited antimicrobial options, clinical decision-making can utilize these readily available parameters to prioritize and intensify organ support and inflammation modulation for patients sustaining the most severe physiological insult—representing a potentially pivotal pathway toward improving outcomes.

As a major manifestation of CRAB infection, bloodstream infection (BSI) and ventilator-associated pneumonia (VAP) represent two common and life-threatening infections in the ICU. Although they differ in primary site and initial clinical presentation, they share significant overlaps in underlying risk factors, pathophysiological mechanisms, and therapeutic challenges. Both are frequently observed in mechanically ventilated, multi-site colonized, and immunocompromised hosts (54, 55) and are associated with alarmingly high mortality rates (56, 57). Notably, the incidence of both CRAB-BSI and CRAB-VAP increased in ICUs during the COVID-19 pandemic (30), indicating they are jointly influenced by host immune status and healthcare environmental pressures. Therefore, in clinical management, risk assessment and early intervention strategies for these severe CRAB infections should be conceptually aligned, with a shared emphasis on monitoring host vulnerability and the systemic inflammatory response. Although developed and validated specifically for BSI, the nomogram presented in this study—which integrates age, consciousness status, corticosteroid use, SOFA score, and CRP—is fundamentally built upon these shared pathophysiological pillars. Thus, it may also provide a valuable conceptual framework and a practical reference for initial risk stratification in patients with other severe, systemic CRAB infections, such as VAP, warranting future investigation in those specific cohorts.

Actually, the developed nomogram effectively translates the identified risk paradigm into a pragmatic bedside tool, demonstrating strong predictive performance that holds immediate potential for clinical application. Its utility lies in leveraging parameters such as age, consciousness status, corticosteroid use, SOFA score, and CRP, which are routinely and rapidly available in the ICU, often within the first hours of admission or at the time of CRAB-BSI suspicion. This facilitates integration into clinical workflows at several junctures. First, for risk stratification and triage, a high-risk score can flag patients for intensified monitoring (such as invasive hemodynamic monitoring) and prompt consideration for early, aggressive organ support in a higher-acuity setting. Second, to inform the initial therapeutic strategy, where the quantified mortality risk can guide empirical antibiotic choices while awaiting definitive susceptibility results, favoring combination regimens or newer agents for patients at the highest risk of adverse outcomes. Third, to enhance prognostic communication, provides an objective, evidence-based estimate to structure discussions with patients’ families regarding the anticipated clinical course and the rationale for invasive interventions. Embedding this algorithm as a clinical decision support module within the electronic health record could automate calculation and provide real-time alerts, thereby standardizing risk assessment and promoting timely, guideline-concordant care.

The nomogram exhibited strong predictive performance within the study cohort, as evidenced by an AUROC = 0.863, excellent calibration, and positive net benefit on decision curve analysis. Certain study limitations should be considered. The retrospective, single-center design of this study inherently carries risks of selection bias and may limit the generalizability of our findings. Specifically, caution is warranted when interpreting certain risk factors. For instance, the strong association with corticosteroid use likely reflects the unique patient mix and prevailing treatment practices at our institution, particularly the surge of COVID-19 pneumonia patients receiving immunomodulatory therapy during the pandemic. The operational definition of ‘glucocorticoid use’ warrants careful consideration. In this retrospective study, it was designed as a pragmatic, binary variable capturing any documented systemic use prior to bacteremia. While this ensures sensitivity in identifying patients exposed to iatrogenic immunosuppression, it inherently encompasses substantial heterogeneity. We lacked granular data on specific agents, cumulative doses, treatment durations, and precise clinical indications (e.g., chronic maintenance therapy versus short-course pulse therapy for acute inflammatory conditions). Consequently, our variable serves best as a robust composite marker of ‘significant immunosuppressive exposure’ rather than a precise pharmacologic measure. The strong association observed (OR = 5.82) underscores the profound risk conferred by such exposure in the context of CRAB-BSI. However, our study design cannot elucidate dose–response relationships or define a minimum risk threshold. Future prospective studies should aim to collect detailed corticosteroid exposure data to refine this risk factor, potentially distinguishing between different patterns of use (e.g., high-dose acute vs. low-dose chronic) for more personalized risk assessment. A more fundamental limitation lies in the model’s scope. By design, it incorporates only baseline clinical variables and excludes detailed data on subsequent antimicrobial therapy (e.g., appropriateness, timing, specific agents) and source control interventions. As underscored by recent research (17, 58), these treatment factors are critical, modifiable determinants of outcome in drug-resistant infections. Consequently, our nomogram is best understood and applied as a tool for initial risk stratification at the time of CRAB-BSI diagnosis, providing a pre-therapeutic risk estimate. It does not account for the impact of subsequent clinical decisions. To translate this tool into broader practice, external validation is imperative. Future studies must assess its performance across diverse geographical and epidemiological settings, particularly in centers with different CRAB prevalence rates and patient demographics, to confirm its transportability before widespread clinical adoption.

In conclusion, this study identifies age, impaired consciousness, prior corticosteroid use, higher SOFA score, and elevated CRP level as independent risk factors for 30-day mortality in critically ill patients with CRAB-BSI. By integrating these five readily available clinical parameters, we developed and internally validated a nomogram that accurately predicts mortality risk. This pragmatic tool may aid in the early bedside risk stratification of patients with this life-threatening infection, helping to guide clinical decision-making. External validation is warranted before widespread clinical application.