Cardiovascular diseases are highly prevalent worldwide, with heart failure (HF) representing one of the most significant public health issues (1). HF is characterized by impaired cardiac function, leading to reduced organ perfusion and potential multi-organ damage. It is the leading cause of death and hospitalization in individuals over 60 years old (1, 2). Due to the severity of HF, a large proportion of hospitalized patients require admission to the intensive care unit (ICU) (3, 4). Identifying risk factors for mortality in ICU patients with HF is crucial for optimizing treatment, resource allocation, and a better understanding of the pathophysiology. To date, methods, such as the Acute Physiology Score III (APS III) and the Sequential Organ Failure Assessment (SOFA) score have been explored for assessing disease severity and mortality risk in critically ill ICU patients (5–7). While the SOFA score has been shown to predict long-term all-cause mortality in acute HF patients (5), specific studies focusing exclusively on the application of APS III in HF patients are limited (7). SOFA's complexity, reliance on numerous indicators, and the variability in assessors’ scores are notable shortcomings (8).

Patients with HF often exhibit acid-base and electrolyte disturbances (9, 10). The anion gap (AG), which reflects the difference between the concentrations of measured cations and anions in serum, is a widely used biomarker for acid-base disorders. Tissue hypoperfusion in HF can lead to anaerobic metabolism, increasing the production of organic acids such as lactate (10), a significant unmeasured anion that contributes to AG elevation. Moreover, renal dysfunction, which frequently coexists with HF, impairs clearance and further promotes the accumulation of unmeasured anions. Although elevated AG has been associated with poor prognosis in HF (11–13), fluctuations in albumin concentration (albumin is a major negatively charged plasma protein) can result in underestimation of AG. For every 10 g/L decrease in serum albumin, the AG decreases by approximately 2.3–2.5 mmol/L (14, 15). Given that hypoalbuminemia is common in HF, the albumin-corrected anion gap (ACAG) is particularly relevant for HF patients, as it adjusts for hypoalbuminemia and more accurately reflects the burden of unmeasured anions and acid-base disturbances. ACAG has been investigated as a potential biomarker for disease severity and mortality in various critical conditions, including HF (16–22).

To date, reports on the ACAG role in predicting mortality of ICU patients with HF are limited (21, 22). A study with a small sample size (205 patients with HF) found that ACAG levels were positively associated with mortality in hospitalized HF patients over an average follow-up period of approximately 6 years (21). Li et al. studied 7,787 ICU patients with HF and reported that patients who died at 365 days had significantly higher ACAG levels than survivors, but they did not evaluate ACAG predictive ability for mortality (22). Although both studies identified ACAG as a prognostic marker, they included patients with multiple ICU admissions, who may experience various metabolic derangements. Long-term metabolic adaptation in these cases could affect the prognostic value of ACAG. To clarify its precise role in mortality prediction, studies on HF patients with a single ICU admission are needed. Additionally, both studies examined the relationship between ACAG levels and medium- or long-term (1 year or 6 years) mortality. Since identifying biomarkers for short-term mortality enables early targeted interventions that may help reduce mortality, we aimed to investigate the role of ACAG in predicting 30-day (short-term) and 365-day (medium-term) mortality in ICU patients with HF. To this end, we extracted and analyzed data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database. Distinct from the two published studies (21, 22), our current study only selected HF patients with a single ICU admission, and we employed propensity score matching (PSM) to minimize confounding effects and further evaluate the association between ACAG levels and mortality. Additionally, we assessed ACAG's predictive ability for mortality using receiver operating characteristic (ROC) curve analysis.

In this study, we focused on patients with a single ICU admission. This selection aimed to recruit a relatively homogeneous cohort to reduce confounding factors related to multiple admissions, such as varying interventions, chronic complications, and changes in clinical status over time, which could obscure the direct relationship between ACAG and outcomes. By analyzing the physiological derangements reflected by ACAG at ICU admission and their impact on prognosis, especially 30-day mortality, while minimizing confounding, we aimed to elucidate the actual role of ACAG as a prognostic biomarker in the group of patients we chose to study. We present the following findings: (1) patients who died at 30 and 365 days had significantly higher ACAG levels compared to survivors; (2) ACAG levels showed a positive association with mortality at both time points; (3) propensity score matching confirmed a positive relationship between ACAG levels and mortality; (4) SAPS II, lactate, bicarbonate, BUN, creatinine, and hematocrit indirectly mediated the association between ACAG and all-cause mortality; (5) ACAG had a higher AUC value than AG in predicting mortality; and (6) SpO₂ and hemoglobin were negatively associated with lactate levels, while heart rate, ALT, AST, creatinine, and blood urea nitrogen were positively associated.

Identifying high-risk ICU patients with HF within 30 days enables urgent interventions, such as intensive monitoring, medication adjustments, and advanced therapies, to reduce mortality. Additionally, many HF patients are readmitted to ICU within 30 days (26), and biomarkers predicting early mortality can guide post-discharge management to reduce the risk of readmission. Therefore, we first compared ACAG levels between patients who died at 30 days and survivors. Our results showed that patients who died at 30 days had significantly higher ACAG levels that were positively associated with mortality. To our knowledge, this is the first report on ACAG predicting short-term mortality in ICU patients with HF.

A few studies have investigated the relationship between ACAG levels and mortality in patients with HF (21, 22). In a retrospective, real-world study of hospitalized HF patients (whether admitted to ICU was not specified), it was found that the death group had significantly higher baseline ACAG levels than the survival group during a mean follow-up period of ∼6 years (21). Another study involving over 7,500 ICU patients with HF reported that ACAG was associated with higher 1-year all-cause mortality (22). This 1-year data, along with ours, suggest a positive association between ACAG levels and short- and medium-term mortality in ICU patients with HF. Compared to the two previous studies (21, 22), which relied solely on multivariate Cox regression, our study employed a more rigorous approach by combining propensity score matching with Cox regression to more effectively minimize confounding effects. Our results confirmed a positive association between ACAG levels and mortality after propensity score matching.

In the present study, SAPS II and BUN were found to be the most significant indirect mediators of the association between ACAG and all-cause mortality, as shown by mediation effect analysis. The SAPS II scoring system incorporates both acute physiological parameters and chronic health conditions, making it a more comprehensive tool for ICU mortality prediction (27). In contrast, the APS III scoring system exclusively evaluates acute physiological parameters to assess the severity of acute illnesses (28). Therefore, SAPS II was selected as a key variable for mediation analysis. It is well documented that comorbidities relate to all-cause mortality in HF patients (29–31). The European Heart Failure Pilot Study showed that diabetes, chronic kidney disease, and anemia were independently associated with increased mortality in HF patients (29). Another study from the UK found that diabetes, chronic kidney disease, and lung disease substantially increased the risk of mortality in HF patients (30). A recent systematic review and meta-analysis study also revealed that diabetes and chronic kidney disease were significant risk factors for mortality in HF patients (31). These previous findings align with the mediating role of SAPS II in the association between ACAG and mortality observed by our study. Therefore, identifying comorbidities in heart failure and initiating prompt, effective treatment is crucial for improving outcomes. We found that BUN, a key marker of kidney function, accounted for 22.68% of the mediated effect. Higher baseline BUN has been reported as a strong predictor of increased post-discharge mortality in patients hospitalized for HF (32), supporting our finding that BUN significantly mediated the association between ACAG and mortality.

Elevated lactate is the most common and clinically significant unmeasured anion, contributing to increased ACAG levels in critical illnesses (25). Lactate accumulation can occur due to tissue hypoperfusion in the context of HF (33). The liver and kidneys are two primary organs responsible for lactate clearance (34). Based on these considerations, heart rate, SpO₂, hemoglobin, ALT, AST, creatinine, and blood urea nitrogen were included in the multivariate regression analysis to examine their associations with lactate levels. The results showed that SpO₂ and hemoglobin were negatively associated with lactate levels, while heart rate, ALT, AST, creatinine, and blood urea nitrogen were positively associated. These findings suggest that tissue hypoxia caused by HF promotes anaerobic metabolism and increases lactate production, while liver and kidney dysfunction impairs lactate metabolism and excretion, ultimately leading to lactate elevation and contributing to increased ACAG levels. This is also reflected in our finding that lactate mediated the association between ACAG and the risk of all-cause mortality. Therefore, optimizing oxygen delivery through increasing hemoglobin levels, maintaining sufficient SpO₂, close monitoring and support of hepatic and renal function are critical to reduce lactate levels and improve outcomes in patients with HF. In the present study, the predictive ability of lactate for mortality was found to be inferior to that of ACAG. While lactate levels were associated with lower SpO₂ and hemoglobin, and higher ALT, AST, creatinine, and blood urea nitrogen, we did not analyze the extent to which lactate contributes to the observed increase in ACAG, which could be an interesting direction for future research.

A previous study assessed the ability of ACAG to predict long-term mortality (approximately 6 years) in hospitalized HF patients, and reported an AUC value of 0.773, indicating a moderate or acceptable level of performance (21). However, these authors studied a small sample size (205 patients) and did not provide details on hospitalization frequency. We found an AUC value of 0.647 for ACAG, which fell within the category of poor or low performance, although it was higher than that of AG. Our findings highlight the limited standalone predictive value of ACAG, support its use alongside other clinical markers to improve prognostic accuracy, and may alert other clinicians and researchers to avoid redundant studies.

This study has the following limitations: (1) it is observational, limiting the ability to infer causality; (2) ACAG data were missing in approximately 40% of patients, which, although likely random, may still introduce bias and reduce statistical power; (3) the data analyzed in the present study were collected from a single medical center, limiting the generalizability of the findings; and (4) as MIMIC-IV includes data only from the Beth Israel Deaconess Medical Center, prior ICU admissions at other institutions could not be accounted for and may represent a source of unmeasured confounding. Prospective studies are needed to validate our findings.

There is a significant positive association between ACAG and short- and medium-term all-cause mortality in HF patients with a single ICU admission. Notably, according to ROC analysis, ACAG had a performance that ranges from poor to, at most, moderate, in predicting mortality. The ACAG index should be combined with other clinical markers to ensure accurate risk stratification, and clinicians should be cautious in solely relying on ACAG for decision-making. Taken together, if anything, our results reaffirm that pathophysiologically uninformed composite scores, when used in isolation, are unlikely to yield strong predictive ability, and that it is questionably whether this meaningfully informs therapeutic intervention targets.