Despite significant improvements in preventive measures, cardiac arrest (CA) remains a serious public health issue [1]. In the United States, approximately 290,000 in-hospital cardiac arrests (IHCA) and 350,000 out-of-hospital cardiac arrests (OHCA) occur annually [2]. The incidence of IHCA ranges from approximately 1 to 17 per 1000 admissions, though the global incidence in adults has not been well characterized [3]. IHCA remains a relatively neglected condition compared to OHCA. A systematic review of 92 randomized CA clinical trials (1995–2014, $\ge$50 patients each) found that only 4 (4%) focused exclusively on IHCA patients [4]. The 2010 American Heart Association (AHA) Emergency Cardiovascular Care (ECC)Committee established goals to improve CA survival rates, aiming to double survival for both OHCA and IHCA by 2020. Current reported 30-day survival rates for patients with OHCA and IHCA are 17% and 24%, respectively [5], underscoring the necessity of focusing on both IHCA and OHCA to improve patient survival.

Acute kidney injury (AKI) occurs in approximately 50% of post-cardiac arrest patients [6] and has been associated with poor clinical outcomes in studies of cardiac intensive care unit patients [7]. AKI is a significant risk factor for unfavorable neurological outcomes and elevated mortality following CA [8, 9]. Previous research has indicated that severe AKI is associated with reduced hospital survival (48% versus 65%, p = 0.006) and reduced 12-month survival with favorable neurobehavioral outcomes (30% versus 53%, p 0.001) [10].

In CA patients, early recognition of AKI is critical, as renal replacement therapy (RRT) becomes the only therapeutic option when there is severe progression of the disease. Wang et al. [11] established and validated a predictive model for early acute kidney injury after cardiac arrest resuscitation through retrospective collection of clinical cases. Similarly, Lin et al. [12] developed and validated a risk prediction model for AKI after cardiac arrest using clinical data. However, both studies share the same limitations: they are single-center retrospective studies with relatively small sample sizes and limited included variables. Therefore, developing robust risk prediction models for cardiac arrest-associated acute kidney injury (CA-AKI) is necessary. Such models would facilitate early and accurate detection of CA-AKI, enabling timely intervention before severe progression occurs.

Using MIMIC-IV data, we developed a well-calibrated CA-AKI prediction model with routinely available clinical variables.

The model incorporated five accessible predictors (weight, SpO₂, sodium, SOFA, OASIS) and demonstrated clinical utility via DCA/CIC analysis, enabling early high-risk CA-AKI identification and improved clinical decisions.

Demographic indicators are important variables in predictive models. This study found that body weight is positively correlated with the occurrence of CA-AKI. Overweight, as the second major risk factor for cardiovascular diseases [15], can significantly increase the risk of conditions such as hypertension, acute coronary syndrome (ACS), diabetes, and dyslipidemia, all of which are important predisposing factors for CA [16]. A recent study revealed that among patients with ACS, the overall incidence of sudden cardiac arrest (SCA) is as high as 17.5% [17]. Obesity can affect renal function through obesity-related glomerulopathy (ORG) and alterations in renal hemodynamics. Activation of the renin-angiotensin-aldosterone system (RAAS), hypertension, renal lipotoxicity mediated by adipokines, and activation of the nucleotide-binding oligomerization domain-like receptor pyrin domain-containing 3 (NLRP3) inflammasome can all exacerbate renal damage [18]. Therefore, an increase in body weight not only elevates the risk of CA but also promotes the occurrence of AKI through multiple mechanisms. A study conducted at Santo Bortolo Hospital in Italy included 142 patients with CA-AKI. The results showed that the body mass index (BMI) of these patients was 25.2 (23.8–27.7), falling within the overweight range [19]. This result is consistent with the conclusion of our study, which indicates that body weight is positively correlated with the occurrence of CA-AKI.

Vital signs are one of the most commonly used and important indicators in predictive models. A growing body of animal experimental and clinical data suggests that immediate inhalation of 100% oxygen CA is prone to causing hyperoxemia. Elevated oxygen partial pressure can increase the production of oxygen-free radicals, exacerbate reperfusion injury, and be associated with unfavorable neurological outcomes. Although some studies have adjusted the target SpO₂ for patients with OHCA after return of spontaneous circulation (ROSC) from 98–100% down to 90–94%, this adjustment has not significantly improved the discharge survival rate [20]. Excessive oxygen administration can also disrupt the balance between the production and clearance of reactive oxygen species within the kidneys, promoting cell apoptosis, inflammatory cascades, and renal tissue fibrosis, thus providing a pathological basis for AKI [21]. A recent study found that in comatose OHCA patients, those who simultaneously received a lower mean arterial pressure and a lenient oxygen delivery strategy had a significantly higher risk of AKI [22]. In this study, we observed a negative correlation between SpO₂ levels and the incidence of CA-AKI, indicating that precise control of oxygenation targets in the management of vital signs for IHCA patients may be a crucial step in reducing renal injury and improving long-term prognosis.

During the construction of predictive models, laboratory indicators are often used as surrogate markers to indirectly reflect clinical outcomes. Sodium metabolism disorders are extremely common in the ICU, yet they are rarely incorporated into the prognostic assessment of CA. Current data show that among comatose CA patients who have achieved ROSC, approximately 20% already exhibit hyponatremia upon admission to the ICU, while hypernatremia is rare. One study has demonstrated that even after adjustments, hyponatremia remains associated with a reduced probability of attaining a favorable functional outcome at 180 days [23]. The underlying mechanism for these findings is that sodium ions are involved in the electrical conduction of myocardial cells. Severe hyponatremia can induce paroxysmal bradyarrhythmias, which may subsequently lead to pulseless electrical activity or even CA [24]. The kidneys are a key organ for regulating water-electrolyte balance. In clinical practice, AKI often coexists with sodium metabolism disorders. When renal function is impaired, both urine dilution and concentration functions may be affected. Abnormalities in urine dilution function, coupled with excessive water intake by the patient, can result in an excess of water in the body, leading to dilutional hyponatremia. Conversely, if the urine concentration function is impaired and the patient has insufficient water intake, excessive water loss from the body can occur, resulting in hypernatremia. These conditions underscore the critical role of the kidneys in regulating water and sodium balance, particularly during AKI, when impaired renal function significantly affects the metabolic equilibrium of sodium and water. Although existing literature has reported a linear relationship between the coefficient of variation in serum sodium levels and the risk of AKI, the causal relationship between the two remains unclear [25]. In cases of AKI, the incidence of sodium metabolism disorders ranges from 22.5% to 24.6%, and patients with such disorders face a significantly higher risk of mortality [26]. Marahrens et al. [27] found that serum sodium levels were associated with the risk of in-hospital mortality in AKI patients, with hypernatremic patients exhibiting a higher mortality risk. Additionally, another study also indicated that both hyponatremia and hypernatremia were associated with poor outcomes in AKI patients [28]. In this study, we observed a negative correlation between serum sodium levels and the risk of CA-AKI. In summary, sodium metabolism disorders are not only related to the occurrence of AKI but also closely associated with the prognosis of CA patients. Therefore, for patients with sodium metabolism disorders, close monitoring of their cardiac rhythm and renal function is crucial to promptly identify and manage potential risks.

In clinical practice, various scoring systems are frequently incorporated as predictive variables into clinical prediction models. The APACHE II score is the most widely used disease severity scoring system in ICUs worldwide [29]. However, it requires the collection of 12 physiological and laboratory indicators, along with 2 disease-related variables [30], making data acquisition time-consuming and not conducive to early, rapid assessment. The SOFA score, on the other hand, is used to quantify organ dysfunction. Studies have shown that the SOFA score is an independent risk factor for mortality in elderly ICU patients [31] and can objectively assess the severity of the post-cardiac arrest syndrome (PCAS) [32]. When combined with blood lactate levels, the SOFA score demonstrates improved predictive performance for the discharge survival rate of PCAS patients [33]. A study conducted by the University of Science and Technology of China, which included 347 patients with CA (197 cases in the CA-AKI group and 150 cases in the no CA-AKI group), revealed that although there was no statistically significant difference in the SOFA scores between the AKI group and the no-AKI group (11.294 $\pm$ 3.400 versus 10.840 $\pm$ 3.595), the mean score in the AKI group was slightly higher than that in the no-AKI group [34]. In this study, we also found a positive correlation between the SOFA score and the risk of CA-AKI. The OASIS score, which does not require laboratory or imaging tests, is simple and easy to use and has been proven to effectively discriminate the condition and prognosis of ICU patients [35]. A multicenter prospective study compared the predictive abilities of OASIS, APACHE II, SAPS II, and SOFA for 28-day mortality in ICU patients with acute kidney injury (ICU-AKI). The results indicated that OASIS performed the best (an OASIS score $>$33 suggested a poor short-term prognosis) [36]. In summary, the predictive value of SOFA and OASIS in patients with AKI, CA, and those in the ICU has been validated. However, their predictive performance for CA-AKI remains unclear. This study confirmed that both scores were positively correlated with the risk of CA-AKI. Ideally, a “gold standard” should be used for clinical diagnosis. However, when CA patients are admitted to the ICU, they often lose consciousness and are subject to numerous objective constraints, making it difficult to immediately conduct gold standard examinations. Under such circumstances, the rational application of the aforementioned scoring systems can assist clinicians in rapidly assessing the patient’s condition and predicting the prognosis.

This prediction model has several advantages in the management and prognosis of patients with CA. First, it is helpful for the early detection of CA-AKI patients. One study have found that whether the condition of patients with severe AKI can improve and recover early has a significant impact on their survival rates, hospitalization rates, and other outcomes [37]. The remission time of AKI is related to the long-term prognosis of patients. Research has shown that if AKI resolved within 7 days, the 1-year survival rate exceeded 90%. In contrast, the 1-year survival rate after discharge was only 77% [38]. Therefore, early identification of CA-AKI through prediction models is crucial for promptly adopting intervention measures. Early intervention can prevent adverse neurological outcomes, reduce disability rates, and improve survival rates. Second, it is helpful for the rational allocation of clinical resources. This prediction model aids in the rational allocation of clinical resources, such as dialysis machines and intensive care unit beds, thereby enabling more efficient use of limited medical resources and improving patient outcomes. Moreover, continuous refinement of the model based on real-world feedback can further enhance its performance and clinical utility.

This research has identified a set of independent predictors closely linked to the occurrence of AKI in patients with IHCA who were admitted to the ICU. These predictors encompassed a range of factors, specifically weight, SpO₂, sodium, SOFA, and OASIS. The prediction model that was developed based on these predictors demonstrated promising predictive results, accurately forecasting the likelihood of the development of AKI. Moreover, it exhibited significant clinical utility in the risk evaluation of AKI among ICU-admitted IHCA patients, aiding the medical staff in making more informed treatment decisions.

Future studies should not only deploy this model across diverse clinical environments but also rigorously validate its efficacy in a broader spectrum of settings to ensure its generalizability and widespread applicability. It would also be worthwhile to investigate the feasibility of integrating this model into electronic health records (EHR) or real-time clinical decision-support systems, which could offer a practical and forward-looking perspective for future research endeavors.

The datasets supporting the findings of this study are available from the corresponding author upon reasonable request.

HA, ZL, YN designed the research study and formulated the research plan. GX, KH, and JD carried out the research, completing the collection of experimental data and conducting all related experimental procedures. TT and GH analyzed the experimental data, interpreted the research results, and provided critical insights into the data discussion. HA drafted the original manuscript. All authors contributed to the critical revision of the manuscript for important intellectual content, and participated in the discussion and refinement of the study’s core conclusions. XY and HF as corresponding authors, supervised the entire research process, provided overall guidance on the study design and manuscript revision, and finalized the research conclusions. All authors read and approved the final version of the manuscript to be published. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work, ensuring that any questions related to the accuracy or integrity of the work are appropriately investigated and resolved.

This study is a secondary analysis of the MIMIC IV database and does not require additional ethical approval. The study was conducted in accordance with the Declaration of Helsinki. The database contains de identified data without any protected health information, and informed consent was waived due to the use ofde-identified data.

We appreciate the researchers at the MIT Laboratory for Computational Physiology for publicly sharing the MIMIC-IV clinical database.

This work was supported by the National Key Research and Development Program of China (2021YFC3002200, 2021YFC3002202), the Postgraduate Research and Innovation Program of Tianjin Municipal Education Commission (2022BKY113), the Chunshui discipline construction fund of China (302-0704000002), and the National Key Research and Development Program of China (2023ZD0505500, 2023ZD0505504).

The authors declare no conflict of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/RCM47434.