INTRODUCTION
Acute cardiogenic pulmonary edema (ACPE) is a critical and life-threatening medical emergency marked by the rapid onset of severe respiratory distress due to fluid accumulation in the lungs [1]. This condition often results from underlying heart failure and poses a significant challenge to healthcare providers, as it requires prompt and precise clinical assessments to mitigate its high early mortality risk and improve patient outcomes [2,3]. ACPE demands urgent intervention because delays in diagnosis or treatment can lead to rapid deterioration, underscoring the need for tools that allow clinicians to quickly and accurately assess risk [3,4].
In emergency settings, where timely decision-making is crucial, the ability to swiftly gauge the likelihood of short-term adverse outcomes is paramount [5–7]. Clinical scores have emerged as essential tools in this context, offering a systematic approach to predicting mortality risk based on easily accessible patient data at presentation [8–11]. These scores not only aid in the rapid assessment of disease severity, but also facilitate informed decision-making regarding treatment strategies, resource allocation, and patient triage [12,13]. Given the high stakes of managing ACPE, an effective clinical scoring system is crucial for guiding clinicians through fast-paced and high-pressure situations.
The utility of clinical scores lies in their simplicity and strong predictive value. By distilling complex clinical information into an easy-to-use tool, clinicians can assess patient risk quickly and confidently, optimizing interventions such as intensive monitoring, aggressive therapy, or advanced diagnostics [14]. However, despite the existence of several clinical scores designed to predict short-term mortality in ACPE, many have notable limitations. Most of these scoring systems fail to incorporate key laboratory findings, which are known to have independent prognostic value in predicting outcomes [8–10,15]. Moreover, many existing tools rely on subjectively assessed clinical parameters, potentially introducing bias and reducing the objectivity of risk stratification [8,11]. Additionally, some models suffer from small sample sizes or lack validation across diverse populations, further limiting their robustness and generalizability [10,11]. Given these deficiencies, there is a pressing need for a more comprehensive and validated scoring system that includes key clinical and laboratory variables, such as comorbidities, etiological factors, and measures of cardiac function [15,16]. Such a tool would offer greater accuracy in predicting mortality risk, improving the management of this complex condition and ensuring that high-risk patients receive the appropriate level of care.
To address these shortcomings, we developed a novel scoring system—the SABIHA score—specifically designed to predict short-term mortality in patients with ACPE. The SABIHA score stands for systolic blood pressure (SBP), age, blood urea nitrogen (BUN), invasive mechanical ventilation (IMV) requirement, heart rate (HR), and anemia—variables selected for their established prognostic significance in ACPE. Each component has been independently linked to outcomes in this patient population, making them ideal contributors to a simplified yet effective risk stratification tool. Importantly, the SABIHA score is built on data from a large, multicenter cohort representing a diversity of patient populations and clinical scenarios, which enhances its generalizability and reliability.
By integrating key clinical and laboratory data, the SABIHA score addresses the limitations of existing models and offers clinicians a straightforward, objective, and memorable tool for predicting short-term mortality in ACPE. In this study, we demonstrate how the SABIHA score can be used as a predictive risk stratification tool to enhance clinical decision-making and patient care in acute settings.
METHODS
Ethics statement
The study protocol was approved by the Ethics Committee of Harran University Faculty of Medicine (No. HRÜ/23.15.07). Informed consent was waived due to the use of de-identified data and the retrospective nature of the study. All study procedures were conducted in accordance with the Declaration of Helsinki.
Study design and setting
The SABIHA score was developed using multicenter data collected from a diverse cohort of patients across several institutions, ensuring that the scoring system reflects a wide range of clinical scenarios and enhancing its generalizability and reliability in various healthcare settings. This retrospective observational cohort study, conducted across six distinct centers between January 2015 and December 2023, aimed to establish a prognostic score based on key risk determinants influencing the short-term prognosis of patients admitted to the emergency department (ED) and hospitalized for ACPE.
ACPE was delineated as the presence of alveolar or interstitial edema, confirmed through chest x-ray and/or evidenced by an oxygen saturation level <90% on room air before initiation of treatment, concomitant with pronounced respiratory distress, auscultatory findings of crackles over the lungs, and orthopnea [17,18]. Inclusion criteria encompassed individuals aged >18 years who presented with a clinical diagnosis of ACPE. Exclusion criteria encompassed patients experiencing cardiogenic shock necessitating urgent invasive interventions, including those with ST-elevation myocardial infarction, those who were dialysis-dependent, a hospital stay <24 hours (refused further treatment), concomitant terminal disease, patients lost to follow-up, and those with missing prognostic data.
Data collection
Patients were sequentially included in the study, and solely the initial presentation within the study evaluation timeframe was incorporated, thereby ensuring the absence of duplicate representations of any individual in the dataset. De-identified data comprising comprehensive patient information extracted from medical records such as demographic details, medical history, fundamental clinical characteristics, initial evaluation parameters, administered treatments, conducted procedures (electrocardiogram, echocardiography, chest radiography, medical treatment, and IMV), and the trajectory of hospitalization from admission to discharge, inclusive of short-term mortality records, were systematically collected.
Conventional therapy, comprising standard administration of diuretics, opiates, and nitrates alongside continuous positive airway pressure or conventional oxygen support for oxygen supplementation, was initiated as the primary intervention. IMV was employed to provide oxygen support for patients unresponsive to conventional treatment within the initial 2-hour timeframe [19,20].
Study endpoint and follow‑up
Death from any cause within 30 days after the ED visit was considered the primary endpoint. Evaluation of the primary outcome measure was conducted by investigators who were blinded to patient status regarding predictor variables. Data from multiple sources were utilized, including ED health records, hospital health records, a computerized hospital patient tracking and record system, and a review of provincial death records. Standardized definitions were used for all patient-related variables, clinical diagnoses, and short-term outcomes.
Statistical analysis
Analyses were performed using the statistical programs IBM SPSS ver. 29.0 (IBM Corp) and R ver. 4.0.0 (R Foundation for Statistical Computing). Kolmogorov-Smirnov test was used to assess if the distribution of continuous variables was normal. Utilizing the expectation-maximization strategy, the few missing values were substituted. Continuous variables exhibiting a normal distribution are reported as mean±standard deviation and were analyzed using Student t-test or the t-test, as applicable, while those demonstrating a non-normal distribution are presented as median (interquartile range) and were compared using the Mann-Whitney test. Categorical variables are expressed as frequencies (percentages) and were compared between groups using the chi-square test or Fisher exact test, as deemed appropriate.
Split-sample testing methodology was employed to develop and validate the risk model. Using computerized randomization, individuals were assigned to derivation and validation cohorts at a 4:3 ratio. In the derivation set, variables exhibiting a significant association (P<0.1) with mortality were included in a multivariable model. Furthermore, variables demonstrating a significance level of P<0.1 in the univariable logistic regression analysis were incorporated into subsequent multivariable analyses employing the forward stepwise selection method to identify significant predictors. Factors independently predictive of outcome (P<0.05) were then selected for the development of a clinical score. To facilitate implementation of this score in clinical practice, a scoring system was developed [21].
The risk score model was recreated by dichotomizing continuous variables that were significant in the multivariable regression analysis according to optimal cutoff values. The optimal cutoff values for each continuous variable were identified using receiver operating characteristic (ROC) curve analysis, and the Youden index was employed to determine the point that maximized the sensitivity and specificity for predicting mortality. We analyzed the performance of the risk score using unweighted variables, i.e., each variable was dichotomized and weighted as either 1 or 0. A weight of 0 was assigned to the category with the lowest mortality for each variable. Logistic regression analysis was conducted on the SABIHA score categories, setting 0 as the reference point, to evaluate the odds of short-term outcomes. Individual scores were summed to obtain a final denoted as the SABIHA score. Initially, this analysis was performed within the derivation dataset, after which the derived predictor score was applied to the validation sample to examine its correlation with short-term outcomes. The Mantel-Haenszel test was selected to assess the linear association between the stratified risk categories of the SABIHA score and the incidence of primary outcome events. The ability to predict the endpoint was assessed by measuring the area under the ROC curve (AUC) with a 95% confidence interval (CI) to evaluate the discrimination of the prognostic score. Discriminatory capacity was assessed using the concordance index (C-index), while internal validation was conducted through the bootstrap method with 2,000 iterations to generate bias-corrected C-index values. The Hosmer-Lemeshow goodness-of-fit test and Brier score were employed for calibration assessment. The developed risk score (SABIHA score) was categorized as low (0–2 points), moderate (3–4 points), or high (5–6 points) according to the total clinical score obtained. Time-to-event data of these three groups were evaluated using Kaplan-Meier curves and compared using the log-rank test. Overall survival time was calculated from the time of admission to the day of death or the last follow-up. A two-sided P<0.05 was deemed statistically significant. Additional methods are provided in Supplementary Material 1.
RESULTS
Population characteristics
Out of 1,129 patients, 41 patients were excluded from the study: 23 presenting with cardiogenic shock, 2 requiring dialysis, 4 discharged within 24 hours, 1 diagnosed with concomitant terminal illness, 2 lost to follow-up, and 9 lacking prognostic data. Ultimately, 1,088 patients were enrolled and allocated randomly into one of two groups: a derivation cohort, consisting of 623 individuals, and a validation cohort, consisting of 465 individuals, at a ratio of 4:3 (Fig. 1). The mortality was 9.0% (56 of 567) in the derivation cohort and 9.2% (43 of 422) in the validation cohort. Survivor demographics, clinical history, comorbidities, physiological parameters, arterial blood gas variables, electrocardiographic findings, laboratory results, echocardiographic assessments, and in-hospital mechanical and pharmacological treatment data were analyzed to identify factors independently predictive of short-term mortality among patients with ACPE, juxtaposed with corresponding data from those who died (Table 1).
Logistic regression analysis and the SABIHA score
Logistic regression analysis was conducted within the derivation cohort (Table 2). Following the univariate analysis, variables with a significance level of P<0.1 were included in the multivariable analysis to identify those most relevant to the study outcomes. Multivariate analysis identified six variables (SBP, age, BUN, IMV requirement, HR, and anemia) that had significant correlations (P<0.05) with the primary outcomes. These variables were retained for inclusion in the final model development. The continuous variables of SBP, age, BUN, and HR were dichotomized and the multivariable regression analysis was repeated; all predictors maintained their clinical significance (all P<0.05) (Table 3). The final selected predictors of short-term outcomes used to develop the scoring system were as follows (continuous variables were dichotomized): 1×(SBP <110 mmHg); 1×(age ≥75 years); 1×(BUN ≥33 mg/dL); 1×(IMV requirement); 1×(HR ≥110 beats per minute); 1×(presence of anemia) (Table 4).
Risk stratification of the SABIHA score in the derivation and validation cohorts
Values were assigned to the risk score model (SABIHA score) obtained from the derivation cohort, with a minimum of 0 and a maximum of 6 points. In the derivation cohort, the incidence of short-term all-cause deaths was observed to increase from 0 to 6 points as follows: 0 points, 1.6% (4 of 250); 1 point, 2.3% (5 of 222); 2 points, 9.8% (8 of 82); 3 points, 36.6% (15 of 41); 4 points, 66.7% (12 of 18); 5 points, 71.4% (5 of 7); and 6 points, 87.5% (7 of 8). The probability of outcome events increased gradually for patients with different risk stratifications (low, 0–2 points; moderate, 3–4 points; and high, 5–6 points) in both derivation and validation cohorts (Fig. 2). The Mantel-Haenszel test demonstrated a significant linear relationship between risk stratification and the occurrence of outcome events (χ2=37.21, P<0.001). ROC curves for baseline and full models are shown in Fig. 3. The impact of the SABIHA score on mortality was examined using a generalized additive model, with the baseline model incorporating SBP, age, BUN, IMV requirement, HR, and anemia to estimate mortality risk. In the derivation cohort, the baseline model demonstrated a strong discriminatory ability for mortality risk (AUC, 0.949), which was similarly high in the validation cohort (AUC, 0.840). This analysis highlights the model’s robustness in both derivation and validation settings.
Kaplan-Meier survival analysis revealed notable distinctions in short-term event outcomes among the risk categories delineated by the SABIHA score (Fig. 4). These differences were found to be statistically significant for both cohorts, as indicated by the results of the log-rank test (P<0.001). Additional results are provided in Supplementary Material 2 and Supplementary Figs. 1–5.
DISCUSSION
ACPE is a critical emergency characterized by the rapid onset of respiratory distress due to fluid accumulation in the lungs. This condition has persistently high early mortality rates, underscoring the need for effective risk stratification strategies [1–7]. This study aimed to address this need by developing and validating the SABIHA score, a practical risk model for predicting short-term mortality in patients with ACPE. Using a retrospective cohort design, data were drawn from six health centers and included 1,088 patients, divided into derivation and validation cohorts, to support robust model development and testing. The SABIHA score was built on key, readily accessible clinical variables—age, BUN, HR, intubation status, anemia, and SBP—identified through multivariable analysis as independent predictors of mortality risk. The score demonstrated strong predictive accuracy, with C-indices of 0.879 in the derivation cohort and 0.863 in the validation cohort, in addition to excellent calibration. By integrating important clinical and laboratory parameters, including BUN and hemoglobin levels often overlooked in previous models, the SABIHA score provides a comprehensive approach to ACPE risk stratification [10,11]. Its validation across diverse patient populations enhances the score’s applicability and generalizability in clinical practice, supporting its use as a reliable tool for mortality risk assessment in ACPE patients.
The SABIHA score comprises six robust parameters—SBP, age, BUN, IMV requirement, HR, and presence of anemia—that are strong predictors of short-term mortality in ACPE, supported by extensive literature and our findings. SBP serves as a vital hemodynamic marker, with lower values indicating greater cardiac compromise and mortality risk [6–11]. Age correlates with overall health and comorbidity burden, increasing vulnerability to adverse outcomes [5,22–24]. Elevated BUN levels indicate renal impairment and systemic congestion, reflecting heart failure severity and a poorer prognosis [5,9,25]. The requirement for IMV signifies severe respiratory compromise and advanced disease, heightening mortality risk [26,27]. Increased HR indicates sympathetic activation and hemodynamic instability, marking heart failure decompensation severity [28–30]. Anemia exacerbates tissue hypoxia and cardiac workload, further predicting adverse outcomes [31,32]. Together, these parameters allow for a comprehensive assessment of physiological derangements in ACPE, enabling effective mortality risk stratification and tailored management strategies.
When comparing the SABIHA score to previous models like the 3CPO (Three Interventions in Cardiogenic Pulmonary Edema) score [9] and the Pulmonary Edema Prognostic Score (PEPS) [10], it is crucial to address the limitations of these earlier scores. The 3CPO score, which is based on a limited set of clinical parameters, does not incorporate critical laboratory markers and comorbidities, potentially reducing its predictive accuracy. In contrast, the SABIHA score integrates a broader range of variables, including laboratory parameters such as BUN and hemoglobin levels, providing a more holistic assessment of mortality risk in ACPE. Notably, admission BUN and anemia are recognized as important predictors of clinical outcomes in acute heart failure [3,5,6,9,25,33]. The 3CPO trial also failed to demonstrate a significant reduction in mortality associated with noninvasive positive pressure ventilation in ACPE cases [7,34–39], while IMV has been identified as an independent predictor of poor outcomes, consistent with our findings [5–7,34–39].
Similarly, the PEPS relies primarily on clinical parameters, which may limit its predictive capability by not incorporating essential laboratory markers and comorbidities [11]. Additionally, PEPS was developed based on a small sample size. The SABIHA score was developed based on a large sample size and a comprehensive variable set, including anemia, intubation status, and BUN levels, ensuring more accurate assessment of mortality risk in ACPE. Both the 3CPO and PEPS scores also demonstrated inadequate validation across diverse patient populations, potentially compromising their clinical generalizability [16]. In contrast, our study utilized a large, diverse cohort from multiple health centers, allowing robust model development and rigorous validation of the SABIHA score. Furthermore, inclusion of a validation cohort enhances the external validity of our findings and the applicability of the SABIHA score in various clinical settings.
The SABIHA score represents a significant advancement in risk stratification of ACPE, effectively addressing the limitations of previous models like 3CPO and PEPS. By incorporating comprehensive clinical and laboratory parameters from a large, diverse patient population, the score provides a practical and reliable tool for predicting short-term mortality, allowing clinicians to identify high-risk patients and allocate resources efficiently. Its simplicity facilitates easy integration into routine practice, enhancing its clinical utility. In conclusion, the SABIHA score is a validated and innovative tool for predicting short-term mortality in ACPE, and future research should aim to further validate it in diverse populations and assess its impact on clinical decision-making and outcomes.
Limitations
There were several limitations to our study. Firstly, its retrospective design may have resulted in the introduction of biases, documentation was incomplete, and causal relationships could not be established. Although efforts were made to address these limitations through rigorous data collection and analysis protocols, retrospective studies remain susceptible to confounding factors that may affect the validity and generalizability of the findings. Secondly, selection bias may have been present due to our study's inclusion criteria, which could have excluded patients with specific characteristics or comorbidities. Additionally, the exclusion of patients with cardiogenic shock requiring urgent interventions and those with missing prognostic data may have influenced the study's results and limited its external validity. Thirdly, the generalizability of our findings may be limited since the study was conducted across six health centers. Variations in patient demographics, clinical practices, and healthcare infrastructure across different regions may impact the applicability of the SABIHA score in diverse clinical settings. Fourth, despite efforts to standardize data collection procedures, variations in the measurement of variables such as blood pressure, HR, and laboratory parameters may have occurred across different healthcare facilities, potentially affecting the accuracy and consistency of the results. Fifthly, while we included a validation cohort to assess the performance of the SABIHA score, external validation in independent cohorts is essential to confirm the robustness and generalizability of our findings. Future studies should aim to validate the SABIHA score in diverse patient populations to ensure its reliability across different clinical settings. Lastly, although the SABIHA score incorporates several clinical and laboratory parameters, other potentially relevant prognostic factors, such as imaging findings, biomarkers, and functional status, were not considered. Furthermore, the retrospective nature of this study limited the availability of consistent and reliable respiratory rate data, preventing its direct inclusion in the SABIHA score and necessitating the use of indirect indicators.
Future research could explore the effect of adding these variables on the predictive accuracy of the risk score. Additionally, the clinical utility of the SABIHA score in guiding treatment decisions and improving patient outcomes requires further investigation. While the SABIHA score was validated using a 4:3 split in a multicenter cohort, this approach does not provide the full rigor of external validation across separate institutions. Future studies could enhance the validity of the score by applying the model to fully independent cohorts. Prospective studies evaluating the impact of risk stratification using the SABIHA score on clinical decision-making and patient outcomes are warranted.
Conclusions
The SABIHA score is a novel prognostic tool designed to predict short-term mortality in ACPE patients. Using data from a large, diverse cohort across six health centers, it incorporates the accessible clinical parameters of SBP, age, BUN, intubation status, HR, and anemia. The score demonstrated excellent predictive accuracy in both derivation and validation cohorts.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Supplement1
Print 
