Evaluation of the C-reactive protein–albumin–lymphocyte (CALLY) index as a prognostic marker in patients with sepsis

Sepsis remains a leading cause of mortality worldwide, and early identification of high-risk patients is essential. The C-reactive protein–albumin–lymphocyte (CALLY) index has emerged as a potential prognostic biomarker but has not been widely studied in emergency department (ED) populations.

We conducted a retrospective cohort study including adult patients diagnosed with sepsis upon ED admission. The CALLY index was calculated as (albumin × lymphocyte count) / C-reactive protein (CRP), using laboratory values obtained at presentation. The primary outcome was 30-day all-cause mortality. Predictive performance was assessed using multivariable logistic regression with least absolute shrinkage and selection operator (LASSO) variable selection, calibration metrics, and Receiver Operating Characteristic (ROC) curve analyses.

A total of 669 patients were included, with a 30-day mortality rate of 23% (n = 156). Non-survivors had significantly higher CRP (107.30 ± 40.97 vs. 94.01 ± 28.09 mg/L, p = 0.002), lower albumin (2.81 ± 0.30 vs. 3.59 ± 0.22 g/dL, p < 0.001), and lower lymphocyte count (0.71 ± 0.31 vs. 1.30 ± 0.40 × 10⁹/L, p < 0.001). The mean CALLY index was markedly elevated in deceased patients (69.7 ± 54.2 vs. 23.1 ± 15.6, p < 0.001). The area under the ROC of the CALLY index was 0.906 (95% CI, 0.879–0.934). At the threshold yielding the maximum Youden Index (31.32), the sensitivity was 0.853, specificity was 0.854, and accuracy was 0.852.

The CALLY index is a simple and accessible biomarker independently associated with 30-day mortality in ED patients with sepsis and may support early risk stratification in acute care.

Not applicable.

Sepsis remains a critical global health concern, defined by an uncontrolled immune reaction to infection that can lead to severe organ failure [1]. Recent estimates indicate that sepsis accounts for approximately 11 million deaths annually, representing nearly 20% of all global deaths [2]. Despite advancements in medical care, the burden of sepsis persists, particularly in low- and middle-income countries where healthcare resources are limited. In some low- and middle-income countries, sepsis mortality rates can exceed 40%, compared to 15–25% in high-income countries [3]. Early identification and timely intervention are crucial, as delays in recognition and treatment are associated with increased mortality [4]. However, the heterogeneity of sepsis presentations and the lack of specific diagnostic tools continue to impede prompt diagnosis and management [5]. Consequently, there is an urgent need for reliable biomarkers and predictive models to enhance early detection and improve patient outcomes.

Isolated laboratory markers such as C-reactive protein (CRP), albumin, and lymphocyte count have long been studied for their association with sepsis outcomes, yet their individual limitations are well recognized [6, 7]. CRP reflects systemic inflammation but lacks prognostic precision when used alone [8]. Low albumin levels frequently accompany critical illness and signal both catabolic stress and endothelial dysfunction [9]. Concurrently, lymphopenia reflects the profound immune dysregulation seen in sepsis, including T cell exhaustion and impaired host defense [10]. Recognizing the need for a more integrated view, recent efforts have focused on composite indices that consolidate multiple biological dimensions into a unified metric. The C-reactive protein–albumin–lymphocyte (CALLY) index combines inflammatory, nutritional, and immune components into a single, easily computable score derived from standard laboratory tests.

Initially proposed for use in oncology, the CALLY index has been shown to predict prognosis in colorectal, gastric and esophageal cancer [11,12,13]. Emerging evidence suggests its potential relevance in critical illness. A recent cohort study of intensive care unit patients with sepsis demonstrated that higher CALLY scores were independently associated with lower short-term mortality and reduced incidence of acute kidney injury [14]. However, data on its prognostic performance in emergency department (ED) populations remain sparse. Given its simplicity and reliance on widely available data, the CALLY index may serve as a practical tool for early risk stratification in patients with sepsis. Its empirical design, based on routinely collected laboratory parameters, supports a practical and potentially generalizable approach for diverse emergency care environments.

Given its derivation from widely available laboratory parameters, the CALLY index may serve as a practical and scalable tool for early risk stratification. Clarifying its role in routine clinical practice could inform decision-making in high-risk patients and help guide triage and resource allocation in acute care settings. Therefore, this study aims to evaluate the prognostic value of the CALLY index in predicting 30-day all-cause mortality among patients diagnosed with sepsis in the ED.

Study design and setting

This retrospective cohort study was conducted at Kartal Dr. Lütfi Kırdar City Hospital, a tertiary care center in İstanbul, Türkiye. Adult patients who presented to the ED between January 1, 2022, and January 1, 2024, and met the criteria for sepsis at the time of admission were included. The study protocol was approved by the Kartal Dr. Lütfi Kırdar City Hospital Ethics Committee (Approval No: 2024/010.99/11/33, Date: 25.12.2024). Due to the retrospective nature of the study, the requirement for informed consent was waived. The study was conducted in accordance with the Declaration of Helsinki and adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Study population

Patients aged 18 years or older who presented with sepsis at ED admission were eligible. Sepsis was defined based on the Sepsis-3 criteria: suspected or confirmed infection accompanied by an acute increase of ≥ 2 points in the Sequential Organ Failure Assessment (SOFA) score. Diagnoses were verified retrospectively using clinical documentation, vital signs, and laboratory data from the ED visit. Patients were excluded if they met any of the following criteria: (1) missing laboratory values required to calculate the CALLY index (CRP, serum albumin, or absolute lymphocyte count), (2) history of end-stage renal disease requiring dialysis, (3) chronic liver disease including cirrhosis, (4) transfer from another hospital, or (5) unavailability of follow-up information to determine 30-day survival status.

Data collection and variables

All clinical data were obtained from the hospital’s electronic medical records (EMR). Demographic information, comorbidities (e.g., hypertension, diabetes, coronary artery disease, chronic kidney disease, chronic obstructive pulmonary disease, malignancy, and dementia), admission vital signs (systolic and diastolic blood pressure, heart rate, respiratory rate, body temperature, peripheral oxygen saturation), and Glasgow Coma Scale (GCS) scores were recorded. Laboratory values were extracted from the first blood draw performed upon arrival to the ED. The SOFA score was manually calculated by two emergency physicians based on recorded parameters at admission. None of the individual parameters required for SOFA score calculation had more than 5% missingness. Multiple imputation using chained equations was applied for variables with isolated missing values. In cases where PaO₂ was unavailable, and the patient had not received supplemental oxygen and maintained a peripheral oxygen saturation > 95%, the respiratory component was conservatively assigned a score of 0, in alignment with established SOFA estimation practices. Following imputation, complete SOFA scores were available for all included patients.

The CALLY index was calculated using the formula: CALLY = (albumin × lymphocyte count) / CRP, where CRP is expressed in mg/L, lymphocyte count in 10⁹/L, and albumin in g/dL. No adjustment was made for hydration status or acute-phase albumin shifts, reflecting routine clinical use. Laboratory data were internally validated, and implausible outliers were cross-checked against raw EMR entries.

Outcome definition

The primary outcome was 30-day all-cause mortality, defined as death occurring within 30 days of the index ED visit. Mortality status was determined using hospital records and cross-referenced with national death registries when available. Patients with insufficient documentation to verify 30-day mortality status were excluded from the analysis.

Analysis

All statistical analyses were performed using R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics were expressed as mean ± standard deviation for normally distributed continuous variables and as counts with percentages for categorical variables. Group comparisons between survivors and deceased patients were conducted using the independent samples t-test for continuous variables and the chi-square test for categorical variables. Variables with p < 0.05 in univariate analysis were considered for further modeling.

To identify variables independently associated with 30-day mortality, least absolute shrinkage and selection operator (LASSO) logistic regression was first applied using 10-fold cross-validation. A total of 23 variables were selected by LASSO and entered into a multivariable logistic regression model. Multicollinearity among predictors was assessed using variance inflation factor (VIF) analysis, and a reduced model including only coronary artery disease (CAD), dementia, and the CALLY index was retained.

The final logistic regression model was evaluated on an independent test set (20% holdout). Model discrimination was assessed using the area under the receiver operating characteristic curve (AUROC). Additional evaluation metrics included accuracy, balanced accuracy, and McNemar’s test for classification agreement. The added value of the CALLY index was examined through permutation testing and likelihood ratio test (LRT) comparing models with and without the index.

Model calibration was assessed with a Hosmer–Lemeshow goodness-of-fit test using decile groups of predicted probabilities and visualized with a calibration plot comparing observed and predicted mortality rates. The diagnostic performance of the CALLY index alone was evaluated by computing the AUROC with 95% confidence intervals using the DeLong method. The cutoff value corresponding to the maximum Youden Index—where the sum of sensitivity and specificity is highest—was used to summarize diagnostic performance, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and classification accuracy, all reported with 95% confidence intervals. Confidence intervals were estimated using 2000 stratified bootstrap replicates where applicable. Pairwise comparisons of AUROCs were performed using the DeLong method. A post hoc sample size adequacy check was performed. With 156 events and three predictors, the events-per-variable (EPV) was 52, which far exceeds the conventional threshold of 10 for logistic regression models [15]. In addition, the AUROC of the CALLY index (0.906; 95% CI, 0.879–0.934) had a narrow confidence interval (width = 0.055), indicating precise estimation. Post hoc power analysis (pROC package) demonstrated > 99% power for detecting the observed discrimination.

A total of 910 patients were screened in the emergency department, and 241 were excluded based on predefined criteria. The final cohort included 669 patients, of whom 156 (23%) died within 30 days (Fig. 1). Baseline demographic characteristics, vital signs, and clinical scores by mortality status are presented in Table 1. Compared to survivors, the deceased group was older and had higher systolic blood pressure, lower diastolic blood pressure, higher pulse rate, and higher respiratory rate (p < 0.001 for all). The mean body temperature was elevated in the deceased group, while SpO₂ was lower (p < 0.001 for both). The SOFA score was statistically significantly higher in the deceased group compared with survivors (p < 0.001). No statistically significant difference was observed in the qSOFA categories or GCS scores.

Patient flowchart

Full size image

Laboratory findings and clinical outcomes are shown in Table 2. The mean CALLY index was substantially higher in deceased patients (p < 0.001). No statistically significant differences were observed between the groups in terms of ICU admission (p = 0.50), blood culture positivity (p = 0.99), or urine culture positivity (p = 0.11). However, comorbidities such as diabetes (p < 0.001), chronic kidney disease (p < 0.001), coronary artery disease (p = 0.001), chronic obstructive pulmonary disease (p < 0.001), malignancy (p < 0.001), and dementia (p < 0.001) were significantly more frequent among deceased patients.

Logistic regression modeling was first performed using a reduced set of variables selected by LASSO and refined through variance inflation factor (VIF) analysis. This base model yielded an AUROC of 0.6151. When the CALLY index was added, the model’s AUROC increased substantially to 0.9516. In the final model, the CALLY index remained independently associated with mortality [odds ratio (OR) = 1.07 per unit increase, p < 0.001] (Table 3). Model calibration was adequate, as indicated by a non-significant Hosmer–Lemeshow test (p = 0.283). Permutation testing further supported the prognostic contribution of the CALLY index, with AUROC decreasing to 0.5642 when its values were randomized. A likelihood ratio test confirmed significant model improvement after adding the CALLY index [χ² = 171.02, p < 0.001].

The diagnostic performance of the CALLY Index in predicting 30-day mortality is summarized in Table 4. The AUROC was 0.906 (95% CI, 0.879–0.934), indicating excellent discriminative ability. The cutoff value corresponding to the maximum Youden Index (31.32), which reflects the point maximizing the combined sensitivity and specificity, yielded a sensitivity of 0.853 (95% CI, 0.776–0.917) and a specificity of 0.854 (95% CI, 0.778–0.903). The positive predictive value was 0.638 (95% CI, 0.551–0.719), and the negative predictive value was 0.949 (95% CI, 0.927–0.969). The overall classification accuracy at this threshold was 0.852 (95% CI, 0.806–0.885).

In additional analysis, the AUROC of the CALLY index was compared with that of CRP (0.582; 95% CI, 0.526–0.638), albumin (0.644; 95% CI, 0.589–0.700), lymphocyte count (0.841; 95% CI, 0.807–0.876), and lactate (0.769; 95% CI, 0.730–0.809). The corresponding receiver operating characteristic curves are shown in Fig. 2. Pairwise comparisons using the DeLong method indicated that the AUROC of the CALLY index was statistically higher than each of the individual biomarkers (Table 5).

Receiver operating characteristic curves for the CALLY index, CRP, albumin, lymphocyte count, lactate, SOFA, and qSOFA in predicting 30-day mortality

Full size image

In this retrospective cohort study of patients presenting with sepsis to the ED, the CALLY index was found to be independently associated with 30-day mortality. When incorporated into a multivariable logistic regression model, the index improved overall prognostic performance and demonstrated good model calibration. These findings support the potential utility of the CALLY index as a simple, accessible biomarker that integrates inflammatory, nutritional, and immune components, and may aid early risk stratification in acute care settings.

The value of the CALLY index in this study is further highlighted by its practicality in the ED, where rapid risk stratification is essential. Unlike prognostic models designed for intensive care units, our study evaluated patients at the point of first hospital contact, reflecting real-world ED conditions. Previous studies have shown that lab-based markers like the CRP-to-albumin ratio (CAR) and lactate-to-albumin ratio (LAR) predict sepsis mortality in ED settings; for instance, Yoo et al. found LAR outperformed SOFA for 28-day mortality, while Sisto et al. reported that CRP/albumin ≥ 32 at triage was independently linked to sepsis diagnosis [16, 17]. In our cohort, higher CALLY scores were strongly associated with increased 30-day mortality, and model discrimination improved markedly with its inclusion. This aligns with findings from Zhang et al., who showed that CALLY predicted mortality and acute kidney injury in ICU sepsis patients, and Ding et al., who observed that lower CALLY quartiles were associated with all-cause mortality in COPD [14, 18]. Similarly, Ji et al. found the index predicted both short- and long-term cardiovascular events in STEMI patients, with an AUROC exceeding 0.81 [19]. These findings suggest the CALLY index is emerging as a reliable, generalizable prognostic tool with strong performance across diverse acute care populations. Its simplicity, low cost, and derivation from routine labs support its potential role in guiding early decision-making in ED settings.

The predictive strength of the CALLY index likely stems from its integration of three key physiological domains central to sepsis: inflammation, nutrition, and immune function. In our cohort, CRP was significantly elevated in non-survivors, consistent with its role as a marker of systemic inflammation. Albumin, which declines in response to inflammation, capillary leakage, and catabolism, was markedly lower among non-survivors, reflecting physiological stress and nutritional compromise. Lymphopenia, another hallmark of poor prognosis in sepsis, was also more pronounced in the deceased group, indicating immune exhaustion and dysfunction. Recent literature supports the relevance of each component: CRP and albumin together have been associated with persistent inflammation and catabolism; lymphocyte depletion has been linked to impaired T cell activity and sepsis-induced immunosuppression [20, 21]. By consolidating these markers into a single score, the CALLY index captures the multidimensional pathophysiology of sepsis more effectively than isolated parameters. This integrated signal likely underpins its strong prognostic performance in our model and reinforces its potential utility in early clinical risk assessment.

The diagnostic performance of the CALLY index in our cohort suggests its potential clinical utility in emergency settings. With an AUROC of 0.906, the index showed strong discriminative ability for predicting 30-day mortality. At the cutoff value of 31.32—corresponding to the maximum Youden Index—the index achieved both high sensitivity and specificity, and its negative predictive value of 0.949 may support early reassurance when assessing lower-risk patients. In emergency departments, such a profile could contribute to early risk stratification and prioritization of higher-risk patients once laboratory data are available. A very recent study by Sarıdaş and Çetinkaya, which also evaluated the CALLY index in an ED sepsis population using a machine learning approach, reported a notably high AUROC of 0.995 for their best-performing model [22]. While that study used advanced modeling techniques, our findings complement theirs by demonstrating strong standalone performance of the CALLY index in a more conventional statistical framework, potentially enhancing its interpretability and usability in routine practice. Previous studies have reported more modest AUROC values for standard scoring systems in ICU populations. For instance, SOFA and APACHE II scores were reported to yield AUROCs of 0.688 and 0.689, respectively, while combining SOFA with albumin improved the AUROC to 0.881 [23, 24]. More complex approaches, such as tracking SOFA trajectories or using transcriptomic classifiers like the 29-host mRNA Inflammatix-Severity-3b (IMX-SEV-3b), have also been explored but require serial data or advanced infrastructure [25, 26]. In contrast, the CALLY index is derived from a single timepoint using three routine laboratory parameters, offering balanced accuracy of 0.824 and overall accuracy of 0.852 in our cohort, making it a potentially useful tool in frontline care.

Recent evidence has increasingly supported the use of composite indices integrating inflammation, nutritional status, and immune function as prognostic markers in sepsis. For instance, Zhang et al. demonstrated in a multicenter cohort of 1,123 septic patients that a higher CALLY index was independently associated with lower 30‑ and 60‑day mortality and reduced acute kidney injury risk (adjusted HR ≈ 0.97) [14]. Similarly, our findings align with Lin et al. (2025), who showed that a combined SOFA-plus-CALLY model outperformed standard organ dysfunction scores in predicting septic shock outcomes [27]. Beyond CALLY, other composite biomarkers like the albumin-to-neutrophil–lymphocyte ratio (ANLR) have shown robust predictive utility—with superior performance compared to individual markers —and the HALP index (hemoglobin, albumin, lymphocyte, platelet) has also been linked to lower mortality in sepsis when combined with SOFA [28, 29]. In contrast, Wu et al. explored the relationship between CALLY and peripheral artery disease in a non-sepsis population, finding a significant negative correlation, which further supports the index’s broader relevance in systemic inflammatory and nutritional states [30]. Taken together, these studies reinforce that the parsimonious yet multidimensional nature of the CALLY index offers both clinical feasibility and strong prognostic strength in sepsis settings.

A notable strength of this study is its use of a rigorous modeling strategy designed to enhance generalizability and minimize overfitting. We employed LASSO regression with 10-fold cross-validation, a technique well-suited for selecting the most predictive variables from high-dimensional datasets [31]. This approach reduced our model to just three variables—coronary artery disease, dementia, and the CALLY index—yielding a parsimonious and interpretable predictive tool. Calibration was assessed both statistically and visually, with the Hosmer–Lemeshow test and calibration plots demonstrating strong agreement between predicted and observed outcomes. To evaluate the added value of the CALLY index, permutation testing showed a sharp decline in model performance when CALLY values were randomized, and a likelihood ratio test confirmed significant model improvement when the index was included. These findings collectively support the robustness and clinical relevance of the CALLY index as an early mortality risk indicator in sepsis. Its ease of calculation, reliance on widely available laboratory parameters, and strong performance in our model suggest it may be a practical adjunct for early prognostication in ED settings, warranting further validation in prospective studies.

This study has several limitations that should be acknowledged. First, its retrospective single-center design may limit generalizability, as patient characteristics, sepsis management protocols, and laboratory practices may vary across institutions and regions. While our cohort was large and diverse, external validation in multi-center or international datasets is required to confirm the broader applicability of the CALLY index. Second, although we used rigorous statistical techniques—including LASSO regression with cross-validation—to mitigate overfitting, residual confounding cannot be fully excluded. Third, the final model included only three predictors (CALLY index, CAD, and dementia). This parsimony improves clarity and bedside applicability, but future research may also explore the performance of the index when compared with broader multivariable models. Fourth, the model relies on baseline laboratory values obtained at ED admission, without incorporating dynamic changes over time. Although this enhances real-time usability, it may overlook prognostically relevant trends from serial measurements. In addition, functional immune markers were not included, which may have provided deeper mechanistic insights into sepsis pathophysiology. Moreover, the mathematical formulation of the CALLY index assumes equal proportional contributions of its components, which may not fully capture the nonlinear interactions in sepsis; alternative formulations or individualized weighting schemes should be evaluated in future research. Finally, as a retrospective study, we could not capture laboratory turnaround times, and we only assessed 30-day mortality, without long-term outcomes such as functional recovery or rehospitalization, which are increasingly recognized as clinically meaningful. Future research should evaluate the CALLY index in longitudinal cohorts and compare its performance against emerging risk stratification tools using decision curve and cost-effectiveness analyses.

This study demonstrates that the CALLY index is independently associated with 30-day mortality among patients with sepsis presenting to the ED. Its strong discriminative performance, practicality, and reliance on routine laboratory parameters support its potential utility for early mortality risk assessment in acute care settings. Incorporation of the CALLY index into a multivariable model significantly improved predictive performance, and the index alone showed excellent diagnostic accuracy. Given its ease of calculation and reliance on commonly ordered laboratory tests, the CALLY index may aid in identifying high-risk patients and informing clinical decisions during early assessment.

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

None.

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Authors and Affiliations

1. Department of Emergency Medicine, Kartal Dr. Lütfi Kırdar City Hospital, Şemsi Denizer Cad. E-5 Karayolu Cevizli Mevkii, Kartal/Istanbul, 34890, Türkiye

   Erdal Yılmaz & Rohat Ak

Contributions

E.Y.: Conceptualization, Methodology, Investigation, Data Curation, Visualization, Writing – Original Draft, Supervision.R.A.: Resources, Formal Analysis, Writing – Review & Editing, Validation.

Corresponding author

Correspondence to Rohat Ak.

Ethics approval and consent to participate

This study was approved by the local ethics committee (Ethics Committee Ruling number: 2024/010.99/11/33). Due to the retrospective nature of the study, the ethics committee waived the requirement for informed consent.

Consent for publication

Not applicable.

Competing interests

Dr. Rohat Ak is a member of the BMC Emergency Medicine editorial board. He was not involved in the journal’s review or decision-making process for this manuscript. The other authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions