Skip to main content
NCBI home page
PLOS One logoLink to PLOS One
. 2026 Jul 30;21(7):e0353124. doi: 10.1371/journal.pone.0353124

Development and internal validation of a prediction model incorporating SOFA, Prognostic Nutritional Index, and Neutrophil-to-Albumin Ratio for early prediction of in-hospital mortality in patients with Sepsis: A single-center retrospective study

Lutian Yi 1,*, Tao Yu 1,2, Xiaolu Zheng 1
Editor: Benjamin M Liu3
PMCID: PMC13422871  PMID: 42531212

Abstract

Objective

Sepsis prognosis is influenced by organ dysfunction, inflammation, and nutritional status. We aimed to develop and internally validate a prediction model incorporating the Sequential Organ Failure Assessment (SOFA) score, prognostic nutritional index (PNI), and neutrophil-to-albumin ratio (NAR) for early risk stratification of in-hospital mortality in patients with sepsis.

Methods

We conducted a single-center retrospective study of 120 patients with sepsis. A multivariable logistic regression model was developed and internally validated using bootstrap resampling. Model performance was evaluated through discrimination, calibration, and decision curve analysis, with incremental predictive value assessed by net reclassification improvement and integrated discrimination improvement.

Results

The SOFA score and NAR were independently associated with in-hospital mortality, while PNI did not reach statistical significance in multivariable analysis. The combined model demonstrated good discrimination with an area under the receiver operating characteristic curve of 0.904, higher than SOFA score alone (0.845) and quick SOFA (0.700), although the difference versus SOFA alone was not statistically significant (P = 0.075). Bootstrap internal validation yielded an optimism-corrected AUC of 0.870. Calibration was acceptable (Hosmer-Lemeshow P = 0.057; Brier score = 0.098). Decision curve analysis demonstrated greater net clinical benefit compared with SOFA score alone. The combined model showed potential improvement in risk reclassification (net reclassification improvement = 0.90; integrated discrimination improvement = 0.15; both P < 0.001), although its improvement in discrimination over the SOFA score alone did not reach statistical significance.

Conclusions

The combined model showed statistically significant improvement in risk reclassification (net reclassification improvement = 0.90; integrated discrimination improvement = 0.15; both P < 0.001), although its improvement in overall discrimination (AUC) over the SOFA score alone did not reach statistical significance (P = 0.075).

Introduction

Sepsis, defined as life-threatening organ dysfunction caused by a dysregulated host response to infection, remains a leading cause of morbidity and mortality in intensive care units (ICUs), particularly respiratory ICUs where pneumonia frequently precipitates both sepsis and respiratory failure [1,2]. Despite substantial advances in critical care management, outcomes remain poor due to the considerable clinical heterogeneity of sepsis, which impedes early risk stratification and personalized therapeutic interventions [3]. According to a 2017 global burden of disease study, sepsis accounted for approximately 48.9 million incident cases and 11.0 million deaths worldwide, representing nearly 20% of all global mortality [4]. Consequently, accurate early identification of high-risk patients upon hospital admission is essential for timely clinical escalation and optimal ICU resource allocation.

Although blood cultures remain the reference standard for identifying causative pathogens in sepsis, they suffer from long turnaround time, limited sensitivity, and difficulty in cultivating fastidious organisms [5,6]. Consequently, culture-independent biomarkers and clinical scoring systems have gained increasing attention for early risk stratification. Conventional markers including C-reactive protein (CRP) and procalcitonin (PCT) facilitate diagnosis but demonstrate limited prognostic utility [7]. Emerging evidence suggests that sepsis prognosis is determined not solely by the inciting pathogen but by the interplay between systemic inflammation and the patient’s physiological reserve. Hypoalbuminemia, prevalent in septic patients, correlates with adverse outcomes and reflects both malnutrition and the systemic inflammatory response [8].

These observations have prompted interest in composite biomarkers that capture multiple pathophysiological dimensions of sepsis. The prognostic nutritional index (PNI)—derived from serum albumin and absolute lymphocyte count—integrates nutritional status and immune competence [9,10]. Since albumin and lymphocyte measurements are routinely obtained at admission, PNI can be rapidly calculated without additional laboratory testing, making it practical for bedside risk assessment. The neutrophil-to-albumin ratio (NAR) similarly combines innate immune activation with nutritional reserve, offering a simple indicator of inflammatory burden relative to nutritional status [11,12]. The CRP/albumin ratio (CAR) has likewise demonstrated prognostic value in inflammatory and malignant conditions [13–15].

However, these inflammatory-nutritional markers do not directly quantify organ dysfunction, which constitutes a cardinal feature of the Sepsis-3 definition [1] and a principal determinant of prognosis. The Sequential Organ Failure Assessment (SOFA) score provides standardized quantification of organ dysfunction severity [1], and its prognostic utility in sepsis has been substantiated [16,17]. The quick SOFA (qSOFA), a simplified bedside version, remains widely adopted for early sepsis screening, although its predictive accuracy for in-hospital mortality varies considerably across cohorts. Recent reviews further support the clinical value of SOFA and related prognostic approaches in critical care [18].

A critical knowledge gap remains regarding whether integrated nutritional-inflammatory markers improve risk stratification beyond SOFA alone. Prior studies have characterized CAR and NAR as independent prognostic tools in sepsis [11–15], yet evidence supporting their combined use with SOFA remains limited. This issue is clinically relevant because patients with equivalent organ dysfunction severity frequently experience different outcomes, suggesting that nutritional reserve and inflammatory-immune balance may provide additional prognostic information. We therefore conducted a single-center retrospective cohort study to develop and internally validate a multivariable prediction model integrating SOFA, PNI, and NAR for early prediction of in-hospital mortality in patients with sepsis, and to assess its incremental prognostic value over SOFA alone.

Materials and Methods

Study design and setting

This single-center retrospective cohort study was conducted at a university-affiliated tertiary hospital. Patients were recruited from the intensive care unit (ICU), the respiratory intensive care unit (RICU), and the department of respiratory medicine. Consecutive adult patients admitted between January 2021 and December 2024 with a diagnosis of sepsis were screened for eligibility. The electronic medical records were accessed for research purposes on 25/3/2025.

Participants

Patients were eligible if they met all of the following criteria: (1) age ≥ 18 years; (2) diagnosis of sepsis according to the Sepsis-3 clinical guidelines [1]; and (3) complete and retrievable electronic medical records. The diagnosis of sepsis was confirmed by experienced clinicians based on documented infection and sepsis-related organ dysfunction in accordance with Sepsis-3 criteria.

Patients were excluded if they met any of the following criteria: (1) pre-existing severe cardiovascular disease, including acute myocardial infarction or New York Heart Association class IV heart failure; or (2) concurrent immunosuppressive therapy or conditions likely to substantially confound outcomes, such as active malignancy, hematologic disorders, or human immunodeficiency virus infection.

Ethical approval

This study was approved by the Institutional Ethics Committee of the hospital (Approval No. 2024 Research 029) and represents a secondary analysis of the approved retrospective cohort. Because of the retrospective observational design, the requirement for written informed consent was waived. All patient identifiers were removed before analysis, and the study was conducted in accordance with the Declaration of Helsinki.

Data collection

Trained investigators extracted data from the electronic medical record system using a standardized data collection instrument. Baseline demographic characteristics, vital signs, infection sources, and comorbidities were systematically recorded. Laboratory parameters, including complete blood count with differential, serum albumin, C-reactive protein (CRP), procalcitonin, brain natriuretic peptide, and lactate, were measured within 24 h of admission, with preference given, when available, to values obtained before initiation of antimicrobial or vasopressor therapy.

The SOFA score was calculated using the worst values recorded during the initial 24 h after admission. The prognostic nutritional index (PNI) was calculated as serum albumin (g/L) + 5 × absolute lymphocyte count (×10⁹/L). The neutrophil-to-albumin ratio (NAR) was calculated as absolute neutrophil count (×10⁹/L)/ serum albumin (g/L). The CRP-to-albumin ratio (CAR) was calculated as serum CRP (mg/L)/ serum albumin (g/L). Complete clinical and laboratory data were available for all participants; therefore, no imputation was required for variables included in the final model.

Outcome definition

The primary endpoint was in-hospital mortality, defined as death from any cause during the index hospitalization. Patients who died during hospitalization were classified as non-survivors, whereas those discharged alive with documented clinical improvement were classified as survivors. Patients transferred to other hospitals or discharged against medical advice were excluded from the analysis.

Model development

Variable selection was based on clinical relevance and mechanistic plausibility. The SOFA score was retained as the reference measure of organ dysfunction severity. PNI and NAR were prespecified predictors because they reflect nutritional-immune homeostasis and inflammatory-nutritional balance, respectively, which are relevant to sepsis pathobiology but not directly captured by SOFA. Other markers, such as CAR, were evaluated descriptively and in univariate analyses but were not incorporated into the primary model.

Given the limited number of outcome events, a parsimonious multivariable logistic regression model was developed. Multicollinearity was assessed using the variance inflation factor (VIF); all values were <5, indicating no significant collinearity. Using the β coefficients from the multivariable logistic regression model (Table 2), the final prediction equation was constructed as follows: logit(p) = −1.842 + 0.581 × SOFA − 0.097 × PNI + 3.348 × NAR, where p denotes the predicted probability of in-hospital mortality. Predicted probability was calculated as p = 1/ (1 + e^ − logit(p)). Univariate logistic regression results for all candidate variables are presented in Supplementary Table 1.

Model performance and internal validation

Discriminatory performance was assessed using the area under the receiver operating characteristic curve (AUC), with pairwise comparisons performed using the DeLong method. The combined model was compared with SOFA and quick Sequential Organ Failure Assessment (qSOFA) scores. Internal validation was performed using bootstrap resampling with 1,000 iterations. In each iteration, the model was refitted in a bootstrap sample and evaluated in the original cohort. Optimism was estimated as the difference between the mean bootstrap performance and the apparent performance, and an optimism-corrected AUC was then derived. Calibration was assessed using calibration plots, the Hosmer-Lemeshow goodness-of-fit test, and the Brier score. Decision curve analysis was performed to evaluate the clinical utility of the combined model relative to SOFA. Incremental predictive value was assessed using continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI). The 95% confidence intervals for NRI and IDI were not calculated; the reported P values were derived from 1,000 bootstrap resamples. The study was reported in accordance with the TRIPOD reporting guidelines [19].

Statistical analysis

Statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA) and R version 4.5.0 (R Foundation for Statistical Computing, Vienna, Austria). The following R packages were used: pROC for receiver operating characteristic analysis, rms for logistic regression modeling and bootstrap internal validation, rmda for decision curve analysis, and boot for bootstrap resampling. Continuous variables are expressed as mean ± standard deviation and compared using the independent samples t-test. Categorical variables are presented as counts and proportions and compared using the chi-square test. Optimal diagnostic thresholds were identified by maximizing the Youden index. A two-tailed P value <0.05 was considered statistically significant.

Results

Patient characteristics

The study cohort comprised 120 patients, with a mean age of 73.1 ± 12.7 years, including 55 men and 65 women. Stratification by in-hospital mortality yielded 80 survivors and 40 non-survivors. Baseline characteristics, including age, sex, vital signs, comorbidities, and infection sites, were generally comparable between groups; however, certain individual comorbidities and primary infection sources differed significantly (Table 1). Non-survivors had a higher admission respiratory rate and longer hospital stay than survivors.

Table 1. Comparison of clinical data between sepsis patients with good and poor in-hospital mortality.

Items Poor prognosis groups

(n = 40)		 Good prognosis groups

(n = 80)		 t or X2 P
Patient characteristics
Age (years) 74.20 ± 12.37 71.91 ± 13.64 0.923 0.359
Gender 0.193 0.661
Male [n(%)] 21 (52.5%) 44 (55.0%)
Female [n(%)] 19 (47.5%) 36 (45.0%)
Hospitalization days (day) 16.08 ± 8.89 11.68 ± 5.65 2.854 0.006
Vital signs
SBP (mmHg) 106.85 ± 18.02 110.45 ± 16.27 −1.065 0.291
DBP (mmHg) 62.48 ± 10.68 65.84 ± 12.05 −1.556 0.123
MAP (mmHg) 77.26 ± 12.54 80.71 ± 12.15 −1.435 0.155
HR (bpm) 101.05 ± 8.63 99.15 ± 11.06 1.031 0.305
RR (bpm) 23.02 ± 3.35 21.65 ± 3.13 2.167 0.034
BT (°C) 37.38 ± 0.53 37.56 ± 0.53 −1.798 0.076
Laboratory data
CRP (mg/L) 200.91 ± 100.71 160.12 ± 76.41 2.257 0.028
PCT (ng/mL) 32.93 ± 36.26 14.20 ± 24.58 2.946 0.005
Neutrophil count (×10⁹/L) 16.74 ± 7.19 11.62 ± 6.48 3.800 <0.001
BNP (pg/mL) 1346.72 ± 1466.48 344.78 ± 313.42 4.273 <0.001
Lac (mmol/L) 6.99 ± 5.20 2.38 ± 1.52 6.019 <0.001
SOFA Score 5.62 ± 3.33 2.51 ± 1.58 5.601 <0.001
qSOFA 1.35 ± 0.86 0.85 ± 0.70 3.182 0.002
CAR 7.83 ± 4.71 5.46 ± 2.84 2.915 0.005
NAR 0.65 ± 0.31 0.39 ± 0.23 4.503 <0.001
PNI 29.50 ± 5.68 34.30 ± 5.21 −4.486 <0.001
Lymphocyte count (×10⁹/L) 0.45 ± 0.35 0.85 ± 0.47 −5.213 <0.001
Albumin (g/L) 27.24 ± 5.44 30.05 ± 4.27 −2.857 0.006
Underlying disease
Hypertension [n(%)] 24 (60.0%) 42 (52.5%) — 0.560
Diabetes [n(%)] 11 (27.5%) 27 (33.8%) — 0.538
Cardio-cerebrovascular diseases [n(%)] 8 (20.0%) 14 (17.5%) — 0.804
Null (no underlying disease) [n(%)] 5 (12.5%) 13 (16.3%) — 0.787
Infection site
Respiratory system [n(%)] 20 (50.0%) 50 (62.5%) — 0.239
Digestive system [n(%)] 3 (7.5%) 10 (12.5%) — 0.540
Urinary system [n(%)] 3 (7.5%) 9 (11.3%) — 0.749
Bloodstream infection [n(%)] 4 (10.0%) 7 (8.8%) — 1.000
Soft tissue [n(%)] 6 (15.0%) 8 (10.0%) — 0.547
Open in a new tab

Continuous variables are presented as mean ± SD and were compared using the independent-samples t-test. Categorical variables were compared using Fisher’s exact test when any expected cell count was < 5; otherwise, the chi-square test was applied. P values are two-tailed. A dash (—) in the test statistic column indicates that Fisher’s exact test was used, which does not produce a traditional test statistic.

Laboratory findings and derived indices

Survivors and non-survivors differed significantly in inflammatory and metabolic parameters (Table 1). The non-survivor group had higher procalcitonin, CRP, neutrophil count, brain natriuretic peptide, and lactate levels, with lower lymphocyte count and serum albumin levels (all P < 0.01).

Regarding prognostic indices, non-survivors had a lower prognostic nutritional index (29.50 ± 5.68 vs. 34.30 ± 5.21, P < 0.001) and higher neutrophil-to-albumin ratio (0.65 ± 0.31 vs. 0.39 ± 0.23, P < 0.001) and CRP-to-albumin ratio (7.83 ± 4.71 vs. 5.46 ± 2.84, P < 0.001). Severity scores were also higher in non-survivors, with SOFA scores of 5.62 ± 3.33 versus 2.51 ± 1.58 (P < 0.001) and qSOFA scores of 1.35 ± 0.86 versus 0.85 ± 0.70 (P = 0.002).

Independent predictors of in-hospital mortality

In multivariable logistic regression analysis, SOFA score (OR = 1.79 per point, 95% CI: 1.36–2.35, P < 0.001) and NAR (OR = 28.44 per unit, 95% CI: 3.03–267.17, P = 0.003) were independently associated with in-hospital mortality. PNI showed an inverse association with mortality but did not reach statistical significance (OR = 0.91 per unit, 95% CI: 0.80–1.03, P = 0.122). CAR was not retained in the final model after multivariable selection (Table 2).

Table 2. Multivariable logistic regression analysis for independent predictors of in-hospital mortality in patients with sepsis.

Variable β coefficient SE Wald χ² P Value Adjusted OR 95% CI for OR
SOFA score 0.581 0.139 17.405 < 0.001 1.787 1.361–2.347
PNI −0.097 0.062 2.393 0.122 0.908 0.803–1.026
NAR 3.348 1.143 8.584 0.003 28.444 3.028–267.166
Constant −1.842 2.377 0.601 0.438 — —
Open in a new tab

Abbreviations: SOFA, Sequential Organ Failure Assessment; PNI, prognostic nutritional index; NAR, neutrophil-to-albumin ratio; SE, standard error; OR, odds ratio; CI, confidence interval.

Model Discrimination

Receiver operating characteristic curve analysis was used to evaluate the discriminative performance of the combined model and its individual components (Fig 1, Table 3). The combined model (SOFA + PNI + NAR) achieved an area under the curve (AUC) of 0.904 (95% CI: 0.833–0.975). It showed better discrimination than qSOFA (AUC = 0.700; DeLong P < 0.001) and a numerically higher AUC than SOFA alone (AUC = 0.845; DeLong P = 0.075), although the latter difference was not statistically significant. Among the individual predictors, SOFA had the highest AUC (0.845), followed by PNI (0.778), NAR (0.762), and CAR (0.675). The optimal cutoff value for the combined model was 0.414, yielding a sensitivity of 83.3% and a specificity of 90.5%.

Fig 1. Receiver operating characteristic curves of the combined model (SOFA + PNI + NAR) and individual markers for predicting in-hospital mortality in patients with sepsis.

Fig 1

Open in a new tab

The combined model showed greater discrimination than the individual markers. AUC, area under the curve; SOFA, Sequential Organ Failure Assessment; PNI, prognostic nutritional index; NAR, neutrophil-to-albumin ratio; CAR, C-reactive protein-to-albumin ratio. The numerical difference in AUC between the combined model and SOFA alone did not reach statistical significance (P = 0.075).

Table 3. Discriminatory performance of individual indicators and the combined model for predicting in-hospital mortality in patients with sepsis.

Indicator AUC 95%CI Cut-off

value		 Sensitivity

(%)		 Specificity

(%)		 Youden index
SOFA 0.845 0.757–0.933 ≥3.5 77.8 82.1 0.599
PNI 0.778 0.675–0.881 <29.63 72.2 84.5 0.567
NAR 0.762 0.667–0.857 ≥0.502 63.9 77.4 0.413
CAR 0.675 0.562–0.788 ≥6.506 58.3 78.6 0.369
qSOFA 0.700 0.604–0.795 ≥2 47.2 82.1 0.294
Combined model 0.904 0.833–0.975 ≥0.414 83.3 90.5 0.738
Open in a new tab

Abbreviations: AUC, area under the receiver operating characteristic curve; CI, confidence interval; SOFA, Sequential Organ Failure Assessment; PNI, prognostic nutritional index; NAR, neutrophil-to-albumin ratio; CAR, C-reactive protein-to-albumin ratio; qSOFA, quick Sequential Organ Failure Assessment.

Notes: The cut-off value for the combined model refers to the optimal predicted probability threshold. The cut-off for qSOFA was set at ≥2 according to its original definition. These cut-off values were derived from the present sample and should be considered exploratory rather than definitive; they are not intended for clinical decision-making without external validation.

Internal validation

Bootstrap internal validation using 1,000 resamples yielded an optimism-corrected AUC of 0.870, compared with an apparent AUC of 0.904 (optimism = 0.034; Table 3), indicating acceptable internal stability.

Calibration

Calibration was evaluated using the Hosmer-Lemeshow goodness-of-fit test, bootstrap-corrected calibration metrics, calibration curves, and the Brier score. Bootstrap internal validation with 1,000 resamples yielded an optimism-corrected calibration slope of 0.92 and intercept of −0.04, indicating close agreement between predicted and observed mortality probabilities. The Hosmer-Lemeshow test showed no significant lack of fit (P = 0.057), the calibration curve demonstrated consistent agreement between predicted and observed outcomes across the entire range of risk (Fig 2), and the Brier score of 0.098 further confirmed favorable predictive accuracy.

Fig 2. Calibration curve of the combined model (SOFA + PNI + NAR) for predicting in-hospital mortality in patients with sepsis.

Fig 2

Open in a new tab

The plot was generated using 1,000 bootstrap resamples to assess agreement between predicted probabilities and observed outcomes. The 45° diagonal line indicates the ideal reference. The model curve showed reasonable agreement with the reference line across the range of predicted risk.

Clinical utility

Decision curve analysis showed that the combined model provided greater net benefit than SOFA alone across a clinically relevant range of threshold probabilities for in-hospital mortality (Fig 3).

Fig 3. Decision curve analysis of the combined model (SOFA + PNI + NAR) versus SOFA alone for predicting in-hospital mortality in patients with sepsis.

Fig 3

Open in a new tab

The y-axis represents net benefit, and the x-axis represents threshold probability. The combined model (red line) showed a higher net benefit than SOFA alone (blue line) across a clinically relevant range of threshold probabilities.

Incremental predictive value

Compared with SOFA alone, the combined model showed improved risk reclassification. The continuous net reclassification improvement (NRI) was 0.90 (P < 0.001), and the integrated discrimination improvement (IDI) was 0.15 (P < 0.001).

Survival analysis

Using ROC-derived optimal cut-off values (SOFA ≥ 3.5, NAR ≥ 0.502, and PNI < 29.63), Kaplan-Meier analysis demonstrated significant stratification of in-hospital mortality risk. Patients with SOFA ≥ 3.5 had lower survival probability than those with SOFA < 3.5 (log-rank P < 0.001). Patients with PNI < 29.63 also had lower in-hospital survival than those with PNI ≥ 29.63 (log-rank P = 0.024). The survival difference between patients with NAR ≥ 0.502 and those with NAR < 0.502 was not statistically significant (log-rank P = 0.068) (Fig 4).

Fig 4. Kaplan–Meier curves of in-hospital survival in patients with sepsis, stratified by the optimal cut-off values of SOFA (≥ 3.5 vs. < 3.5), NAR (≥ 0.502 vs. < 0.502), and PNI (< 29.63 vs. ≥ 29.63).

Fig 4

Open in a new tab

P values were calculated using the log-rank test.

Discussion

This single-center retrospective study developed and internally validated a combined model integrating SOFA, PNI, and NAR to predict in-hospital mortality in septic patients. The model demonstrated acceptable discrimination and calibration with meaningful net benefit on decision curve analysis. Relative to SOFA alone, incorporation of PNI and NAR improved risk reclassification (NRI = 0.90, IDI = 0.15), suggesting that nutritional and inflammatory biomarkers contribute information beyond organ dysfunction scores.

The three components address distinct pathophysiological dimensions. SOFA quantifies organ dysfunction—a central determinant of sepsis prognosis. PNI integrates albumin and lymphocyte count, reflecting both nutritional reserve and adaptive immune capacity. NAR captures the ratio of systemic inflammation to nutritional status. By design, no single marker fully represents this complexity; their combination allowed the model to capture multifactorial risk that SOFA alone cannot.

SOFA is the reference standard for sepsis severity and consistently predicts mortality across populations [16,17]. In our cohort, SOFA was the single strongest predictor. Yet SOFA inherently lags: it reflects organ dysfunction that has already manifested clinically. The model’s additional components—PNI and NAR—operate on a different timescale, capturing nutritional depletion and inflammatory activation that may precede overt organ failure. Earlier detection of these processes could allow more timely intervention.

Prior work links PNI to sepsis outcomes through its dual reflection of nutritional and immune status [20–23]. Elevated NAR similarly associates with mortality [11,24]. In our multivariable analysis, PNI did not independently reach significance, yet its inclusion improved overall reclassification. NAR remained an independent predictor, though its wide confidence interval counsels caution. In a sensitivity analysis comparing the full model with a simplified two-variable model (SOFA + NAR), the full model showed only a minimal numerical increase in AUC (0.904 vs. 0.896), and neither NRI nor IDI indicated a statistically significant improvement (NRI = 0.17, P = 0.362; IDI = 0.02, P = 0.11). These results suggest that the incremental contribution of PNI beyond SOFA and NAR was limited in this cohort. Nevertheless, PNI was retained as a prespecified predictor because of its biological and clinical plausibility. The lack of statistical significance for NAR in univariable Kaplan–Meier analysis likely reflects loss of signal from dichotomization and small event count. CAR was excluded from the final model, a finding consistent with substantial overlap of its information with the three retained variables [13–15].

A key observation was that the combined model did not significantly exceed SOFA’s discrimination (DeLong P = 0.075). However, the NRI and IDI results—both statistically significant—indicate that risk reclassification occurred despite the marginal AUC difference. This dissociation reflects a real phenomenon: improvements in individual risk stratification need not always translate to large increments in overall discrimination. Conversely, when compared against qSOFA (a simpler bedside tool), our model performed substantially better (AUC 0.904 vs 0.700, P < 0.001), establishing that the complexity added meaningful discrimination relative to simplified screening scores. The plateau relative to SOFA likely reflects sample size constraints, the modest number of deaths, and potential redundancy among organ dysfunction, nutritional, and inflammatory measures in predicting this outcome.

We selected PNI over more complex nutritional indices such as mNUTRIC for practical reasons: it requires only baseline laboratory values routinely available at admission. Similarly, NAR demands no specialized assays. This design choice prioritized bedside applicability without sacrificing information gain.

Decision curve analysis reinforced the discriminative and reclassification findings: the combined model accrued net clinical benefit across the range of reasonable decision thresholds. Kaplan–Meier stratification was significant for SOFA and PNI independently, but not for NAR alone—an observation attributable to power limitations and the continuous nature of NAR in its native form. Collectively, these results position the combined model as a potential tool for earlier and more granular risk assessment in sepsis. The model may be applied at hospital admission to identify patients at elevated risk of in-hospital death who may warrant closer monitoring, although its role in guiding specific management decisions requires prospective validation. For bedside nurses in respiratory and critical care settings, this model may assist in early identification of high-risk septic patients, facilitating more intensive monitoring and timely communication with the medical team. Replication in external cohorts remains essential.

Several limitations warrant acknowledgment. First, the single-center design with modest sample size and outcome frequency constrains generalizability. With 40 in-hospital deaths and three prespecified predictors, the model development was broadly consistent with conventional EPV recommendations, although a larger number of events would still be desirable for more precise estimation. Internal validation mitigated but did not eliminate optimism bias. Second, external validation in independent cohorts is a mandatory next step. Third, the marginal statistical advantage over SOFA demands confirmation in larger samples before claims of superiority can be sustained. Fourth, a small subset of patients presented with SOFA scores of 0–1, reflecting real-world diagnostic heterogeneity at hospital entry. The exclusion criteria, while intended to reduce confounding, may limit the applicability of the model to sepsis patients with major comorbidities, who represent a substantial proportion of the real-world sepsis population. Fifth, comprehensive data for APACHE II or SAPS II were unavailable, precluding comparison with these alternative severity instruments. Sixth, the model was trained on baseline values only; dynamic biomarker trajectories may hold additional prognostic signal. PNI was retained in the model based on clinical plausibility and prior evidence linking nutritional status to sepsis outcomes, rather than on statistical significance alone. However, its independent association with in-hospital mortality was not confirmed in this cohort (P = 0.122). The wide confidence interval for NAR likely reflects the limited number of events, so this effect estimate should be interpreted cautiously. Finally, these findings remain exploratory and should not guide clinical decisions without prospective validation in multicenter cohorts.

Conclusions

We evaluated SOFA, PNI, and NAR for predicting in-hospital mortality in septic patients. A combined model incorporating these three markers achieved acceptable discrimination (AUC 0.904) and demonstrated meaningful risk reclassification over SOFA alone (NRI = 0.90, IDI = 0.15). However, the gain in discrimination over SOFA alone was not statistically significant; nonetheless, the significant NRI and IDI indicate that PNI and NAR capture clinically relevant reclassification. Since all three variables are extractable from baseline admission data, the model should be easy to apply at admission. Single-center design and modest sample size limit generalizability; prospective multicenter validation is needed before clinical adoption.

Supporting information

S1 Dataset. Anonymized dataset for the prediction model.

The dataset includes all variables used in the analysis (age, sex, vital signs, laboratory parameters, SOFA score, PNI, NAR, CAR, and in-hospital mortality outcome). Patient identifiers have been removed to protect privacy.

(CSV)

pone.0353124.s001.csv (25.7KB, csv)

Acknowledgments

The authors extend their sincere gratitude to all the staff at the participating hospital who assisted with data collection and to the patients whose data made this study possible.

Data Availability

All relevant data are within the manuscript and its Supporting Information files. The anonymized dataset used for this analysis is provided as Supporting Information (S1 Dataset).

Funding Statement

This work was supported by the Zhejiang Provincial Medical and Health Research Program (Grant No. 2024KY369 to TY). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  • 1.Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315(8):801–10. doi: 10.1001/jama.2016.0287 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Sieswerda E, Bax HI, Hoogerwerf JJ, de Boer MGJ, Boermeester M, Bonten MJM, et al. The 2021 Dutch Working Party on Antibiotic Policy (SWAB) guidelines for empirical antibacterial therapy of sepsis in adults. BMC Infect Dis. 2022;22(1):687. doi: 10.1186/s12879-022-07653-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Lu LY, Yong Y, Song JG. New progress in clinical research of immunotherapy for sepsis. Shanghai Med J. 2019;45(9):648–52. [Google Scholar]
  • 4.Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395(10219):200–11. doi: 10.1016/S0140-6736(19)32989-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Zeng X, Lloyd KM, Hamdy RF, Shapiro CA, Fisher MA, Lin J, et al. Identification and characterization of an invasive, hyper-aerotolerant Campylobacter jejuni strain causing bacteremia in a pediatric leukemia patient. ASM Case Rep. 2025;1(3):e00060-24. doi: 10.1128/asmcr.00060-24 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Liu BM, Carlisle CP, Fisher MA, Shakir SM. The Brief Case: Capnocytophaga sputigena Bacteremia in a 94-Year-Old Male with Type 2 Diabetes Mellitus, Pancytopenia, and Bronchopneumonia. J Clin Microbiol. 2021;59(7):e0247220. doi: 10.1128/JCM.02472-20 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Chen Y, Yan A, Zhang L, Hu X, Chen L, Cui J, et al. Comparative analysis of inflammatory biomarkers for the diagnosis of neonatal sepsis: IL-6, IL-8, SAA, CRP, and PCT. Open Life Sci. 2025;20(1):20221005. doi: 10.1515/biol-2022-1005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Wang B, Sha W, Gui S, Zou Z-Y. Efficacy of albumin and loop diuretic Co-administration in reducing mortality among septic patients with hypoalbuminemia. Eur J Pharmacol. 2025;999:177692. doi: 10.1016/j.ejphar.2025.177692 [DOI] [PubMed] [Google Scholar]
  • 9.Al-Shami I, Hourani HMA, Alkhatib B. The use of prognostic nutritional index (PNI) and selected inflammatory indicators for predicting malnutrition in COVID-19 patients: A retrospective study. J Infect Public Health. 2023;16(2):280–5. doi: 10.1016/j.jiph.2022.12.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Toscano A, Bellone F, Maggio N, Cinquegrani M, Spadaro F, Bueti FM, et al. Unlocking the Predictive Power of Nutritional Scores in Septic Patients. Nutrients. 2025;17(3):545. doi: 10.3390/nu17030545 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Hwang YJ, Chung SP, Park YS, Chung HS, Lee HS, Park JW, et al. Newly designed delta neutrophil index-to-serum albumin ratio prognosis of early mortality in severe sepsis. Am J Emerg Med. 2015;33(11):1577–82. doi: 10.1016/j.ajem.2015.06.012 [DOI] [PubMed] [Google Scholar]
  • 12.Zhao TP, Chen HG. Prognostic value of the ratio of neutrophil to albumin in severe sepsis in children. Zhejiang Med J. 2019;41(13):1387–90. [Google Scholar]
  • 13.Zhao Y, Xia W, Lu Y, Chen W, Zhao Y, Zhuang Y. Predictive value of the C-reactive protein/albumin ratio in severity and prognosis of acute pancreatitis. Front Surg. 2023;9:1026604. doi: 10.3389/fsurg.2022.1026604 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Fan Z, Fan K, Gong Y, Huang Q, Yang C, Cheng H, et al. The CRP/Albumin Ratio Predicts Survival And Monitors Chemotherapeutic Effectiveness In Patients With Advanced Pancreatic Cancer. Cancer Manag Res. 2019;11:8781–8. doi: 10.2147/CMAR.S211363 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Basile-Filho A, Lago AF, Menegueti MG, Nicolini EA, Rodrigues LA de B, Nunes RS, et al. The use of APACHE II, SOFA, SAPS 3, C-reactive protein/albumin ratio, and lactate to predict mortality of surgical critically ill patients: A retrospective cohort study. Medicine (Baltimore). 2019;98(26):e16204. doi: 10.1097/MD.0000000000016204 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Ranzani OT, Singer M, Salluh JIF, Shankar-Hari M, Pilcher D, Berger-Estilita J, et al. Development and Validation of the Sequential Organ Failure Assessment (SOFA)-2 Score. JAMA. 2025;334(23):2090–103. doi: 10.1001/jama.2025.20516 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Ko BS, Ryoo SM, Han E, Chang H, Yune CJ, Lee HJ, et al. Modified Cardiovascular Sequential Organ Failure Assessment Score in Sepsis: External Validation in Intensive Care Unit Patients. J Korean Med Sci. 2023;38(50):e418. doi: 10.3346/jkms.2023.38.e418 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Yu J, Sun K, Zhou Y, Fan Y, Zhang X, Chen H, et al. Update of the sequential organ failure assessment score: current status and challenges?. Front Med (Lausanne). 2026;12:1733090. doi: 10.3389/fmed.2025.1733090 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Collins GS, Moons KGM, Dhiman P, Riley RD, Beam AL, Van Calster B, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. doi: 10.1136/bmj-2023-078378 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Denstaedt SJ, Cano J, Wang XQ, Donnelly JP, Seelye S, Prescott HC. Blood count derangements after sepsis and association with post-hospital outcomes. Front Immunol. 2023;14:1133351. doi: 10.3389/fimmu.2023.1133351 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Saavedra-Torres JS, Pinzón-Fernández MV, Nati-Castillo HA, Cadena Correa V, Lopez Molina LC, Gaitán JE, et al. Immunodynamic Disruption in Sepsis: Mechanisms and Strategies for Personalized Immunomodulation. Biomedicines. 2025;13(9):2139. doi: 10.3390/biomedicines13092139 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Xie M, Huang L, Li L, Qin Y, Feng B, Cai Q, et al. Association between prognostic nutritional index and 90-day mortality in septic patients: a retrospective cohort study. BMC Nutr. 2025;11(1):232. doi: 10.1186/s40795-025-01212-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Yang Y-L, Huang Y-S, Bai Y-L, Huang X-X, Fu Z-Y, Huang Z-W. Development and validation of a nutrition-integrated nomogram for predicting 28-day mortality in sepsis patients. Front Nutr. 2026;12:1726151. doi: 10.3389/fnut.2025.1726151 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Yoo KH, Choi SH, Suh GJ, Chung SP, Choi HS, Park YS, et al. The usefulness of lactate/albumin ratio, C-reactive protein/albumin ratio, procalcitonin/albumin ratio, SOFA, and qSOFA in predicting the prognosis of patients with sepsis. Sci Rep. 2024;14(1):11234. doi: 10.1016/j.ajem.2023.12.028 [DOI] [PubMed] [Google Scholar]

Decision Letter 0

Benjamin Liu

14 Jun 2026

-->PONE-D-26-25132-->-->Development and Internal Validation of a Prediction Model Incorporating SOFA, Prognostic Nutritional Index, and Neutrophil-to-Albumin Ratio for Early Prediction of In-Hospital Mortality in Patients with Sepsis: A Single-Center Retrospective Study-->-->PLOS One

Dear Dr. Yi,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

ACADEMIC EDITOR'S COMMENTS: -->--> -->-->The authors should add introduction that blood cultures play a central diagnostic role in detecting sepsis, particularly when bloodstream infection (bacteremia or fungemia) is suspected but blood cultures suffer from long turnaround time, low sensitivity, requirement of specific culture conditions and difficulties of growing fastidious organisms. More references should be cited, with this two (PMID: 41246370; PMID: 34142857) as examples (citing is optional).

Please submit your revised manuscript by Jul 14 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:-->

  • A letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

  • A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

  • An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

-->

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only  the individual author can complete the verification step; PLOS staff cannot  verify ORCID iDs on behalf of authors.

We look forward to receiving your revised manuscript.

Kind regards,

Benjamin M. Liu, PhD, D(ABMM), MB(ASCP)

Academic Editor

PLOS One

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please note that PLOS One has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“This work was supported by the Zhejiang Provincial Medical and Health Research Program (Grant No. 2024KY369). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“This work was supported by the Zhejiang Provincial Medical and Health Research Program (Grant No. 2024KY369).”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

4. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

5. Thank you for stating the following financial disclosure:

“This work was supported by the Zhejiang Provincial Medical and Health Research Program (Grant No. 2024KY369).”

Please state what role the funders took in the study.  If the funders had no role, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

If this statement is not correct you must amend it as needed.

Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

6. In the online submission form you indicate that your data is not available for proprietary reasons and have provided a contact point for accessing this data. Please note that your current contact point is a co-author on this manuscript. According to our Data Policy, the contact point must not be an author on the manuscript and must be an institutional contact, ideally not an individual. Please revise your data statement to a non-author institutional point of contact, such as a data access or ethics committee, and send this to us via return email. Please also include contact information for the third party organization, and please include the full citation of where the data can be found.

7. PLOS requires an ORCID iD for the corresponding author in Editorial Manager on papers submitted after December 6th, 2016. Please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field. This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager.

8. Your ethics statement should only appear in the Methods section of your manuscript. If your ethics statement is written in any section besides the Methods, please move it to the Methods section and delete it from any other section. Please ensure that your ethics statement is included in your manuscript, as the ethics statement entered into the online submission form will not be published alongside your manuscript.

9. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

10. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. -->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->2. Has the statistical analysis been performed appropriately and rigorously? -->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: The manuscript is scientifically and technically sound and well written. The statistical analysis has been adequately performed and the statistics are presented in a simple fashion. The only problem I found with the manuscript is that the authors have not adequately explained or declared as to how they came up with the formula for the logistic regression (lines 131-132). I feel like this should be explained in the final manuscript.

Reviewer #2: The technique suggested by author can be accepted as a method for sepsis validation the data set is adequate. Such studies are a benefit to critical I'll hospital patients. Very few articles available on above subject.

**********

-->6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review?  For information about this choice, including consent withdrawal, please see our Privacy Policy.-->

Reviewer #1: No

Reviewer #2: Yes: Jayant balani

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures

You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation.

NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications.

-->

PLoS One. 2026 Jul 30;21(7):e0353124. doi: 10.1371/journal.pone.0353124.r002

Author response to Decision Letter 1

16 Jun 2026

Manuscript ID: PONE-D-26-25132

Title: Development and Internal Validation of a Prediction Model Incorporating SOFA, Prognostic Nutritional Index, and Neutrophil-to-Albumin Ratio for Early Prediction of In-Hospital Mortality in Patients with Sepsis: A Single-Center Retrospective Study

Corresponding Author: Lutian Yi

Dear Dr. Benjamin M. Liu and the Reviewers,

Thank you for your thoughtful and constructive comments on our manuscript. We have carefully addressed all the points raised by the Academic Editor and the reviewers. Below we provide a point-by-point response. All changes in the revised manuscript have been highlighted using track changes.

Response to Academic Editor

Comment:

The authors should add introduction that blood cultures play a central diagnostic role in detecting sepsis, particularly when bloodstream infection (bacteremia or fungemia) is suspected but blood cultures suffer from long turnaround time, low sensitivity, requirement of specific culture conditions and difficulties of growing fastidious organisms. More references should be cited, with this two (PMID: 41246370; PMID: 34142857) as examples (citing is optional).

Response:

We thank the Editor for this valuable suggestion. We agree that the limitations of blood cultures provide important context for the use of rapid, culture-independent biomarkers and clinical scoring systems. Accordingly, we have revised the Introduction by replacing the original opening sentence of the biomarker paragraph (“Biomarkers play a key role in assessing infection severity and prognosticating sepsis outcomes.”) with the following sentences (inserted immediately after the Sepsis-3 definition and before the discussion of conventional markers such as CRP and PCT):

“Although blood cultures remain the reference standard for identifying causative pathogens in sepsis, they suffer from long turnaround time, limited sensitivity, and difficulty in cultivating fastidious organisms [5,6]. Consequently, culture-independent biomarkers and clinical scoring systems have gained increasing attention for early risk stratification. Conventional markers including C-reactive protein (CRP) and procalcitonin (PCT) facilitate diagnosis but demonstrate limited prognostic utility [7]. Emerging evidence suggests that sepsis prognosis is determined not solely by the inciting pathogen but by the interplay between systemic inflammation and the patient’s physiological reserve. Hypoalbuminemia, prevalent in septic patients, correlates with adverse outcomes and reflects both malnutrition and the systemic inflammatory response [8].”

We have incorporated the two suggested references into the revised Introduction. All subsequent reference numbers have been updated throughout the manuscript accordingly. The revised text is highlighted in red in the tracked-changes version.

Response to Reviewer #1

Comment:

The technique suggested by author can be accepted as a method for sepsis validation the data set is adequate. Such studies are a benefit to critical ill hospital patients. Very few articles available on above subject. The only problem I found with the manuscript is that the authors have not adequately explained or declared as to how they came up with the formula for the logistic regression (lines 131-132). I feel like this should be explained in the final manuscript.

Response:

We thank the reviewer for this important comment. In the original manuscript, the logistic regression equation was presented without explicitly stating how it was derived. As shown in Table 2, the β coefficients for all model terms, including the intercept, were already reported. To make the derivation transparent, we have revised the Methods section (Model development subsection) in the marked-up manuscript, where the changes are highlighted in red. The revised text reads as follows:

“Given the limited number of outcome events, a parsimonious multivariable logistic regression model was developed. Multicollinearity was assessed using the variance inflation factor (VIF); all values were <5, indicating no significant collinearity. Using the β coefficients from the multivariable logistic regression model (Table 2), the final prediction equation was constructed as follows: logit(p) = −1.842 + 0.581 × SOFA − 0.097 × PNI + 3.348 × NAR, where p denotes the predicted probability of in-hospital mortality. Predicted probability was calculated as p = 1 / (1 + e^−logit(p)). Univariate logistic regression results for all candidate variables are presented in Supplementary Table 1.”

The coefficients in the equation are identical to those reported in Table 2. We believe this revision fully addresses the reviewer’s concern.

Response to Reviewer #2

Comment:

The technique suggested by author can be accepted as a method for sepsis validation the data set is adequate. Such studies are a benefit to critical ill hospital patients. Very few articles available on above subject.

Response:

We sincerely thank the reviewer for the positive and encouraging comments. We are pleased that the reviewer found our study and dataset appropriate for this purpose and recognized the potential clinical value of the model in critically ill patients with sepsis. We agree that further external validation will be important before routine clinical application, and we hope our work may serve as a useful foundation for future studies in this area.

Response to Journal Requirements

1.PLOS ONE style requirements and file naming

We have checked the PLOS ONE style requirements and file naming conventions and confirm that the revised files have been prepared accordingly.

2. Code sharing guidelines

We have reviewed PLOS ONE’s code sharing policy. No author-generated custom code was developed for this study. All statistical analyses were performed using standard functions in SPSS version 25.0 (IBM Corp.) and R version 4.5.0 with publicly available packages (pROC, rms, rmda, boot). Therefore, code sharing does not apply to this submission.

3. Funding statement removal from Acknowledgments

We have removed all funding-related text from the manuscript. The Acknowledgments section now contains only non-financial acknowledgments and does not include any funding information. Funding information appears only in the Funding Statement in the online submission system, as required.

4. Matching grant information

We confirm that the correct grant number is 2024KY369 from the Zhejiang Provincial Medical and Health Research Program. The same grant information has been provided consistently in the submission materials.

5. Role of funder

The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This statement has been included in the cover letter as requested.

6. Data availability statement

The anonymized dataset used for this analysis has been provided as Supporting Information (S1 Dataset). No third-party contact is required.

7. ORCID iD for corresponding author

The corresponding author has created and validated an ORCID iD in Editorial Manager via the “Update my Information → Fetch/Validate” link.

8. Ethics statement location

We confirm that the ethics statement appears only in the Methods section under the subheading “Ethical Approval.” The duplicate ethics statement previously located in the Declarations section has been removed.

9. Citation of recommended works

We have cited the two references recommended by the Academic Editor (PMID: 41246370 and PMID: 34142857) in the revised Introduction (now numbered as [5] and [6]). These references are relevant to the added discussion on blood culture limitations. No other reviewer-recommended citations were requested.

10. Reference list completeness and retractions

We have carefully reviewed the reference list and made the necessary corrections to ensure completeness and accuracy. We also verified that none of the cited references have been retracted.

We believe the revised manuscript is now suitable for publication in PLOS ONE. We thank the Academic Editor and reviewers again for their time and valuable input.

Sincerely,

Lutian Yi, MD

Corresponding author

Attachment

Submitted filename: Response to Reviewers.docx

pone.0353124.s003.docx (16.8KB, docx)

Decision Letter 1

Benjamin Liu

18 Jun 2026

Development and Internal Validation of a Prediction Model Incorporating SOFA, Prognostic Nutritional Index, and Neutrophil-to-Albumin Ratio for Early Prediction of In-Hospital Mortality in Patients with Sepsis: A Single-Center Retrospective Study

PONE-D-26-25132R1

Dear Dr. Yi,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Benjamin M. Liu, PhD, D(ABMM), MB(ASCP)

Academic Editor

PLOS One

Additional Editor Comments (optional):

Reviewers' comments:

Acceptance letter

Benjamin Liu

PONE-D-26-25132R1

PLOS One

Dear Dr. Yi,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Benjamin M. Liu

Academic Editor

PLOS One

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    S1 Dataset. Anonymized dataset for the prediction model.

    The dataset includes all variables used in the analysis (age, sex, vital signs, laboratory parameters, SOFA score, PNI, NAR, CAR, and in-hospital mortality outcome). Patient identifiers have been removed to protect privacy.

    (CSV)

    pone.0353124.s001.csv (25.7KB, csv)
    Attachment

    Submitted filename: Response to Reviewers.docx

    pone.0353124.s003.docx (16.8KB, docx)

    Data Availability Statement

    All relevant data are within the manuscript and its Supporting Information files. The anonymized dataset used for this analysis is provided as Supporting Information (S1 Dataset).

    Articles from PLOS One are provided here courtesy of PLOS

    RESOURCES