Abstract
Objective
The aim of this study is to investigate risk factors for acute respiratory distress syndrome (ARDS) in severe acute brain injury (SABI) patients and construct a nomogram-based predictive model.
Methods
A retrospective analysis was conducted on 200 SABI patients admitted to Lishui Hospital between January 2021 and April 2025, who were randomly allocated into training group (n = 140) and validation group (n = 60). ARDS risk factors were identified and incorporated into a predictive model. Model performance was evaluated via Receiver operating characteristic curve (ROC curve), calibration plots, and decision curve analysis (DCA).
Results
Multivariate logistic regression revealed three independent predictors of ARDS in SABI patients: Sepsis, PaO₂/FiO₂, Pulmonary infection, (all P < 0.05). The area under the ROC curve (AUC) was 0.778 for the training set and 0.754 for the validation set. Calibration curves demonstrated good predicted-observation agreement, while DCA confirmed the clinical utility of the nomogram.
Conclusion
This study developed and validated a nomogram prediction model incorporating three variables: Sepsis, PaO₂/FiO₂ and Pulmonary infection. The model demonstrated good discriminative ability and calibration in predicting the risk of ARDS in patients with SABI, thereby facilitating early risk stratification and supporting clinical decision-making.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12883-026-04708-9.
Keywords: Severe acute brain injury, Acute respiratory distress syndrome, Nomogram, Prediction model
Introduction
Severe acute brain injury (SABI) represents a serious public health concern due to its high incidence, mortality, and frequent neurological sequelae in survivors [1–3]. Clinical and experimental evidence demonstrates that SABI, as a complex hierarchical event involving multiple pathophysiological mechanisms, extends beyond intracranial injuries to affect organs outside the central nervous system [4]. Among them, acute respiratory distress syndrome (ARDS) is a representative complication in SABI patients. Featuring acute and diffuse inflammatory responses in the lungs, it seriously threatens the lives of patients [5, 6]. Epidemiological data indicate ARDS develops in up to 40% of patients with SABI, with these patients exhibiting prolonged ICU stays, elevated mortality rates, and poorer neurological outcomes compared to those without ARDS [7, 8].
In SABI patients, the pathophysiology of brain-lung interplay involves complex mechanisms including inflammatory cascade responses, immunosuppression, and autonomic nervous system dysfunction [9, 10]. SABI can both induce and exacerbate ARDS through multiple pathways, while ARDS-related hypoxemia and systemic inflammation further aggravate secondary brain injury. Similarly, therapeutic approaches for SABI and ARDS often present conflicting priorities. SABI management may require hypertonic therapy, volume resuscitation, and blood pressure elevation, which potentially contradict ARDS-focused conservative fluid strategies [9]. Conversely, ARDS treatments such as fluid restriction, prone position ventilation, and lung-protective ventilation may inadequately support brain injury or even exacerbate neurological damage [11, 12]. Notably, a majority of randomized controlled trials (RCTs) validating ARDS treatment protocols exclude patients with neurological injuries or higher intracranial pressure (ICP), restricting the generalization of their results to these patients [13, 14] .
The complex interplay between lung and brain physiology presents significant clinical challenges in managing patients suffering from SABI and ARDS. Therefore, exploring potential risk factors for ARDS after SABI and identifying patients at higher risk for ARDS early after onset are crucial for developing optimal treatment strategies. This includes precise positive end-expiratory pressure (PEEP) titration based on lung compliance, strict MAP maintenance, and rigorous volume management to balance oxygenation and cerebral perfusion requirements. Current research trends emphasize ARDS prevention and early pharmacological intervention over treatment alone, aiming to reduce both ARDS incidence and mortality rates [15]. While risk factor-based prediction models have been widely applied in critical care [16], ARDS prediction models have yielded unsatisfactory results due to the complexity and heterogeneity of the etiology. This retrospective study aims to identify risk factors for ARDS in SABI patients, develop a corresponding prediction model, and evaluate its predictive performance, ultimately facilitating ARDS risk assessment in this patient population.
Materials and methods
Objects of study
This retrospective study included 235 patients with SABI admitted to Lishui People’s Hospital between January 2021 and April 2025. Approval was granted by the Scientific Research Ethics Committee of Lishui Municipal People’s Hospital (Approval Number: 20250603). All procedures in this study followed the ethical guidelines of the Helsinki Declaration of the World Medical Association.As a retrospective analysis utilizing de-identified medical records, patient privacy risks were minimal, and obtaining individual consent for historical cases was impractical. Therefore, the ethics committee approved an exemption from the requirement to obtain informed consent from subjects. This study strictly adheres to the principle of confidentiality, and all data processing is conducted in a secure environment.
The inclusion criteria included: (1) SABI diagnosis confirmed by medical history, clinical symptoms, signs, head imaging examination, meeting the definition of SABI (sudden-onset of traumatic or non-traumatic brain injury leading to a decline in consciousness levels, usually accompanied by other serious neurological deficits) [17]; (2) Age ≥ 18 years old.
The exclusion criteria included: (1) Severe systemic diseases (e.g., heart failure, hepatic failure, severe pneumonia); (2) Death or discharge within 48 h of admission; (3) Incomplete clinical data on admission.
A total of 35 patients were excluded (eight patients with severe systemic diseases, 10 patients who died or were discharged within 48 h of admission, and 17 patients with incomplete admission data), yielding 200 patients in the statistical analysis at last (Fig. 1).
Fig. 1.
Patient inclusion process and study schematic diagram
Collection of clinical data
(1) Baseline characteristics and medical history: Data on demographics and chronic conditions present at admission were collected, including sex, age, and pre-existing comorbidities (e.g., hypertension, heart disease [including chronic coronary artery disease, cardiomyopathy, valvular heart disease, etc.], chronic liver or kidney diseases). (2) Clinical and laboratory parameters: Vital signs and laboratory test results obtained within 24 h of admission were recorded. (3) Complications (newly developed during the current hospitalization prior to ARDS onset): Complications such as hydrocephalus, pulmonary infection, sepsis, and cardiac insufficiency were documented. All patients were randomized into the training group and the validation group, and fell into ARDS and non-ARDS groups based on whether or not they developed ARDS during hospitalization, with risk factors for ARDS being screened. Given that SABI encompasses various etiologies, patients were classified into four clinicopathological subgroups based on the principal admission diagnosis: Acute Ischemic Stroke (AIS), Subarachnoid Hemorrhage (SAH), Intracerebral Hemorrhage (ICH), and Traumatic Brain Injury (TBI).
The ARDS definition consisted of 4 components according to the Berlin criteria: (1) new onset of clinical symptoms in the lungs within one week; (2) hypoxemia (PaO2/FiO2 ≤ 300mmHg); (3) the presence of bilateral lung shadows on imaging; (4) exclusion of heart failure or fluid overload as primary explanations. ARDS severity is stratified based on hypoxemia: mild (P/F ratio 201–300 mmHg), moderate (101–200 mmHg), and severe (≤ 100 mmHg and PEEP ≥ 5 cm H2O) [17].
Statistical analysis methods
The prior sample size estimation was conducted using the G*Power 3.1.9 software, Statistical analysis was performed using SPSS 26.0 and R software (version 4.3.0). Normality of data distribution was assessed by the K-S test. Normally distributed continuous variables were presented as mean ± standard deviation (x ± s) and compared between groups using independent t-test. Non-normally distributed data were described as interquartile range and compared using the rank-sum test. Categorical variables were expressed as n (%) and analyzed with χ² or Fisher’s exact tests. Candidate variables were preliminarily screened using LASSO regression combined with 10-fold cross-validation. Subsequently, the selected variables were incorporated into a multivariable logistic regression model to identify independent predictors, and their odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. Utilizing the “rms” package in R, the study developed a predictive line chart considering independent risk factors. Model performance was evaluated through calibration curves, ROC analysis (AUC), and decision curve analysis (DCA). Statistical significance was set at P < 0.05.
Results
Clinical characteristics of patients
The study comprised 200 patients, with 140 in the training group and 60 in the validation group. ARDS incidence was 38.57% (54/140) in the training group and 41.67% (25/60) in the validation group, with no statistically significant intergroup differences in clinical characteristics (P > 0.05) (Table 1).
Table 1.
Clinical characteristics of patients in the training set and validation set
| Clinical data | Training set (n = 140) | Validation set (n = 60) | t/z/x² | P |
|---|---|---|---|---|
| SBP(mmHg) | 135.34 ± 28.13 | 134.52 ± 29.65 | 0.19 | 0.852 |
| DBP(mmHg) | 79.90 ± 17.76 | 76.58 ± 18.18 | 1.20 | 0.231 |
| Heart rate(bpm) | 84.53 ± 18.94 | 89.05 ± 21.54 | -1.48 | 0.139 |
| PO2 (mmHg) | 103.54 ± 39.65 | 107.60 ± 41.61 | -0.65 | 0.514 |
| PaO₂/FiO₂(mmHg) | 231.07 ± 93.64 | 233.63 ± 89.95 | -0.18 | 0.858 |
| Blood glucose(mmol/l) | 10.97 ± 3.95 | 11.59 ± 4.66 | -0.96 | 0.336 |
| WBC(10^9/L) | 12.19 ± 5.67 | 12.56 ± 6.24 | -0.41 | 0.684 |
| Neutrophil relative value | 0.78 ± 0.17 | 0.80 ± 0.15 | -0.78 | 0.435 |
| PLT(10^9/L) | 186.29 ± 82.49 | 196.98 ± 101.58 | -0.78 | 0.435 |
| Albumin (g/L) | 33.56 ± 6.61 | 33.58 ± 6.22 | -0.02 | 0.982 |
| FIB (g/L) | 3.58 ± 2.10 | 3.33 ± 1.55 | 0.83 | 0.409 |
| Lymphocyte relative value | 0.09 (0.06, 0.18) | 0.08 (0.04, 0.23) | -1.13 | 0.258 |
| PCT(ng/ml) | 0.33 (0.13, 1.03) | 0.23 (0.14, 1.20) | -0.08 | 0.939 |
| IL-6 (pg/ml) | 66.50 (21.43, 335.00) | 89.50 (33.00, 224.75) | -0.81 | 0.419 |
| CRP (mg/L) | 3.45 (1.70, 21.08) | 4.65 (2.18, 22.35) | -0.55 | 0.581 |
| Cr (umol/L) | 71.00 (58.00, 90.00) | 81.00 (60.50, 113.25) | -1.90 | 0.057 |
| BUN(mmol/L) | 5.74 (4.52, 7.62) | 6.67 (5.08, 8.55) | -1.82 | 0.069 |
| PT(s) | 13.55 (12.20, 14.83) | 13.10 (12.05, 14.50) | -0.83 | 0.405 |
| APTT(s) | 31.25 (28.50, 35.85) | 30.85 (27.80, 36.50) | -0.30 | 0.767 |
| DD (ug/ml) | 3.42 (1.05, 16.01) | 3.81 (1.08, 21.13) | -0.62 | 0.538 |
| DD/FIB | 0.95 (0.28, 6.45) | 0.94 (0.32, 7.10) | -0.63 | 0.532 |
| Male, n(%) | 92 (65.71) | 35 (58.33) | 0.99 | 0.320 |
| Hypertension, n(%) | 71 (50.71) | 37 (61.67) | 2.03 | 0.154 |
| Heart disease, n(%) | 13 (9.29) | 10 (16.67) | 2.25 | 0.134 |
| Liver diseases, n(%) | 12 (8.57) | 3 (5.00) | 0.34 | 0.558 |
| Kidney disease, n(%) | 14 (10.00) | 6 (10.00) | 0.00 | 1.000 |
| Lung diseases, n(%) | 14 (10.07) | 5 (8.33) | 0.15 | 0.702 |
| History of stroke, n(%) | 21 (15.00) | 5 (8.33) | 1.65 | 0.199 |
| Epilepsy, n(%) | 18 (12.86) | 7 (11.67) | 0.05 | 0.816 |
| Hydrocephalus, n(%) | 11 (7.86) | 8 (13.33) | 1.46 | 0.226 |
| Pulmonary infection, n(%) | 108 (77.14) | 43 (71.67) | 0.68 | 0.409 |
| Cardiac insufficiency, n(%) | 22 (15.71) | 11 (18.33) | 0.21 | 0.647 |
| Sepsis, n(%) | 38 (27.14) | 17 (28.33) | 0.03 | 0.863 |
| ARDS, n(%) | 54 (38.57) | 25 (41.67) | 0.17 | 0.682 |
| CRRT, n(%) | 16 (11.43) | 12 (20.00) | 2.56 | 0.109 |
Note: SBP Systolic blood pressure, DBP Diastolic blood pressure, WBC White blood cell, PLT platelet, FIB Fibrinogen, PCT Procalcitonin, IL-6 interleukin-6, CRP C-reactive protein, Cr creatinine, BUN blood urea nitrogen, PT Prothrombin time, APTT Activated partial thromboplastin time, DD D-dimer, ARDS Acute respiratory distress syndrome, CRRT Continuous renal replacement therapy
Baseline characteristics in the training group
Compared the distribution of SABI subtypes between patients with and without ARDS in the baseline characteristics and found no significant difference.Descriptive analysis revealed significant differences between the non-ARDS and ARDS groups in lung infection (P < 0.001), cardiac insufficiency (P = 0.031), sepsis (P < 0.001), continuous renal replacement therapy (P = 0.008), partial pressure of oxygen (P = 0.008), PaO₂/FiO₂ (P = 0.003), white blood cell count (P = 0.019), D-dimer (P = 0.017), and D-dimer/fibrinogen (P = 0.007). All comparisons demonstrated an obvious difference (P < 0.05) (Tables 2 and 3).
Table 2.
Comparison of clinical characteristics between the non-ARDS and ARDS groups
| Clinical data | Non-ARDS (n = 86) | ARDS (n = 54) | t/z/x² | P |
|---|---|---|---|---|
| Etiology | 3.488 | 0.322 | ||
| Acute Ischemic Stroke | 18 (20.930) | 6 (11.111) | ||
| Intracerebral Hemorrhage | 37 (43.023) | 22 (40.741) | ||
| Subarachnoid Hemorrhage | 13 (15.116) | 13 (24.074) | ||
| Traumatic Brain Injury | 18 (20.930) | 13 (24.074) | ||
| Underlying diseases | ||||
| Hypertension, n(%) | 47 (54.65) | 24 (44.44) | 1.38 | 0.240 |
| Heart disease, n(%) | 9 (10.47) | 4 (7.41) | 0.37 | 0.544 |
| Liver disease, n(%) | 5 (5.81) | 7 (12.96) | 1.35 | 0.246 |
| Kidney disease, n(%) | 10 (11.63) | 4 (7.41) | 0.66 | 0.418 |
| Lung diseases, n(%) | 7 (8.14) | 7 (13.21) | 0.93 | 0.335 |
| History of stroke, n(%) | 12 (13.95) | 9 (16.67) | 0.19 | 0.662 |
| Complications | ||||
| Epilepsy, n(%) | 10 (11.63) | 8 (14.81) | 0.30 | 0.583 |
| Hydrocephalus, n(%) | 8 (9.30) | 3 (5.56) | 0.23 | 0.632 |
| Pulmonary infection, n(%) | 57 (66.28) | 51 (94.44) | 14.92 | < 0.001 |
| Cardiac insufficiency, n(%) | 9 (10.47) | 13 (24.07) | 4.64 | 0.031 |
| Sepsis, n(%) | 14 (16.28) | 24 (44.44) | 13.31 | < 0.001 |
| CRRT, n(%) | 5 (5.81) | 11 (20.37) | 6.94 | 0.008 |
Note: CRRT Continuous renal replacement therapy
Table 3.
Comparison of laboratory indicators between non-ARDS and ARDS groups
| Clinical data | Non-ARDS (n = 86) | ARDS (n = 54) | t/z/x² | P |
|---|---|---|---|---|
| SBP(mmHg) | 138.99 ± 24.83 | 129.54 ± 32.10 | 1.84 | 0.068 |
| DBP(mmHg) | 79.93 ± 15.95 | 79.85 ± 20.48 | 0.02 | 0.981 |
| Heart rate(bpm) | 84.71 ± 19.46 | 84.24 ± 18.24 | 0.14 | 0.887 |
| PO2 (mmHg) | 110.56 ± 45.09 | 92.35 ± 25.59 | 2.70 | 0.008 |
| PaO₂/FiO₂ | 249.81 ± 100.47 | 201.22 ± 73.02 | 3.08 | 0.003 |
| Blood sugar(mmol/L) | 10.67 ± 3.98 | 11.44 ± 3.90 | -1.11 | 0.269 |
| WBC(10^9/L) | 11.30 ± 5.22 | 13.61 ± 6.11 | -2.38 | 0.019 |
| Neutrophil relative value | 0.78 ± 0.16 | 0.77 ± 0.19 | 0.55 | 0.581 |
| PLT(10^9/L) | 192.70 ± 79.42 | 176.09 ± 86.94 | 1.16 | 0.248 |
| Albumin (g/L) | 33.94 ± 5.80 | 32.95 ± 7.74 | 0.86 | 0.393 |
| FIB (g/L) | 3.79 ± 2.14 | 3.25 ± 2.01 | 1.48 | 0.141 |
| Lymphocyte relative value | 0.10 (0.06, 0.17) | 0.09 (0.05, 0.24) | -0.09 | 0.925 |
| PCT(ng/ml) | 0.33 (0.13, 0.78) | 0.33 (0.12, 1.14) | -0.18 | 0.854 |
| IL-6 (pg/ml) | 67.00 (31.75, 336.75) | 63.00 (12.00, 99.75) | -1.13 | 0.259 |
| CRP (mg/L) | 3.40 (2.02, 22.25) | 3.60 (1.52, 20.58) | -0.04 | 0.971 |
| Cr (umol/L) | 71.50 (59.75, 88.00) | 71.00 (55.25, 00.75) | -0.02 | 0.981 |
| BUN(mmol/L) | 5.69 (4.37, 7.42) | 5.79 (4.77, 8.23) | -0.83 | 0.404 |
| PT(s) | 13.55 (12.12, 14.67) | 13.70 (12.20, 16.47) | -1.25 | 0.210 |
| APTT(s) | 31.20 (28.52, 33.80) | 32.25 (28.35, 42.92) | -1.13 | 0.257 |
| DD (ug/ml) | 2.25 (0.84, 11.20) | 5.75 (1.90, 27.22) | -2.39 | 0.017 |
| DD/FIB | 0.66 (0.24, 3.84) | 1.54 (0.54, 10.96) | -2.71 | 0.007 |
Note: SBP Systolic blood pressure, DBP Diastolic blood pressure, WBC White blood cell, PLT platelet, FIB Fibrinogen, PCT Procalcitonin, IL-6 interleukin-6, CRP C-reactive protein, Cr creatinine, BUN blood urea nitrogen, PT Prothrombin time, APTT Activated partial thromboplastin time, DD D-dimer
Independent risk factor analysis
Variables were initially selected using LASSO regression with 10-fold cross-validation. The optimal lambda value was determined to be 0.056, which identified six potential predictors: Pulmonary infection, Sepsis, WBC, FIB, APTT, and PaO₂/FiO₂. These variables were subsequently entered into a logistic regression model for further analysis. Ultimately, three independent factors remained statistically significant: Pulmonary infection, Sepsis, and PaO₂/FiO₂. (Figures 2 and 3; Table 4)
Fig. 2.
Variable selection using LASSO regression with 10-fold cross-validation. (A) Coefficient paths across log(λ). (B) Cross-validated binomial deviance
Fig. 3.
Relative importance of candidate predictors selected by LASSO regression
Table 4.
Single factor and multivariable logistic analysis
| Variable | Multifactorial | |||
|---|---|---|---|---|
| SE | Z | OR (95%CI) | P | |
| Pulmonary infection | 0.665 | 3.046 | 7.577 (2.335, 34.331) | 0.002 |
| Sepsis | 0.437 | 2.389 | 2.837 (1.221, 6.823) | 0.017 |
| PaO₂/FiO₂ | 0.003 | -3.059 | 0.992 (0.986, 0.997) | 0.002 |
Note: SE standard error, OR odds ratio, CI confidence interval. The OR for the PaO₂/FiO₂ ratio corresponds to a decrease of 1 mmHg in this parameter. The model included all candidate variables selected via LASSO regression
Predictive modeling for the occurrence of ARDS nomogram in SABI patients
A risk prediction model was constructed using statistically significant variables. Variable scores were summed to obtain a total score, corresponding to the ‘risk’ axis value for estimating the probability of ARDS in patients. Higher total scores indicated higher risk of ARDS, and vice versa (Figs. 2 and 4).
Fig. 4.
Line chart for ARDS risk prediction in SABI patients
Performance of column line charts
The column-line graph demonstrated good discriminative performance, with an AUC of 0.778 (95% CI: 0.700–0.857) in the training set and 0.754 (95% CI: 0.625–0.884) in the validation set (Fig. 5). Figure 6 compares the predictive performance of the three-factor nomogram with that of single-predictor models in the training set. The combined model (AUC = 0.778;) outperformed each single-factor model. Furthermore, As shown in sFigure 2, the predicted probability density distributions based on the training set (A) and the validation set (B) demonstrate a distinguishable bimodal pattern between ARDS and non-ARDS patients, indicating that the model possesses good discriminatory ability. Meanwhile, the calibration curve basically varies around the diagonal, proving that the model fitting effect is relatively ideal (Figs. 7 and 8). The decision curve analysis (DCA) demonstrated the good clinical utility value of the model (Figs. 9 and 10).
Fig. 5.
ROC curves of the nomogram for ARDS prediction in SABI patients
Fig. 6.
Comparison of ROC curves between the three-factor nomogram and single-factor models in the training set
Fig. 7.
The calibration curves of predictive nomogram for ARDS in the training set
Fig. 8.
The calibration curve in the validation set
Fig. 9.
Decision curve in the training set
Fig. 10.
Decision curves in the validation set
Discussion
The development of acute respiratory distress syndrome (ARDS) following severe acute brain injury (SABI) presents a complex and challenging scenario in neurocritical care, with its high incidence strongly linked to poor clinical outcomes [9, 18]. Through a retrospective analysis, this study identified pulmonary infection, sepsis, and a low oxygenation index (PaO₂/FiO₂) as independent risk factors for ARDS in SABI patients. Based on these factors, we constructed an intuitive nomogram prediction model. Internal validation demonstrated that the model achieved moderate and stable discriminative ability in both the training and validation sets (AUCs of 0.778 and 0.754, respectively). Calibration curves and decision curve analysis (DCA) further confirmed its good calibration and potential for positive clinical net benefit.Our AUC values were comparable with the performance range (AUC 0.76–0.83) reported in a recent study that used machine learning to predict ARDS in TBI patients [19]. For ARDS prediction models in general critical care or specific etiologies like pneumonia, the reported AUC values vary considerably. This variation depends on model complexity, variable selection, and data types, with basic clinical models typically ranging from 0.60 to 0.80, while models integrating multimodal data or advanced machine learning techniques can exceed 0.85 [20, 21]. Our model targets the specific SABI population. Notably, by employing only three routine clinical variables, it attained a discriminative performance comparable to that of more complex machine learning models. This balance between predictive efficacy and practical clinical utility offers a disease-specific risk assessment tool for this challenging management problem.
Our study identified pulmonary infection as the strongest independent predictor of ARDS in SABI patients. This aligns with established clinical observations that pneumonia is one of the most frequent direct triggers for ARDS [22, 23]. In the context of SABI, this association is markedly amplified. Patients often present with impaired consciousness, leading to a weakened cough reflex and diminished airway clearance. Combined with the frequent need for mechanical ventilation, this renders them highly susceptible to aspiration, hospital-acquired pneumonia, and ventilator-associated pneumonia (VAP) [24, 25]. Evidence indicates that VAP develops in approximately 20% of traumatic brain injury patients around five days post-injury, on average [26]. Notably, pathogenic bacterial growth can be detected in lower respiratory tract specimens from some patients as early as 24 h after intubation [27]. Consequently, pulmonary infection in SABI is not merely a common complication; it acts as a critical “second hit” that triggers uncontrolled pulmonary inflammation, disrupts the alveolar-capillary barrier, and initiates the pathophysiological cascade of ARDS. This finding underscores the vital importance of extremely early, proactive prevention, monitoring, and treatment of pulmonary infections in the management of SABI patients.
Sepsis emerged as another critical predictor in our analysis. While both localized pneumonia and systemic sepsis can precipitate ARDS, they share a common terminal pathway: a dysregulated inflammatory response coupled with immunothrombosis. In pulmonary infection, activated immune cells within the alveoli release a surge of inflammatory mediators, causing direct injury to the alveolar epithelium. Systemic sepsis, conversely, assaults the pulmonary capillary endothelium via a circulating cytokine storm [26].
These processes act synergistically, dramatically accelerating the development of permeability edema across the alveolar-capillary membrane. Furthermore, the activated coagulation system during infection promotes the formation of “immunothrombi” within the pulmonary microvasculature, which further impairs gas exchange and intensifies inflammation [28]. Besides, This vulnerability is compounded by the fact that SABI itself can induce a significant systemic inflammatory response syndrome (SIRS), a process linked to the “blast injury” hypothesis and centrally mediated immune dysregulation [29]. The literature consistently identifies sepsis as the most common extra-pulmonary etiology of ARDS, responsible for 31–34% of all cases, and notes that sepsis-associated ARDS carries a worse prognosis and higher mortality compared to ARDS from other causes [30, 31]. Therefore, sepsis signifies a profoundly amplified state of systemic pro-inflammation and pro-coagulation, which markedly heightens the lung’s susceptibility to injury in patients with SABI.
The third independent predictor identified was a low PaO₂/FiO₂ ratio measured early after admission. As the core metric for quantifying the severity of hypoxemia, the PaO₂/FiO₂ ratio forms the cornerstone of the Berlin definition for diagnosing ARDS [32]. Our study confirms that a lower PaO₂/FiO₂ value obtained within the first 24 h of admission, even if it does not yet meet the formal diagnostic threshold for ARDS, strongly signals a high risk of progression to ARDS. Table 4 indicates that for every 50 mmHg decrease in the PaO₂/FiO₂ ratio, the risk of developing ARDS increases by approximately 50%.This predictive value can be interpreted through several interrelated mechanisms. An early low ratio may reflect underlying, subclinical lung injury, such as neurogenic pulmonary edema, micro-atelectasis, or undetected infection. Furthermore, the state of hypoxemia itself can exacerbate pulmonary endothelial damage through mechanisms like hypoxic vasoconstriction and oxidative stress. Finally, emerging evidence suggests that combining the oxygenation index with respiratory mechanics parameters, such as driving pressure, improves the prediction of mortality risk in ARDS patients [33]. This underscores that the trajectory of the PaO₂/FiO₂ ratio carries significant pathophysiological information beyond a single snapshot.Therefore, the early admission PaO₂/FiO₂ serves not merely as a diagnostic criterion but as a potent early warning sign. It alerts clinicians to the need for heightened vigilance, closer monitoring, and potentially augmented respiratory support.
The three independent predictors identified in this study: pulmonary infection, sepsis, and a low PaO₂/FiO₂, suggest that the development of ARDS after SABI follows a multi-layered and dynamically evolving pathological network.The initial severe brain injury acts as the “first hit,” capable of inducing systemic endothelial damage and immune activation through neuroinflammation and a catecholamine surge [7, 34]. The subsequent development of either pulmonary infection or sepsis then constitutes a potent “second hit,” dramatically amplifying inflammatory and coagulatory cascades that directly assault the alveolar-capillary barrier [35]. In addition, a low PaO₂/FiO₂ ratio serves both as a marker of this early micro-injury and, through hypoxia-driven oxidative stress and cellular apoptosis, may establish a positive feedback loop that further aggravates the process.This conceptual model of “neuro-immune-pulmonary” cross-talk shares a fundamental principle with recent frameworks proposed in related fields to explain complex comorbidities—namely, that intricate clinical phenotypes often arise from shared biological pathways acting across organ systems and biological scales [36, 37].
We fully acknowledge that center-specific clinical practices—encompassing ventilation strategies, fluid management protocols, infection control standards, and neuro-monitoring or sedation-analgesia regimens—can influence both the predictors themselves and the incidence of ARDS. Among the three predictors, the PaO₂/FiO₂ ratio, as an objective and continuously measured physiological variable, is likely the most stable across different settings. The diagnosis of pulmonary infection, while subject to minor interpretive variations, is based on relatively standardized criteria, suggesting its predictive role may remain robust. In contrast, the diagnosis and management of sepsis exhibit greater heterogeneity between ICUs, making it a key variable requiring careful attention during future external validation of the model.
Limitations
Several limitations of this study should be acknowledged. Its single-center, retrospective design, coupled with a relatively modest sample size, inherently carries the risk of unforeseen confounding factors that may influence the results. However, to enhance the methodological rigor of our modeling process, we adhered to established statistical practices for prediction model development and validation, as exemplified in recent large-scale studies [38]. Future research should focus on multi-center, prospective clinical trials to further validate the effects of these risk factors and to assess the generalizability of our predictive model, thereby providing more robust evidence to guide clinical decision-making for this complex patient population.
Conclusion
In summary, pulmonary infection, sepsis, and a low PaO₂/FiO₂ ratio were identified as independent risk factors for ARDS in patients with SABI. Based on these predictors, we developed a well-performing and parsimonious nomogram prediction model. This tool may serve as a practical bedside aid, assisting clinicians in the early identification of high-risk patients. Such early recognition could enable the timely initiation of targeted monitoring and preventive strategies, with the potential to ultimately improve outcomes in this complex patient population.
Supplementary Information
Supplementary Material 1. sFigure 1. Power analysis for sample size estimation using G*Power software. sFigure 2. Distribution of predicted probabilities from the nomogram for ARDS in the training and validation sets.
Acknowledgements
The authors acknowledge the financial support from the Zhejiang Provincial Medical and Health Science and Technology Program Project. We also thank Chang su for Data collection and JunLong xu for helpful comments on the early draft of the manuscript.
Abbreviations
- ARDS
Acute respiratory distress syndrome
- SABI
Severe acute brain injury
- PEEP
positive end-expiratory pressure
- ROC
receiver operating characteristic
- AUC
area under curve
- OR
odds ratio
- CI
confidence interval
- WBC
white blood cell
- PLT
platelet
- BUN
blood urea nitrogen
- Cr
creatinine
- IL-6
interleukin-6
- CRP
C-reactive protein
Authors’ contributions
All authors made substantial contributions to this work. Qiaoqiao You conceived and designed the study, developed the methodology, and wrote the initial manuscript. Wen Hu, Xian Zhang was responsible for data collection, curation, and analysis. Jianping Ye, Chang Su, Miaomiao Chen contributed to interpretation of results, review, editing, and visualization. Jun Long Xu provided supervision, acquired funding, and administered the project. All authors reviewed and approved the final version of the manuscript for submission.
Funding
This research was funded by the Zhejiang Provincial Medical and Health Science and Technology Program Project (Project No. 2025KY1991).
Data availability
The original contributions introduced in the research are provided in the article/supplementary materials, and further inquiries are available directly by contacting the corresponding author.
Declarations
Ethics approval and consent to participate
All procedures in this study followed the ethical guidelines of the Helsinki Declaration of the World Medical Association.Ethical approval for this retrospective study was obtained from the Scientific Research Ethics Committee of Lishui Municipal People’s Hospital (Approval No.: 20250603). The committee waived the need for individual informed consent as the study utilized anonymized retrospective data, and all patient identifiers were removed to ensure privacy.
Consent for publication
All participants provided written informed consent for publication.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Qiaoqiao You, Chang Su and Jianping Ye contributed equally to this work.
References
- 1.Robba C, Bonatti G, Pelosi P, Citerio G. Extracranial complications after traumatic brain injury: targeting the brain and the body. Curr Opin Crit Care. 2020;26(2):137–46. [DOI] [PubMed] [Google Scholar]
- 2.Koutsoukou A, Katsiari M, Orfanos SE, et al. Respiratory mechanics in brain injury: A review. World J Crit Care Med. 2016;5(1):65–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.McDonald SJ, Sharkey JM, Sun M, et al. Beyond the brain: peripheral interactions after traumatic brain injury. J Neurotrauma. 2020;37(5):770–81. [DOI] [PubMed] [Google Scholar]
- 4.Carney N, Totten AM, O’reilly C, et al. Guidelines for the management of severe traumatic brain injury. Fourth Ed Neurosurg. 2017;80(1):6–15. [DOI] [PubMed] [Google Scholar]
- 5.Kaku S, Nguyen CD, Htet NN, et al. Acute respiratory distress syndrome: Etiology, Pathogenesis, and summary on management. J Intensive Care Med. 2020;35(8):723–37. [DOI] [PubMed] [Google Scholar]
- 6.Bos LDJ, Ware LB. Acute respiratory distress syndrome: causes, pathophysiology, and phenotypes. Lancet. 2022;400(10358):1145–56. [DOI] [PubMed] [Google Scholar]
- 7.Matin N, Sarhadi K, Crooks CP, et al. Brain-Lung crosstalk: management of concomitant severe acute brain injury and acute respiratory distress syndrome. Curr Treat Options Neurol. 2022;24(9):383–408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kim JA, Wahlster S, Labuzetta JN, et al. Focused management of patients with severe acute brain injury and ARDS. Chest. 2022;161(1):140–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ziaka M, Exadaktylos A. Brain-lung interactions and mechanical ventilation in patients with isolated brain injury. Crit Care. 2021;25(1):358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ziaka M, Exadaktylos A. Pathophysiology of acute lung injury in patients with acute brain injury: the triple-hit hypothesis. Crit Care. 2024;28(1):71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Beqiri E, Smielewski P, Guérin C, et al. Neurological and respiratory effects of lung protective ventilation in acute brain injury patients without lung injury: brain vent, a single centre randomized interventional study. Crit Care. 2023;27(1):115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hourmant Y, Huard D, Demeure Dit Latte D, et al. Effect of continuous infusion of hypertonic saline solution on survival of patients with brain injury: a systematic review and meta-analysis. Anaesth Crit Care Pain Med. 2023;42(2):101177. [DOI] [PubMed] [Google Scholar]
- 13.Wright JM, Gerges C, Shammassian B, et al. Prone position ventilation in neurologically ill patients: A systematic review and proposed protocol. Crit Care Med. 2021;49(3):e269–78. [DOI] [PubMed] [Google Scholar]
- 14.Sharie SA, Almari R, Azzam S, et al. Brain protective ventilation strategies in severe acute brain injury. Curr Neurol Neurosci Rep. 2025;25(1):68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gong MN, Thompson BT. Acute respiratory distress syndrome: shifting the emphasis from treatment to prevention. Curr Opin Crit Care. 2016;22(1):21–37. [DOI] [PubMed] [Google Scholar]
- 16.Wu Z, Liu Y, Xu J, et al. A Ventilator-associated pneumonia prediction model in patients with acute respiratory distress syndrome. Clin Infect Dis. 2020;71(Suppl 4):S400–8. [DOI] [PubMed] [Google Scholar]
- 17.Frisvold S, Coppola S, Ehrmann S, et al. Respiratory challenges and ventilatory management in different types of acute brain-injured patients. Crit Care. 2023;27(1):247. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Peng J, Pan F, Xu Y, et al. Mesenchymal stem cell transplantation alleviated TBI-induced lung injury by inhibiting PAD4-dependent NET formation. J Adv Res. 2025:S2090-1323(25)00438-2. [DOI] [PMC free article] [PubMed]
- 19.Wang R, Cai L, Zhang J, et al. Prediction of acute respiratory distress syndrome in traumatic brain injury patients based on machine learning algorithms. Med (Kaunas). 2023;59(1):171. [DOI] [PMC free article] [PubMed]
- 20.Wei S, Zhang H, Li H, et al. Establishment and validation of predictive model of ARDS in critically ill patients. J Transl Med. 2025;23(1):64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Yang J, Zeng S, Cui S, et al. Predictive modeling of acute respiratory distress syndrome using machine learning: systematic review and Meta-Analysis. J Med Internet Res. 2025;27:e66615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Xie R, Tan D, Liu B et al. Acute respiratory distress syndrome (ARDS): from mechanistic insights to therapeutic strategies. MedComm (2020). 2025;6(2):e70074. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ferguson ND, Frutos-Vivar F, Esteban A, et al. Clinical risk conditions for acute lung injury in the intensive care unit and hospital ward: a prospective observational study. Crit Care. 2007;11(5):R96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hu PJ, Pittet JF, Kerby JD, et al. Acute brain trauma, lung injury, and pneumonia: more than just altered mental status and decreased airway protection. Am J Physiol Lung Cell Mol Physiol. 2017;313(1):L1–15. [DOI] [PubMed] [Google Scholar]
- 25.Ziaka M, Exadaktylos A. ARDS associated acute brain injury: from the lung to the brain. Eur J Med Res. 2022;27(1):150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Robba C, Rebora P, Banzato E, et al. Incidence, risk Factors, and effects on outcome of Ventilator-Associated pneumonia in patients with traumatic brain injury: analysis of a Large, Multicenter, Prospective, observational longitudinal study. Chest. 2020;158(6):2292–303. [DOI] [PubMed] [Google Scholar]
- 27.Harrell KN, Lee WB, Rooks HJ, et al. Early pneumonia diagnosis decreases ventilator-associated pneumonia rates in trauma population. J Trauma Acute Care Surg. 2023;94(1):30–5. [DOI] [PubMed] [Google Scholar]
- 28.Lim EY, Lee SY, Shin HS, Kim GD. Reactive oxygen species and strategies for antioxidant intervention in acute respiratory distress syndrome. Antioxid (Basel). 2023;12(11):2016. [DOI] [PMC free article] [PubMed]
- 29.Mascia L. Acute lung injury in patients with severe brain injury: a double hit model. Neurocrit Care. 2009;11(3):417–26. [DOI] [PubMed] [Google Scholar]
- 30.Jiang L, Yu C, Xie C, et al. Enhancing early mortality prediction for sepsis-associated acute respiratory distress syndrome patients via optimized machine learning algorithm: development and multiple databases’ validation of the SAFE-Mo. Int J Surg. 2025;111(9):6008–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Xu H, Sheng S, Luo W, et al. Acute respiratory distress syndrome heterogeneity and the septic ARDS subgroup. Front Immunol. 2023;14:1277161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distress syndrome: the Berlin definition. JAMA. 2012;307(23):2526–33. [DOI] [PubMed] [Google Scholar]
- 33.Yu YY, Ou WF, Wu JJ, et al. A combination of oxygenation and driving pressure can provide valuable information in predicting the risk of mortality in ARDS patients. PLoS ONE. 2023;18(12):e0295261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Chacón-Aponte AA, Durán-Vargas ÉA, Arévalo-Carrillo JA, et al. Brain-lung interaction: a vicious cycle in traumatic brain injury. Acute Crit Care. 2022;37(1):35–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Mascia L, D’albo R, Cavalli I, et al. Organ crosstalk: brain-lung interaction. Front Med (Lausanne). 2025;12:1655813. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Kuang N, Liu Z, Yu G, et al. Neurodevelopmental risk and adaptation as a model for comorbidity among internalizing and externalizing disorders: genomics and cell-specific expression enriched morphometric study. BMC Med. 2023;21(1):291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Wu X, Zhang K, Kuang N, et al. Author correction: developing brain asymmetry shapes cognitive and psychiatric outcomes in adolescence. Nat Commun. 2025;16(1):6325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Yu G, Liu Z, Wu X, et al. Common and disorder-specific cortical thickness alterations in internalizing, externalizing and thought disorders during early adolescence: an adolescent brain and cognitive development study. J Psychiatry Neurosci. 2023;48(5):E345–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Material 1. sFigure 1. Power analysis for sample size estimation using G*Power software. sFigure 2. Distribution of predicted probabilities from the nomogram for ARDS in the training and validation sets.
Data Availability Statement
The original contributions introduced in the research are provided in the article/supplementary materials, and further inquiries are available directly by contacting the corresponding author.