Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice BMC Geriatr . 2026 Mar 4;26:506. doi: 10.1186/s12877-026-07241-z Search in PMC Search in PubMed View in NLM Catalog Add to search Thoracic muscle loss increases the use of mechanical ventilation in elderly patients with pulmonary embolism: constructing and validating a machine learning model on a two-center cohort Zexiang Deng Zexiang Deng 1 Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, No. 1 Minde Road, Nanchang, 330006 China 2 Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, 330006 China Find articles by Zexiang Deng 1, 2, # , Dan Luo Dan Luo 1 Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, No. 1 Minde Road, Nanchang, 330006 China 2 Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, 330006 China Find articles by Dan Luo 1, 2, # , Jingjing Zhou Jingjing Zhou 1 Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, No. 1 Minde Road, Nanchang, 330006 China 2 Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, 330006 China Find articles by Jingjing Zhou 1, 2, ✉ , Xiaomei Zhang Xiaomei Zhang 3 Department of Radiology, The First Affiliated Hospital of Gannan Medical University, Ganzhou Economic and Technological Development Zone, No. 128, Jinling West Road, Ganzhou, 341001 China Find articles by Xiaomei Zhang 3 , Haibo Ren Haibo Ren 1 Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, No. 1 Minde Road, Nanchang, 330006 China 2 Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, 330006 China Find articles by Haibo Ren 1, 2 , Weiwei Deng Weiwei Deng 4 Clinical and Technical Support, Philips Healthcare, No.718 Lingshi Road, Jing’an District, Shanghai, China Find articles by Weiwei Deng 4 , Lianggeng Gong Lianggeng Gong 1 Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, No. 1 Minde Road, Nanchang, 330006 China 2 Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, 330006 China Find articles by Lianggeng Gong 1, 2, ✉ Author information Article notes Copyright and License information 1 Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, No. 1 Minde Road, Nanchang, 330006 China 2 Jiangxi Provincial Key Laboratory of Intelligent Medical Imaging, Nanchang, 330006 China 3 Department of Radiology, The First Affiliated Hospital of Gannan Medical University, Ganzhou Economic and Technological Development Zone, No. 128, Jinling West Road, Ganzhou, 341001 China 4 Clinical and Technical Support, Philips Healthcare, No.718 Lingshi Road, Jing’an District, Shanghai, China ✉ Corresponding author. # Contributed equally. Received 2024 Oct 10; Accepted 2026 Feb 23; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13067536 PMID: 41782086 Abstract Objective This study aimed to develop and validate a machine learning (ML) model to predict the need for mechanical ventilation (MV) in elderly patients diagnosed with acute pulmonary embolism (APE). Materials and methods The study included a cohort of 321 patients from two centers. Center A contributed 261 patients for the development of four ML models: Random Forest, XGBoost, Logistic Regression (LR), and Support Vector Classifier (SVC). The remaining 60 patients from Center B were used for external testing. Feature selection incorporated CT histogram features related to muscular density and area of the pectoralis muscles on CT images at the level of the fourth thoracic vertebra, common geriatric comorbidities, and routine laboratory tests for APE. The area under the curve (AUC) was used to evaluate the predictive performance of the models; calibration curves were employed to assess calibration performance, and the sPESI score served as a baseline comparator. Shapley Additive exPlanations (SHAP) plots were utilized to visualize the importance of each feature. Results The final set of features included low oxygen saturation, smoke status, CT_PMA_10th, CT_PMA_90th, CT_PMI_75th, CT_PMA_Fat_ratio, chest pain, syncope, diabetes, gender, chronic heart failure, NT-proBNP/BNP positive, and D-dimer. In the internal validation set, the four models performed well and exhibited similar performance, with AUC values exceeding 0.80. Among the models evaluated, the LR model demonstrated the best performance on the external test set, with an AUC of 0.837, an accuracy of 0.817, a recall of 0.750, a specificity of 0.833, a precision of 0.529, and an F1-score of 0.621. The SHAP plot revealed that low oxygen saturation, CT_PMA_10th, and CT_PMI_75th were highly important features. Conclusion Loss of pectoral muscle may be associated with the need for MV in elderly patients with APE. The prediction model developed in this study, which includes this factor, could aid in identifying high-risk individuals and may inform future efforts to improve early risk stratification and personalized management in this population. Supplementary Information The online version contains supplementary material available at 10.1186/s12877-026-07241-z. Keywords: Acute pulmonary embolism, CT histogram, Machine learning, Mechanical ventilation, Older adults Introduction Acute pulmonary embolism (APE) presents a significant challenge in contemporary society due to its high prevalence and mortality rates [ 1 , 2 ], particularly among the elderly. Elderly individuals are especially vulnerable to APE due to multiple risk factors, including decreased physical activity and the presence of comorbidities such as chronic obstructive pulmonary disease (COPD) and cancer [ 3 , 4 ]. Therefore, developing a tailored management tool for elderly patients is essential for improving care and optimizing treatment strategies. In developing such management tools, attention must be paid to geriatric-specific conditions that influence outcomes. Skeletal muscle atrophy has also become a significant health concern for elderly patients [ 5 , 6 ], as aging is often accompanied by systemic chronic inflammation and a reduced capacity for skeletal muscle repair, leading to muscle fibrosis and fat deposition [ 7 ]. This phenomenon is reflected in diminished skeletal muscle area and density on CT images. The reduction in pectoralis muscle mass, including major respiratory accessory muscles such as the pectoralis major (PMA) and pectoralis minor (PMI), is associated with the severity of airflow limitation and deteriorating respiratory function [ 8 , 9 ], which contributes to a poor prognosis in patients with pneumonia [ 10 – 12 ]. Furthermore, recent studies have demonstrated that CT-defined pectoralis muscle mass and density predict 30-day mortality in patients with APE [ 13 ]. This finding suggests that pectoral muscle loss may adversely affect treatment outcomes. Various imaging techniques have been proposed for evaluating skeletal muscle status [ 14 – 16 ], with CT histogram analysis providing quantitative data on pectoral muscle density and distribution. It is important to note that the incidence of acute respiratory failure (ARF) is increasing among elderly patients with APE [ 17 ]. Factors such as persistent thrombus, acute pulmonary hypertension, and right heart failure, combined with ongoing perfusion defects, may lead to severe hypoxemia. In such cases, mechanical ventilation (MV), whether invasive or non-invasive, may become necessary to preserve the patient's life. While muscle loss is a known consequence of MV [ 18 ], it remains unclear whether pre-existing sarcopenia is itself an independent risk factor that heightens the likelihood of requiring MV in the first place. Machine learning (ML) models are increasingly being deployed to predict outcomes in both respiratory and cardiovascular diseases, offering significant potential for enhanced clinical applicability and patient management [ 19 ]. The integration of these advanced computational approaches aims to move beyond traditional statistical methods to identify complex, non-linear relationships within vast clinical datasets. However, previous research has primarily focused on the shock resulting from acute right heart failure in APE patients [ 20 , 21 ], and there is few ML models available for predicting the need for MV in patients with APE. This study will utilize CT histogram analysis to assess skeletal muscle status, integrating common geriatric diseases and routine laboratory tests for APE into the development and validation of ML models. Methods Patients This study enrolled patients aged 65 years and older with APE from two tertiary referral hospitals: Center A (a multi-specialty hospital), from which patients were enrolled between January 2020 and April 2024, and Center B (a general hospital), from which patients were collected between January 2024 and March 2024. Inclusion criteria were: (1) a confirmed diagnosis of APE by CT pulmonary angiography (CTPA) [ 1 ], (2) a non-enhanced CT scan performed prior to CTPA within 24 h, and (3) hospitalization. Exclusion criteria included: (1) patients who had undergone pectoral muscle resection, (2) patients with incomplete clinical data or laboratory test, and (3) patients with poor-quality CT images (Fig. 1 ). Fig. 1. Open in a new tab Inclusion and exclusion flowchart. CTPA: Computed Tomography Pulmonary Angiography The primary outcome, MV, is defined as the use of either non-invasive ventilation or invasive mechanical ventilation at any point during the index hospitalization following confirmed PE admission. CT scan A standardized non-enhanced CT scan was performed with the patient in the prone position, ensuring both arms were elevated above the head to minimize beam-hardening artifacts and enhance visualization of thoracic structures. Three different CT scanners were used for image acquisition: Philips Brilliance iCT, Philips IQon CT, and GE Revolution CT, with a tube voltage of 120 kV, tube current of 30 mA or 35 mA, and pitch of 1.0 mm. The scan range extended from the level of the lung apices to the costophrenic angles. Both the slice thickness and reconstruction interval were 1 mm for all scanners. CTPA scans were performed using the same scanners, with intravenous administration of an iodine-based contrast medium at a rate of 4.0–4.5 mL/s through a peripheral venous line. Automatic bolus tracking was conducted in the pulmonary trunk using a trigger threshold of 100 Hounsfield units (HU). The scanning equipment and parameters for Center B are detailed in Supplement Table 1. Radiological features A radiologist with 3 years of experience in chest CT used ITK-SNAP 4.0 to delineate the bilateral PMA and PMI at the axial level of the T4 vertebra [ 8 ]. Subsequently, another radiologist with 3 years of chest CT experience randomly selected 50% of the patients to redraw the lesions for an inter-observer consistency test. Histogram features, including the 10th, 25th, 50th, 75th, and 90th percentiles of the PMA and PMI, were extracted using Python (version 3.12.4) with the libraries numpy, nibabel, and scipy.stats (Fig. 2 ). A novel metric, the intermuscular fat ratio, was introduced to characterize the degree of muscle degeneration. This metric is defined as the ratio of the area of fat within the pectoral muscles (HU < −30) to the total area of the pectoral muscles [ 22 ]. The fat ratio was calculated separately for the PMA and PMI muscles, and patients were classified into three categories: Mild (< 10%), Moderate (10%−20%), and Severe (> 20%). Fig. 2. Open in a new tab CT features report of the PMA and PMI for a representative patient Ultimately, these CT features were utilized for subsequent feature screening, including the 10th, 25th, 50th, 75th, and 90th percentiles of CT values for the pectoralis major and pectoralis minor muscles, the fat ratios of the pectoralis major and minor muscles, and the total muscle areas of the pectoralis muscles. Clinical features In this study,low blood oxygen saturation was defined as a measurement below 90% without supplemental oxygen or below 95% with supplemental oxygen, which is considered a predictive factor clinically. Additionally, COPD and smoke are known to be associated with pectoral muscle degeneration [ 8 ]. Individuals with COPD and smokers are at a higher risk of experiencing respiratory adverse events. Common geriatric comorbidities, such as hypertension, diabetes, chronic heart failure (CHF—encompassing chronic coronary artery disease, cardiomyopathy, and cor pulmonale), cancer, and common symptoms of APE, including dyspnea, chest pain, and syncope [ 1 ], were included in the clinical feature selection process. All clinical features included in the analysis were collected at admission and represent baseline values. Laboratory features Routine laboratory tests for APE, including D-dimer, NT-proBNP or BNP, and cTnI, were also included. NT-proBNP and BNP levels are assessed using specific thresholds to indicate acute heart failure. The NT-proBNP threshold is age-dependent: 1800 pg/mL for patients under 75 years old and 900 pg/mL for patients aged 75 years and older, according to the ICON-RELOADED study [ 23 ]. For BNP, the threshold is uniformly set at 100 pg/mL, with values above this threshold considered indicative of heart failure. cTnI levels > 0.05 ng/mL are considered positive, indicating acute myocardial injury. All laboratory features included in the analysis were collected at admission and represent baseline values. Models development The modeling process began with the careful curation of patient data from Center A, which were randomly partitioned into training and testing sets in a 7:3 ratio. To ensure balanced representation of positive and negative cases, a stratified sampling technique was implemented, maintaining consistency across both training and validation sets. Additionally, weighted algorithms were integrated during model development to address potential biases in the data. Embedded within the fivefold cross-validation on the training set, feature selection via Lasso regression was performed independently in each iteration. Specifically, for every fold, features were selected using the four training folds before evaluating the model on the remaining validation fold. This approach enabled the identification of the most informative radiological, clinical, and laboratory features that significantly contributed to the prediction task. The selected features were then assembled into a feature matrix, with highly correlated features meticulously pruned to mitigate multicollinearity and improve model interpretability. Utilizing Python libraries such as sklearn and xgboost, we developed four ML models: Random Forest (RF), Extreme Gradient Boosting (XGB), Logistic Regression (LR), and Support Vector Machine Classifier (SVC). Each model underwent rigorous hyperparameter tuning through a fivefold cross-validation grid search to ensure optimal performance within the training domain. To address class imbalance and improve predictive performance for the minority class, the class_weight = 'balanced'parameter was consistently applied during model instantiation. After training and fitting the models on the training set, they were evaluated on both the validation and test sets to assess their generalization capabilities. The performance of the models was evaluated using the receiver operating characteristic area under the curve (ROC-AUC), F1 score, precision, recall (sensitivity), and specificity. For all models, the decision threshold was set at 0.5 for evaluation. Model calibration was assessed using calibration curves.. Additionally, the simplified Pulmonary Embolism Severity Index (sPESI) score, with a cutoff value of ≥ 1, served as a benchmark for comparison, providing insights into the models' clinical relevance. To mitigate the "black box" effect of ML models, Shapley Additive exPlanations (SHAP) plots were utilized to visualize the importance of each feature in the classification outcomes generated by the four ML models. Statistical analysis Continuous variables with a normal distribution were expressed as mean ± standard deviation (SD) and compared using an independent samples t-test. For variables not following a normal distribution, the median (M), first quartile (Q1), and third quartile (Q3) were reported, and comparisons were made using the Mann–Whitney U test. Categorical variables were compared using the χ 2 test or Fisher’s exact test. Inter-observer agreement for chest muscle characteristics was assessed using the intraclass correlation coefficient (ICC). A p-value of less than 0.05 was considered statistically significant. Statistical analysis and modeling were performed using Python. Continuous variables were standardized prior to modeling. Predictive models were developed using algorithms from the scikit-learn library. All statistical plots and result visualizations were generated using Python and Origin 2024. Result Patients A total of 321 patients from two centers were included in this study: 261 from Center A and 60 from Center B. In Center A, 50 patients required MV during hospitalization, while in Center B, this number was 12. Analysis of the baseline clinical characteristics revealed a significant difference only in the prevalence of hypertension between the training and internal validation sets, with no significant differences observed between the two centers (see Table 1 ). These findings imply that there are negligible differences in the baseline clinical characteristics of patients from the two centers. Table 1. Clinical characteristics of the subjects at baseline Database Center A Center B p Total ( n = 261) Training set ( n = 182) Validation set ( n = 79) p Test set ( n = 60) Population Gender:female[counts(%)] 139(53.3%) 95(52.2%) 44(55.7%) 0.603 33(55.0%) 0.615 Age,years [mean ± SD] 75.2 ± 6.8 75.5 ± 6.7 74.5 ± 6.9 0.242 74.3 ± 6.3 0.342 Symptoms Dyspnea [counts(%)] 141(54%) 94(51.6%) 47(59.5%) 0.243 30(50.0%) 0.573 Chest pain [counts(%)] 13(5%) 7(3.8%) 6(7.6%) 0.222 # 5(8.3%) 0.348 # Syncope [counts(%)] 16(6%) 12(6.6%) 4(5.1%) 0.783 2(3.3%) 0.543 # Low oxygen saturation a 46(18%) 31(17%) 15(19%) 0.703 9(15.5%) 0.701 Smoke 0.231 0.083 Never [counts(%)] 210(80.5%) 150(82.4%) 60(75.9%) 41(71.7%) Ex [counts(%)] 19(7.3%) 10(5.5%) 9(11.4%) 3(5.0%) Current [counts(%)] 32(12.3%) 22(12.1%) 10(12.7%) 14(23.3%) Complication Hypertension [counts(%)] 135(51.7%) 85(46.7%) 50(63.3%) 0.014 30 (50.0%) 0.810 Diabetes[counts(%)] 29(11.1%) 20(11.0%) 9(11.4%) 0.924 9(15.0%) 0.400 COPD[counts(%)] 57(21.8%) 38(20.9%) 19(24.1%) 0.569 8(13.8%) 0.169 CHF[counts(%)] 89(34.1%) 67(36.8%) 22(27.8%) 0.016 # 22(36.7%) 0.268 Active cancer[counts(%)] 26(10.0%) 20(11.0%) 6(7.6%) 0.400 11(18.3%) 0.067 Laboratory test D-dimer,mg/L [Median (IQR)] 6.1(3.1, 10.0) 6.5(3.5.4.2) 4.9(2.3,9.6) 0.196 6.2(4.0,10.1) 0.586 cTnI (+) b [counts(%)] 26 (10%) 19(10.4%) 7(8.7%) 0.696 8(13%) 0.444 NT-proBNP/BNP (+) c [counts(%)] 129 (49.4%) 90(49.5%) 39(49.4%) 0.998 37 (61.7%) 0.087 sPESI 0.363 0.087 = 0[counts(%)] 95(36.4%) 63(34.6%) 32(40.5%) 29(48.3%) ≥ 1[counts(%)] 166(63.6%) 119(65.4%) 47(59.5%) 31(51.7%) Open in a new tab COPD Chronic Obstructive Pulmonary Disease, CHF Chronic heart failure, IQR Interquartile Range, NT-proBNP N-Terminal pro-B-type Natriuretic Peptide, BNP B-type Natriuretic Peptide, sPESI Simple Pulmonary Embolism Severity Score # Fisher’s test a blood oxygen saturation < 90% without supplemental oxygen or < 95% with supplemental oxygen b > 0.05 pg/mL c For NT-proBNP, the threshold is age-dependent: 1800 pg/mL for patients > 75 years old, and 900 pg/mL for patients ≤ 75 years old. For BNP, the threshold is uniformly set at 100 pg/mL, with values above this threshold considered positive (+) The inter-observer agreement for histogram characteristics of the chest muscles was almost perfect (ICC: 0.80–0.88), while the agreement for the chest muscle area was substantial (ICC: 0.72–0.77) (see Table 2 ). Table 2. Inter-observer consistency of radiological features Features ICC (95% CI) Total Pectoral Muscle Area 0.75 (0.66, 0.82) PMA CT_PMA_Fat_ratio 0.80 (0.72,0.85) CT_PMA_10th 0.81 (0.75,0.86) CT_PMA_25th 0.85 (0.80,0.89) CT_PMA_50th 0.88 (0.83,0.91) CT_PMA_75th 0.87 (0.82,0.90) CT_PMA_90th 0.83 (0.77,0.88) PMI CT_PMI_Fat_ratio 0.82 (0.75,0.87) CT_PMI_10th 0.86 (0.81,0.90) CT_PMI_25th 0.83 (0.77,0.88) CT_PMI_50th 0.88 (0.83,0.91) CT_PMI_75th 0.87 (0.82,0.91) CT_PMI_90th 0.84 (0.78,0.88) Open in a new tab ICC Intraclass Correlation Coefficient Features selection A total of 16 radiological features related to the pectoral muscles, along with 15 clinical characteristics (excluding the sPESI score, as outlined in Table 1 ), were standardized and analyzed using fivefold cross-validated Lasso regression in the training set. This process identified 13 features with non-zero coefficients (Fig. 3 ). The Spearman correlation heatmap showed no strong correlations among the identified features (Fig. 4 ). Consequently, all 13 features were incorporated into the analysis, including low oxygen saturation, smoke status, CT_PMA_10th, CT_PMA_90th, CT_PMI_75th, CT_PMA_Fat_ratio, chest pain, syncope, diabetes, gender, chronic heart failure, NT-proBNP/BNP (+), and D-dimer levels. Fig. 3. Open in a new tab Feature selection and hyperparameter tuning for the LASSO regression model. A The trajectory of LASSO coefficients for all candidate variables. Each colored line represents the change in the coefficient estimate of one variable across a single cross-validation fold as the regularization penalty increases (i.e., as log(λ) decreases from left to right). Variables whose coefficients shrink to zero across folds are effectively excluded by the model, demonstrating the feature selection property of LASSO. B Ten-fold cross-validation curve based on binomial deviance. The red solid line indicates the mean cross-validation error across all folds, with the gray shaded area representing the standard error of the mean. The optimal lambda value is selected at the point where the mean error is minimized, balancing model complexity and predictive performance Fig. 4. Open in a new tab Spearman correlation heatmap for selected clinical and imaging features. CHF: Chronic heart failure; NT-proBNP: N-Terminal pro-B-type Natriuretic Peptide; BNP: B-type Natriuretic Peptide Development and validation of models In the training set, the ensemble models, RF and XGB, outperformed the SVC and LR models (Table 3 -a, Fig. 5 ). In the validation set, all four ML models demonstrated comparable AUC values (Table 3 -b, Fig. 5 ). However, in the external test set, the LR model consistently outperformed the other models across all performance metrics, achieving an AUC of 0.837 (95% CI: 0.696 to 0.977), an accuracy of 0.817(0.719–0.915), a recall of 0.750(0.505–0.995), a specificity of 0.833(0.728–0.939), a precision of 0.529(0.292–0.767), and an F1-score of 0.621(0.370–0.866) (Table 3 -c, Fig. 5 ). The DeLong test revealed that the AUC of the LR model was significantly higher than those of the RF and SVC; however, no statistically significant difference was observed when compared with the XGB ( Table 3 ) . Based on the calibration curves, which demonstrated that the LR model's curve was closest to the ideal diagonal on both the validation and test sets (Fig. 5 -c). Considering the factors above, we selected LR as the optimal model. Table 3. Results of the DeLong test between LR and other models Model AUC (95%CI) z p LR 0.837 (0.696–0.977) - - SVC 0.781 (0.553–0.891) 2.920 0.004** XGB 0.771 (0.619–0.906) 1.835 0.067 RF 0.731 (0.560–0.894) 2.332 0.020* Open in a new tab SVC Support Vector Classifier, LR Logistic Regression, RF Random Forest, XGB eXtreme Gradient Boosting Fig. 5. Open in a new tab The performance of various ML models and sPESI in predicting MV; ( a ) ROC curves for each model across the training, validation, and test datasets. b Performance metrics for each model across the three datasets. c Calibration curves of each model.SVC: Support Vector Classifier; LR: Logistic Regression; RF: Random Forest; XGB: eXtreme Gradient Boosting; sPESI: simplified Pulmonary Embolism Severity Index Furthermore, the performance of all ML models exceeded that of the sPESI across all datasets (Table 4 , Fig. 5 ). Table 4. Performance of various models and sPESI in classification Training set Models Recall(95%CI) Specificity (95%CI) Accuracy (95%CI) Precise (95%CI) F1-score (95%CI) SVC 0.686(0.532–0.840) 0.803(0.738–0.867) 0.780(0.720–0.840) 0.453(0.319–0.587) 0.545(0.399–0.691) LR 0.686(0.532–0.840) 0.796(0.731–0.861) 0.775(0.714–0.835) 0.444(0.312–0.577) 0.539(0.393–0.684) RF 0.971(0.916–1.000) 0.905(0.857–0.952) 0.918(0.878–0.958) 0.708(0.580–0.837) 0.819(0.710–0.911) XGB 0.914(0.822–1.000) 0.864(0.809–0.919) 0.874(0.825–0.922) 0.615(0.483–0.748) 0.736(0.608–0.856) sPESI 0.829(0.704–0.953) 0.381(0.302–0.459) 0.467(0.395–0.540) 0.242(0.165–0.318) 0.374(0.267–0.477) Validation set SVC 0.667(0.428–0.905) 0.797(0.698–0.895) 0.772(0.680–0.865) 0.435(0.232–0.637) 0.526(0.301–0.748) LR 0.733(0.510–0.957) 0.750(0.644–0.856) 0.747(0.651–0.843) 0.407(0.222–0.593) 0.524(0.309–0.732) RF 0.533(0.281–0.786) 0.812(0.717–0.908) 0.759(0.665–0.854) 0.400(0.185–0.615) 0.457(0.223–0.690) XGB 0.667(0.428–0.905) 0.781(0.680–0.883) 0.759(0.665–0.854) 0.417(0.219–0.614) 0.513(0.290–0.732) sPESI 0.867(0.695–1.000) 0.500(0.378–0.623) 0.570(0.460–0.679) 0.289(0.156–0.421) 0.433(0.255–0.593) Test set sSVC 0.583(0.304–0.862) 0.833(0.728–0.939) 0.783(0.679–0.888) 0.467(0.214–0.719) 0.519(0.251–0.784) LR 0.750(0.505–0.995) 0.833(0.728–0.939) 0.817(0.719–0.915) 0.529(0.292–0.767) 0.621(0.370–0.866) RF 0.500(0.217–0.783) 0.833(0.728–0.939) 0.767(0.660–0.874) 0.429(0.169–0.688) 0.462(0.190–0.732) XGB 0.500(0.217–0.783) 0.812(0.702–0.923) 0.750(0.640–0.860) 0.400(0.152–0.648) 0.444(0.179–0.709) sPESI 0.583(0.304–0.862) 0.500(0.359–0.641) 0.517(0.390–0.643) 0.226(0.079–0.373) 0.326(0.125–0.521) Open in a new tab SVC Support Vector Classifier, LR Logistic Regression, RF Random Forest, XGB eXtreme Gradient Boosting, sPESI Simplified Pulmonary Embolism Severity Index SHAP summary plots for all four models consistently identified low oxygen saturation, CT_PMA_10th, and CT_PMI_75th as highly important features (Fig. 6 ). Additionally, smoke and male gender were notably significant in the LR model. To illustrate the decision-making processes of the models, SHAP force plots were generated for two representative patients from the external test set. Fig. 6. Open in a new tab SHAP bar and beeswarm plots for various models: ( A ) LR model, ( B ) SVC model, ( C ) RF model, and ( D ) XGB model. E – F SHAP force plots for two patients in the external validation set using the LR model. Patient “E” experienced symptomatic improvement and was discharged from the hospital after receiving anticoagulant therapy. Patient “F” developed respiratory failure and required MV after admission. CHF: Chronic heart failure; NT-proBNP: N-Terminal pro-B-type Natriuretic Peptide; BNP: B-type Natriuretic Peptide For improved accessibility and to facilitate clinical implementation, the prediction model is publicly available as an online tool. You can use our model via the following link: http://6smv42w8qol6hvnt6b2y7r.streamlit.app/ . Discussion In this study, we developed four ML models to predict the need for MV during hospitalization in elderly patients with APE. Each model outperformed the sPESI. Among these, the LR model demonstrated the highest performance and robustness. Additionally, we used SHAP plots to visualize feature importance in the LR model. This analysis helped identify key predictors of MV requirements in elderly APE patients [ 24 ], which may help generate hypotheses for future studies. The results indicated that low oxygen saturation, CT_PMA_10th, CT_PMI_75th, and smoke history were among the more important predictors. To facilitate clinical application, we have deployed the LR model as a web-based interactive tool, which can be integrated into clinical workflow to assist in early risk stratification—when an elderly patient presents with acute pulmonary embolism, clinicians can input key predictors to promptly estimate the likelihood of mechanical ventilation requirement, thereby supporting timely clinical decision-making. Elderly individuals are not only more susceptible to APE but also face increased vulnerability due to multiple comorbidities, resulting in higher mortality rates compared to younger patients [ 25 ]. Our model aims to enhance healthcare outcomes for this demographic. Skeletal muscle atrophy, prevalent among the elderly, is associated with various respiratory diseases [ 16 , 26 – 28 ] and is characterized by reductions in both muscle area and attenuation on CT scans. Decreased pectoralis muscle density correlates with reduced hydration and increased intramuscular fat [ 29 ], potentially impairing respiratory function and increasing the risk of severe respiratory failure in elderly APE patients. Our study indicates that increased intermuscular fat is not a primary factor driving the need for MV in APE patients, as the fat ratio defined in our study had a negligible effect on the ML model's prediction, as evidenced by the SHAP plots. This aligns with existing literature, which shows that pathological muscle fibrosis is a significant cause of skeletal muscle atrophy and functional decline, in addition to the increase in intermuscular fat [ 30 , 31 ]. In patients with PE who develop severe respiratory failure, MV (invasive or non‑invasive) serves as an essential adjunctive therapy alongside thrombolysis, though it increases the risk of adverse events, including heart failure. [ 1 , 32 , 33 ].. Our developed model facilitates the rapid identification of such patients, enabling timely thrombolysis or high-flow oxygen therapy to potentially avoid the need for MV. Regarding model development, the sPESI score itself was excluded from feature selection to avoid multicollinearity, as its individual components were already included as separate variables. This allowed our ML algorithms to independently assess the importance of each raw clinical variable for predicting MV need. However, sPESI served as a critical performance benchmark due to its clinical relevance as a guideline-endorsed tool for risk stratification. The comparison highlights a key conceptual difference: unlike the sPESI's prescribed weighting for mortality prediction, our ML models employed a data-driven, learned weighting scheme optimized specifically for predicting MV requirement. While the sPESI score performs reasonably well in predicting mortality [ 34 ], its performance in predicting MV requirements is limited. Our research, demonstrating that a tailored, learned model can outperform this generalized clinical score for a specific critical outcome, can be integrated into a risk assessment system for APE. Thoracic muscle loss is increasingly recognized as an independent risk factor for impaired respiratory mechanics, reduced cough efficacy, delayed weaning, and ventilator dependency in aging populations and those with critical illness [ 35 – 37 ]. The impact of thoracic muscle loss extends beyond PE, influencing outcomes in various acute respiratory conditions [ 38 ]. For example, acute muscle wasting is associated with increased 90-day mortality in older adults with severe community-acquired pneumonia [ 39 ]. Low skeletal muscle mass has also been linked to severe in elderly patients undergoing non-small cell lung cancer surgery. This underscores that thoracic muscle metrics reflect a generalizable physiological reserve, influencing a wide range of acute and chronic respiratory conditions. Sarcopenia directly impairs respiratory muscle strength and endurance, leading to reduced vital capacity and weakened airway clearance ability. This thereby increases the risk of respiratory failure, potentially necessitating MV support. Multiple imaging modalities can be used to indicate sarcopenia [ 14 – 16 ]. CT-defined radiological features of the pectoralis muscle are associated with exacerbation or mortality in conditions such as pneumonia, lung cancer, or COPD [ 27 ]. Research by Meyer et al. indicated that the average density of the pectoralis muscle from CTPA images is an independent predictor of 30-day mortality in APE patients [ 13 ]. However, previous studies have rarely utilized non-contrast CT to assess chest muscle attenuation and area [ 16 , 25 – 27 ], and the impact of intramuscular arterioles during the contrast phase on pectoralis muscle attenuation remains unclear. Therefore, we utilized recent non-contrast CT images to identify appropriate parameters that characterize the degree of muscle degradation. Additionally, we introduced CT histogram analysis, including the percentiles of CT attenuation for the PMA and PMI in our ML model. Muscle density is not always uniformly distributed, especially in elderly patients (as shown in Fig. 2 ), and average density may not accurately reflect the degree of degeneration. Our study also found that the inter-observer agreement for the histogram characteristics of the chest muscles was higher than that for the chest muscle area, indicating that CT histogram features are more stable biological markers. In addition, low oxygen saturation, smoke, and male gender also rank highly in importance. During an episode of APE, ventilation-perfusion mismatch is almost inevitable [ 31 ], often resulting in reduced blood oxygen saturation. As the cardinal sign of hypoxemia [ 40 ], low oxygen saturation compromises cellular energy production. Oxygen is vital for aerobic respiration, the primary pathway for ATP synthesis [ 41 ]. Acute severe hypoxemia leads to a shift towards anaerobic metabolism, resulting in lactic acid accumulation and metabolic acidosis. This energy deficit directly impairs the function of highly metabolically active organs, including the respiratory muscles. It predicts future MV risk by promoting systemic inflammation, and cardiac strain, even increasing the likelihood of difficult weaning and prolonged ventilator dependence [ 42 ]. Smoke has been consistently linked with various respiratory diseases [ 43 ]. Long-term smokers may develop chronic airway inflammation [ 44 ], which can be exacerbated by APE, worsening their pre-existing chronic respiratory conditions [ 45 ]. Moreover, smoke can alter fibrin characteristics, making thrombi more resistant to lysis [ 46 ], potentially leading to anticoagulation failure and persistent ventilation/perfusion defects. Male gender is also a risk factor, although it may not be as significant as the previously mentioned factors. Recent studies have shown that, compared to female APE patients, males are more likely to require MV during hospitalization [ 47 ]. This trend is also observed in other respiratory conditions, such as acute exacerbation of COPD and community-acquired pneumonia [ 48 , 49 ], potentially related to differences in airway structure between genders. This suggests that, in managing elderly male patients with APE, greater attention should be given to the risk of ARF. Syncope and chest pain are pivotal symptoms in prognosticating APE [ 1 ]. Syncope, affecting 10–20% of patients, signals severe hemodynamic compromise from right ventricular failure and is strongly linked to elevated mortality, myocardial injury, and right ventricular dysfunction [ 50 – 52 ]. Chest pain, especially angina-like pain, suggests concurrent right ventricular ischemia due to increased myocardial oxygen demand and impaired supply [ 1 ]. In contrast, dyspnea is a common but nonspecific symptom arising from ventilation/perfusion mismatch [ 53 ]. Its presence in various cardiopulmonary conditions and weak association with specific PE severity limits its utility in risk stratification models. Therefore, both syncope and chest pain were included as candidate features in our model development process. Their considerable predictive importance was subsequently confirmed by high SHAP values in the final LR model. Limitation This study has several limitations. Firstly, to ensure clinical practicality, we extracted CT features of the pectoralis muscles only at the level of the fourth thoracic vertebra. A more comprehensive assessment utilizing the entire thoracic level would provide a full representation of muscle degeneration; however, this approach is both time-consuming and labor-intensive. Future investigations could benefit from integrating automated segmentation models based on deep learning to enhance this approach. Secondly, due to the constraints of retrospectively collected data, the majority of patients in this study did not have height measurements available for calculating the skeletal muscle index (SMI). Thirdly, although this study involves two centers, both are located in a southern province of China, which may introduce regional bias, and its sample size and diversity remained limited. This may lead to performance degradation when the model is applied to patients from other regions, different ethnic backgrounds, or distinct healthcare systems, indicating potential regional bias. Future research should focus on external validation of the model using broader, multi-center, multi-regional, and multinational cohorts to verify its generalizability. Another limitation of this study is the absence of established imaging severity scores, such as the Qanadli score or the right ventricle-to-left ventricle (RV/LV) diameter ratio, in the model development. Finally, variations in CT numbers across scanners from different manufacturers may limit the generalizability of the model. Conclusion Loss of pectoral muscle may be associated with the need for MV in elderly APE patients. The prediction model developed in this study, which incorporates this factor, could help identify high-risk individuals and may provide a foundation for future research aimed at optimizing early risk stratification and personalized management in this vulnerable population. Supplementary Information Supplementary Material 1. (17.1KB, docx) Acknowledgements Not applicable. Abbreviations APE Acute pulmonary embolism ARF Acute respiratory failure AUC Areas under the curve CHF Chronic heart failure COPD Chronic obstructive pulmonary disease ICC Intraclass Correlation Coefficient LR Logistic regression ML Machine learning MV Mechanical ventilation PMA Pectoralis major PMA_area The area of pectoralis major PMI Pectoralis minor PMI_area The area of pectoralis minor RF Random forest SVC Support Vector Classifier SHAP SHapley Additive exPlanations sPESI Simplified Pulmonary Embolism Severity Index XGB Extreme Gradient Boosting CTPA Computed Tomography Pulmonary Angiography CT_PMA_Fat_Ratio The intermuscular fat ratio of the PMA CT_PMI_Fat_Ratio The intermuscular fat ratio of the PMI CT_PMA_10 th The 10th percentile of the CT attenuation values of PMA CT_PMI_10 th The 10th percentile of the CT attenuation values of PMI CT_PMA_25 th The 25th percentile of the CT attenuation values of PMA CT_PMI_25 th The 25th percentile of the CT attenuation values of PMIss CT_PMA_50 th The 50th percentile of the CT attenuation values of PMA CT_PMI_50 th The 50th percentile of the CT attenuation values of PMI CT_PMA_75 th The 75th percentile of the CT attenuation values of PMA CT_PMI_75 th The 75th percentile of the CT attenuation values of PMI CT_PMA_90 th The 90th percentile of the CT attenuation values of PMA CT_PMI_90 th The 90th percentile of the CT attenuation values of PMI Authors’ contributions (I) Conception and design: L.G.; Z.D.; W.D. (II) Administrative support: L.G.; H.R. (III) Provision of study materials or patients: L.G.; Z.D.; D.L.; X.Z. (IV) Collection and assembly of data: L.G.; Z.D.; D.L.; X.Z.; J.Z. (V) Data analysis and interpretation: L.G.; Z.D.; J.Z.; D.L.; W.D. (VI) Manuscript writing: L.G.; Z.D.; D.L.; J.Z.; W.D. (VII) Final approval of manuscript: L.G.; J.Z.; H.R.; D.L.; Z.D.; X.Z.; W.D. Funding Science and Technology Plan of Jiangxi Provencal Health Commission (202310508) and Internal Funding Project of the Second Affiliated Hospital, Jiangxi Medical College, Nanchang University (2024efyB04) supported this study supported this study. Data availability The datasets used and/or analyzed during the current study not publicly available due to protect participant privacy but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This retrospective study was conducted at The Second Affiliated Hospital of Nanchang University and was approved by its Institutional Review Board (IRB) on October 13, 2023, which also waived the requirement for informed consent. The study protocol was registered and filed with the National Medical Research Registration and Filing Information System (approval ID: MR-36–24-042480). Written informed consent was waived by the Institutional Review Board (IRB). Consent for publication Personal information about the study participants will not be published. 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. Zexiang Deng and Dan Luo contributed equally to this work. Contributor Information Jingjing Zhou, Email: [email protected]. Lianggeng Gong, Email: [email protected]. References 1. Konstantinides SV, Meyer G, Becattini C, Bueno H, Geersing GJ, Harjola VP, et al. 2019 ESC Guidelines for the diagnosis and management of acute pulmonary embolism developed in collaboration with the European Respiratory Society (ERS). EUR HEART J. 2020;41:543–603. 10.1093/eurheartj/ehz405. [ DOI ] [ PubMed ] [ Google Scholar ] 2. Zhai Z, Wang D, Lei J, Yang Y, Xu X, Ji Y, Yi Q, Chen H, Hu X, Liu Z, Mao Y, Zhang J, Shi J, Zhang Z, Wu S, Gao Q, Tao X, Xie W, Wan J, Zhang Y, Zhang S, Zhen K, Zhang Z, Fang B, Wang C. Trends in risk stratification, in-hospital management and mortality of patients with acute pulmonary embolism: an analysis from the China pUlmonary thromboembolism REgistry Study (CURES). EUR RESPIR J. 2021; 58. 10.1183/13993003.02963-2020. [ DOI ] [ PubMed ] 3. Brink A, Elf J, Svensson PJ, Engstrom G, Melander O, Zoller B. Sex-specific risk factors for deep venous thrombosis and pulmonary embolism in a population-based historical cohort study of middle-aged and older individuals. J Am Heart Assoc. 2023;12:e27502. 10.1161/JAHA.122.027502. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Spirk D, Husmann M, Hayoz D, Baldi T, Frauchiger B, Engelberger R, et al. Predictors of in-hospital mortality in elderly patients with acute venous thrombo-embolism: the SWIss Venous ThromboEmbolism Registry (SWIVTER). Eur Heart J. 2012;33:921–6. 10.1093/eurheartj/ehr392. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Miyazaki S, Tamaki A, Wakabayashi H, Arai H. Definition, diagnosis, and treatment of respiratory sarcopenia. Curr Opin Clin Nutr Metab Care. 2024;27:210–8. 10.1097/MCO.0000000000001003. [ DOI ] [ PubMed ] [ Google Scholar ] 6. Kera T, Kawai H, Ejiri M, Ito K, Hirano H, Fujiwara Y, et al. Respiratory sarcopenia is a predictor of all-cause mortality in community-dwelling older adults-the Otassha Study. J Cachexia Sarcopenia Muscle. 2023;14:1894–9. 10.1002/jcsm.13266. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Mann CJ, Perdiguero E, Kharraz Y, Aguilar S, Pessina P, Serrano AL, et al. Aberrant repair and fibrosis development in skeletal muscle. Skelet Muscle. 2011;1:21. 10.1186/2044-5040-1-21. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Zhou K, Wu F, Zhao N, Zheng Y, Deng Z, Yang H, Wen X, Xiao S, Yang C, Chen S, Zhou Y, Ran P. Association of pectoralis muscle area on computed tomography with airflow limitation severity and respiratory outcomes in COPD: A population-based prospective cohort study. Pulmonology. 2023. 10.1016/j.pulmoe.2023.02.004. [ DOI ] [ PubMed ] 9. Sato S, Miyazaki S, Tamaki A, Yoshimura Y, Arai H, Fujiwara D, et al. Respiratory sarcopenia: a position paper by four professional organizations. Geriatr Gerontol Int. 2023;23:5–15. 10.1111/ggi.14519. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Besutti G, Pellegrini M, Ottone M, Cantini M, Milic J, Bonelli E, et al. The impact of chest CT body composition parameters on clinical outcomes in COVID-19 patients. PLoS ONE. 2021;16:e251768. 10.1371/journal.pone.0251768. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Surov A, Kardas H, Besutti G, Pellegrini M, Ottone M, Onur MR, et al. Prognostic role of the pectoralis musculature in patients with COVID-19. A multicenter study. Acad Radiol. 2023;30:77–82. 10.1016/j.acra.2022.05.003. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Nakagawara K, Shiraishi Y, Chubachi S, Tanabe N, Maetani T, Asakura T, et al. Integrated assessment of computed tomography density in pectoralis and erector spinae muscles as a prognostic biomarker for coronavirus disease 2019. Clin Nutr. 2024;43:815–24. 10.1016/j.clnu.2024.02.004. [ DOI ] [ PubMed ] [ Google Scholar ] 13. Meyer HJ, Kardas H, Schramm D, Bär C, Wienke A, Borggrefe J, et al. CT-defined pectoralis muscle mass and muscle density are associated with mortality in acute pulmonary embolism. A multicenter analysis. Clin Nutr. 2023;42:1036–40. 10.1016/j.clnu.2023.04.022. [ DOI ] [ PubMed ] [ Google Scholar ] 14. Peng F, Xu H, Song Y, Xu K, Li S, Cai X, et al. Longitudinal study of multi-parameter quantitative magnetic resonance imaging in Duchenne muscular dystrophy: hyperresponsiveness of gluteus maximus and detection of subclinical disease progression in functionally stable patients. J Neurol. 2023;270:1439–51. 10.1007/s00415-022-11470-8. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Erley J, Roedl K, Ozga AK, de Heer G, Schubert N, Breckow J, Burdelski C, Tahir E, Kluge S, Huber TB, Yamamura J, Adam G, Molwitz I. Dual-Energy CT muscle fat fraction as a new imaging biomarker of body composition and survival predictor in critically ill patients. Eur Radiol. 2024. 10.1007/s00330-024-10779-4. [ DOI ] [ PMC free article ] [ PubMed ] 16. Hocaoglu E, Ors S, Yildiz O, Inci E. Correlation of pectoralis muscle volume and density with severity of COVID-19 pneumonia in adults. Acad Radiol. 2021;28:166–72. 10.1016/j.acra.2020.11.017. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Bikdeli B, Wang Y, Jimenez D, Parikh SA, Monreal M, Goldhaber SZ, et al. Pulmonary embolism hospitalization, readmission, and mortality rates in US older adults, 1999-2015. JAMA J Am Med Assoc. 2019;322:574–6. 10.1001/jama.2019.8594. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Jiang T, Lin T, Shu X, Song Q, Dai M, Zhao Y, Huang L, Tu X, Yue J. Prevalence and prognostic value of preexisting sarcopenia in patients with mechanical ventilation: a systematic review and meta-analysis. CRIT CARE. 2022; 26. 10.1186/s13054-022-04015-y. [ DOI ] [ PMC free article ] [ PubMed ] 19. Ross HJ, Peikari M, Vishram-Nielsen J, Fan CS, Hearn J, Walker M, et al. Predicting heart failure outcomes by integrating breath-by-breath measurements from cardiopulmonary exercise testing and clinical data through a deep learning survival neural network. Eur Heart J Digit Health. 2024;5:324–34. 10.1093/ehjdh/ztae005. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Pancani R, Villari L, Aquilini F, Palla A, Carrozzi L, Celi A. Prognostic role of respiratory failure in acute pulmonary embolism: a prospective multicenter study. Thromb Res. 2022;217:33–5. 10.1016/j.thromres.2022.07.002. [ DOI ] [ PubMed ] [ Google Scholar ] 21. Vedovati MC, Cimini LA, Pierpaoli L, Vanni S, Cotugno M, Pruszczyk P, et al. Prognostic value of respiratory index in haemodynamically stable patients with acute pulmonary embolism: the Respiratory Index model study. Eur Heart J Acute Cardiovasc Care. 2020;9:286–92. 10.1177/2048872620913849. [ DOI ] [ PubMed ] [ Google Scholar ] 22. Sabatino A, Avesani CM, Regolisti G, Adinolfi M, Benigno G, Delsante M, et al. Sarcopenic obesity and its relation with muscle quality and mortality in patients on chronic hemodialysis. Clin Nutr. 2023;42:1359–68. 10.1016/j.clnu.2023.06.032. [ DOI ] [ PubMed ] [ Google Scholar ] 23. Nohara Y, Matsumoto K, Soejima H, Nakashima N. Explanation of machine learning models using shapley additive explanation and application for real data in hospital. Comput Methods Programs Biomed. 2022;214:106584. 10.1016/j.cmpb.2021.106584. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Pauley E, Orgel R, Rossi JS, Strassle PD. Age-stratified national trends in pulmonary embolism admissions. Chest. 2019;156:733–42. 10.1016/j.chest.2019.05.021. [ DOI ] [ PubMed ] [ Google Scholar ] 25. Kaluźniak-Szymanowska A, Krzymińska-Siemaszko R, Deskur-Śmielecka E, Lewandowicz M, Kaczmarek B, Wieczorowska-Tobis K. Malnutrition, Sarcopenia, and Malnutrition-Sarcopenia Syndrome in Older Adults with COPD. NUTRIENTS. 2021; 14. 10.3390/nu14010044. [ DOI ] [ PMC free article ] [ PubMed ] 26. van Bakel S, Gietema HA, Stassen PM, Gosker HR, Gach D, van den Bergh JP, et al. CT scan-derived muscle, but not fat, area independently predicts mortality in COVID-19. Chest. 2023;164:314–22. 10.1016/j.chest.2023.02.048. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Sun C, Anraku M, Kawahara T, Karasaki T, Kitano K, Nagayama K, et al. Respiratory strength and pectoralis muscle mass as measures of sarcopenia: relation to outcomes in resected non-small cell lung cancer. J Thorac Cardiovasc Surg. 2022;163:779–87. 10.1016/j.jtcvs.2020.10.133. [ DOI ] [ PubMed ] [ Google Scholar ] 28. Pasco JA, Sui SX, West EC, Anderson KB, Rufus-Membere P, Tembo MC, et al. Fatty liver index and skeletal muscle density. Calcif Tissue Int. 2022;110:649–57. 10.1007/s00223-021-00939-9. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Lemos DR, Babaeijandaghi F, Low M, Chang CK, Lee ST, Fiore D, et al. Nilotinib reduces muscle fibrosis in chronic muscle injury by promoting TNF-mediated apoptosis of fibro/adipogenic progenitors. Nat Med. 2015;21:786–94. 10.1038/nm.3869. [ DOI ] [ PubMed ] [ Google Scholar ] 30. Zoico E, Corzato F, Bambace C, Rossi AP, Micciolo R, Cinti S, et al. Myosteatosis and myofibrosis: relationship with aging, inflammation and insulin resistance. Arch Gerontol Geriatr. 2013;57:411–6. 10.1016/j.archger.2013.06.001. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Santolicandro A, Prediletto R, Fornai E, Formichi B, Begliomini E, Giannella-Neto A, et al. Mechanisms of hypoxemia and hypocapnia in pulmonary embolism. Am J Respir Crit Care Med. 1995;152:336–47. 10.1164/ajrccm.152.1.7599843. [ DOI ] [ PubMed ] [ Google Scholar ] 32. Friedman O, Horowitz JM, Ramzy D. Advanced cardiopulmonary support for pulmonary embolism. Tech Vasc Interv Radiol. 2017;20:179–84. 10.1053/j.tvir.2017.07.007. [ DOI ] [ PubMed ] [ Google Scholar ] 33. Lee ES, Baltsen CD, Stubblefield WB, Granfeldt A, Andersen A, Stannek K, et al. Intubation and mechanical ventilation in patients with acute pulmonary embolism: a scoping review. J Intensive Care Med. 2026;41:3–11. 10.1177/08850666241285862. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Jiménez D, Aujesky D, Moores L, Gómez V, Lobo JL, Uresandi F, et al. Simplification of the pulmonary embolism severity index for prognostication in patients with acute symptomatic pulmonary embolism. Arch Intern Med. 2010;170:1383–9. 10.1001/archinternmed.2010.199. [ DOI ] [ PubMed ] [ Google Scholar ] 35. Sergeeva VA, Runikhina NK. Respiratory sarcopenia: aspects of pathogenesis, approaches to diagnosis. Pulmonologiia. 2024;34:869–78. 10.18093/0869-0189-2024-4271. [ Google Scholar ] 36. Morisawa T, Kunieda Y, Koyama S, Suzuki M, Takahashi Y, Takakura T, et al. The relationship between sarcopenia and respiratory muscle weakness in community-dwelling older adults. Int J Environ Res Public Health. 2021;18:13257. 10.3390/ijerph182413257. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Moon SW, Kim SY, Choi JS, Leem AY, Lee SH, Park MS, Kim YS, Chung KS. Thoracic Skeletal Muscle Quantification Using Computed Tomography And Prognosis Of Elderly ICU Patients. Research Square. 2021. 10.21203/rs.3.rs-144315/v1. [ DOI ] [ PMC free article ] [ PubMed ] 38. Yoshida Y, Taniyama M, Fujihara D, Yamada A, Fukui M, Doi N, Takahashi R, Okayama S, Iwama H. The relationship between sarcopenia and respiratory muscle weakness undergoing phase 2 cardiac rehabilitation in elder patients. EUR HEART J. 2025;46(1):ehaf784.3930. 10.1093/eurheartj/ehaf784.3930 39. Shang N, Li Q, Ji W, Liu H, Guo S. Acute muscle wasting is associated with poor prognosis in older adults with severe community-acquired pneumonia. Eur Geriatr Med. 2023;15:73–82. 10.1007/s41999-023-00895-7. [ DOI ] [ PubMed ] [ Google Scholar ] 40. Lagina M, Valley TS. Diagnosis and management of acute respiratory failure. Crit Care Clin. 2024;40:235–53. 10.1016/j.ccc.2024.01.002. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Le Page LM, Rider OJ, Lewis AJ, Noden V, Kerr M, Giles L, Ambrose LJA, Ball V, Mansor L, Heather LC, Tyler DJ. Assessing the effect of hypoxia on cardiac metabolism using hyperpolarized 13 C magnetic resonance spectroscopy. NMR BIOMED. 2019; 32. 10.1002/nbm.4099. [ DOI ] [ PMC free article ] [ PubMed ] 42. Sterling LH, Fernando SM, Lawler PR, Price S, Fan E, Goligher E, et al. Navigating hypoxemic respiratory failure in critically ill cardiac patients. JACC: Advances. 2025;4:101616. 10.1016/j.jacadv.2025.101616. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Chan KH, Wright N, Xiao D, Guo Y, Chen Y, Du H, et al. Tobacco smoking and risks of more than 470 diseases in China: a prospective cohort study. Lancet Public Health. 2022;7:e1014–26. 10.1016/S2468-2667(22)00227-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Che KF, Tufvesson E, Tengvall S, Lappi-Blanco E, Kaarteenaho R, Levänen B, et al. The neutrophil-mobilizing cytokine interleukin-26 in the airways of long-term tobacco smokers. Clin Sci. 2018;132:959–83. 10.1042/CS20180057. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Bertoletti L, Quenet S, Mismetti P, Hernández L, Martín-Villasclaras JJ, Tolosa C, et al. Clinical presentation and outcome of venous thromboembolism in COPD. Eur Respir J. 2012;39:862–8. 10.1183/09031936.00058811. [ DOI ] [ PubMed ] [ Google Scholar ] 46. Undas A, Topór-Madry R, Tracz W, Pasowicz M. Effect of cigarette smoking on plasma fibrin clot permeability and susceptibility to lysis. Thromb Haemost. 2009;102:1289–91. 10.1160/TH09-03-0187. [ DOI ] [ PubMed ] [ Google Scholar ] 47. Sathianathan S, Meili Z, Romero CM, Juarez JJ, Bashir R. Racial and gender disparities in the management of acute pulmonary embolism. J Vasc Surg Venous Lymphat Disord. 2024;12:101817. 10.1016/j.jvsv.2024.101817. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. De-Miguel-Díez J, Jiménez-García R, Hernández-Barrera V, Zamorano-Leon JJ, Villanueva-Orbaiz R, Albaladejo-Vicente R, et al. Trends in mechanical ventilation use and mortality over time in patients receiving mechanical ventilation in Spain from 2001 to 2015. Eur J Intern Med. 2020;74:67–72. 10.1016/j.ejim.2019.11.023. [ DOI ] [ PubMed ] [ Google Scholar ] 49. de Miguel-Díez J, Jiménez-García R, Hernández-Barrera V, Puente-Maestu L, Ji Z, de Miguel-Yanes JM, et al. Ventilatory support use in hospitalized patients with community-acquired pneumonia. Fifteen-year trends in Spain (2001-2015). Arch Bronconeumol. 2020;56:792–800. 10.1016/j.arbres.2019.12.008. [ DOI ] [ PubMed ] [ Google Scholar ] 50. de Winter MA, van Bergen EDP, Welsing PMJ, Kraaijeveld AO, Kaasjager KHAH, Nijkeuter M. The prognostic value of syncope on mortality in patients with pulmonary embolism: a systematic review and meta-analysis. Ann Emerg Med. 2020;76:527–41. 10.1016/j.annemergmed.2020.03.026. [ DOI ] [ PubMed ] [ Google Scholar ] 51. Zhang S, Xu X, Ji Y, Yang Y, Yi Q, Chen H, Hu X, Liu Z, Mao Y, Zhang J, Shi J, Lei J, Wang D, Zhang Z, Wu S, Gao Q, Tao X, Xie W, Wan J, Zhang Y, Zhang M, Shao X, Zhang Z, Fang B, Yang P, Zhai Z, Wang C, Undefined U. Clinical Phenotypes With Prognostic Implications in Pulmonary Embolism Patients With Syncope. Front Cardiovasc Med. 2022; 9. 10.3389/fcvm.2022.836850. [ DOI ] [ PMC free article ] [ PubMed ] 52. Natanzon SS, Beigel R, Chernomordik F, Mazin I, Herscovici R, Goitein O, Ben-Zekry S, Grupper A, Matetzky S. P6460The significance of syncope as a presenting symptom in patients with intermediate risk pulmonary embolism. EUR HEART J. 2019; 40. 10.1093/eurheartj/ehz746.1052. 53. Miniati M, Cenci C, Monti S, Poli D. Clinical presentation of acute pulmonary embolism: survey of 800 cases. PLoS ONE. 2012. 10.1371/journal.pone.0030891. [ 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. (17.1KB, docx) Data Availability Statement The datasets used and/or analyzed during the current study not publicly available due to protect participant privacy but are available from the corresponding author on reasonable request. Articles from BMC Geriatrics are provided here courtesy of BMC ACTIONS View on publisher site PDF (4.2 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top