Development and evaluation of a clinical prediction model for in-hospital mortality in patients with acute exacerbation of chronic obstructive pulmonary disease

Introduction

Chronic obstructive pulmonary disease (COPD) is a common, preventable, and treatable heterogeneous lung disease characterized by persistent, progressive respiratory symptoms (dyspnea, cough, sputum production, and/or acute exacerbations with airflow limitation, typically triggered by exposure to large amounts of harmful particles or gases) due to abnormalities in the airways and/or alveoli (1). According to the World Health Organization noted, COPD was the third leading cause of death globally as of March 2023, was responsible for 3.23 million deaths in 2019, and is the seventh leading cause of poor health worldwide. Patients with COPD experience 0.5 to 3.5 acute exacerbations annually. These exacerbations are influenced by numerous factors and are also closely related to inadequate early prevention and standardized treatment (2). Acute exacerbation of chronic obstructive pulmonary disease (AECOPD) refers to a sudden worsening of respiratory symptoms (worsening dyspnea and/or cough, increased sputum production), which may be accompanied by tachypnea and/or tachycardia, necessitating additional treatment (3).

Forced expiratory volume in 1 second (FEV1) and body mass index (BMI) are well-established indicators for assessing COPD severity and long-term prognosis and are commonly featured as risk factors in clinical models for predicting long-term survival outcomes in stable patients with COPD (4,5). However, FEV1 does not accurately reflect COPD severity in patients with AECOPD, thus presenting limitations in the prediction of short-term outcomes for these patients (6).

The 2023 Global Initiative for Chronic Obstructive Lung Disease (GOLD) report states that the goals of treating AECOPD include minimizing the negative impact of the exacerbation and preventing similar subsequent events. Early prevention, timely detection, and active interventional treatment can significantly reduce readmission rates and mortality among patients with COPD, markedly decrease the years of life lost, and substantially improve patients’ quality of life after discharge. Therefore, a crucial task during treatment is the early identification of high-risk patients with poor prognosis following an AECOPD, the active adjustment of treatment strategies, and the improvement of short-term outcomes (7). Numerous clinical prognostic models have been developed primarily for stable COPD populations, such as the Dyspnea, Eosinopenia, Consolidation, Acidemia, and Atrial Fibrillation (DECAF) score, Community-Acquired Pneumonia Severity Score (CAPS), Body mass index, airflow Obstruction, Dyspnea, and Exercise capacity (BODE) index, National Early Warning Score 2 (NEWS2) (8), and Age, Dyspnea, and airflow Obstruction (ADO) index, which are primarily used to predict readmission risk and long-term mortality risk in stable patients (9-11). However, the DECAF and Blood urea nitrogen, Age, Pressure (systolic), 65 years old and above (BAP-65) Score are currently the most widely recognized prognostic models for AECOPD patients. Additionally, the predictive performance of these models when combined with AECOPD, age, sex, partial pressure of carbon dioxide (PaCO₂), lactate level, creatinine level, high-sensitivity troponin T (hscTnT) level, heart rate, and diastolic blood pressure remains unclear (12). Consequently, the purpose of this study was to integrate current research on laboratory indicators, investigate their value in predicting short-term prognosis in patients with AECOPD, construct a clinical prediction model for the short-term prognosis of patients with AECOPD, and evaluate the model’s predictive discrimination and calibration. The overall aim of this work is to provide a reliable tool for the clinical prediction of short-term prognosis outcomes in patients with AECOPD. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0166/rc).

Methods

Study participants

A retrospective analysis was conducted on 297 patients with AECOPD admitted to the Emergency Department and Respiratory and Critical Care Medicine Department of The Sixth People’s Hospital of Nantong between February 2022 and February 2025. After two patients with acute severe stroke and six patients with advanced/end-stage lung cancer were excluded, 289 patients were ultimately included in the analysis.

The inclusion criteria were as follows: (I) meeting the GOLD guideline diagnostic criteria for AECOPD (13); (II) a diagnosis of AECOPD in the Emergency Department or the Respiratory and Critical Care Medicine Department of The Sixth People’s Hospital of Nantong (if patients were admitted via the outpatient clinic, the admission time was calculated from the day of the outpatient or emergency visit); and (III) complete clinical data. Meanwhile, the exclusion criteria were as follows: (I) accompanying severe cardiac, hepatic, or renal insufficiency; (II) presence of malignant tumors or autoimmune diseases; and (III) hospitalization for reasons other than AECOPD. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Affiliated Hospital of Nantong University (No. 2025-L094) and informed consent was taken from all the patients.

Data collection

In this retrospective cohort study, all data were extracted from the hospital’s electronic medical record system and collected through use of standardized case report forms. Preset data variables included baseline characteristics, comorbidities, and comprehensive laboratory and functional examination results upon admission. The primary outcome of the study was all-cause mortality during hospitalization.

Handling of missing data

Among the 56 candidate included predictor variables, a portion had missing values. Analysis revealed that the missingness patterns for variables such as nasal cannula flow rate and carcinoembryonic antigen were significantly associated with patients’ clinical status (e.g., age, inflammatory markers, and PaCO₂), indicating nonrandom missingness.

To minimize selection bias and preserve statistical power, multiple imputation was applied via chained equations to handle all missing data. Subsequent statistical analyses were performed on the completed datasets generated after multiple imputation.

Statistical analysis and model construction

All statistical analyses were performed with R language (The R Foundation for Statistical Computing, Vianna, Austria). Continuous variables are expressed as the mean ± standard deviation, and intergroup comparisons of these variables were conducted via the t-test; meanwhile, categorical variables are expressed as frequency and percentage, and intergroup comparisons of these variables were conducted with the Chi-squared test.

Model construction was divided into two stages: variable screening and model building. For variable screening, 33 variables with P<0.1 in the univariate analysis were included in the candidate pool. Subsequently, highly correlated variables were excluded based on clinical redundancy, and variables with severe multicollinearity, specifically, a variance inflation factor (VIF) >5, were excluded; ultimately 18 variables were retained for the final model.

For model building, given the limited sample size and the relatively small number of outcome events, Firth’s penalized maximum likelihood estimation method was used for logistic regression analysis to avoid model overfitting and the “quasi-complete separation” problem. During modeling, core clinical variables (age and gender) were forced into the model, and a backward stepwise method based on the profile likelihood ratio test was used to select the remaining independent predictors.

The final constructed prediction model included eight variables: age, gender, PaCO₂, lactate, creatinine, hscTnT, heart rate, and diastolic blood pressure.

Model validation and evaluation

K-fold cross-validation was used for rigorous internal validation of the model’s performance. The whole dataset was randomly divided into five subsets. Iteratively, four subsets were used as the training set to build the model, and the remaining subset was used as the validation set for evaluation. This process was repeated five times, and the average of the performance metrics was recorded. Model performance was evaluated in terms of discrimination and calibration. Discrimination was assessed by plotting of the receiver operating characteristic curve and calculation of the area under the curve (AUC). Calibration, the agreement between predicted probabilities and actual observed risk, was assessed via calibration curves and the Brier score.

Results

Baseline characteristics of patients

The study initially collected data from 289 patients. After data cleaning, which involved excluding two patients with missing outcome information, the final study cohort consisted of 287 patients. Among these, 24 (8.36%) patients died and 263 (91.64%) showed improvement. The mean age of the improved group was 76.17±8.06 years, and the mean age of the death group was 74.29±8.30 years. Univariate analysis revealed statistically significant differences (P<0.05) between the death group and the improvement group among 33 variables. These variables included BMI, white blood cell count, neutrophil percentage, eosinophil count (14), high-sensitivity C-reactive protein (hs-CRP) levels, liver function markers [total bilirubin, alanine aminotransferase (ALT) levels, aspartate aminotransferase (AST) levels, lactate dehydrogenase (LDH)/albumin ratio (15)], renal function markers (urea and creatinine levels), blood glucose levels, arterial blood gas parameters (pH, PaCO₂, and lactate levels) (16), myocardial injury markers (hscTnT and N-terminal B-type natriuretic peptide precursor), inflammatory markers (procalcitonin and D-dimer levels), vital signs (heart rate, respiratory rate, and oxygen saturation), and comorbidities [cardiac insufficiency, pulmonary encephalopathy, cor pulmonale, and type 2 diabetes mellitus (17)]. Detailed results are presented in Table 1.

Handling of missing data

To construct the prediction model, this study first systematically screened candidate variables. Variables containing identification information, those with poor data quality, and those with a missing rate exceeding 50% were excluded, resulting in 56 variables being included for subsequent analysis. For a portion of the variables, there were missing values, with the missing rates for nasal cannula flow (49.5%), carcinoembryonic antigen (34.5%), immunoglobulin E (25.4%), buffer base (16.4%), and glucose (11.1%) all exceeding 10%.

Analysis revealed that the missingness patterns had systematic characteristics. Missingness in the nasal cannula flow variable was associated with younger age (r=−0.25), higher PaCO₂ (r=0.13), and mortality outcome (r=0.15), reflecting the clinical practice in which critically ill patients are more likely to receive advanced respiratory support rather than standard nasal cannula oxygen. Similarly, missingness in the carcinoembryonic antigen variable was negatively correlated with inflammatory markers, suggesting that physicians tend to order more comprehensive tests for patients with more severe inflammatory responses (Figures 1-3). These findings confirm that the data were not missing completely at random but were related to the patients’ clinical status.

Figure 1 Relationship between missing NC flow (L/min) data and age (years). NC, nasal cannula.

Figure 2 Relationship between missing CEA (ng/mL) and high-sensitivity CRP level. CEA, carcinoembryonic antigen; CRP, C-reactive protein.

Figure 3 Relationship between missing IgE (IU/mL) data and PaCO₂ level. IgE, immunoglobulin E; PaCO₂, partial pressure of arterial carbon dioxide.

To avoid the selection bias that could arise from a complete-case analysis, multiple imputation by chained equations was employed to handle the missing data. All subsequent statistical analyses were conducted on the imputed complete datasets to ensure the robustness of the results.

Finally, with in-hospital mortality serving as the outcome variable, eight variables—age, gender, PaCO₂, lactate, creatinine, hscTnT, heart rate, and diastolic blood pressure—were incorporated. Firth’s penalized logistic regression was used to build the prediction model in order to address potential model instability issues due to the small sample size.

Model construction

Thirty-three variables with a P value less than 0.1 from the univariate analysis were included in the initial candidate pool. To address collinearity among variables and enhance the model’s clinical utility, the candidate pool was preprocessed. The processing strategy included (I) removing variables with high clinical redundancy (e.g., retaining only the total white blood cell count over its subtypes) and (II) prioritizing the retention of more objective and quantitative biomarkers (e.g., retaining N-terminal pro-B-type natriuretic peptide over the diagnosis of cardiac insufficiency). Concurrently, variables with a VIF >5 were iteratively identified and removed to eliminate multicollinearity. This screening process resulted in 18 variables being retained for the final model building.

Firth penalty logistic regression analysis (a backward stepwise selection method based on likelihood ratio test) was used to screen the variables with significant independent predictive value from the remaining candidate variables. This process identified 13 variables: BMI, white blood cell count, total bilirubin level, creatinine level, glucose level, PaCO₂, lactate level, hscTnT level, procalcitonin level, heart rate, respiratory rate, oxygen saturation, and diastolic blood pressure.

The final multivariate analysis established an eight-variable model for predicting in-hospital mortality in patients with AECOPD. The results indicated that age, gender, PaCO₂ (mmHg), lactate level (mmol/L), creatinine level (µmol/L), hscTnT (ng/mL), heart rate (beats/min), and diastolic blood pressure (mmHg) were independent risk factors for mortality in patients with AECOPD. The basic characteristics and parameters of the variables in the final model are detailed in Table 2.

Based on the aforementioned eight independent risk factors, a clinical prediction model was constructed. The model parameters are shown in Table 2. According to the binary logistic regression function, where P represents the probability of in-hospital mortality outcomes in AECOPD patients, β0∼βp and others are the regression coefficients of variables in the model, and x1,……,xp, are the independent variables. Y=β0+β1x1+…+βpxp, after functional transformation, the prediction equation is: P = 1/[1 + e^(−Y)]. The calculation formula for Y is as follows: Y = −0.014 × (age) − 2.167 × (gender) + 0.028 × (creatinine) + 0.041 × (PaCO₂)² + 0.641 × (lactate) + 0.003 × (hscTnT) + 0.023 × (heart rate) − 0.007 × (diastolic blood pressure) − 9.3543. The above formula can be calculated by directly substituting the values of each indicator, and its units must be consistent with those in Table 2. Here, e represents the natural logarithm base.

Multicollinearity diagnostics indicated that the VIF for each independent variable in the final model was less than 5, and the tolerance was greater than 0.1 (Table 3), indicating no severe multicollinearity issues in the model.

However, during the diagnostic process, the standard logistic regression model indicated the presence of (quasi-)complete separation in the data. Given the relatively small sample size of this study (n=280), this phenomenon was expected. It typically occurs when certain strong predictor variables can perfectly or nearly perfectly distinguish the minority outcome events (e.g., death). This further supported the rationale for our choice of Firth’s penalized logistic regression as the final analytical method, as it is specifically designed to address such data separation issues caused by small sample sizes.

To enhance the clinical utility of the model, a nomogram was developed for visual representation of the final model, as detailed in Figure 4. Using this nomogram, clinicians can quickly and intuitively estimate a patient’s corresponding risk probability of in-hospital mortality by plotting points on the graph based on the patient’s eight indicator values and summing the total points.

Figure 4 Nomogram of the clinical prediction model for prognostic outcomes in patients with acute exacerbation of chronic obstructive pulmonary disease. BP, blood pressure; hs-cTnT, high-sensitivity troponin T; PaCO₂, partial pressure of arterial carbon dioxide.

Validation of the prediction model

The model’s discrimination was evaluated via AUC analysis, and the results indicated that the model achieved a mean AUC of 0.9385 [95% confidence interval (CI): 0.9375–0.9396] for internal validation.

The model’s calibration was assessed via calibration curves and the Brier score. During internal validation, the model’s Brier score was 0.037. A value this close to 0 indicates good agreement between the predicted probabilities and the actual observed outcomes (Figure 5). In the calibration curve (Figure 6), the model’s predicted probability curve closely followed the ideal 45° diagonal line, suggesting that the predicted risk probabilities were in good agreement with the actual observed mortality risk among patients.

Figure 5 Internal validation of the prediction model in the training set: discrimination assessment. AUC, area under the curve; FPR, false positive rate; TPR, true positive rate.

Figure 6 Internal validation of the prediction model in the training set: calibration assessment.

Evaluation of diagnostic performance

To investigate the model’s clinical practical value, we analyzed its diagnostic performance at different predicted probability cutoff points (Table 4). Based on the Youden index calculation, the optimal cutoff value for this model was determined to be 0.080. At this threshold, the corresponding sensitivity was 87.5%, the specificity was 90.1%, the positive predictive value (PPV) was 44.7%, and the negative predictive value (NPV) was as high as 98.8%. Decision curve analysis (DCA) showed that across a wide range of threshold probabilities (from approximately 5% to 80%), the net benefit of using this model for decision-making was higher than that of the extreme strategies of treating all patients as high-risk or treating all patients as low-risk. This indicates that in most clinical scenarios, risk stratification based on this model can yield greater clinical benefit as compared to traditional “one-size-fits-all” approaches.

Example cases of clinical application

Case 1 (low risk): a 78-year-old female with AECOPD had a creatinine level of 63.3 µmol/L, an admission heart rate of 100 beats/min, a diastolic blood pressure of 78 mmHg, a PaCO₂ of 51.3 mmHg, a lactate level of 1.78 mmol/L, and a hscTnT level of 10.87 ng/mL. Input of this information into the model calculation yielded an in-hospital mortality risk probability of 0.002 (0.2%) for this patient. At a cutoff value of 0.1, the probability that this patient would not experience in-hospital mortality was determined to be 99%. This patient was predicted to be at low risk, suggesting a favorable prognosis.

Case 2 (high risk): a 78-year-old male with AECOPD had a creatinine level of 170.6 µmol/L, an admission heart rate of 98 beats/min, a diastolic blood pressure of 55 mmHg, a PaCO₂ of 36.4 mmHg, a lactate level of 6.56 mmol/L, and a significantly elevated hscTnT level (66.1 ng/mL). Input of this information into the model calculation yielded an in-hospital mortality risk probability of 0.897 (89.7%) for this patient. At a cutoff value of 0.7, the probability that this patient would experience in-hospital mortality was determined to be 100%. This patient was predicted to be at high risk, indicating critical illness requiring close monitoring and active intervention.

DCA showed that across a wide range of threshold probabilities (from approximately 5% to 80%), the net benefit of using this model for decision-making was higher than the extreme strategies of treating all patients as high risk or treating all patients as low risk. This indicates that in most clinical scenarios, risk stratification based on this model can yield greater clinical benefit compared to traditional “one-size-fits-all” approaches.

Discussion

This study successfully developed and validated an in-hospital mortality risk prediction model incorporating eight predictors, based on clinical data from 287 patients with AECOPD. The final identified predictive indicators included age, gender, PaCO₂, lactate level, creatinine level, hscTnT level, heart rate, and diastolic blood pressure. This model comprehensively reflects the pathophysiological disturbances in patients across multiple dimensions, including gas exchange, metabolic status, organ function, and cardiovascular load, providing a practical quantitative tool for the early identification of high-risk patients.

Advanced age and male gender were identified as significant risk factors in this study. Our data show that deceased patients were generally over 70 years old (18), aligning with the notion that COPD is a disease of “accelerated lung aging”, involving various aging-related pathological pathways (19,20). The decline in organ functional reserve and immunosenescence associated with advanced age may collectively contribute to poor prognosis. Furthermore, males constituted the vast majority in this study cohort, and their increased risk might be primarily linked to smoking, a well-established independent risk factor for AECOPD (21,22). Smoking is also a well-known factor for chronic inflammation, impaired immunity, and reduced lung reserve (23). Smoking, which occurs at a higher prevalence among males, may exacerbate disease severity, also highlighting the importance of targeted smoking cessation education (24). Meanwhile, the role of gender difference itself in COPD susceptibility, disease course, and prognosis warrants attention (25), and future studies with larger samples are needed to further determine its relevance.

PaCO₂ is a core indicator of ventilatory function (26). The significantly higher PaCO₂ in the deceased group in this study is consistent with previous research indicating that acute hypercapnia during AECOPD is an important prognostic factor for long-term mortality (27-29). It signifies the failure of compensatory mechanisms in the respiratory system. Elevated lactate levels directly reflect inadequate tissue perfusion and a hypoxic state (30). Our data confirmed significantly higher lactate levels in the deceased group, consistent with findings from previous work (31). Hyperlactatemia is a sensitive marker of tissue hypoxia, suggesting the disease has progressed to a more critical stage.

Creatinine, is key indicator for assessing renal function, and its elevation suggests the occurrence of acute kidney injury. The significantly higher creatinine levels in the deceased group in our study suggests to us that the mechanism may be related to renal injury triggered or exacerbated by tissue hypoxia due to AECOPD (32). Elevated creatinine levels in patients with AECOPD might indicate the future development of high-risk conditions such as cor pulmonale, although this remains to be confirmed with larger sample data.

hscTnT is a biomarker of myocardial injury. Its significant elevation in the deceased group in out confirms its important value as a predictor of all-cause mortality in patients with COPD (33,34). This indicates that cardiovascular events [e.g., acute myocardial injury (35)] are a significant driver of mortality in patients with AECOPD and cannot be overlooked.

Heart rate and diastolic blood pressure together reflect the patient’s cardiovascular load status. An increased resting heart rate has been demonstrated to be associated with COPD severity and all-cause mortality (36). Moreover, AECOPD and blood pressure fluctuations influence one another. Therefore, when AECOPD patients receive treatment from respiratory specialists (37), it is recommended to adopt a comprehensive cardiopulmonary therapy regimen. Clinicians must closely monitor and stabilize cardiovascular function, and correct respiratory failure.

The strength of this study lies in the application of the advanced Firth’s penalized logistic regression method, which effectively addressed instability issues in modeling with small-sample data, and the demonstration of the model’s robustness through rigorous internal validation. The comprehensive selection of indicators also provides a multi-dimensional perspective for risk assessment. However, this study also involved several limitations. First, we employed a single-center, retrospective design with a limited sample size, and in particular, there was a small number of female patients and death events, which might reduce the model’s generalizability to broader populations. Second, the conclusions regarding the mechanisms related to certain variables (e.g., the link between creatinine and prognosis) lacks direct support from large-scale clinical data. Future research should involve multicenter, prospective, large-sample cohort studies for external validation and further clarify the related pathophysiological mechanisms.

Conclusions

The in-hospital mortality risk prediction model for patients with AECOPD developed in this study demonstrated good discrimination, calibration, and clinical utility. The eight indicators incorporated into the model are all routine clinical tests, making them easily accessible and amenable to clinical translation. Applying this model—or the nomogram derived from it—can assist clinicians in the early and rapid identification of high-risk patients, thereby enabling the timely adjustment of treatment strategies, rational allocation of medical resources, and thus the improvement of patient outcomes.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0166/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0166/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0166/prf

Funding: This study was supported by the National Natural Science Foundation of China (No. 32471191), the Jiangsu Planned Projects for Postdoctoral Research Fund (No. 2021K031A), the Deputy General Manager Project of Science and Technology of Jiangsu Province (No. FZ20210652), the Science and Technology Program of Nantong (No. JCZ2023024), and the Research Project of Nantong Municipal Health Commission (No. MS2023076).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0166/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Affiliated Hospital of Nantong University (No. 2025-L094) and informed consent was taken from all the patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Celli B, Fabbri L, Criner G, et al. Definition and Nomenclature of Chronic Obstructive Pulmonary Disease: Time for Its Revision. Am J Respir Crit Care Med 2022;206:1317-25. [Crossref] [PubMed]
2. World Health Organization. The top 10 causes of death (2020, December 9). Available online: https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death
3. Celli BR, Fabbri LM, Aaron SD, et al. An Updated Definition and Severity Classification of Chronic Obstructive Pulmonary Disease Exacerbations: The Rome Proposal. Am J Respir Crit Care Med 2021;204:1251-8. [Crossref] [PubMed]
4. Celli BR, Cote CG, Marin JM, et al. The body-mass index, airflow obstruction, dyspnea, and exercise capacity index in chronic obstructive pulmonary disease. N Engl J Med 2004;350:1005-12. [Crossref] [PubMed]
5. Abu Hussein N, Ter Riet G, Schoenenberger L, et al. The ADO index as a predictor of two-year mortality in general practice-based chronic obstructive pulmonary disease cohorts. Respiration 2014;88:208-14. [Crossref] [PubMed]
6. Jones PW, Agusti AG. Outcomes and markers in the assessment of chronic obstructive pulmonary disease. Eur Respir J 2006;27:822-32. [Crossref] [PubMed]
7. Global initiative for chronic obstructive lung disease. Global strategy for the diagnosis management and prevention of chronic obstructive pulmonary disease 2023 report. Available online: http://goldCOPD.org/2023-gold-report-2
8. Zhang C, Ling W, Pan H, et al. Prediction of 12-month exacerbation-related readmission in hospitalized patients with COPD: a single-center study in China. Eur J Med Res 2025;30:773. [Crossref] [PubMed]
9. Xu Z, Li F, Xin Y, et al. Prognostic risk prediction model for patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD): a systematic review and meta-analysis. Respir Res 2024;25:410. [Crossref] [PubMed]
10. Bai JS, Liu JX, Zhang Q, et al. A Predictive Model Based on Blood Indicators for Admission to the ICU with AECOPD. Clin Lab 2022;68. [Crossref] [PubMed]
11. Ding CW, Huang SS, Xu YH, et al. Lactate dehydrogenase to albumin ratio and prognosis in patients with acute exacerbation of chronic obstructive pulmonary disease: a retrospective cohort study. BMC Pulm Med 2025;25:154. [Crossref] [PubMed]
12. Li X, Yi Q, Luo Y, et al. Clinical characteristics and prognosis prediction in patients with AECOPD and type 2 diabetes mellitus: A multicenter observational study. Chron Respir Dis 2025;22:14799731251325251. [Crossref] [PubMed]
13. Qian Y, Cai C, Sun M, et al. Analyses of Factors Associated with Acute Exacerbations of Chronic Obstructive Pulmonary Disease: A Review. Int J Chron Obstruct Pulmon Dis 2023;18:2707-23. [Crossref] [PubMed]
14. Castro-Portillo E, Soriano JB, López-Izquierdo R, et al. Accuracy of early warning scores for predicting clinical worsening in COPD patients. ERJ Open Res 2025;11:00744-2024. [Crossref] [PubMed]
15. Ellingsen J, Janson C, Bröms K, et al. Clinical phenotypes predict exacerbations of COPD: the TIE cohort study. Respir Med 2025;244:108182. [Crossref] [PubMed]
16. Shafi HMH, Ali B, Dayam F, et al. Risk Factors and Outcomes in Patients With Acute Exacerbation of Chronic Obstructive Pulmonary Disease. Cureus 2025;17:e90919. [Crossref] [PubMed]
17. Guan L, Li J, Liang L, et al. Blood Eosinophil Stability Predicts Clinical Outcomes in Hospitalized Patients with Acute Exacerbations of COPD. Int J Chron Obstruct Pulmon Dis 2025;20:2913-23. [Crossref] [PubMed]
18. Zhu Y, Wang M, Gu XN, et al. Development and validation of the machine learning model for acute exacerbation of chronic obstructive pulmonary disease prediction based on inflammatory biomarkers. Front Med (Lausanne) 2025;12:1616712. [Crossref] [PubMed]
19. Easter M, Bollenbecker S, Barnes JW, et al. Targeting Aging Pathways in Chronic Obstructive Pulmonary Disease. Int J Mol Sci 2020;21:6924. [Crossref] [PubMed]
20. Yoon YS, Jin M, Sin DD. Accelerated lung aging and chronic obstructive pulmonary disease. Expert Rev Respir Med 2019;13:369-80. [Crossref] [PubMed]
21. Halpin DMG. Mortality of patients with COPD. Expert Rev Respir Med 2024;18:381-95. [Crossref] [PubMed]
22. Liao QQ, Mo YJ, Zhu KW, et al. Platelet-to-Lymphocyte Ratio (PLR), Neutrophil-to-Lymphocyte Ratio (NLR), Monocyte-to-Lymphocyte Ratio (MLR), and Eosinophil-to-Lymphocyte Ratio (ELR) as Biomarkers in Patients with Acute Exacerbation of Chronic Obstructive Pulmonary Disease (AECOPD). Int J Chron Obstruct Pulmon Dis 2024;19:501-18. [Crossref] [PubMed]
23. Qi S, Li X, Jiang Y, et al. Analysis of risk factors and development of predictive model for acute myocardial injury in patients with acute exacerbation of chronic obstructive pulmonary disease. J Thorac Dis 2025;17:1977-90. [Crossref] [PubMed]
24. Jimenez-Ruiz CA, Pascual Lledó JF, Cícero Guerrero A, et al. Analysis of quality of life in patients with Chronic Obstructive Pulmonary Disorder (COPD) who give up smoking. Semergen 2018;44:310-5. [Crossref] [PubMed]
25. DeMeo DL. Sex, Gender, and COPD. Annu Rev Physiol 2025;87:471-90. [Crossref] [PubMed]
26. Balzanelli MG, Distratis P, Lazzaro R, et al. The importance of arterial blood gas analysis as a systemic diagnosis approach in assessing and preventing chronic diseases, from emergency medicine to the daily practice. Eur Rev Med Pharmacol Sci 2023;27:11653-63. [Crossref] [PubMed]
27. Ai-Ping C, Lee KH, Lim TK. In-hospital and 5-year mortality of patients treated in the ICU for acute exacerbation of COPD: a retrospective study. Chest 2005;128:518-24. [Crossref] [PubMed]
28. Galli JA, Krahnke JS, James Mamary A, et al. Home non-invasive ventilation use following acute hypercapnic respiratory failure in COPD. Respir Med 2014;108:722-8. [Crossref] [PubMed]
29. Chu CM, Chan VL, Lin AW, et al. Readmission rates and life threatening events in COPD survivors treated with non-invasive ventilation for acute hypercapnic respiratory failure. Thorax 2004;59:1020-5. [Crossref] [PubMed]
30. Vavřička J, Brož P, Follprecht D, et al. Modern Perspective of Lactate Metabolism. Physiol Res 2024;73:499-514. [Crossref] [PubMed]
31. Nichol AD, Egi M, Pettila V, et al. Relative hyperlactatemia and hospital mortality in critically ill patients: a retrospective multi-centre study. Crit Care 2010;14:R25. [Crossref] [PubMed]
32. Freund O, Melloul A, Fried S, et al. Management of acute exacerbations of COPD in the emergency department and its associations with clinical variables. Intern Emerg Med 2024;19:2241-8. [Crossref] [PubMed]
33. Kaier TE, Alaour B, Marber M. Cardiac troponin and defining myocardial infarction. Cardiovasc Res 2021;117:2203-15. [Crossref] [PubMed]
34. Nilsson U, Vanfleteren LEGW. Troponin as a biomarker for mortality in stable COPD. Eur Respir J 2020;55:1902447. [Crossref] [PubMed]
35. Fatima I, Dass J, Aslam ZR, et al. Role of Arterial Blood Gas in Risk Stratification of Patients With Acute Exacerbation of Chronic Obstructive Pulmonary Disease in the Emergency Department: A Systematic Review. Cureus 2025;17:e94369. [Crossref] [PubMed]
36. Jensen MT, Marott JL, Lange P, et al. Resting heart rate is a predictor of mortality in COPD. Eur Respir J 2013;42:341-9. [Crossref] [PubMed]
37. Melloul A, Freund O, Tiran B, et al. Respiratory Specialist Visits Before Admissions with COPD Exacerbation are Linked to Improved Management and Outcomes. Int J Chron Obstruct Pulmon Dis 2024;19:2387-96. [Crossref] [PubMed]

(English Language Editor: J. Gray)