Title
Author
DOI
Article Type
Special Issue
Volume
Issue
A novel model for predicting the short-term prognosis of acute pulmonary embolism: immuno-inflammatory age-specific shock index model (LogSII-SIA)
1Department of Geriatrics, The Affiliated Yongchuan Hospital of Chongqing Medical University, 402160 Chongqing, China
2Chongqing Municipality Clinical Research Center for Geriatric Diseases (Yongchuan Hospital, Chongqing Medical University), 402160 Chongqing, China
3Central Laboratory, The Affiliated Yongchuan Hospital of Chongqing Medical University, 402160 Chongqing, China
4Department of Clinical Pharmacy, West China Hospital of Sichuan University, 610041 Chengdu, Sichuan, China
5Department of Neurology, Chongqing Changshou People’s Hospital, 401220 Chongqing, China
*Corresponding Author(s):700416@hospital.cqmu.edu.cn (Fei He)
| History | Submitted: 09 March 2025 | Accepted: 08 May 2025 | Published: 08 December 2025 |
| Copyright: | ©2025 The Author(s). Published by MRE Press. |
Background: To evaluate the forecasting reliability of the systemic immuno-inflammatory index (SII) in combination with the age-specific shock index (SIA) for the short-term prognosis of acute pulmonary embolism (APE). Methods: The base-10 logarithm was applied to SII, denoted as LogSII. Models were constructed using univariate analysis, Least Absolute Shrinkage and Selection Operator (LASSO) regression for variable selection, and binary logistic regression with 5-fold cross-validation, followed by efficacy analysis. Receiver operator characteristic (ROC) curves and area under the curves (AUC) were plotted, and the forecasting reliability of each indicator was assessed using DeLong’s test. Results: In the original binary logistic regression model (4 variables), LogSII (Odds Ratio (OR) = 6.969, 95% confidence interval (CI): 2.676–18.150, p < 0.001) and SIA (OR = 1.040, 95% CI: 1.013–1.068, p = 0.004) were significant predictors. A simplified model, referred to collectively as Immuno-Inflammatory Age-Specific Shock Index (LogSII-SIA), was subsequently derived: logit (p) = −10.071 + 1.732 × LogSII + 0.052 × SIA. The Hosmer-Lemeshow test indicated a good fit (p = 0.241). The model demonstrated good predictive performance (mean AUC: 0.835 ± 0.080, AUC range: 0.726–0.923), accurate probabilistic prediction (mean Brier score: 0.103) and a low risk of overfitting. The AUC of the simplified pulmonary embolism severity index (sPESI) score was 0.742 (95% CI: 0.687–0.792, p < 0.001). DeLong’s test analysis indicated that the difference between the AUCs of the sPESI score and LogSII-SIA was statistically significant (Z = 1.991, p = 0.0464). Conclusions: LogSII and SIA served as standalone risk factors for the near-term prognosis of APE cases. LogSII-SIA has a higher predictive value for the short-term prognosis of APE than sPESI, and its generalizability needs to be further verified.
Cite this article
Lingwei Huang, Juan Liao, Meimei Yang, Ruixue Liu, Dan Lv, Ying Li, Fei He. A novel model for predicting the short-term prognosis of acute pulmonary embolism: immuno-inflammatory age-specific shock index model (LogSII-SIA). Signa Vitae. 2025; 21(12): 81-89. doi: 10.22514/sv.2025.191
1. Introduction
Acute pulmonary embolism (APE) involves a blockage of the pulmonary arteries caused by various emboli, with thrombus emboli being the most common [1]. Classified as the third most frequent cardiovascular disorder, it carries high morbidity and fatality rates [2]. The adverse outcomes of APE vary widely among at-risk patients, ranging from asymptomatic cases to severe presentations, including shock [3]. Current management of pulmonary embolism emphasizes the benefits of early-stage diagnosis, particularly the prompt assessment of disease severity and patient prognosis [4]. Klok et al. [5] concluded that the early development of targeted therapeutic strategies and return-to-activity programs can minimize the likelihood of severe issues in APE individuals, which depends on effective risk stratification and prognostic evaluation.
Risk stratification and prognostic assessment of patients with APE are predominantly conducted using clinical parameters such as hemodynamics, right ventricular echocardiography, computed tomography pulmonary angiography, cardiac injury markers, and the pulmonary embolism severity index (PESI) or simplified pulmonary embolism severity index (sPESI) [6, 7]. These risk assessment models are often criticized for their complexity and high cost. Clinically, it is often impossible to apply these models without delay, in light of the patient’s condition, medical constraints and other circumstances [8]. The extensive crosstalk between inflammation, coagulation and immunity holds a major position in the pathophysiology of pulmonary embolism [9]. The systemic immuno-inflammatory index (SII) provides comprehensive information on neutrophil, lymphocyte, and platelet counts and is readily derived from routine blood tests [10]. The shock index (SI), which reflects the hemodynamic stability of the patient, is a simple, valid, and widely utilized tool for the assessment of acute and critical conditions [11] and has given rise to the age-specific shock index (SIA) [12].
Although a small number of studies have found that SII [13] and SIA [14] may help to project the prognosis for APE individuals, no studies have combined SII and SIA to evaluate the short-range forecast for APE patients. To further simplify the assessment process and explore a more effective prognostic model, we aim to examine the clinical effectiveness of combining SII with SIA in assessing the short-term prognosis of APE patients. This approach seeks to offer clinicians a theoretical foundation and a novel method to accurately evaluate pulmonary embolism severity and develop personalized treatment plans early in the clinical course.
2. Methods
2.1 Study population
Patients admitted to The Affiliated Yongchuan Hospital of Chongqing Medical University from January 2018 to June 2024 and diagnosed with APE were consecutively enrolled. Inclusion criteria: ① Computed Tomography Pulmonary Angiography (CTPA)-confirmed pulmonary embolism; ② Over the age of 18; ③ Symptom onset within less than two weeks. Exclusion criteria: ① Relapsed pulmonary embolism; ② Patients lacking complete clinical information; ③ Patients treated with anticoagulation or thrombolysis prior to hospital admission; ④ Patients with chronic thromboembolic pulmonary hypertension. The patient inclusion process is shown in Fig. 1.
Fig. 1.Patient inclusion flowchart. CTPA: Computed Tomography Pulmonary Angiography.
2.2 Data collection
Patients’ clinical data comprising gender, age, cerebrovascular disease, chronic lung disease, diabetes, malignancy, surgery, heart failure, lower extremity venous thrombosis, sPESI score, sPESI risk classification, and 30-day prognosis of confirmed pulmonary embolism were collected through a review of medical records. Laboratory markers obtained within 24 hours of APE diagnosis included serum creatinine, troponin, N-terminal pro-brain natriuretic peptide (NT-proBNP), lymphocyte count, leukocyte count, erythrocyte count, platelet count, neutrophil count, hemoglobin, hematocrit, D-dimer, and fibrinogen. Vital signs recorded at the onset of APE diagnosis comprising arterial oxygen saturation (SpO2), heart rate, respiratory rate, and systolic blood pressure. All test samples from patients were collected and measured at room temperature (25 °C). Calculations were performed to determine SII (platelet count × neutrophil count/lymphocyte count), SI (heart rate/systolic blood pressure) and SIA (age × SI). The sPESI scoring and grading criteria [15] are shown in Table 1.
| Relevant factor | Score |
| Age >80 yr | 1 point |
| Systolic blood pressure <100 mmHg | 1 point |
| Combined chronic heart failure or chronic lung disease | 1 point |
| Heart rate >110 beats/min | 1 point |
| SpO2 <90% | 1 point |
| Malignant tumor | 1 point |
| Note: sPESI: simplified pulmonary embolism severity index; SpO2: arterial oxygen saturation. Patients with a score of 0 were labeled as low risk, and those with a score of 1 or higher were labeled as high risk. |
2.3 Statistical assessment
All continuous variables were evaluated for normality using the Kolmogorov-Smirnov test. Normally distributed variables were presented as the mean ± standard deviation, though data deviating from normality were expressed as the median (interquartile range (IQR)). Unpaired t-tests and Mann-Whitney U-tests were employed to compare normally and non-normally distributed variables between the two groups, respectively. The chi-square test was utilized to examine categorical variables across the study groups. Variables showing statistically significant differences (p < 0.05) in the univariate analysis were subsequently analyzed using Least Absolute Shrinkage and Selection Operator (LASSO) regression to identify key predictors [16]. Continuous variables (SII and NT-proBNP) with right-skewed distributions and large differences (in the thousands) between their maximum and minimum values were log-transformed using base 10 to compress the range and mitigate the impact of extreme values, denoted as LogSII and LogNT-proBNP, respectively [17]. Binary logistic regression was used to construct the original model (Model 1) and a simplified version (Model 2). The variance inflation factor (VIF) was used to assess multicollinearity, using a threshold of 5. A 5-fold cross-validation was performed to verify model stability and assess the risk of overfitting [18]. The joint LogSII-SIA score was defined as a weighted formula. The efficacy of the study sample was assessed. ROC curves were plotted for LogSII, SIA, sPESI and LogSII-SIA. A comparison of the ROC curve AUCs was conducted using DeLong’s test. Statistical analyses and visualizations were conducted using SPSS 26.0 (IBM Corp., Armonk, NY, USA), MedCalc 22.0 (MedCalc Software Ltd., Ostend, Belgium) and RStudio 4.4.2. p-values of < 0.05 were judged to be statistically significant.
3. Results
3.1 General data
Among the 280 patients, 52% were admitted through emergency admission, 37% were admitted to the Intensive Care Unit (ICU) and 27% had massive pulmonary embolism. A total of 237 patients (84.6%) survived, including 140 males (59.1%) and 97 females (40.9%), with a mean age of 65.03 ± 13.09 years, while 43 patients (15.4%) died, comprising 31 males (72.1%) and 12 females (27.9%), with a mean age of 71.67 ± 14.99 years, during the 30-day follow-up period. The age, leukocyte count, neutrophil count, NT-proBNP, troponin, creatinine, heart rate, respiratory rate, D-dimer, SII, SI and SIA in the deceased group were notably greater than those in the survival group. Conversely, the lymphocyte count, erythrocyte count, hemoglobin, hematocrit, systolic blood pressure, and arterial oxygen saturation were notably reduced in the deceased group. These differences were statistically significant (p < 0.05). The incidence of malignant tumors, cerebrovascular disease, heart failure, chronic lung disease, sPESI scores and sPESI risk classifications were greater in the deceased group versus the survival group, demonstrating notable statistical disparities (p < 0.05). The remaining metrics did not exhibit statistically significant variations (p > 0.05). A comparison of general data is shown in Table 2.
| Variables | All patients (n = 280) | Death group (n = 43) | Survival group (n = 237) | Statistic | p value |
| Age (IQR, yr) | 68 (56, 76) | 73 (65, 85) | 67 (56, 75) | Z = −2.936 | 0.003* |
| Gender (Male (n (%))) | 171 (61.1) | 31 (72.1) | 140 (59.1) | χ2 = 2.596 | 0.107 |
| Cerebrovascular disease (n (%)) | 92 (32.9) | 21 (48.8) | 71 (30.0) | χ2 = 5.880 | 0.015* |
| Surgery (n (%)) | 34 (12.1) | 4 (9.3) | 30 (12.7) | χ2 = 0.134 | 0.714 |
| Chronic lung disease (n (%)) | 95 (33.9) | 23 (53.5) | 72 (30.4) | χ2 = 8.670 | 0.003* |
| Congestive heart failure (n (%)) | 78 (27.9) | 24 (55.8) | 54 (22.8) | χ2 = 19.757 | <0.001* |
| Lower extremity venous thrombosis (n (%)) | 151 (53.9) | 27 (62.8) | 124 (52.3) | χ2 = 1.606 | 0.205 |
| Diabetes (n (%)) | 51 (18.2) | 8 (18.6) | 43 (18.1) | χ2 = 0.005 | 0.943 |
| Malignant tumor (n (%)) | 53 (18.9) | 13 (30.2) | 40 (16.9) | χ2 = 4.230 | 0.040* |
| Heart rate (IQR, Times/min) | 89 (80, 105) | 102 (90, 108) | 88 (79, 103) | Z = −3.665 | <0.001* |
| Respiratory (IQR, Times/min) | 23 (20, 28) | 29 (25, 31) | 22 (20, 26) | Z = −5.588 | <0.001* |
| Systolic blood pressure (x̄ ± s, mmHg) | 125.26 ± 21.397 | 113.19 ± 22.727 | 127.45 ± 20.443 | T = −4.137 | <0.001* |
| SpO2 (IQR, %) | 96 (93, 98) | 93 (86, 95) | 97 (94, 98) | Z = −5.436 | <0.001* |
| Creatinine (IQR, μmoL/L) | 70.00 (57.00, 89.75) | 91.00 (64.00, 110.00) | 69.00 (56.00, 85.00) | Z = −3.803 | <0.001* |
| Troponin (IQR, ng/mL) | 0.013 (0.011, 0.048) | 0.069 (0.028, 0.223) | 0.012 (0.010, 0.033) | Z = −6.097 | <0.001* |
| NT-proBNP (IQR, pg/mL) | 558.50 (205.25, 2107.50) | 2810.00 (1150.00, 7790.00) | 444.00 (158.00, 1325.50) | Z = −6.595 | <0.001* |
| Lymphocyte count (IQR, ×109/L) | 1.14 (0.79, 1.60) | 0.84 (0.50, 1.45) | 1.19 (0.84, 1.63) | Z = −3.137 | 0.002* |
| Leucocyte count (IQR, ×109/L) | 8.70 (6.33, 11.83) | 14.40 (10.10, 19.30) | 8.00 (6.15,10.30) | Z = −6.314 | <0.001* |
| Erythrocyte count (IQR, ×1012/L) | 4.19 (3.55, 4.59) | 3.71 (2.98, 4.45) | 4.20 (3.68, 4.60) | Z = −2.587 | 0.010* |
| Neutrophil count (IQR, ×109/L) | 6.28 (4.37, 9.77) | 12.03 (8.22, 15.60) | 5.75 (4.17, 8.21) | Z = −6.377 | <0.001* |
| Platelet count (IQR, ×109/L) | 187.50 (141.25, 252.50) | 204.00 (125.00, 300.00) | 184.00 (142.00, 242.00) | Z = −1.191 | 0.233 |
| Hemoglobin (IQR, g/L) | 125.50 (108.00, 140.75) | 113.00 (88.00, 134.00) | 127.00 (112.00, 143.00) | Z = −3.266 | 0.001* |
| Hematocrit (IQR, %) | 0.38 (0.33, 0.42) | 0.36 (0.30, 0.41) | 0.39 (0.33, 0.43) | Z = −2.420 | 0.016* |
| D-dimer (IQR, mg/L) | 7.94 (3.52, 12.95) | 12.84 (5.40, 21.48) | 7.49 (3.13, 12.18) | Z = −3.790 | <0.001* |
| Fibrinogen (IQR, g/L) | 3.30 (2.50, 4.50) | 3.90 (2.80, 4.80) | 3.20 (2.40, 4.45) | Z = −1.508 | 0.132 |
| SII (IQR) | 906.41 (568.81, 1927.39) | 2934.77 (1465.89, 6708.91) | 844.88 (560.98, 1423.83) | Z = −5.325 | <0.001* |
| SI (IQR) | 0.719 (0.609, 0.883) | 0.907 (0.726, 1.083) | 0.699 (0.602, 0.833) | Z = −4.985 | <0.001* |
| SIA (IQR) | 45.88 (38.56, 59.86) | 62.78 (53.18, 76.08) | 44.03 (37.10, 57.31) | Z = −5.789 | <0.001* |
| sPESI score (IQR, point) | 1 (0, 2) | 2 (2, 3) | 1 (0, 2) | Z = −5.243 | <0.001* |
| sPESI risk classification (n (%)) | χ2 = 15.761 | <0.001* | |||
| Low risk | 97 (34.6) | 3 (7.0) | 94 (39.7) | ||
| High risk | 183 (65.4) | 40 (93.0) | 143 (60.3) | ||
| Note: SpO2: arterial oxygen saturation; NT-proBNP: N-terminal pro-brain natriuretic peptide; SII: systemic immuno-inflammatory index; SI: shock index; SIA: age-specific shock index; sPESI: simplified pulmonary embolism severity index; IQR: interquartile range. *p < 0.05. |
3.2 Multifactorial logistic regression analysis
The LASSO regression identified twelve key variables: Leucocyte, SpO2, Respiratory, Systolic blood pressure, Chronic lung disease, SII, SIA, hemoglobin, D-dimer, creatinine, NT-proBNP and malignant tumor (Supplementary material 1). Due to the limited number of deaths (n = 43), four core predictors with minimal cross-validation errors SIA (reflecting circulatory failure), LogSII (reflecting inflammation), Hemoglobin (associated with oxygen delivery) and LogNT-proBNP (a marker predictive of heart failure)—were retained to construct the original multivariate logistic regression model: logit (p) = −13.585 − 0.023 × Hemoglobin + 1.942 × LogSII + 1.999 × LogNT-proBNP + 0.039 × SIA (Table 3). VIF values for all included variables were below 5 (LogSII: 1.065; SIA: 1.161; LogNT-proBNP: 1.129; Hemoglobin: 1.040), suggesting the absence of significant multicollinearity. The Hosmer-Lemeshow test indicated good calibration (p = 0.982). The five-fold cross-validation demonstrated excellent predictive performance (mean AUC: 0.896 ± 0.057, AUC range: 0.833–0.987) and high accuracy in probability estimation (mean Brier score: 0.088), with a controlled risk of overfitting (Supplementary material 2). LogSII was a strong predictor of risk (OR = 6.969, 95% CI: 2.676–18.150, p < 0.001), while SIA also showed a significant association (OR = 1.040, 95% CI: 1.013–1.068, p = 0.004). Both LogSII and SIA were independent risk factors in the multivariate model, and their significance was held after controlling for other variables (e.g., NT-proBNP, Hemoglobin). The prediction model was further simplified by incorporating only two variables, LogSII and SIA, to construct a simplified model (Model 2), referred to as LogSII-SIA: logit (p) = −10.071 + 1.732 × LogSII + 0.052 × SIA (Table 4). The Hosmer-Lemeshow test showed good calibration (p = 0.241). The results of the five-fold cross-validation demonstrated good predictive performance (mean AUC: 0.835 ± 0.080, AUC range: 0.726–0.923) and good probabilistic accuracy (mean Brier score: 0.103), with a low risk of overfitting (Supplementary material 3). Compared with the original model, the simplified model exhibited a 6.1% reduction in AUC and a 17% increase in Brier score but required 50% fewer variables and offered improved clinical applicability (Table 5). The formula for calculating the weights of the joint LogSII-SIA score, derived from our study sample, was LogSII-SIA Score = 0.444 × (LogSII − 3.0)/0.4 + 0.556 × (SIA − 50.3)/16.7 (see Supplementary material 4 for a detailed explanation of the calculation steps). Based on post hoc efficacy analysis (α = 0.05, effect size f2 = 0.35), the statistical power of the current sample size (n = 280) was 92%, indicating that the model was sufficiently powered to detect a significant effect of the predictor variables on the outcome. This analysis reinforces the study’s methodological robustness, eliminating the need for an expanded sample size.
| Variables | β | SE | Wald | OR (95% CI) | p |
| Hemoglobin | −0.023 | 0.009 | 6.643 | 0.977 (0.961–0.995) | 0.010* |
| LogSII# | 1.942 | 0.488 | 15.806 | 6.969 (2.676–18.150) | <0.001* |
| LogNT-proBNP# | 1.999 | 0.408 | 23.959 | 7.379 (3.315–16.426) | <0.001* |
| SIA | 0.039 | 0.014 | 8.454 | 1.040 (1.013–1.068) | 0.004* |
| Note: SII: systemic immuno-inflammatory index; NT-proBNP: N-terminal pro-brain natriuretic peptide; SIA: age-specific shock index; CI: confidence interval; SE: Standard Error; OR: Odds Ratio. #: SII and NT-proBNP were logarithmized with a base of 10; *: p < 0.05. |
| Variables | β | SE | Wald | OR (95% CI) | p |
| LogSII# | 1.732 | 0.412 | 17.650 | 5.653 (2.520–12.684) | <0.001* |
| SIA | 0.052 | 0.011 | 22.134 | 1.053 (1.031–1.076) | <0.001* |
| Note: SII: systemic immuno-inflammatory index; SIA: age-specific shock index; CI: confidence interval; SE: standard error; OR: odds ratio. #: SII was logarithmized with a base of 10; *: p < 0.05. |
| Model | AUC | Brier scores | Hosmer-Lemeshow test, p-value | Number of variables | Clinical operability |
| Original model 1 (4 variables) | 0.896 ± 0.057 | 0.088 | 0.982 | 4 | Higher (more variables required) |
| Simplified model 2 (LogSII-SIA) | 0.835 ± 0.080 | 0.103 | 0.241 | 2 | Better (bedside fast) |
| LogSII-SIA: Immuno-Inflammatory Age-Specific Shock Index; AUC: area under the curves. |
3.3 ROC curve
The ROC curve showed that the AUC of the sPESI score, LogSII, SIA and LogSII-SIA were 0.742, 0.755, 0.777 and 0.832, and all demonstrated predictive value for the immediate prognosis of individuals suffering from acute pulmonary embolism (p < 0.05). DeLong’s test analysis suggested that the difference in AUC between the sPESI score and LogSII-SIA was statistically significant (Z = 1.991, p < 0.05), indicating that LogSII combined with SIA had a greater ability to predict 30-day all-cause mortality in patients with APE compared to the sPESI score. ROC curve and analysis are shown in Tables 6,7 and Fig. 2. AUC values derived from RStudio 4.4.2 (5-fold cross-validation) and MedCalc 22.0 (ROC curve analysis) may exhibit minor numerical variations due to platform-specific computational algorithms.
| Variables | AUC Value | 95% CI | Youden Index | Optimal cut-off value | Sensitivity (%) | Specificity (%) | Z value | p value |
| LogSII | 0.755 | 0.701–0.804 | 0.5354 | 3.16 | 76.74 | 76.79 | 5.103 | <0.001* |
| SIA | 0.777 | 0.724–0.825 | 0.5165 | 51.10 | 83.72 | 67.93 | 7.922 | <0.001* |
| sPESI score | 0.742 | 0.687–0.792 | 0.4299 | 1.00 | 76.74 | 66.24 | 6.707 | <0.001* |
| LogSII-SIA | 0.832 | 0.782–0.873 | 0.5968 | 0.12 | 88.37 | 71.31 | 10.534 | <0.001* |
| Note: APE: acute pulmonary embolism; SII: systemic immuno-inflammatory index; SIA: age-specific shock index; sPESI: simplified pulmonary embolism severity index; LogSII: SII was logarithmized with a base of 10; CI: confidence interval; AUC: area under the curves. *: p < 0.001. |
| Verify by comparing two cases | Statistic | p value |
| LogSII vs. SIA | Z = 0.388 | 0.6982 |
| LogSII vs. sPESI score | Z = 0.197 | 0.8436 |
| LogSII vs. LogSII-SIA | Z = 1.868 | 0.0617 |
| SIA vs. sPESI score | Z = 0.840 | 0.4012 |
| SIA vs. LogSII-SIA | Z = 2.503 | 0.0123* |
| sPESI score vs. LogSII-SIA | Z = 1.991 | 0.0464* |
| Note: SII: systemic immuno-inflammatory index; SIA: age-specific shock index; sPESI: simplified pulmonary embolism severity index; LogSII: SII was logarithmized with a base of 10; *: p < 0.05. |
Fig. 2.ROC Curves of LogSII, SIA, sPESI Score and LogSII-SIA. Note: SII: systemic immuno-inflammatory index; SIA: age-specific shock index; sPESI: simplified pulmonary embolism severity index; LogSII: SII was logarithmized with a base of 10; AUC: area under the curves.
4. Discussion
Due to the complexity of APE, clinicians must assess the severity and progression of pulmonary embolism through multiple complex steps, which are difficult to perform quickly and may delay effective treatments such as thrombolysis or intervention for high-risk patients. Such delays can ultimately result in patient mortality [19]. Studies have shown [19] that mortality due to pulmonary embolism is higher within 30 days of diagnosis compared to longer follow-up periods, with half of patients succumbing to causes other than pulmonary embolism by 90 days. The main advantage of PESI and sPESI is the ability to reliably identify patients with a low 30-d mortality rate. It is therefore the current mainstream method for predicting death occurring within a 30-day period in APE patients and is recommended by international guidelines [20, 21]. However, several limitations of sPESI have been noted. First, sPESI has high sensitivity but low specificity [7]. Second, the fact that sPESI is based only on the subjective personal history of illness and does not take into account objective imaging examination, laboratory indices, and other parameters may be a serious limitation [20, 21]. In addition, the score has limitations in terms of the authenticity of the medical history and difficulty in data collection. In this study, an attempt was made to find a simple and easy, more objective, and accurate method to predict the short-term prognosis of patients with APE. Therefore, we constructed a combined SII and SIA model the Immuno-Inflammatory Age-Specific Shock Index (LogSII-SIA) which demonstrated superior predictive performance for short-term prognosis in acute pulmonary embolism compared with the sPESI score. While its performance was slightly lower than that of the original model, it offered substantial clinical convenience, particularly in resource-limited settings.
Studies have shown [14] that SI can be used to determine the prognosis of patients with APE. As people age, their blood pressure increases, and their heart rate slows down [22]. Catecholamine levels and the application of vasoactive substances considerably influence heart rate and systolic blood pressure [23]. When peripheral perfusion is inadequate, catecholamine secretion increases, leading to arterial vasoconstriction and an elevated heart rate to sustain basic tissue metabolism [23]. This physiological phenomenon may reduce the accuracy of SI as a prognostic indicator. Age is a stable indicator that is not disturbed by other factors. Therefore, age, a risk factor, was included in addition to heart rate and blood pressure to enhance the accuracy of patient prognosis evaluation [24]. We find that SIA effectively forecasts adverse outcomes in APE patients. For each one-unit increase in SIA, the risk of death increased by 5.3% (OR = 1.053, 95% CI: 1.031–1.076). Although the effect size is relatively modest, its status as a continuous variable enables meaningful risk stratification around critical thresholds (e.g., SIA >50 vs. ≤50). When combined with LogSII (OR = 5.653), the joint model demonstrates a multiplicative effect on risk. Moreover, the simple, non-invasive, rapidly obtainable, and accurate nature of SIA as a risk index for circulatory failure is one of its key strengths, making it a practical tool.
Our findings identified SII as an independent predictor of APE severity. Each one-unit increase in LogSII was associated with a 5.653-fold increase in the risk of death (OR = 5.653, 95% CI: 2.520–12.684), indicating strong predictive performance. Additionally, SII is derived from budget-friendly and routine complete blood counts, making it a practical tool. The mechanism is considered as follows: Serum levels of procoagulant and proinflammatory cells such as leukocytes, platelets, and endothelial cells are elevated following the occurrence of APE [9]. The release of procoagulant and proinflammatory particles triggers increased platelet activation and neutrophil recruitment. This process disrupts the vascular wall’s anticoagulant and anti-inflammatory phenotypes, fosters thrombus formation, and generates oxygen radicals, leading to tissue damage [9, 25, 26]. This is tied to an adverse prognosis and a greater recent mortality risk in patients diagnosed with pulmonary embolism. In addition, neutrophils and macrophages in APE patients can infiltrate the walls of the pulmonary arteries and the right ventricle, leading to right ventricular dysfunction, myocardial inflammation, and necrosis [9, 27]. Lymphocytes are immune cells, and infection and immunity play critical roles in pulmonary embolism [28]. The secretion of epinephrine and glucocorticoids during the sympathetic nervous system activation elicited by APE may lead to a decrease in lymphocyte counts [29]. Lymphopenia leads to a weakened immune response, contributing to a poor prognosis [29].
However, elevated LogSII is limited in specificity and can be influenced by various factors such as trauma, infections, acute coronary syndromes, malignancies, renal insufficiency, and chronic cardiopulmonary disease [30]. Comorbidities were adjusted for the fact that LogSII remained an independent predictor of 30-day mortality. Furthermore, the combination of LogSII and SIA for predicting the prognosis of pulmonary embolism enhances prognostic accuracy and reduces false-positive rates, making it an important research focus. However, no relevant studies have been conducted to date. In this study, we found that LogSII, SIA and LogSII-SIA demonstrated strong forecasting potential for the near-term prognosis of APE patients. In comparison, LogSII-SIA increased sensitivity and demonstrated significantly greater predictive efficacy compared to the sPESI score. This suggests that in clinical practice, it is necessary to make full use of all important clinical data and adopt a multi-indicator (subjective and objective) joint prediction model to enhance prognostic efficacy and provide a reliable basis for early risk stratification and individualized treatment of APE.
There are some limitations to this study: (1) This study was a single-center, retrospective analysis with a relatively small sample size drawn exclusively from our hospital. It provided preliminary evidence that LogSII-SIA outperformed sPESI in internal validation (AUC: 0.832 vs. 0.742), although its generalizability remains to be established. Large-scale, multicenter, prospective studies are needed to validate these findings further. (2) The possibility of selection bias and unaccounted confounders cannot be fully excluded. Elevated LogSII has limited specificity and can be influenced by factors such as malignant tumors, cerebrovascular disease, and chronic cardiopulmonary conditions. Comorbidities (cerebrovascular disease, chronic lung disease, heart failure, cancer, diabetes) were included as covariates in the logistic regression model. Even after adjustment, LogSII remained an independent predictor of 30-day mortality. (3) Blood samples were collected within 24 hours of diagnosis but not at standardized time points. Parameters such as SII may fluctuate over time due to dynamic physiological responses post-APE (e.g., platelet activation, inflammatory cascades). Variability in sampling timing could introduce bias into SII measurements, potentially affecting the model’s prognostic accuracy. Future studies should standardize blood collection protocols to minimize this confounder. (4) Heart rate and blood pressure values are influenced by numerous factors, some of which were not assessable or controllable within this study population. Additionally, heart rate and blood pressure measurements were limited to a short period at diagnosis and may have been affected by prehospital medications or interventions. These limitations warrant further investigation in future research.
5. Conclusions
In a study involving 280 APE patients from The Affiliated Yongchuan Hospital of Chongqing Medical University, we found that both LogSII and SIA served as standalone risk factors for the near-term prognosis of APE cases. Our findings indicated that the combination of SII and SIA, termed the Immuno-Inflammatory Age-Specific Shock Index (LogSII-SIA), outperformed sPESI in predicting the near-term prognosis of APE patients. Its generalizability requires validation through large-scale prospective studies.
Abbreviations
APE, acute pulmonary embolism; SII, systemic immuno-inflammatory index; LogSII, systemic immuno-inflammatory index (SII) was logarithmized with a base of 10; SI, shock index; SIA, age-specific shock index; PESI, pulmonary embolism severity index; sPESI, simplified pulmonary embolism severity index; ROC curve, receiver operator characteristic curve; AUC, area under the curve; CI, confidence interval; LASSO, Least Absolute Shrinkage and Selection Operator; SpO2, arterial oxygen saturation; NT-proBNP, N-terminal pro-brain natriuretic peptide; LogSII-SIA, the systemic immuno-inflammatory index (SII) in combination with the age-specific shock index (SIA); Or Immuno-Inflammatory Age-Specific Shock Index Model; VIF, variance inflation factor; IQR, interquartile range; OR, Odds Ratio; CTPA, Computed Tomography Pulmonary Angiography; ICU, Intensive Care Unit.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Author contributions
LH and FH—designed the research study. LH, JL, MY, RL, DL, YL and FH—performed the research. LH and JL—analyzed the data. LH, JL, MY and RL—wrote the manuscript. All authors read and approved the final manuscript.
Ethics approval and consent to participate
The study was conducted by the guidelines set out in the Declaration of Helsinki. The study was approved by The Medical Ethics Committee of The Affiliated Yongchuan Hospital of Chongqing Medical University (No. 2024EC0045) and was carried out in strict accordance with the requirements. This study is a retrospective study that involves only the collection of data, files and medical records that are already in existence. Therefore, the Medical Ethics Committee of The Affiliated Yongchuan Hospital of Chongqing Medical University gave a waiver of informed consent. We respect and protect the rights and privacy of the participants and ensure the confidentiality of their personal information, and the privacy and personally identifiable information of the participants has been de-informatized and does not involve destructive and sensitive research on the participants of the study.
Acknowledgment
Not applicable.
Funding
This study was supported by the joint project of Science and Health of Chongqing Municipal Health Commission in 2021, Grant No.: 2022MSXM177, and the patent for technological innovation and application development of Yongchuan District Science and Technology Bureau in 2021, Grant No.: 2021yc-cxfz30017.
Conflict of interest
The authors declare no conflict of interest.
Supplementary material
Supplementary material associated with this article can be found, in the online version, at https://oss.signavitae.com/mre-signavitae/article/1997923503871803392/attachment/Supplementary%20material.zip.
