Boqun Shi

Boqun Shi

Auto-generated profile, based on publicly available information

Boqun Shi

Cardiometabolic Medicine Center, Department of Cardiology, Fuwai Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College/National Center for Cardiovascular Diseases, Beijing 100037, China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 April 2023 in Journal of the American Heart Association
Verified
Journal Article
Open access
BSBoqun Shi
Boqun Shi
HWHao-Yu Wang
Hao-Yu Wang
JLJinpeng Liu
Jinpeng Liu
ZCZhongxing Cai
Zhongxing Cai
CSChenxi Song
Chenxi Song
KDKe-Fei Dou
Ke-Fei Dou
Background: The PRAISE (Prediction of Adverse Events Following an Acute Coronary Syndrome) score is a machine‐learning‐based model for predicting 1‐year all‐cause death, myocardial infarction, and Bleeding Academic Research Consortium (BARC) type 3/5 bleeding. Its utility in an unselected Asian population undergoing percutaneous coronary intervention for acute coronary syndrome remains unknown. We aimed to validate the PRAISE score in a real‐world Asian population. Methods and Results: A total of 6412 consecutive patients undergoing percutaneous coronary intervention for acute coronary syndrome were prospectively included. The PRAISE scores were compared with established scoring systems (GRACE [Global Registry of Acute Coronary Events] 2.0, PRECISE‐DAPT (Predicting Bleeding Complications in Patients Undergoing Stent Implantation and Subsequent Dual Antiplatelet Therapy), and PARIS [Patterns of Non‐Adherence to Anti‐Platelet Regimen in Stented Patients]) to evaluate their discrimination, calibration, and reclassification. The risk of all‐cause mortality (hazard ratio [HR], 12.24 [95% CI, 5.32–28.15]) and recurrent acute myocardial infarction (HR, 3.92 [95% CI, 1.76–8.73]) was greater in the high‐risk group than in the low‐risk group. The C‐statistics for death, myocardial infarction, and major bleeding were 0.75 (0.67–0.83), 0.61 (0.52–0.69), and 0.62 (0.46–0.77), respectively. The observed to expected ratio of death, myocardial infarction, and major bleeding was 0.427, 0.260, and 0.106, respectively. Based on the decision curve analysis, the PRAISE score displayed a slightly greater net benefit for the 1‐year risk of death (5%–10%) than the GRACE score did. Conclusions: The PRAISE score showed limited potential for risk prediction in our validation cohort with acute coronary syndrome. As a result, new prediction models or model refitting are required with improved discrimination and accuracy in risk prediction.

12 May 2025 in Journal of Medical Internet Research
Verified
Journal Article
Open access
BSBoqun Shi
Boqun Shi
LCLiangguo Chen
Liangguo Chen
SPShuo Pang
Shuo Pang
YWYue Wang
Yue Wang
SWShen Wang
Shen Wang
XWXiaofan Wu
Xiaofan Wu
Background: Accurate mortality risk prediction is crucial for effective cardiovascular risk management. Recent advancements in artificial intelligence (AI) have demonstrated potential in this specific medical field. Qwen-2 and Llama-3 are high-performance, open-source large language models (LLMs) available online. An artificial neural network (ANN) algorithm derived from the SWEDEHEART (Swedish Web System for Enhancement and Development of Evidence-Based Care in Heart Disease Evaluated According to Recommended Therapies) registry, termed SWEDEHEART-AI, can predict patient prognosis following acute myocardial infarction (AMI). Objective: This study aims to evaluate the 3 models mentioned above in predicting 1-year all-cause mortality in critically ill patients with AMI. Methods: The Medical Information Mart for Intensive Care IV (MIMIC-IV) database is a publicly available data set in critical care medicine. We included 2758 patients who were first admitted for AMI and discharged alive. SWEDEHEART-AI calculated the mortality rate based on each patient’s 21 clinical variables. Qwen-2 and Llama-3 analyzed the content of patients’ discharge records and directly provided a 1-decimal value between 0 and 1 to represent 1-year death risk probabilities. The patients’ actual mortality was verified using follow-up data. The predictive performance of the 3 models was assessed and compared using the Harrell C-statistic (C-index), the area under the receiver operating characteristic curve (AUROC), calibration plots, Kaplan-Meier curves, and decision curve analysis. Results: SWEDEHEART-AI demonstrated strong discrimination in predicting 1-year all-cause mortality in patients with AMI, with a higher C-index than Qwen-2 and Llama-3 (C-index 0.72, 95% CI 0.69-0.74 vs C-index 0.65, 0.62-0.67 vs C-index 0.56, 95% CI 0.53-0.58, respectively; all P<.001 for both comparisons). SWEDEHEART-AI also showed high and consistent AUROC in the time-dependent ROC curve. The death rates calculated by SWEDEHEART-AI were positively correlated with actual mortality, and the 3 risk classes derived from this model showed clear differentiation in the Kaplan-Meier curve (P<.001). Calibration plots indicated that SWEDEHEART-AI tended to overestimate mortality risk, with an observed-to-expected ratio of 0.478. Compared with the LLMs, SWEDEHEART-AI demonstrated positive and greater net benefits at risk thresholds below 19%. Conclusions: SWEDEHEART-AI, a trained ANN model, demonstrated the best performance, with strong discrimination and clinical utility in predicting 1-year all-cause mortality in patients with AMI from an intensive care cohort. Among the LLMs, Qwen-2 outperformed Llama-3 and showed moderate predictive value. Qwen-2 and SWEDEHEART-AI exhibited comparable classification effectiveness. The future integration of LLMs into clinical decision support systems holds promise for accurate risk stratification in patients with AMI; however, further research is needed to optimize LLM performance and address calibration issues across diverse patient populations.

27 October 2021 in JMIR mHealth and uHealth
Verified
Journal Article
Open access
BSBoqun Shi
Boqun Shi
XLXi Liu
Xi Liu
QDQiuting Dong
Qiuting Dong
YYYuxiu Yang
Yuxiu Yang
ZCZhongxing Cai
Zhongxing Cai
WSWeihua Song
Weihua Song
Background: In China, ischemic heart disease is the main cause of mortality. Having cardiac rehabilitation and a secondary prevention program in place is a class IA recommendation for individuals with coronary artery disease. WeChat-based interventions seem to be feasible and efficient for the follow-up and management of chronic diseases. Objective: This study aims to evaluate the effectiveness of a tertiary A-level hospital, WeChat-based telemedicine intervention in comparison with conventional community hospital follow-up on medication adherence and risk factor control in individuals with stable coronary artery disease. Methods: In this multicenter prospective study, 1424 patients with stable coronary artery disease in Beijing, China, were consecutively enrolled between September 2018 and September 2019 from the Fuwai Hospital and 4 community hospitals. At 1-, 3-, 6-, and 12-month follow-up, participants received healthy lifestyle recommendations and medication advice. Subsequently, the control group attended an offline outpatient clinic at 4 separate community hospitals. The intervention group had follow-up visits through WeChat-based telemedicine management. The main end point was medication adherence, which was defined as participant compliance in taking all 4 cardioprotective medications that would improve the patient’s outcome (therapies included antiplatelet therapy, β-blockers, statins, and angiotensin-converting-enzyme inhibitors or angiotensin-receptor blockers). Multivariable generalized estimating equations were used to compare the primary and secondary outcomes between the 2 groups and to calculate the relative risk (RR) at 12 months. Propensity score matching and inverse probability of treatment weighting were performed as sensitivity analyses, and propensity scores were calculated using a multivariable logistic regression model. Results: At 1 year, 88% (565/642) of patients in the intervention group and 91.8% (518/564) of patients in the control group had successful follow-up data. We matched 257 pairs of patients between the intervention and control groups. There was no obvious advantage in medication adherence with the 4 cardioprotective drugs in the intervention group (172/565, 30.4%, vs 142/518, 27.4%; RR 0.99, 95% CI 0.97-1.02; P=.65). The intervention measures improved smoking cessation (44/565, 7.8%, vs 118/518, 22.8%; RR 0.48, 95% CI 0.44-0.53; P<.001) and alcohol restriction (33/565, 5.8%, vs 91/518, 17.6%; RR 0.47, 95% CI 0.42-0.54; P<.001). Conclusions: The tertiary A-level hospital, WeChat-based intervention did not improve adherence to the 4 cardioprotective medications compared with the traditional method. Tertiary A-level hospital, WeChat-based interventions have a positive effect on improving lifestyle, such as quitting drinking and smoking, in patients with stable coronary artery disease and can be tried as a supplement to community hospital follow-up. Trial Registration: ClinicalTrials.gov NCT04795505; https://clinicaltrials.gov/ct2/show/NCT04795505

Verified
Journal Article
Open access
PDF file
BSBoqun Shi
Boqun Shi
HWHao-Yu Wang
Hao-Yu Wang
DYDong Yin
Dong Yin
CZCheng-Gang Zhu
Cheng-Gang Zhu
LFLei Feng
Lei Feng
KDKe-Fei Dou
Ke-Fei Dou
Background and aims: Our goals in the study were to (1) quantify the discordance in LDL-C levels between equations (the Friedewald, Sampson, and Martin/Hopkins equations) and compare them with direct LDL-C (dLDL-C); and (2) explore the proportion of misclassified patients by calculated LDL-C using these three different equations. Methods: A total of 30,349 consecutive patients with angiographically confirmed coronary artery disease (CAD) were prospectively enrolled. Concordance was defined as if the LDL-C was <1.8 mmol/L with each pairwise comparison of LDL-C equations. Estimated LDL-C that fell into the same category as dLDL-C at the following levels: <1.4, 1.4 to 1.7, 1.8 to 2.5, 2.6 to 2.9, and ≥3.0 mmol/L was considered to have been correctly categorized. Results: The concordance was 96.3% (Sampson vs. Martin/Hopkins), 95.0% (Friedewald vs. Sampson), and 91.4% (Friedewald vs. Martin/Hopkins), respectively. This proportion fell to 82.4% in those with hypertriglyceridemia (TG ≥ 1.7 mmol/L). With an accurate classification rate of 73.6%, the Martin/Hopkins equation outperformed the Sampson equation (69.5%) and the Friedewald equation (59.3%) by a wide margin. Conclusions: Comparing it to the validated Martin/Hopkins equation, the Friedewald equation produced the lowest levels of LDL-C, followed by the Sampson equation. In the classification of LDL-C, the Martin/Hopkins equation has also been shown to be more accurate. There is a significant difference between the equations and the direct measurement method, which may lead to overtreatment or undertreatment.

Verified
Journal Article
Open access
PDF file
BSBoqun Shi
Boqun Shi
XTXieraili Tiemuerniyazi
Xieraili Tiemuerniyazi
RZRui Zhang
Rui Zhang
CSChenxi Song
Chenxi Song
KCKongyong Cui
Kongyong Cui
KDKe-Fei Dou
Ke-Fei Dou
Background: Clinical guidelines recommend surgical intervention when left ventricular thrombus (LVT) is complicated with left ventricular aneurysm (LVA). Objectives: This study aimed to review the changes in the treatment of LVT combined with LVA over the past 12 years at our center and to compare the efficacy of medical therapy and surgical treatment on patient outcomes. Methods: Between January 2009 and June 2021, 723 patients with LVT combined with LVA were enrolled, of whom 205 received surgical ventricular reconstruction (SVR) therapy and 518 received medical therapy. The following clinical outcomes were gathered via observation: all-cause death, cardiovascular death, and major adverse cardiovascular and cerebrovascular events (MACCEs; defined as the composite of cardiovascular death, ischemic stroke, and acute myocardial infarction). The median follow-up time was 1403 [707, 2402] days. Results: The proportion of SVR dropped yearly in this group of patients, from a peak of 64.5% in 2010 to 7.5% in 2021 (p for trend < 0.001). Meanwhile, the proportion of anticoagulant use increased quickly, from 8.0% in 2016 to 67.9% in 2021 (p for trend < 0.001). The incidence rates of all-cause mortality, cardiovascular death, and MACCEs were 12.9% (n = 93), 10.5% (n = 76), and 14.7% (n = 106), respectively. In the multivariable analysis, there were no significant differences in all-cause death (HR of 0.60, 95% CI of 0.32–1.13, p = 0.11), cardiovascular death (HR of 0.79, 95% CI of 0.41–1.50, p = 0.5), and MACCEs (HR of 0.82, 95% CI of 0.49–1.38, p = 0.5) between the two groups. The competing risk regression performed in the propensity score matching (PSM) and inverse probability of treatment weighting (IPTW) analyses was in line with the unmatched analysis. Conclusions: The rate of SVR dropped significantly among patients with both LVT and LVA, while there was an improvement in oral anticoagulant utilization. SVR with thrombus removal did not improve all-cause mortality and cardiovascular outcomes in patients with LVT and LVA. Ventricular aneurysm with thrombus may not be an indication for surgery.

6 October 2024 in JMIR Preprints
Verified
Preprint
Open access
BSBoqun Shi
Boqun Shi
LCLiangguo Chen
Liangguo Chen
SPShuo Pang
Shuo Pang
YWYue Wang
Yue Wang
SWShen Wang
Shen Wang
XWXiaofan Wu
Xiaofan Wu
BACKGROUND Accurate mortality risk prediction is crucial for effective cardiovascular risk management. Recent advancements in artificial intelligence (AI) have demonstrated potential in this specific medical field. Qwen-2 and Llama-3 are high-performance, open-source large language models (LLMs) available online. An artificial neural network (ANN) algorithm derived from the SWEDEHEART (Swedish Web System for Enhancement and Development of Evidence-Based Care in Heart Disease Evaluated According to Recommended Therapies) registry, termed SWEDEHEART-AI, can predict patient prognosis following acute myocardial infarction (AMI). OBJECTIVE This study aims to evaluate the 3 models mentioned above in predicting 1-year all-cause mortality in critically ill patients with AMI. METHODS The Medical Information Mart for Intensive Care IV (MIMIC-IV) database is a publicly available data set in critical care medicine. We included 2758 patients who were first admitted for AMI and discharged alive. SWEDEHEART-AI calculated the mortality rate based on each patient’s 21 clinical variables. Qwen-2 and Llama-3 analyzed the content of patients’ discharge records and directly provided a 1-decimal value between 0 and 1 to represent 1-year death risk probabilities. The patients’ actual mortality was verified using follow-up data. The predictive performance of the 3 models was assessed and compared using the Harrell C-statistic (C-index), the area under the receiver operating characteristic curve (AUROC), calibration plots, Kaplan-Meier curves, and decision curve analysis. RESULTS SWEDEHEART-AI demonstrated strong discrimination in predicting 1-year all-cause mortality in patients with AMI, with a higher C-index than Qwen-2 and Llama-3 (C-index 0.72, 95% CI 0.69-0.74 vs C-index 0.65, 0.62-0.67 vs C-index 0.56, 95% CI 0.53-0.58, respectively; all P<.001 for both comparisons). SWEDEHEART-AI also showed high and consistent AUROC in the time-dependent ROC curve. The death rates calculated by SWEDEHEART-AI were positively correlated with actual mortality, and the 3 risk classes derived from this model showed clear differentiation in the Kaplan-Meier curve (P<.001). Calibration plots indicated that SWEDEHEART-AI tended to overestimate mortality risk, with an observed-to-expected ratio of 0.478. Compared with the LLMs, SWEDEHEART-AI demonstrated positive and greater net benefits at risk thresholds below 19%. CONCLUSIONS SWEDEHEART-AI, a trained ANN model, demonstrated the best performance, with strong discrimination and clinical utility in predicting 1-year all-cause mortality in patients with AMI from an intensive care cohort. Among the LLMs, Qwen-2 outperformed Llama-3 and showed moderate predictive value. Qwen-2 and SWEDEHEART-AI exhibited comparable classification effectiveness. The future integration of LLMs into clinical decision support systems holds promise for accurate risk stratification in patients with AMI; however, further research is needed to optimize LLM performance and address calibration issues across diverse patient populations.