Yeijin Kim

Yeijin Kim

Auto-generated profile, based on publicly available information

Yeijin Kim

College of Pharmacy, Kyung Hee University, Seoul, Republic of Korea

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
17 April 2025 in Cancer Control
Verified
Journal Article
Open access
NLNguyen Le
Nguyen Le
SHSola Han
Sola Han
AKAhmed Saleh Kenawy
Ahmed Saleh Kenawy
YKYeijin Kim
Yeijin Kim
CPChanhyun Park
Chanhyun Park
Background: Hospitalized patients with blood cancer face an elevated risk for cardiovascular diseases caused by cardiotoxic cancer therapies, which can lead to cardiovascular-related unplanned readmissions. Objective: We aimed to develop a machine learning (ML) model to predict 90-day unplanned readmissions for major adverse cardiovascular events (MACE) in hospitalized patients with blood cancers. Design: A retrospective population-based cohort study. Methods: We analyzed patients aged ≥18 with blood cancers (leukemia, lymphoma, myeloma) using the Nationwide Readmissions Database. MACE included acute myocardial infarction, ischemic heart disease, stroke, heart failure, revascularization, malignant arrhythmias, and cardiovascular-related death. Six ML algorithms (L2-Logistic regression, Support Vector Machine, Complement Naïve Bayes, Random Forest, XGBoost, and CatBoost) were trained on 2017-2018 data and tested on 2019 data. The SuperLearner algorithm was used for stacking models. Cost-sensitive learning addressed data imbalance, and hyperparameters were tuned using 5-fold cross-validation with Optuna framework. Performance metrics included the Area Under the Receiver Operating Characteristics Curve (ROCAUC), Precision-Recall AUC (PRAUC), balanced Brier score, and F2 score. SHapley Additive exPlanations (SHAP) values assessed feature importance, and clustering analysis identified high-risk subpopulations. Results: Among 76 957 patients, 1031 (1.34%) experienced unplanned 90-day MACE-related readmissions. CatBoost achieved the highest ROCAUC (0.737, 95% CI: 0.712-0.763) and PRAUC (0.040, 95% CI: 0.033-0.050). The SuperLearner algorithm achieved slight improvements in most performance metrics. Four leading predictive features were consistently identified across algorithms, including older age, heart failure, coronary atherosclerosis, and cardiac dysrhythmias. Twenty-three clusters were determined with the highest-risk cluster (mean log odds of 1.41) identified by nonrheumatic/unspecified valve disorders, coronary atherosclerosis, and heart failure. Conclusions: Our ML model effectively predicts MACE-related readmissions in hospitalized patients with blood cancers, highlighting key predictors. Targeted discharge strategies may help reduce readmissions and alleviate the associated healthcare burden.

1 September 2025 in Cancer Control
Verified
Journal Article
Open access
YKYeijin Kim
Yeijin Kim
YLYan Liu
Yan Liu
HSHae Sun Suh
Hae Sun Suh
CPChanhyun Park
Chanhyun Park
Introduction: Despite improved outcomes with PD-1 inhibitors (PD-1i) in advanced non–small cell lung cancer (NSCLC), the determinants of cancer-, cardiovascular disease (CVD)-, and other-cause mortality remain poorly characterized in older adults. This study applies competing-risks methods to estimate cause-specific mortality and to identify predictors for each cause in this population. Methods: A retrospective cohort study was conducted using the 2014-2019 SEER-Medicare linked database, including patients aged 65 years or older with advanced NSCLC who received PD-1i (nivolumab or pembrolizumab). Mortality outcomes included deaths from CVD, NSCLC, other cancers, and other diseases. Predictors included treatment-related factors (e.g., PD-1i type), demographic factors (e.g., age, sex), socioeconomic status (e.g., Medicaid dual eligibility), cancer-related factors (e.g., NSCLC stage), and comorbidities (e.g., congestive heart failure [CHF]). Competing risk analyses were performed using the Fine-Gray model, with cause-specific Cox models for sensitivity analyses. Results: Among 5076 patients, 68.36% received nivolumab and 31.64% received pembrolizumab. Of 3746 deaths, most were from NSCLC (85.34%), while 2.80% were from CVD. No significant difference in CVD mortality risk was observed between the two PD-1i (sub-distribution HR [sHR] = 1.08; 95% CI: 0.63-1.84), but NSCLC mortality was associated with a lower risk in the pembrolizumab group (sHR = 0.67; 95% CI: 0.60-0.74) compared to the nivolumab group. A history of CHF (sHR = 2.10; 95% CI: 1.37-3.21) and Medicaid dual eligibility (sHR = 2.70 vs private insurance; 95% CI: 1.28-5.56) were associated with increased CVD mortality. NSCLC mortality was higher in Stage IV/distant than in Stage IIIB/regional (sHR = 1.24; 95% CI: 1.13-1.35) and males (sHR = 1.11; 95% CI: 1.03-1.20). Conclusions: These results highlight the potential value of integrated cardio-oncology models and strategies addressing socioeconomic inequities among older adults receiving PD-1i. However, conclusions should be tempered by the SEER–Medicare data structure, including lack of biomarker information and restricted applicability to younger or non-Medicare populations.