HKHeejun Kang

Heejun Kang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Heejun Kang has a total of 48 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Cardiovascular Outcomes of Early LDL-C Goal Achievement in Patients with Very-High-Risk ASCVD

    by Ah-Ram Kim, Ji Ye Han, Minkyoung Kim, Hojoon Lee, Yeon-Hee Baek, Imjin Ahn, Hee Jun Kang, Ji Seon Oh, Ju Hyeon Kim, Osung Kwon, et al.

    Introduction Clinical trial evidence supports early low-density lipoprotein cholesterol (LDL-C) goal achievement in patients with atherosclerotic cardiovascular disease (ASCVD), but real-world evidence in Asia is lacking. We investigated the effects of early LDL-C goal achievement on recurrent major cardiovascular events (MACEs) among patients with very-high-risk ASCVD in South Korea. Methods We included adult patients hospitalized with ASCVD (acute coronary syndrome [ACS], stable angina, ischemic stroke, transient ischemic attack, peripheral arterial disease, or asymptomatic coronary artery disease) at a major Korean tertiary hospital from 2000 to 2020. Patients were categorized into early or non-early target LDL-C groups based on LDL-C measured 4–12 weeks post-discharge (< 55 vs. ≥ 55 mg/dl). The primary outcome was recurrent MACEs, including myocardial infarction, ischemic stroke, all-cause mortality, hospitalization for unstable angina, and coronary revascularization. The secondary outcome was health resource use (frequency/length of stay [LOS]). Results During follow-up (mean: 5 years), the early target LDL-C group (n = 5702) had a lower risk of recurrent MACEs compared with the non-early target LDL-C group (n = 11,232; adjusted hazard ratio [95% CI]: 0.89 [0.82–0.96]). The effect was most pronounced in patients with ACS (0.73 [0.63–0.85]), particularly for recurrent MACEs within 6 months (0.61 [0.44–0.87]). The early target LDL-C group had 19% lower frequency and 31% shorter LOS for cardiovascular-related hospitalizations than the non-early group. Conclusions Early LDL-C goal achievement was associated with lower recurrent MACEs in patients with very-high-risk ASCVD in South Korea, especially in patients with ACS. These findings underscore the importance of early LDL-C management in improving cardiovascular outcomes.


  • Fine-Tuning LLMs with Medical Data: Can Safety Be Ensured?

    by Minkyoung Kim, Yunha Kim, Hee Jun Kang, Hyeram Seo, Heejung Choi, JiYe Han, Gaeun Kee, Seohyun Park, Soyoung Ko, HyoJe Jung, et al.


  • Top Cited Papers
    Open Access

    Task-Specific Transformer-Based Language Models in Health Care: Scoping Review

    by Ha Na Cho, Tae Joon Jun, Young-Hak Kim, Heejun Kang, Imjin Ahn, Hansle Gwon, Yunha Kim, Jiahn Seo, Heejung Choi, Minkyoung Kim, et al.

    Background: Transformer-based language models have shown great potential to revolutionize health care by advancing clinical decision support, patient interaction, and disease prediction. However, despite their rapid development, the implementation of transformer-based language models in health care settings remains limited. This is partly due to the lack of a comprehensive review, which hinders a systematic understanding of their applications and limitations. Without clear guidelines and consolidated information, both researchers and physicians face difficulties in using these models effectively, resulting in inefficient research efforts and slow integration into clinical workflows. Objective: This scoping review addresses this gap by examining studies on medical transformer-based language models and categorizing them into 6 tasks: dialogue generation, question answering, summarization, text classification, sentiment analysis, and named entity recognition. Methods: We conducted a scoping review following the Cochrane scoping review protocol. A comprehensive literature search was performed across databases, including Google Scholar and PubMed, covering publications from January 2017 to September 2024. Studies involving transformer-derived models in medical tasks were included. Data were categorized into 6 key tasks. Results: Our key findings revealed both advancements and critical challenges in applying transformer-based models to health care tasks. For example, models like MedPIR involving dialogue generation show promise but face privacy and ethical concerns, while question-answering models like BioBERT improve accuracy but struggle with the complexity of medical terminology. The BioBERTSum summarization model aids clinicians by condensing medical texts but needs better handling of long sequences. Conclusions: This review attempted to provide a consolidated understanding of the role of transformer-based language models in health care and to guide future research directions. By addressing current challenges and exploring the potential for real-world applications, we envision significant improvements in health care informatics. Addressing the identified challenges and implementing proposed solutions can enable transformer-based language models to significantly improve health care delivery and patient outcomes. Our review provides valuable insights for future research and practical applications, setting the stage for transformative advancements in medical informatics.


  • Open Access

    Forecasting hospital room and ward occupancy using static and dynamic information concurrently (Preprint)

    by Hyeram Seo, Imjin Ahn, Hansle Gwon, Hee Jun Kang, Yunha Kim, Heejung Choi, Minkyoung Kim, Jiye Han, Gaeun Kee, Seohyun Park, et al.

    Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: Predicting bed occupancy rate (BOR) is essential for efficient hospital resource management, long-term budget planning, and patient care planning. Although macro-level BOR prediction for the entire hospital is crucial, predicting occupancy at a detailed level, such as specific wards and rooms, is more practical and useful for hospital scheduling. Objective: The aim of this study was to develop a web-based support tool that allows hospital administrators to grasp BOR for each ward and room according to different time periods. Methods: We trained long short-term memory (LSTM) time-series models using hourly aggregated individual bed data on a daily basis to predict BOR for each ward and room in the hospital. Wards were trained with two models based on 7- and 30-day time windows, and rooms were trained with 3- and 7-day time windows for shorter-term planning. To further improve prediction performance, we also added two models trained by concatenating dynamic data with static data representing room-specific details. Results: The ward-level prediction model with a mean absolute error (MAE) of 0.057, a mean squared error (MSE) of 0.007, a root mean squared error (RMSE) of 0.082, and an R2 score of 0.582. Among the room-level prediction models, the model that combined static data exhibited superior performance with an MAE of 0.123, an MSE of 0.051, an RMSE of 0.226, and an R2 score of 0.320. Model results can be displayed on an electronic dashboard for easy access via the web. Conclusions: We propose predictive BOR models for individual wards and rooms that demonstrate high performance. This result can be visualized through a web-based dashboard, aiding hospital administrators in bed operation planning. This contributes to resource optimization and the reduction of hospital resource utilization.


  • Open Access

    Forecasting Hospital Room and Ward Occupancy Using Static and Dynamic Information Concurrently: Retrospective Single-Center Cohort Study (Preprint)

    by Hyeram Seo, Imjin Ahn, Hansle Gwon, Heejun Kang, Yunha Kim, Heejung Choi, Minkyoung Kim, Jiye Han, Gaeun Kee, Seohyun Park, et al.

    BACKGROUND Predicting the bed occupancy rate (BOR) is essential for efficient hospital resource management, long-term budget planning, and patient care planning. Although macro-level BOR prediction for the entire hospital is crucial, predicting occupancy at a detailed level, such as specific wards and rooms, is more practical and useful for hospital scheduling. OBJECTIVE The aim of this study was to develop a web-based support tool that allows hospital administrators to grasp the BOR for each ward and room according to different time periods. METHODS We trained time-series models based on long short-term memory (LSTM) using individual bed data aggregated hourly each day to predict the BOR for each ward and room in the hospital. Ward training involved 2 models with 7- and 30-day time windows, and room training involved models with 3- and 7-day time windows for shorter-term planning. To further improve prediction performance, we added 2 models trained by concatenating dynamic data with static data representing room-specific details. RESULTS We confirmed the results of a total of 12 models using bidirectional long short-term memory (Bi-LSTM) and LSTM, and the model based on Bi-LSTM showed better performance. The ward-level prediction model had a mean absolute error (MAE) of 0.067, mean square error (MSE) of 0.009, root mean square error (RMSE) of 0.094, and R2 score of 0.544. Among the room-level prediction models, the model that combined static data exhibited superior performance, with a MAE of 0.129, MSE of 0.050, RMSE of 0.227, and R2 score of 0.600. Model results can be displayed on an electronic dashboard for easy access via the web. CONCLUSIONS We have proposed predictive BOR models for individual wards and rooms that demonstrate high performance. The results can be visualized through a web-based dashboard, aiding hospital administrators in bed operation planning. This contributes to resource optimization and the reduction of hospital resource use.

  • Open Access

    Task-Specific Transformer-Based Language Models in Health Care: Scoping Review (Preprint)

    by Ha Na Cho, Tae Joon Jun, Young-Hak Kim, Heejun Kang, Imjin Ahn, Hansle Gwon, Yunha Kim, Jiahn Seo, Heejung Choi, Minkyoung Kim, et al.

    BACKGROUND Transformer-based language models have shown great potential to revolutionize health care by advancing clinical decision support, patient interaction, and disease prediction. However, despite their rapid development, the implementation of transformer-based language models in health care settings remains limited. This is partly due to the lack of a comprehensive review, which hinders a systematic understanding of their applications and limitations. Without clear guidelines and consolidated information, both researchers and physicians face difficulties in using these models effectively, resulting in inefficient research efforts and slow integration into clinical workflows. OBJECTIVE This scoping review addresses this gap by examining studies on medical transformer-based language models and categorizing them into 6 tasks: dialogue generation, question answering, summarization, text classification, sentiment analysis, and named entity recognition. METHODS We conducted a scoping review following the Cochrane scoping review protocol. A comprehensive literature search was performed across databases, including Google Scholar and PubMed, covering publications from January 2017 to September 2024. Studies involving transformer-derived models in medical tasks were included. Data were categorized into 6 key tasks. RESULTS Our key findings revealed both advancements and critical challenges in applying transformer-based models to health care tasks. For example, models like MedPIR involving dialogue generation show promise but face privacy and ethical concerns, while question-answering models like BioBERT improve accuracy but struggle with the complexity of medical terminology. The BioBERTSum summarization model aids clinicians by condensing medical texts but needs better handling of long sequences. CONCLUSIONS This review attempted to provide a consolidated understanding of the role of transformer-based language models in health care and to guide future research directions. By addressing current challenges and exploring the potential for real-world applications, we envision significant improvements in health care informatics. Addressing the identified challenges and implementing proposed solutions can enable transformer-based language models to significantly improve health care delivery and patient outcomes. Our review provides valuable insights for future research and practical applications, setting the stage for transformative advancements in medical informatics.

  • Open Access

    Machine Learning Models to Predict the Warfarin Discharge Dosage Using Clinical Information of East Asian Inpatients: Retrospective Cohort Study (Preprint)

    by Heejung Choi, Young-Hak Kim, Tae Joon Jun, Hee Jun Kang, Imjin Ahn, Hansle Gwon, Yunha Kim, Hyeram Seo, Ha Na Cho, Minkyoung Kim, et al.

    BACKGROUND As warfarin has a narrow therapeutic window and obvious response variability among individuals, it is difficult to rapidly determine personalized warfarin dosage. Adverse drug events resulting from warfarin overdose can be critical. Our study aimed to develop a machine learning (ML) model that predicts the appropriate discharge dosage of warfarin using electronic medical records from a large hospital. Additionally, we externally validated the model to ensure its accuracy. OBJECTIVE This study aimed to develop a machine learning model that predicts individual warfarin dosage based on clinical data within 2 days of hospitalization. METHODS During this retrospective study, adult patients who were prescribed warfarin at a large hospital between January 1, 2018, and October 31, 2020, were recruited as a model development cohort (n=3,168). We externally validated the models using the Medical Information Mart for Intensive Care III (n=891). Variables for the warfarin dosage prediction were selected according to the clinical practice experience of cardiovascular physicians. The study outcome was the warfarin discharge dosage. Four ML models that predicted the proper warfarin discharge dosage were developed. We evaluated the model performance using the mean absolute error (MAE) and prediction accuracy. Finally, we compared the accuracy of the predictions of our models and the predictions of physicians to determine their clinical relevance. RESULTS The MAEs obtained using the internal validation set were as follows: XGBoost, 0.9; neural network, 0.9; random forest, 1.0; linear regression, 1.0; and physicians, 1.3. These values showed that our models had better prediction accuracy than the physicians, who have difficulty determining the warfarin discharge dosage using clinical information obtained within 2 days of hospitalization. CONCLUSIONS Our ML model could help physicians rapidly predict and decide the proper warfarin discharge dosage during hospitalization. Further work is required to determine model generalizability.

  • Open Access

    Explainable predictions of a machine learning model to forecast the postoperative length of stay for severe patients: Machine Learning Model Development and Evaluation (Preprint)

    by Ha Na Cho, Tae Joon Jun, Young-Hak Kim, Hee Jun Kang, Imjin Ahn, Hansle Gwon, Yunha Kim, Hyeram Seo, Heejung Choi, Minkyoung Kim, et al.

    BACKGROUND Understanding the length of stay of severe patients who require general anesthesia is key to enhancing health outcomes. OBJECTIVE Here, we aim to discover how machine learning can support resource allocation management and decision-making resulting from the length of stay prediction. METHODS A retrospective cohort study was conducted from January 2018 to October 2020. A total cohort of 240,000 patients’ medical records was collected. The data were collected exclusively for preoperative variables to accurately analyze the predictive factors impacting the postoperative length of stay. The main outcome of this study is an analysis of the length of stay (in days) after surgery until discharge. The prediction was performed with ridge regression, random forest, XGBoost, and multi-layer perceptron neural network models. The XGBoost resulted in the best performance with an average error within 3 days. Moreover, we explain each feature’s contribution over the XGBoost model and further display distinct predictors affecting the overall prediction outcome at the patient level. RESULTS The risk factors that most importantly contributed to the stay after surgery were as follows: a direct bilirubin laboratory test, department change, calcium chloride medication, gender, and diagnosis with the removal of other organs. Our results suggest that healthcare providers take into account the risk factors such as the laboratory blood test, distributing patients, and the medication prescribed prior to the surgery. We successfully predicted the length of stay after surgery and provide explainable models with supporting analyses. CONCLUSIONS In summary, we demonstrate the interpretation with the XGBoost model presenting insights on preoperative features and defining higher risk predictors to the length of stay outcome. Our development in explainable models supports the current in-depth knowledge for the future length of stay prediction on electronic medical records that aids the decision-making and facilitation of the operation department.

  • Open Access

    Heterogeneous Semantic Graph Construction and HinSAGE Learning from Electronic Medical Records (Preprint)

    by Ha Na Cho, Imjin Ahn, Hansle Gwon, Heejun Kang, Yunha Kim, Heejung Choi, Minkyoung Kim, Jiye Han, Gaeun Kee, Tae Joon Jun, et al.

    BACKGROUND Graph representations learning is a method for introducing how to effectively construct and learn patient embeddings using electronic medical records (EMR) datasets. Adapting the integration limits will support and advance the previous methods to predict the prognosis of patients in network models. OBJECTIVE This study aimed to address the challenge of implementing complex and large EMR, including the following: (1) demonstrating how to build a multi-attributed and multi-relational graph model (2) and applying a downstream disease prediction task of patient’s prognosis using HinSAGE algorithm. METHODS A total of 61,789 patients diagnosed with angina who visited the Asan Medical Center between 2000 and 2016 were included in the study. We illustrated two bipartite graph representations from real-world EMR data sets, to effectively visualize the patients’ representation and apply the heterogeneous graph neural network using the HinSAGE algorithm. Then, we performed the link prediction task for predicting the event occurrence of myocardial infarction, stroke, heart failure and death on binary edge attributes to evaluate the network performance. RESULTS The first constructed graph model contained 492,886 nodes with attributes and 439,045 edges for displaying 10 nodes and 9 edge types. The patient-centric graph database was then used to illustrate a query of a predictive network. Moreover, the second graph model is composed of 107,682 nodes and 53,841 edges with node and edge attributes for displaying 2 node and 2 edge types. Next, the model object underwent an experiment using the HinSAGE algorithm. As a result, the performance evaluation indicated that our heterogeneous graph model outperformed other baseline methods, achieving the area under a receiver operating characteristic curve (AUROC) and the area under precision-recall curve (AUPRC) measurements of 0.69 and 0.17, respectively. CONCLUSIONS The proposed implemented models successfully demonstrated the graph construction and graph representation learning on the EMR dataset to empower three roles: broaden EMR research, decision making in diagnosis, and personalized medicine. Future works may integrate additional data sources alongside the node and edge types to further improve the performance evaluation.

  • Open Access

    Electronic Medical Record–Based Machine Learning Approach to Predict the Risk of 30-Day Adverse Cardiac Events After Invasive Coronary Treatment: Machine Learning Model Development and Validation

    by Osung Kwon, Wonjun Na, Heejun Kang, Tae Joon Jun, Jihoon Kweon, Gyung-Min Park, YongHyun Cho, Cinyoung Hur, Jungwoo Chae, Do-Yoon Kang, et al.

    Background: Although there is a growing interest in prediction models based on electronic medical records (EMRs) to identify patients at risk of adverse cardiac events following invasive coronary treatment, robust models fully utilizing EMR data are limited. Objective: We aimed to develop and validate machine learning (ML) models by using diverse fields of EMR to predict the risk of 30-day adverse cardiac events after percutaneous intervention or bypass surgery. Methods: EMR data of 5,184,565 records of 16,793 patients at a quaternary hospital between 2006 and 2016 were categorized into static basic (eg, demographics), dynamic time-series (eg, laboratory values), and cardiac-specific data (eg, coronary angiography). The data were randomly split into training, tuning, and testing sets in a ratio of 3:1:1. Each model was evaluated with 5-fold cross-validation and with an external EMR-based cohort at a tertiary hospital. Logistic regression (LR), random forest (RF), gradient boosting machine (GBM), and feedforward neural network (FNN) algorithms were applied. The primary outcome was 30-day mortality following invasive treatment. Results: GBM showed the best performance with area under the receiver operating characteristic curve (AUROC) of 0.99; RF had a similar AUROC of 0.98. AUROCs of FNN and LR were 0.96 and 0.93, respectively. GBM had the highest area under the precision-recall curve (AUPRC) of 0.80, and the AUPRCs of RF, LR, and FNN were 0.73, 0.68, and 0.63, respectively. All models showed low Brier scores of 0.95, contributing majorly to the excellent results. Conclusions: Exploiting the diverse fields of the EMR data set, the ML-based 30-day adverse cardiac event prediction models demonstrated outstanding results, and the applied framework could be generalized for various health care prediction models.