Joonki Lee
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Joonki Lee has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Aesun Shin
Department of Preventive Medicine, Seoul National University College of Medicine
8 shared publications - Sunho Choe
Department of Preventive Medicine, Seoul National University College of Medicine, Seoul, Korea
3 shared publications - Daehee Kang
Department of Preventive Medicine, Seoul National University College of Medicine, Seoul, Korea
1 shared publication - Dan Huang
Department of Preventive Medicine, Seoul National University College of Medicine, Seoul 03080, Korea
1 shared publication - Daye Park
Department of Internal Medicine, Severance Hospital, Yonsei University College of Medicine, Seoul, Korea
1 shared publication Dongwook Shin
Department of Family Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea
1 shared publication
Latest Publications
- Open Access
- 1 January 2026
- Published by XMLink in Yonsei Medical Journal
- Vol. 67 (5) , 400-407
Epidemiological Trends in Inflammatory Bowel Disease in Korea: Age-Related Shifts and Urban-Rural Disparities in a Nationwide Study, 2004–2015
Lee J, et al. Yonsei Med J. 2026 ;67:e20. https://doi.org/10.3349/ymj.2025.0327
- 1 January 2026
- Published by XMLink in Yonsei Medical Journal
- Vol. 67 (5) , 400-407
Open AccessEpidemiological Trends in Inflammatory Bowel Disease in Korea: Age-Related Shifts and Urban-Rural Disparities in a Nationwide Study, 2004–2015
Lee J, et al. Yonsei Med J. 2026 ;67:e20. https://doi.org/10.3349/ymj.2025.0327
- Open Access
- 12 November 2025
- Published by Springer Nature in Discover Oncology
- Vol. 16 (1)
Prognostic factors in gastric cancer survivors comparing Cox proportional hazards and machine learning using a Korean National cohort
Background As the prognosis of gastric cancer has improved, the exploration of prognostic factors has become increasingly important. This study aimed to identify prognostic factors of gastric cancer using machine learning and statistical methods and to compare the effectiveness of different methodologies in identifying prognostic factors. Methods We conducted a retrospective cohort study of cancer research data from survivors of gastric cancer in Korea. Patients were followed up from the date of curative treatment of gastric cancer to the date of recurrence, cancer-specific death, or censoring. The Cox proportional hazards, random survival forest, XGBoost, and DeepSurv models were used to calculate the risk of recurrence and cancer-specific death. All the models were trained on 80% of the training set, and the concordance index was used for comparison with 20% of the test set. The SHAP value was used for variable interpretation in the machine learning models. Results A total of 11,029 gastric cancer survivors with a median follow-up time of 6.19 years were included. Remnant stomach after gastric cancer treatment, T stage and N stage were the most important features for recurrence and mortality according to both the Cox model and the machine learning model. All the models had a concordance index greater than 0.7 without large differences. Conclusions The machine learning model is not inferior to conventional statistical analysis models and offers greater flexibility, especially when statistical assumptions are violated. The key prognostic factors identified through this approach include residual stomach after treatment and cancer stage.
- 12 November 2025
- Published by Springer Nature in Discover Oncology
- Vol. 16 (1)
Open AccessPrognostic factors in gastric cancer survivors comparing Cox proportional hazards and machine learning using a Korean National cohort
Background As the prognosis of gastric cancer has improved, the exploration of prognostic factors has become increasingly important. This study aimed to identify prognostic factors of gastric cancer using machine learning and statistical methods and to compare the effectiveness of different methodologies in identifying prognostic factors. Methods We conducted a retrospective cohort study of cancer research data from survivors of gastric cancer in Korea. Patients were followed up from the date of curative treatment of gastric cancer to the date of recurrence, cancer-specific death, or censoring. The Cox proportional hazards, random survival forest, XGBoost, and DeepSurv models were used to calculate the risk of recurrence and cancer-specific death. All the models were trained on 80% of the training set, and the concordance index was used for comparison with 20% of the test set. The SHAP value was used for variable interpretation in the machine learning models. Results A total of 11,029 gastric cancer survivors with a median follow-up time of 6.19 years were included. Remnant stomach after gastric cancer treatment, T stage and N stage were the most important features for recurrence and mortality according to both the Cox model and the machine learning model. All the models had a concordance index greater than 0.7 without large differences. Conclusions The machine learning model is not inferior to conventional statistical analysis models and offers greater flexibility, especially when statistical assumptions are violated. The key prognostic factors identified through this approach include residual stomach after treatment and cancer stage.
- Open Access
- 15 July 2021
- Published by The Editorial Office of Gut and Liver in Gut and Liver
- Vol. 15 (4) , 537-545
A Nationwide Cohort Study Shows a Sex-Dependent Change in the Trend of Peptic Ulcer Bleeding Incidence in Korea between 2006 and 2015
Yong Sung Kim, Joonki Lee, Aesun Shin, Jung Min Lee, Jong Heon Park, and Hwoon-Yong Jung. Gut and Liver 2021;15:537-45. https://doi.org/10.5009/gnl20079
- 15 July 2021
- Published by The Editorial Office of Gut and Liver in Gut and Liver
- Vol. 15 (4) , 537-545
Open AccessA Nationwide Cohort Study Shows a Sex-Dependent Change in the Trend of Peptic Ulcer Bleeding Incidence in Korea between 2006 and 2015
Yong Sung Kim, Joonki Lee, Aesun Shin, Jung Min Lee, Jong Heon Park, and Hwoon-Yong Jung. Gut and Liver 2021;15:537-45. https://doi.org/10.5009/gnl20079
- Open Access
- 30 September 2020
- Published by Korean Society of Cancer Prevention in Journal of Cancer Prevention
- Vol. 25 (3) , 164-172
Gallstones, Cholecystectomy and the Risk of Hepatobiliary and Pancreatic Cancer: A Nationwide Population-based Cohort Study in Korea
Dan Huang, Joonki Lee, Nan Song, Sooyoung Cho, Sunho Choe, Aesun Shin. J Cancer Prev 2020;25:164-72. https://doi.org/10.15430/JCP.2020.25.3.164
- 30 September 2020
- Published by Korean Society of Cancer Prevention in Journal of Cancer Prevention
- Vol. 25 (3) , 164-172
Open AccessGallstones, Cholecystectomy and the Risk of Hepatobiliary and Pancreatic Cancer: A Nationwide Population-based Cohort Study in Korea
Dan Huang, Joonki Lee, Nan Song, Sooyoung Cho, Sunho Choe, Aesun Shin. J Cancer Prev 2020;25:164-72. https://doi.org/10.15430/JCP.2020.25.3.164
National cancer screening program for gastric cancer in Korea: Nationwide treatment benefit and cost
Background: The purpose of this study was to evaluate the nationwide benefit and cost of the national cancer screening program (NCSP) for gastric cancer treatment. Methods: For this nationwide, population‐based study, the Korean National Health Insurance Big Data Base, which included gastric cancer–related treatment information and the costs for all patients with gastric cancer who were 40 years old or older between 2004 and 2013, was restructured. Patients with gastric cancer who participated in the NCSP at least once (the screening group) were compared with those who did not participate in the NCSP (the nonscreening group). Results: The screening group (n = 116,775) spent significantly less on medical care expenses than the nonscreening group (n = 74,927) during the 5 years since the initial treatment ( P < .0001). The screening group presented a significantly better prognosis for 5 and 9 years than the nonscreening group ( P < .0001). The screening group revealed a 41% decreased hazard ratio ( P < .0001) for death in comparison with the nonscreening group; the prognostic benefit became more obvious when treatment was started within the first 4 months after screening. The age‐standardized mortality rate ratio of the screening group versus the nonscreening group was 0.62 ( P < .0001). The NCSP for gastric cancer required an average of 22,169,769 Korean Republic won (US $20,309) for 1 life‐year saved, which was less than the average gross domestic product (GDP) per capita in Korea. Conclusions: The screening group had significantly lower medical care expenses and showed a significantly better prognosis than the nonscreening group. On the basis of the GDP per capita, the NCSP for gastric cancer was cost‐effective for treatment prognosis.
National cancer screening program for gastric cancer in Korea: Nationwide treatment benefit and cost
Background: The purpose of this study was to evaluate the nationwide benefit and cost of the national cancer screening program (NCSP) for gastric cancer treatment. Methods: For this nationwide, population‐based study, the Korean National Health Insurance Big Data Base, which included gastric cancer–related treatment information and the costs for all patients with gastric cancer who were 40 years old or older between 2004 and 2013, was restructured. Patients with gastric cancer who participated in the NCSP at least once (the screening group) were compared with those who did not participate in the NCSP (the nonscreening group). Results: The screening group (n = 116,775) spent significantly less on medical care expenses than the nonscreening group (n = 74,927) during the 5 years since the initial treatment ( P < .0001). The screening group presented a significantly better prognosis for 5 and 9 years than the nonscreening group ( P < .0001). The screening group revealed a 41% decreased hazard ratio ( P < .0001) for death in comparison with the nonscreening group; the prognostic benefit became more obvious when treatment was started within the first 4 months after screening. The age‐standardized mortality rate ratio of the screening group versus the nonscreening group was 0.62 ( P < .0001). The NCSP for gastric cancer required an average of 22,169,769 Korean Republic won (US $20,309) for 1 life‐year saved, which was less than the average gross domestic product (GDP) per capita in Korea. Conclusions: The screening group had significantly lower medical care expenses and showed a significantly better prognosis than the nonscreening group. On the basis of the GDP per capita, the NCSP for gastric cancer was cost‐effective for treatment prognosis.
- Open Access
- 30 November 2019
- Published by Korean Society for Preventive Medicine in Journal of Preventive Medicine and Public Health
- Vol. 52 (6) , 377-383
Validity of Self-reported Stroke and Myocardial Infarction in Korea: The Health Examinees (HEXA) Study
Self-reported disease history is often used in epidemiological studies. In this study, we acquired the hospital records of subjects who self-reported stroke or myocardial infarction (MI) and evaluated the validity of the participants' self-reported disease history. We also determined the level of agreement between specialists and non-specialists. Among the participants in the Health Examinees study, 1488 subjects self-reported stroke or MI during 2012-2017, and medical records were acquired for the 429 subjects (28.8%) who agreed to share their medical information. Each record was independently assigned to 2 medical doctors for review. The records were classified as 'definite,' 'possible,' or 'not' stroke or MI. If the doctors did not agree, a third doctor made the final decision. The positive predictive value (PPV) of self-reporting was calculated with the doctors' review as the gold standard. Kappa statistics were used to compare the results between general doctors and neurologists or cardiologists. Medical records from 208 patients with self-reported stroke and 221 patients with self-reported MI were reviewed. The PPV of self-reported disease history was 51.4% for stroke and 32.6% for MI. If cases classified as 'possible' were counted as positive diagnoses, the PPV was 59.1% for stroke and 33.5% for MI. Kappa statistics showed moderate levels of agreement between specialists and non-specialists for both stroke and MI. The validity of self-reported disease was lower than expected, especially in those who reported having been diagnosed with MI. Proper consideration is needed when using these self-reported data in further studies.
- 30 November 2019
- Published by Korean Society for Preventive Medicine in Journal of Preventive Medicine and Public Health
- Vol. 52 (6) , 377-383
Open AccessValidity of Self-reported Stroke and Myocardial Infarction in Korea: The Health Examinees (HEXA) Study
Self-reported disease history is often used in epidemiological studies. In this study, we acquired the hospital records of subjects who self-reported stroke or myocardial infarction (MI) and evaluated the validity of the participants' self-reported disease history. We also determined the level of agreement between specialists and non-specialists. Among the participants in the Health Examinees study, 1488 subjects self-reported stroke or MI during 2012-2017, and medical records were acquired for the 429 subjects (28.8%) who agreed to share their medical information. Each record was independently assigned to 2 medical doctors for review. The records were classified as 'definite,' 'possible,' or 'not' stroke or MI. If the doctors did not agree, a third doctor made the final decision. The positive predictive value (PPV) of self-reporting was calculated with the doctors' review as the gold standard. Kappa statistics were used to compare the results between general doctors and neurologists or cardiologists. Medical records from 208 patients with self-reported stroke and 221 patients with self-reported MI were reviewed. The PPV of self-reported disease history was 51.4% for stroke and 32.6% for MI. If cases classified as 'possible' were counted as positive diagnoses, the PPV was 59.1% for stroke and 33.5% for MI. Kappa statistics showed moderate levels of agreement between specialists and non-specialists for both stroke and MI. The validity of self-reported disease was lower than expected, especially in those who reported having been diagnosed with MI. Proper consideration is needed when using these self-reported data in further studies.
- Open Access
- 30 November 2018
- Published by Korean Society for Preventive Medicine in Journal of Preventive Medicine and Public Health
- Vol. 51 (6) , 281-288
The Risk of Colorectal Cancer After Cholecystectomy or Appendectomy: A Population-based Cohort Study in Korea
We investigated the association between cholecystectomy or appendectomy and the subsequent risk of colorectal cancer (CRC) in the Korean population. A retrospective cohort study was conducted with the National Health Insurance Service-National Sample Cohort of Korea; this sample was followed up from January 1, 2002, until the date of CRC incidence, loss to follow-up, or December 31, 2015. The exposure status of cholecystectomy and appendectomy was treated as a time-varying covariate. The calculated risk of CRC was stratified by follow-up period, and the association between these surgical procedures and CRC was investigated by a Cox regression model applying appropriate lag periods. A total of 707 663 individuals were identified for analysis. The study population was followed up for an average of 13.66 years, and 4324 CRC cases were identified. The hazard ratio (HR) of CRC was elevated in the first year after cholecystectomy (HR, 1.71; 95% confidence interval [CI], 1.01 to 2.89) and in the first year and 2-3 years after appendectomy (HR, 4.22; 95% CI, 2.87 to 6.20; HR, 2.34; 95% CI, 1.36 to 4.03, respectively). The HRs of CRC after applying 1 year of lag after cholecystectomy and 3 years of lag after appendectomy were 0.80 (95% CI, 0.57 to 1.13) and 0.77 (95% CI, 0.51 to 1.16), respectively. The risk of CRC increased in the first year after cholecystectomy and appendectomy, implying the possibility of bias. When appropriate lag periods after surgery were applied, no association was found between cholecystectomy or appendectomy and CRC.
- 30 November 2018
- Published by Korean Society for Preventive Medicine in Journal of Preventive Medicine and Public Health
- Vol. 51 (6) , 281-288
Open AccessThe Risk of Colorectal Cancer After Cholecystectomy or Appendectomy: A Population-based Cohort Study in Korea
We investigated the association between cholecystectomy or appendectomy and the subsequent risk of colorectal cancer (CRC) in the Korean population. A retrospective cohort study was conducted with the National Health Insurance Service-National Sample Cohort of Korea; this sample was followed up from January 1, 2002, until the date of CRC incidence, loss to follow-up, or December 31, 2015. The exposure status of cholecystectomy and appendectomy was treated as a time-varying covariate. The calculated risk of CRC was stratified by follow-up period, and the association between these surgical procedures and CRC was investigated by a Cox regression model applying appropriate lag periods. A total of 707 663 individuals were identified for analysis. The study population was followed up for an average of 13.66 years, and 4324 CRC cases were identified. The hazard ratio (HR) of CRC was elevated in the first year after cholecystectomy (HR, 1.71; 95% confidence interval [CI], 1.01 to 2.89) and in the first year and 2-3 years after appendectomy (HR, 4.22; 95% CI, 2.87 to 6.20; HR, 2.34; 95% CI, 1.36 to 4.03, respectively). The HRs of CRC after applying 1 year of lag after cholecystectomy and 3 years of lag after appendectomy were 0.80 (95% CI, 0.57 to 1.13) and 0.77 (95% CI, 0.51 to 1.16), respectively. The risk of CRC increased in the first year after cholecystectomy and appendectomy, implying the possibility of bias. When appropriate lag periods after surgery were applied, no association was found between cholecystectomy or appendectomy and CRC.
- Open Access
- 24 September 2018
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 13 (9) , e0204413
Use of sedative-hypnotics and the risk of Alzheimer’s dementia: A retrospective cohort study
There has been a growing interest in the relationship between sedative-hypnotics use and the risk of Alzheimer’s dementia (AD) risk. This study aimed to evaluate the risk of AD associated with the use of sedative-hypnotics. A retrospective cohort study was conducted with randomly selected 5% samples from ≥50 years old beneficiaries of National Health Insurance Service (NHIS) of Korea from January 2002 to December 2015. The exposure to sedative-hypnotics was defined when prescribed over 30 defined daily dose (DDD) after January 2004 and it was categorized by prescribed dosage, types and half-lives of benzodiazepines. Time-dependent Cox regression model with a lag period of 5-years was used to evaluate the association between use of sedative-hypnotics and the risk of subsequent AD. Sensitivity analysis was performed for restricting sedative-hypnotics only when prescribed with insomnia. A total of 268,170 subjects were identified and subjects exposed to sedative-hypnotics showed a higher risk of AD (HR: 1.79; 95% CI: 1.72–1.86) than those who were not. There was an increased risk of AD among subjects exposed to benzodiazepines or zolpidem (HR: 1.75; 95% CI: 1.67–1.82) and antidepressants or low-dose antipsychotics (HR: 1.63; 95% CI: 1.42–1.87). The risk of AD was increased regardless of dose of sedative-hypnotics and half-life among benzodiazepines, especially in exposure to more than 360 DDD of sedative-hypnotics (HR: 1.78; 95% CI: 1.60–1.99) and the long-acting benzodiazepine (HR:1.77; 95% CI: 1.65–1.89).
- 24 September 2018
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 13 (9) , e0204413
Open AccessUse of sedative-hypnotics and the risk of Alzheimer’s dementia: A retrospective cohort study
There has been a growing interest in the relationship between sedative-hypnotics use and the risk of Alzheimer’s dementia (AD) risk. This study aimed to evaluate the risk of AD associated with the use of sedative-hypnotics. A retrospective cohort study was conducted with randomly selected 5% samples from ≥50 years old beneficiaries of National Health Insurance Service (NHIS) of Korea from January 2002 to December 2015. The exposure to sedative-hypnotics was defined when prescribed over 30 defined daily dose (DDD) after January 2004 and it was categorized by prescribed dosage, types and half-lives of benzodiazepines. Time-dependent Cox regression model with a lag period of 5-years was used to evaluate the association between use of sedative-hypnotics and the risk of subsequent AD. Sensitivity analysis was performed for restricting sedative-hypnotics only when prescribed with insomnia. A total of 268,170 subjects were identified and subjects exposed to sedative-hypnotics showed a higher risk of AD (HR: 1.79; 95% CI: 1.72–1.86) than those who were not. There was an increased risk of AD among subjects exposed to benzodiazepines or zolpidem (HR: 1.75; 95% CI: 1.67–1.82) and antidepressants or low-dose antipsychotics (HR: 1.63; 95% CI: 1.42–1.87). The risk of AD was increased regardless of dose of sedative-hypnotics and half-life among benzodiazepines, especially in exposure to more than 360 DDD of sedative-hypnotics (HR: 1.78; 95% CI: 1.60–1.99) and the long-acting benzodiazepine (HR:1.77; 95% CI: 1.65–1.89).