Miryoung Kim

Prof. Miryoung Kim

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
19 February 2026 in Pharmaceuticals
Verified
Journal Article
Open access
PDF file
TKTae Woo Kim
Tae Woo Kim
SPSihyeon Park
Sihyeon Park
MKMiryoung Kim
Miryoung Kim
Background/Objectives: Artificial intelligence (AI) is increasingly applied to drug safety evaluation, yet evidence is dispersed across lifecycle stages and tasks. This scoping review aimed to (1) map how AI supports safety- and treatment-related decision types across the drug lifecycle, and (2) examine evaluation strategies used to assess model reliability for clinical or regulatory use. Methods: Using Arksey and O’Malley’s framework, we searched a major database for studies published in the past decade that applied AI or machine learning to drug safety or medication-related decisions. After screening, we extracted data on lifecycle stage, decision type, AI methods, data sources, and evaluation strategies. A lifecycle–decision matrix was constructed to characterize application patterns. Results: AI applications were concentrated in real-world clinical care × patient-level safety prediction and post-marketing × safety surveillance, using EHRs, spontaneous reporting systems, and clinical text. Common methods included gradient boosting, deep neural networks, graph neural networks, and natural language processing models. This concentration reflects structural incentives favoring safety-oriented applications with readily available data and lower decision liability. Evidence for treatment optimization, regulatory decision modeling, and evidence synthesis was limited. Most studies used internal validation; external validation and real-world deployment were uncommon, indicating early methodological maturity and limited translational readiness. Conclusions: AI demonstrates strong potential to enhance drug safety—particularly in risk prediction and pharmacovigilance—but its use remains uneven across the lifecycle. By situating AI applications within explicit lifecycle stages and decision contexts, this review clarifies where progress has advanced, where translation has stalled, and why these gaps persist. Limited external validation and minimal real-world testing constrain clinical and regulatory adoption. These findings suggest that external validation and real-world testing may contribute to further advances in AI for drug safety.

Verified
Journal Article
Open access
PDF file
HHHyun Jin Han
Hyun Jin Han
MKMiryoung Kim
Miryoung Kim
JLJi Yeon Lee
Ji Yeon Lee
HSHae Sun Suh
Hae Sun Suh
Background: Multiple myeloma (MM) is associated with high morbidity and mortality, with elevated rates of arterial thrombosis and venous thromboembolism (VTE) and ischemic stroke (IS). We aimed to estimate the incidence of VTE and IS categorized by the VTE risk grade among individuals with MM in Korea. Additionally, we explored the potential of the IMPEDE VTE score as a tool for assessing IS risk in patients with MM. Methods: This retrospective cohort study comprised 37,168 individuals aged ≥ 18 years newly diagnosed with MM between January 2008 and December 2021 using the representative claims database of the Korean population. The risk of the incidence of VTE and IS within 6 months after MM diagnosis was stratified based on high-risk (IMPEDE VTE score ≥ 8) and low-risk (<8) categories. The hazard ratios (HRs) were estimated using Cox proportional hazard models. Results: The VTE incidence was 120.4 per 1000 person-years and IS incidence was 149.3 per 1000 person-years. Statistically significant differences were observed in the cumulative incidence rates of VTE between groups with high and low VTE scores (p < 0.001) and between individuals aged ≤ 65 years (p < 0.001) and those with a Charlson comorbidity index (CCI) ≥ 3 compared to lower scores (p < 0.001). Additionally, the cumulative incidence rate of IS differed significantly across all groups (p < 0.001). The HR for the high-risk group in VTE and IS occurrence was 1.59 (95% CI, 1.26–2.00) and 3.47 (95% CI, 2.99–4.02), respectively. Conclusions: It is important to screen and manage high-risk groups for the early development of VTE or IS in patients with newly diagnosed MM.

20 January 2026 in Nutrition, Metabolism and Cardiovascular Diseases
Verified
Journal Article
Open access
MJMinseol Jang
Minseol Jang
MKMiryoung Kim
Miryoung Kim
HSHae Sun Suh
Hae Sun Suh
Sodium-glucose cotransporter-2 (SGLT2) inhibitors, a widely used class of oral antidiabetic agents provide additional cardioprotective and renoprotective benefits, but their use during pregnant remains limited. This study aimed to evaluate the association between SGLT2 inhibitor exposure during pregnancy and the risk of congenital malformations. This population-based cohort study utilized data from the National Health Insurance Service Database (2016-2022). Pregnancies with known teratogen exposure were excluded. SGLT2 inhibitor or insulin (as active comparator) use during the first trimester was defined as exposure. The primary outcomes were congenital malformations and heart defects. Propensity score matching controlled for confounders, and generalized linear regression estimated relative risks (RRs) with 95 % confidence intervals (CIs). Negative control outcomes were employed to assess residual confounding. Among 536,654 pregnancies, 121 pregnancies were exposed to SGLT2 inhibitors (mean [SD] age: 35.07 [4.26]), and 2007 to insulin (mean [SD] age: 34.89 [4.28]). Adjusted RRs (95 % CIs) for congenital malformations and heart defects were 0.88 (0.52-1.46) and 0.83 (0.44-1.58), respectively. In sensitivity analysis restricted to the organogenesis period (gestational weeks 4-10), risk of congenital heart defects was 2.79 (1.16-7.06). No residual confounding detected in negative control outcome. In this study, SGLT2 inhibitor use during the first trimester was not associated with an increased risk of congenital malformations. Nonetheless, the observed increased risk of congenital heart defects during the organogenesis period highlights the importance of exposure timing and warrants cautious interpretation. These findings provide evidence to guide clinical decision-making regarding antidiabetic medication use during pregnancy.

6 July 2026 in JMIR Preprints
Verified
Preprint
Open access
MKMiryoung Kim
Miryoung Kim
BACKGROUND Synthetic population and persona datasets can provide artificial individual-level records without exposing personal data. However, for surveillance use, they must preserve the joint profiles that define vulnerable groups, not only marginal distributions. OBJECTIVE This study evaluated whether a public statistics–based Korean synthetic persona dataset is consistent with Korea Community Health Survey (CHS) and Korean Statistical Information Service (KOSIS) benchmarks for auxiliary small-area public health surveillance. METHODS We compared 1,000,000 synthetic persona rows with 1,156,122 adult respondents from the 2021-2025 CHS and selected KOSIS municipal indicators. Core variables were age group, sex, marital status, living-alone status, and education; province or municipal area replaced marital status in the regional profiles. Economic activity and occupation were treated as sensitivity domains. We compared marginal and joint distributions using percentage-point differences and total variation distance (TVD) and quantified CHS sampling uncertainty using 500-replicate stratified cluster percentile bootstrap intervals and compared high-dimensional TVD with CHS split-half references. Sparse-cell, KOSIS correlation, and regional composition-standardized analyses assessed robustness and potential downstream use. RESULTS Persona rows were close to CHS for sex (TVD<0.001, 95% bootstrap percentile interval 0.000-0.001), province (TVD 0.006, bootstrap percentile interval 0.005-0.006), living-alone status (TVD 0.008, bootstrap percentile interval 0.007-0.009), and age group (TVD 0.013, bootstrap percentile interval 0.012-0.014). Differences were larger for marital status (TVD 0.039, bootstrap percentile interval 0.038-0.040) and education (TVD 0.059, bootstrap percentile interval 0.058-0.061). The national core profile had a TVD of 0.106 (bootstrap percentile interval 0.105-0.107), compared with a CHS split-half median of 0.009 (2.5th-97.5th percentile 0.008-0.011). The municipal profile had an observed TVD of 0.117, a bootstrap mean of 0.122 (95% bootstrap percentile interval 0.120-0.123), and a split-half median of 0.082 (2.5th-97.5th percentile 0.080-0.084). After restriction to CHS cells with n≥100, the municipal TVD decreased to 0.091. At the municipal level, CHS- and persona-composition standardized unmet health care rates differed by a mean of 0.109 percentage points (Pearson r=0.567), but standardization with national profile-specific probabilities did not reproduce observed local CHS rates (Pearson r=0.004 for persona composition). The conditional KOSIS living-alone comparison was modest (Pearson r=0.547). CONCLUSIONS The synthetic persona data showed close agreement for several basic demographic margins but less consistent agreement for high-dimensional sociodemographic profiles. They may support composition-based scenario development and dashboard prototyping but should not be used as stand-alone sources for local surveillance or health-outcome inference without additional calibration and validation.

9 November 2026 in JMIR Public Health and Surveillance
Verified
Journal Article
Open access
MKMiryoung Kim
Miryoung Kim
Background: Synthetic population and persona datasets can provide artificial individual-level records without exposing personal data. However, for surveillance use, they must preserve the joint profiles that define vulnerable groups, not only marginal distributions. Objective: This study evaluated whether a public statistics–based Korean synthetic persona dataset is consistent with Korea Community Health Survey (CHS) and Korean Statistical Information Service (KOSIS) benchmarks for auxiliary small-area public health surveillance. Methods: We compared 1,000,000 synthetic persona rows with 1,156,122 adult respondents from the 2021-2025 CHS and selected KOSIS municipal indicators. Core variables were age group, sex, marital status, living-alone status, and education; province or municipal area replaced marital status in the regional profiles. Economic activity and occupation were treated as sensitivity domains. We compared marginal and joint distributions using percentage-point differences and total variation distance (TVD) and quantified CHS sampling uncertainty using 500-replicate stratified cluster percentile bootstrap intervals and compared high-dimensional TVD with CHS split-half references. Sparse-cell, KOSIS correlation, and regional composition-standardized analyses assessed robustness and potential downstream use. Results: Persona rows were close to CHS for sex (TVD<0.001, 95% bootstrap percentile interval 0.000-0.001), province (TVD 0.006, bootstrap percentile interval 0.005-0.006), living-alone status (TVD 0.008, bootstrap percentile interval 0.007-0.009), and age group (TVD 0.013, bootstrap percentile interval 0.012-0.014). Differences were larger for marital status (TVD 0.039, bootstrap percentile interval 0.038-0.040) and education (TVD 0.059, bootstrap percentile interval 0.058-0.061). The national core profile had a TVD of 0.106 (bootstrap percentile interval 0.105-0.107), compared with a CHS split-half median of 0.009 (2.5th-97.5th percentile 0.008-0.011). The municipal profile had an observed TVD of 0.117, a bootstrap mean of 0.122 (95% bootstrap percentile interval 0.120-0.123), and a split-half median of 0.082 (2.5th-97.5th percentile 0.080-0.084). After restriction to CHS cells with n≥100, the municipal TVD decreased to 0.091. At the municipal level, CHS- and persona-composition standardized unmet health care rates differed by a mean of 0.109 percentage points (Pearson r=0.567), but standardization with national profile-specific probabilities did not reproduce observed local CHS rates (Pearson r=0.004 for persona composition). The conditional KOSIS living-alone comparison was modest (Pearson r=0.547). Conclusions: The synthetic persona data showed close agreement for several basic demographic margins but less consistent agreement for high-dimensional sociodemographic profiles. They may support composition-based scenario development and dashboard prototyping but should not be used as stand-alone sources for local surveillance or health-outcome inference without additional calibration and validation.

1 September 2026 in Journal of Comparative Effectiveness Research
Verified
Journal Article
Open access
HHHyun Jin Han
Hyun Jin Han
MKMiryoung Kim
Miryoung Kim
EWEyoung Wang
Eyoung Wang
MKMin Ji Kim
Min Ji Kim
HSHae Sun Suh
Hae Sun Suh
Aim: This nationwide cohort study primarily aimed to descriptively characterize real-world treatment patterns, including treatment initiation, adherence, treatment duration and discontinuation of the first treatment drug, among patients who experienced a recurrent osteoporotic fracture. Materials & methods: This population-based, retrospective cohort study used data from the Health Insurance Review and Assessment Database of South Korea. It included male and female patients aged 55 years and older who experienced a recurrent osteoporotic fracture within 2 years of their index fracture in 2013. Results: The study period was from 1 January 2012 to 31 December 2021, with an outcome assessment period from 1 January 2013 to 31 December 2017. A total of 34,558 patients with recurrent osteoporotic fractures were observed, and just over half (n = 18,462, 53.4%) received osteoporosis medication, with a significant difference in medication rates between males (23.7%) and females (58.8%) (p < 0.001). The estimated median duration of osteoporosis medication was 146 days (interquartile range: 61–365 days). The discontinuation rate of medication within 2 years of follow-up after their recurrent fracture was 80.5%, with the highest discontinuation rate (90.4%) observed among patients taking daily oral bisphosphonate medication. Conclusion: Although patients with recurrent fractures require intensive management, our results showed a significant unmet need in the initiation and persistence of osteoporosis medication. Therefore, it is crucial to establish a strategic treatment approach to address the treatment gap in managing osteoporosis among very high-risk patients. Dataset name: Health Insurance Review and Assessment Database of South Korea.