Background: Large language models (LLMs) have emerged as promising tools for addressing global disparities in mental health care. However, cloud-based proprietary models raise concerns about data privacy and limited adaptability to local health care systems. In contrast, open-source LLMs offer several advantages, including enhanced data security, the ability to operate offline in resource-limited settings, and greater adaptability to non-English clinical environments. Nevertheless, their performance in psychiatric applications involving non-English language inputs remains largely unexplored. Objective: This study aimed to systematically evaluate the clinical reasoning capabilities and diagnostic accuracy of a locally deployable open-source LLM in both Korean and English psychiatric contexts. Methods: The openbuddy-mistral-7b-v13.1 model, fine-tuned from Mistral 7B to enable conversational capabilities in Korean, was selected. A total of 200 deidentified psychiatric interview notes, documented during initial assessments of emergency department patients, were randomly selected from the electronic medical records of a tertiary hospital in South Korea. The dataset included 50 cases each of schizophrenia, bipolar disorder, depressive disorder, and anxiety disorder. The model translated the Korean notes into English and was prompted to extract 5 clinically meaningful diagnostic clues and generate the 2 most likely diagnoses using both the original Korean and translated English inputs. The hallucination rate and clinical relevance of the generated clues were manually evaluated. Top-1 and top-2 diagnostic accuracy were assessed by comparing the model’s prediction with the ground truth labels. Additionally, the model’s performance on a structured diagnostic task was evaluated using the psychiatry section of the Korean Medical Licensing Examination and its English-translated version. Results: The model generated 997 clues from Korean interview notes and 1003 clues from English-translated notes. Hallucinations were more frequent with Korean input (n=301, 30.2%) than with English (n=134, 13.4%). Diagnostic relevance was also higher in English (n=429, 42.8%) compared to Korean (n=341, 34.2%). The model showed significantly higher top-1 diagnostic accuracy with English input (74.5% vs 59%; P<.001), while top-2 accuracy was comparable (89.5% vs 90%; P=.56). Across 115 questions from the medical licensing examination, the model performed better in English (n=53, 46.1%) than in Korean (n=37, 32.2%), with superior results in 7 of 11 diagnostic categories. Conclusions: This study provides an in-depth evaluation of an open-source LLM in multilingual psychiatric settings. The model’s performance varied notably by language, with English input consistently outperforming Korean. These findings highlight the importance of assessing LLMs in diverse linguistic and clinical contexts. To ensure equitable mental health artificial intelligence, further development of high-quality psychiatric datasets in underrepresented languages and culturally adapted training strategies will be essential.

Background/Objectives: Evidence on the renal benefits and safety of sodium–glucose cotransporter 2 inhibitors (SGLT2i) in the Asia region is still lacking. This study aimed to evaluate the renal and safety outcomes of SGLT2is compared with dipeptidyl peptidase-4 inhibitors (DPP4i) using real-world data. Methods: A retrospective cohort study was conducted using the nationwide claims data in Republic of Korea. We evaluated kidney outcomes (any new-onset kidney events, acute kidney injury (AKI), chronic kidney disease (CKD), and kidney failure) as primary outcomes and safety outcomes (infection, hemodynamic adverse events, and fracture). Propensity score matching was used to adjust confounders, and the hazard ratios were calculated using the Cox proportional hazards model. Results: The study included 13,649 patients in the SGLT2i group and 35,043 in the DPP4i group after the matching. The SGLT2i group had a lower risk of kidney diseases, AKI, and CKD (HR 0.88 [0.61–0.74]) than the DPP4i group. For secondary outcomes, the risk of genital infection was higher (HR 2.38 [2.12–2.68]), and the risk of hyperkalemia was lower in the SGLT2i group than in the DPP4i group (HRs 0.49 [0.36–0.67]). Conclusions: The SGLT2 inhibitors had a lower risk of new-onset kidney outcomes and CKD than the DPP4 inhibitors. A high incidence of genital infection and a low incidence of hyperkalemia were shown in the SGLT2 inhibitor.

Less frequent adverse drug reactions are usually discovered after a drug's release to the market, making effective and timely communication of regulatory post‐market advisories essential for preventing emerging adverse effects. Time series analysis is a key study design for assessing the impact of post‐market safety advisories. However, most previous studies have focused on narrow evaluations, limiting systematic assessment of how different safety advisories affect prescribing practices. This study aimed to investigate changes in prescribing practices following regulatory post‐market safety advisories in Korea. Interrupted time series analyses were conducted using nationwide claims data from 2018 to 2021 and hospital datasets covering the period from 2 years before and 3 years after post‐market safety advisories. We categorized the selected drugs into two groups: safety warning through letters and real‐time safety alarms (contraindications or requiring attention). Twelve post‐market safety advisories were analyzed, including four safety warnings and eight safety alarms, which showed an overall relative reduction (safety warning: relative change [95% confidence interval]: −8.06% [−10.23% to −5.84%], safety alarm on contraindication: −92.65% [−95.65% to −87.59%], and safety alarm on requiring attention: −8.04% [−9.98% to −6.05%]). All types of regulatory interventions reduced the prescribing of targeted drugs; however, the magnitude of these effects differed substantially depending on the type of intervention. By identifying and comparing the influence of regulatory post‐market safety advisories, we can enhance these measures to better protect patient health. Continuous monitoring and systematic assessment of safety‐related regulatory advisories, with ensured reproducibility, are warranted to optimize effectiveness and ensure safe medication practices.

Background: Conventional prediction methods such as logistic regression and gradient boosting have been widely utilized for disease onset prediction for their reliability and interpretability. Deep learning methods promise enhanced prediction performance by extracting complex patterns from clinical data, but face challenges like data sparsity and high dimensionality. Methods: This study compares conventional and deep learning approaches to predict lung cancer, dementia, and bipolar disorder using observational data from eleven databases from North America, Europe, and Asia. Models were developed using logistic regression, gradient boosting, ResNet, and Transformer, and validated both internally and externally across the data sources. Discrimination performance was assessed using AUROC, and calibration was evaluated using Eavg. Findings: Across 11 datasets, conventional methods generally outperformed deep learning methods in terms of discrimination performance, particularly during external validation, highlighting their better transportability. Learning curves suggest that deep learning models require substantially larger datasets to reach the same performance levels as conventional methods. Calibration performance was also better for conventional methods, with ResNet showing the poorest calibration. Interpretation: Despite the potential of deep learning models to capture complex patterns in structured observational healthcare data, conventional models remain highly competitive for disease onset prediction, especially in scenarios involving smaller datasets and if lengthy training times need to be avoided. The study underscores the need for future research focused on optimizing deep learning models to handle the sparsity, high dimensionality, and heterogeneity inherent in healthcare datasets, and find new strategies to exploit the full capabilities of deep learning methods.

To extract information from free-text in clinical records due to the patient's protected health information PHI in the records pre-processing of de-identification is required. Therefore we aimed to identify PHI list and fine-tune the deep learning BERT model for developing de-identification model. The result of fine-tuning the model is strict F1 score of 0.924. Due to the convinced score the model can be used for the development of a de-identification model.

Background: Rosuvastatin and atorvastatin are indicated for cardiovascular protection in patients with type 2 diabetes (T2D) but may differ in clinical effectiveness and safety. This study compared real-world cardiovascular and safety outcomes associated with rosuvastatin versus atorvastatin in patients with T2D, with an emphasis on older adults. Methods: This retrospective cohort study used electronic health records from 10 Korean hospitals standardized to the Common Data Model. Adults (≥18 years) with T2D who were newly prescribed rosuvastatin or atorvastatin were included. After propensity score matching, primary outcomes (myocardial infarction [MI], heart failure, stroke, cardiac arrest, and in-hospital death), secondary outcomes (peripheral arterial disease [PAD] and glaucoma), and safety outcomes (acute kidney injury, cataract, and myalgia) were compared. Subgroup and sensitivity analyses were conducted among patients aged ≥65 years. Results: Among 49,034 patients (rosuvastatin, 16,123; atorvastatin, 32,911), baseline characteristics were well balanced. Across all participating hospitals, the comparative analyses showed no meaningful differences in cardiovascular or safety outcomes between the two statins. However, among patients aged ≥65 years, rosuvastatin was associated with a higher risk of PAD (hazard ratio [HR] 1.19; 95% confidence interval [CI] 1.03–1.38), and this finding was consistent in sensitivity analysis (HR 1.31; 95% CI 1.01–1.70). Conclusions: Rosuvastatin and atorvastatin demonstrated comparable cardiovascular effectiveness and safety. Although rosuvastatin was associated with a modestly higher incidence of PAD in older adults, the difference was small. From a clinical perspective, these findings underscore the importance of individualized statin therapy tailored to patient-specific factors such as age, comorbidity, and vascular health. Overall, both statins provide overall therapeutic equivalence for cardiovascular prevention.