Claude 3 Opus and ChatGPT With GPT-4 in Dermoscopic Image Analysis for Melanoma Diagnosis: Comparative Performance Analysis.

BACKGROUND: Recent advancements in artificial intelligence (AI) and large language models (LLMs) have shown potential in medical fields, including dermatology. With the introduction of image analysis capabilities in LLMs, their application in dermatological diagnostics has garnered significant interest. These capabilities are enabled by the integration of computer vision techniques into the underlying architecture of LLMs. OBJECTIVE: This study aimed to compare the diagnostic performance of Claude 3 Opus and ChatGPT with GPT-4 in analyzing dermoscopic images for melanoma detection, providing insights into their strengths and limitations. METHODS: We randomly selected 100 histopathology-confirmed dermoscopic images (50 malignant, 50 benign) from the International Skin Imaging Collaboration (ISIC) archive using a computer-generated randomization process. The ISIC archive was chosen due to its comprehensive and well-annotated collection of dermoscopic images, ensuring a diverse and representative sample. Images were included if they were dermoscopic images of melanocytic lesions with histopathologically confirmed diagnoses. Each model was given the same prompt, instructing it to provide the top 3 differential diagnoses for each image, ranked by likelihood. Primary diagnosis accuracy, accuracy of the top 3 differential diagnoses, and malignancy discrimination ability were assessed. The McNemar test was chosen to compare the diagnostic performance of the 2 models, as it is suitable for analyzing paired nominal data. RESULTS: In the primary diagnosis, Claude 3 Opus achieved 54.9% sensitivity (95% CI 44.08%-65.37%), 57.14% specificity (95% CI 46.31%-67.46%), and 56% accuracy (95% CI 46.22%-65.42%), while ChatGPT demonstrated 56.86% sensitivity (95% CI 45.99%-67.21%), 38.78% specificity (95% CI 28.77%-49.59%), and 48% accuracy (95% CI 38.37%-57.75%). The McNemar test showed no significant difference between the 2 models (P=.17). For the top 3 differential diagnoses, Claude 3 Opus and ChatGPT included the correct diagnosis in 76% (95% CI 66.33%-83.77%) and 78% (95% CI 68.46%-85.45%) of cases, respectively. The McNemar test showed no significant difference (P=.56). In malignancy discrimination, Claude 3 Opus outperformed ChatGPT with 47.06% sensitivity, 81.63% specificity, and 64% accuracy, compared to 45.1%, 42.86%, and 44%, respectively. The McNemar test showed a significant difference (P<.001). Claude 3 Opus had an odds ratio of 3.951 (95% CI 1.685-9.263) in discriminating malignancy, while ChatGPT-4 had an odds ratio of 0.616 (95% CI 0.297-1.278). CONCLUSIONS: Our study highlights the potential of LLMs in assisting dermatologists but also reveals their limitations. Both models made errors in diagnosing melanoma and benign lesions. These findings underscore the need for developing robust, transparent, and clinically validated AI models through collaborative efforts between AI researchers, dermatologists, and other health care professionals. While AI can provide valuable insights, it cannot yet replace the expertise of trained clinicians.

Surface chlorination of IrO2(110) by HCl.

The ability to controllably chlorinate metal-oxide surfaces can provide opportunities for designing selective oxidation catalysts. In the present study, we investigated the surface chlorination of IrO2(110) by HCl using temperature programmed reaction spectroscopy (TPRS), x-ray photoelectron spectroscopy (XPS), and density functional theory (DFT) calculations. We find that exposing IrO2(110) to HCl, followed by heating to 650 K in ultrahigh vacuum, produces nearly equal quantities of on-top and bridging Cl atoms on the surface, Clt and Clbr, where the Clbr atoms replace O-atoms that are removed from the surface by H2O formation. After HCl adsorption at 85 K, only H2O desorbs at low Cl coverages during TPRS, but HCl begins to desorb in increasing yields as the Cl coverage is increased above about 0.5 monolayer (ML). The desorption of Cl2 was not observed under any conditions, in good agreement with the high barrier for this reaction predicted by DFT. A maximum Cl coverage of 1 ML, with nearly equal coverages of Clt and Clbr atoms, could be generated by reacting HCl with IrO2(110) in UHV. Our results suggest that a kinetic competition between recombinative HCl and H2O desorption under the conditions studied limits the saturation Cl coverage to a value less than the 2 ML maximum predicted by thermodynamics. XPS further shows that the partitioning of Cl between the Clt and Clbr states can be altered by subjecting partially chlorinated IrO2(110) to reductive or oxidative treatments, demonstrating that the Cl site population can change dynamically in response to the gas environment. Our results provide insights for understanding the chlorination of IrO2(110) by HCl and can enable future experimental studies to determine how Cl-modification alters the surface chemical reactivity of IrO2(110) and potentially enhances selectivity toward partial oxidation chemistry.

Accomplishment of canine cloning through in vitro matured oocytes: a pioneering milestone.

The in vitro maturation (IVM) rate of canine oocytes remains low compared to other mammals due to their unique reproductive characteristics. This study aimed to explore the effect of hormone supplementation during the IVM of canine immature oocytes on nuclear maturation and subsequently assess its potential application in canine somatic cell nuclear transfer (SCNT). Immature oocytes were collected and cultured in an IVM medium supplemented with hormones (follicle-stimulating hormone [FSH] and progesterone [P4]) or without hormones (control) for 24 hours. The maturation rates of oocytes in the hormone-treated group (94.92 ± 3.15%) were significantly higher than those in the control group (61.01 ± 4.23%). Both in vitro and in vivo matured oocytes underwent NT to evaluate their utility, and the fusion rates were higher in the in vitro matured group than those in the vivo matured group, not significant between in vivo and in vitro matured group (73.28% and 82.35%, respectively). As a result, 14 fused embryos from the in vitro matured group were transferred into two surrogates, with one surrogate achieving a successful pregnancy and delivering four puppies. Whereas in the in vivo matured group, 85 fused embryos were transferred to 8 surrogate mothers, leading to three surrogates becoming pregnant and delivering one, four, and two puppies. The pregnancy rates were not significant between both groups (50% and 37.50%), but the number of offspring exhibited a significant difference (28.57% and 8.23%). In conclusion, we achieved a remarkable milestone by successfully producing cloned puppies using in vitro matured oocytes, underscoring the feasibility of canine cloning from in vitro recovered oocytes. It is important to note that this study focused only on immature oocytes after ovulation and only during the estrus stage. Further research targeting other stages of the estrous cycle could potentially enhance canine cloning efficiency.

ARHGEF5 binds Drebrin and affects α-tubulin acetylation to direct neuronal morphogenesis and migration during mouse brain development.

Rho guanine nucleotide exchange factors (Rho GEFs) activate Rho GTPases, which act as molecular switches regulating various essential cellular functions. This study investigated the role of ARHGEF5, a Rho GEF known for its involvement in cell migration and invasion processes, in the context of brain development. We found that ARHGEF5 is essential for dendrite development during the early stages of neuronal growth. We also discovered that ARHGEF5 binds to Drebrin E, which is vital for coordinating actin and microtubule dynamics, and facilitates the interaction between Drebrin E and Cyclin-dependent kinase 5, which phosphorylates Drebrin E. Notably, ARHGEF5 deficiency resulted in a decrease in acetylated α-tubulin levels, and the expression of an α-tubulin acetylation mimetic mutant (K40Q) rescued the defects in dendrite development and neuronal migration, suggesting ARHGEF5's role in modulating microtubule stability. Additionally, ARHGEF5 was shown to influence Golgi positioning in the leading processes of migrating cortical neurons during brain development. Our study suggests that ARHGEF5 plays a crucial role in integrating cytoskeletal dynamics with neuronal morphogenesis and migration processes during brain development.

Development of Indocyanine Green/Methyl-ß-cyclodextrin Complex-Loaded Liposomes for Enhanced Photothermal Cancer Therapy.

Photothermal therapy (PTT) is a promising cancer therapeutic approach due to its spatial selectivity and high potency. Indocyanine green (ICG) has been considered a biocompatible PTT agent. However, ICG has several challenges to hinder its clinical use including rapid blood clearance and instability to heat, light, and solvent, leading to a loss of photoactivation property and PTT efficacy. Herein, we leveraged stabilizing components, methyl-ß-cyclodextrin and liposomes, in one nanoplatform (ICD lipo) to enhance ICG stability and the photothermal therapeutic effect against cancer. Compared to ICG, ICD lipo displayed a 4.8-fold reduction in degradation in PBS solvent after 30 days and a 3.4-fold reduction in photobleaching after near-infrared laser irradiation. Moreover, in tumor-bearing mice, ICD lipo presented a 2.7-fold increase in tumor targetability and inhibited tumor growth 9.6 times more effectively than did ICG without any serious toxicity. We believe that ICD lipo could be a potential PTT agent for cancer therapeutics.

A Comprehensive Study on a Deep-Learning-Based Electrocardiography Analysis for Estimating the Apnea-Hypopnea Index.

This study introduces a deep-learning-based automatic sleep scoring system to detect sleep apnea using a single-lead electrocardiography (ECG) signal, focusing on accurately estimating the apnea-hypopnea index (AHI). Unlike other research, this work emphasizes AHI estimation, crucial for the diagnosis and severity evaluation of sleep apnea. The suggested model, trained on 1465 ECG recordings, combines the deep-shallow fusion network for sleep apnea detection network (DSF-SANet) and gated recurrent units (GRUs) to analyze ECG signals at 1-min intervals, capturing sleep-related respiratory disturbances. Achieving a 0.87 correlation coefficient with actual AHI values, an accuracy of 0.82, an F1 score of 0.71, and an area under the receiver operating characteristic curve of 0.88 for per-segment classification, our model was effective in identifying sleep-breathing events and estimating the AHI, offering a promising tool for medical professionals.

Synergistic effects of copper and oxygen vacancies in enhancing the efficacy of partially crystalline CuMnxOy catalyst for ozone decomposition.

To tackle the challenge of ground-level ozone pollution, this study proposed a potential catalytic design approach for ozone decomposition using Cu-Mn bimetallic oxide. This approach is grounded in an understanding of the intrinsic reactivity for catalyst and incorporates a novel potassium-driven low-temperature oxidation process for catalyst synthesis. The research highlights the creation of a highly reactive Cu-Mn oxide phase with extensive defect coverage, leading to significantly increased reaction rates. It also identifies the MnO2(100) facet as a crucial active phase, where oxygen vacancies simultaneously enhance O3 adsorption and decomposition, albeit with a concurrent risk of O2 poisoning due to the stabilization of adsorbed O2. Crucially, the incorporation of Cu offsets the effects of oxygen vacancies, influencing conversion rates and lessening O2 poisoning. The synergistic interplay between Cu and oxygen vacancies elevates the performance of the defect-rich Cu-Mn oxide catalyst. By combining computational and experimental methods, this study not only advances the understanding of the Cu-Mn oxide system for ozone decomposition but also contributes valuable insights into developing more efficient catalysts to mitigate ozone pollution.

Floating electrode-dielectric barrier discharge-based plasma promotes skin regeneration in a full-thickness skin defect mouse model.

Wound healing involves a complex and dynamic interplay among various cell types, cytokines, and growth factors. Macrophages and transforming growth factor-ß1 (TGF-ß1) play an essential role in different phases of wound healing. Cold atmospheric plasma has a wide range of applications in the treatment of chronic wounds. Hence, we aimed to investigate the safety and efficacy of a custom-made plasma device in a full-thickness skin defect mouse model. Here, we investigated the wound tissue on days 6 and 12 using histology, qPCR, and western blotting. During the inflammation phase of wound repair, macrophages play an important role in the onset and resolution of inflammation, showing decreased F4/80 on day 6 of plasma treatment and increased TGF-ß1 levels. The plasma-treated group showed better epidermal epithelialization, dermal fibrosis, collagen maturation, and reduced inflammation than the control group. Our findings revealed that floating electrode-dielectric barrier discharge (FE-DBD)-based atmospheric-pressure plasma promoted significantly faster wound healing in the plasma-treated group than that in the control group with untreated wounds. Hence, plasma treatment accelerated wound healing processes without noticeable side effects and suppressed pro-inflammatory genes, suggesting that FE-DBD-based plasma could be a potential therapeutic option for treating various wounds.

Chiral Materials for Optics and Electronics: Ready to Rise?

Chiral materials have gained burgeoning interest in optics and electronics, beyond their classical application field of drug synthesis. In this review, we summarize the diverse chiral materials developed to date and how they have been effectively applied to optics and electronics to get an understanding and vision for the further development of chiral materials for advanced optics and electronics.

Comparison between Onyx and coil embolization for persistent type 2 endoleaks after endovascular aneurysm repair.

Purpose: Type 2 endoleaks (T2EL) are the most common form of endoleaks after endovascular aneurysm repair (EVAR). Several studies on the feasibility of embolization using ethylene vinyl alcohol copolymer (Onyx, Medtronic) for T2EL have been reported. The purpose of this study was to compare coil and Onyx embolization for T2EL treatment after EVAR. Methods: Between August 2005 and July 2022, 46 patients underwent endovascular embolization for treatment of T2EL (15 Onyx and 31 coils). The primary endpoint was endoleaks resolution or significant aneurysm sac growth of >5 mm in maximal diameter after T2EL embolization. In addition, periprocedural factors, reintervention, sac rupture, and survival analysis were assessed. Results: The follow-up period after embolization was significantly shorter in the Onyx group (11.6 months vs. 34.7 months, P = 0.016), and there was no difference in aneurysm sac growth rate between both groups (20.0% vs. 51.6%; P = 0.472, log-rank test). However, cases with multiple endoleak origins tended to be treated with Onyx (P = 0.002). When applying Onyx, there was no significant difference in results between the transarterial and translumbar approaches. Conclusion: There appears to be no significant difference in the results of Onyx and coil embolization for T2EL treatment, although it is difficult to evaluate effectiveness due to the small number of cases and short follow-up period. However, in cases of multiple origin endoleaks or when the transarterial approach is not feasible, the Onyx by translumbar approach may be a more effective method.

Schisandrin B enhances embryo competence and potentially mitigates endoplasmic reticulum stress during porcine preimplantation development.

Endoplasmic reticulum (ER) stress induced by agents such as tunicamycin (TM) substantially impedes the developmental progression of porcine embryos. Lignan compounds such as Schisandrin B (Sch-B), may have the potential to mitigate this stress. However, there are few studies on the effects of Sch-B on embryo development. To address this research gap, this study evaluates the protective efficacy of Sch-B against TM-induced ER stress during pivotal stages of porcine embryogenesis. Notably, embryos treated with Sch-B exhibited pronounced resistance to TM-induced developmental arrest, particularly at the 4-cell stage, facilitating progression to the 8-cell stage and subsequent blastocyst formation. It was also observed that Sch-B effectively reduced reactive oxygen species (ROS) levels and improved mitochondrial membrane potential (MMP). Furthermore, Sch-B positively influenced the expression of several stress-related genes. These findings highlight the promising role of Sch-B in improving porcine embryo development and mitigating ER stress.

Early Prediction of Mortality for Septic Patients Visiting Emergency Room Based on Explainable Machine Learning: A Real-World Multicenter Study.

BACKGROUND: Worldwide, sepsis is the leading cause of death in hospitals. If mortality rates in patients with sepsis can be predicted early, medical resources can be allocated efficiently. We constructed machine learning (ML) models to predict the mortality of patients with sepsis in a hospital emergency department. METHODS: This study prospectively collected nationwide data from an ongoing multicenter cohort of patients with sepsis identified in the emergency department. Patients were enrolled from 19 hospitals between September 2019 and December 2020. For acquired data from 3,657 survivors and 1,455 deaths, six ML models (logistic regression, support vector machine, random forest, extreme gradient boosting [XGBoost], light gradient boosting machine, and categorical boosting [CatBoost]) were constructed using fivefold cross-validation to predict mortality. Through these models, 44 clinical variables measured on the day of admission were compared with six sequential organ failure assessment (SOFA) components (PaO2/FIO2 [PF], platelets (PLT), bilirubin, cardiovascular, Glasgow Coma Scale score, and creatinine). The confidence interval (CI) was obtained by performing 10,000 repeated measurements via random sampling of the test dataset. All results were explained and interpreted using Shapley's additive explanations (SHAP). RESULTS: Of the 5,112 participants, CatBoost exhibited the highest area under the curve (AUC) of 0.800 (95% CI, 0.756-0.840) using clinical variables. Using the SOFA components for the same patient, XGBoost exhibited the highest AUC of 0.678 (95% CI, 0.626-0.730). As interpreted by SHAP, albumin, lactate, blood urea nitrogen, and international normalization ratio were determined to significantly affect the results. Additionally, PF and PLTs in the SOFA component significantly influenced the prediction results. CONCLUSION: Newly established ML-based models achieved good prediction of mortality in patients with sepsis. Using several clinical variables acquired at the baseline can provide more accurate results for early predictions than using SOFA components. Additionally, the impact of each variable was identified.

Assessment of the Reliability and Clinical Applicability of ChatGPT's Responses to Patients' Common Queries About Rosacea.

Objective: Artificial intelligence chatbot, particularly ChatGPT (Chat Generative Pre-trained Transformer), is capable of analyzing human input and generating human-like responses, which shows its potential application in healthcare. People with rosacea often have questions about alleviating symptoms and daily skin-care, which is suitable for ChatGPT to response. This study aims to assess the reliability and clinical applicability of ChatGPT 3.5 in responding to patients' common queries about rosacea and to evaluate the extent of ChatGPT's coverage in dermatology resources. Methods: Based on a qualitative analysis of the literature on the queries from rosacea patients, we have extracted 20 questions of patients' greatest concerns, covering four main categories: treatment, triggers and diet, skincare, and special manifestations of rosacea. Each question was inputted into ChatGPT separately for three rounds of question-and-answer conversations. The generated answers will be evaluated by three experienced dermatologists with postgraduate degrees and over five years of clinical experience in dermatology, to assess their reliability and applicability for clinical practice. Results: The analysis results indicate that the reviewers unanimously agreed that ChatGPT achieved a high reliability of 92.22% to 97.78% in responding to patients' common queries about rosacea. Additionally, almost all answers were applicable for supporting rosacea patient education, with a clinical applicability ranging from 98.61% to 100.00%. The consistency of the expert ratings was excellent (all significance levels were less than 0.05), with a consistency coefficient of 0.404 for content reliability and 0.456 for clinical practicality, indicating significant consistency in the results and a high level of agreement among the expert ratings. Conclusion: ChatGPT 3.5 exhibits excellent reliability and clinical applicability in responding to patients' common queries about rosacea. This artificial intelligence tool is applicable for supporting rosacea patient education.

Exploring the exponential sensitivity of risk perception in the COVID-19 pandemic.

Individual's risk perception regarding specific hazards is a dynamic process that evolves over time. This study analyzed the relationship between the number of COVID-19 cases and the South Korean public's risk perceptions from the outset of the pandemic to the recent past. More than 70 repeated cross-sectional surveys were conducted biweekly to measure individuals' risk perception. An autoregressive integrated moving average with explanatory variable time series analysis was used to characterize the relationship between the number of COVID-19 cases and level of risk perceptions. It revealed that individuals' risk perception and the number of COVID-19 cases were not linearly related but were logarithmically correlated. This finding can be understood as a psychic numbing effect, suggesting that people's perception of risk is not linear but rather exponentially sensitive to changes. The findings also revealed a significant influence of individuals' trust in local governments on their risk perceptions, highlighting the substantial role played by local governments in direct risk management during the COVID-19 pandemic.

Large tandem duplications in cancer result from transcription and DNA replication collisions.

Despite the abundance of somatic structural variations (SVs) in cancer, the underlying molecular mechanisms of their formation remain unclear. Here, we use 6,193 whole-genome sequenced tumors to study the contributions of transcription and DNA replication collisions to genome instability. After deconvoluting robust SV signatures in three independent pan-cancer cohorts, we detect transcription-dependent replicated-strand bias, the expected footprint of transcription-replication collision (TRC), in large tandem duplications (TDs). Large TDs are abundant in female-enriched, upper gastrointestinal tract and prostate cancers. They are associated with poor patient survival and mutations in TP53, CDK12, and SPOP. Upon inactivating CDK12, cells display significantly more TRCs, R-loops, and large TDs. Inhibition of G2/M checkpoint proteins, such as WEE1, CHK1, and ATR, selectively inhibits the growth of cells deficient in CDK12. Our data suggest that large TDs in cancer form due to TRCs, and their presence can be used as a biomarker for prognosis and treatment.

A Deep Learning-Based Approach for Prediction of Vancomycin Treatment Monitoring: Retrospective Study Among Patients With Critical Illness.

BACKGROUND: Vancomycin pharmacokinetics are highly variable in patients with critical illnesses, and clinicians commonly use population pharmacokinetic (PPK) models based on a Bayesian approach to dose. However, these models are population-dependent, may only sometimes meet the needs of individual patients, and are only used by experienced clinicians as a reference for making treatment decisions. To assist real-world clinicians, we developed a deep learning-based decision-making system that predicts vancomycin therapeutic drug monitoring (TDM) levels in patients in intensive care unit. OBJECTIVE: This study aimed to establish joint multilayer perceptron (JointMLP), a new deep-learning model for predicting vancomycin TDM levels, and compare its performance with the PPK models, extreme gradient boosting (XGBoost), and TabNet. METHODS: We used a 977-case data set split into training and testing groups in a 9:1 ratio. We performed external validation of the model using 1429 cases from Kangwon National University Hospital and 2394 cases from the Medical Information Mart for Intensive Care-IV (MIMIC-IV). In addition, we performed 10-fold cross-validation on the internal training data set and calculated the 95% CIs using the metric. Finally, we evaluated the generalization ability of the JointMLP model using the MIMIC-IV data set. RESULTS: Our JointMLP model outperformed other models in predicting vancomycin TDM levels in internal and external data sets. Compared to PPK, the JointMLP model improved predictive power by up to 31% (mean absolute error [MAE] 6.68 vs 5.11) on the internal data set and 81% (MAE 11.87 vs 6.56) on the external data set. In addition, the JointMLP model significantly outperforms XGBoost and TabNet, with a 13% (MAE 5.75 vs 5.11) and 14% (MAE 5.85 vs 5.11) improvement in predictive accuracy on the inner data set, respectively. On both the internal and external data sets, our JointMLP model performed well compared to XGBoost and TabNet, achieving prediction accuracy improvements of 34% and 14%, respectively. Additionally, our JointMLP model showed higher robustness to outlier data than the other models, as evidenced by its higher root mean squared error performance across all data sets. The mean errors and variances of the JointMLP model were close to zero and smaller than those of the PPK model in internal and external data sets. CONCLUSIONS: Our JointMLP approach can help optimize treatment outcomes in patients with critical illnesses in an intensive care unit setting, reducing side effects associated with suboptimal vancomycin administration. These include increased risk of bacterial resistance, extended hospital stays, and increased health care costs. In addition, the superior performance of our model compared to existing models highlights its potential to help real-world clinicians.

Ligand-Mediated Mechanical Enhancement in Protein Complexes at Nano- and Macro-Scale.

Protein self-assembly plays a vital role in a myriad of biological functions and in the construction of biomaterials. Although the physical association underlying these assemblies offers high specificity, the advantage often compromises the overall durability of protein complexes. To address this challenge, we propose a novel strategy that reinforces the molecular self-assembly of protein complexes mediated by their ligand. Known for their robust noncovalent interactions with biotin, streptavidin (SAv) tetramers are examined to understand how the ligand influences the mechanical strength of protein complexes at the nanoscale and macroscale, employing atomic force microscopy-based single-molecule force spectroscopy, rheology, and bioerosion analysis. Our study reveals that biotin binding enhances the mechanical strength of individual SAv tetramers at the nanoscale. This enhancement translates into improved shear elasticity and reduced bioerosion rates when SAv tetramers are utilized as cross-linking junctions within hydrogel. This approach, which enhances the mechanical strength of protein-based materials without compromising specificity, is expected to open new avenues for advanced biotechnological applications, including self-assembled, robust biomimetic scaffolds and soft robotics.

In-Advance Prediction of Pressure Ulcers via Deep-Learning-Based Robust Missing Value Imputation on Real-Time Intensive Care Variables.

Pressure ulcers (PUs) are a prevalent skin disease affecting patients with impaired mobility and in high-risk groups. These ulcers increase patients' suffering, medical expenses, and burden on medical staff. This study introduces a clinical decision support system and verifies it for predicting real-time PU occurrences within the intensive care unit (ICU) by using MIMIC-IV and in-house ICU data. We develop various machine learning (ML) and deep learning (DL) models for predicting PU occurrences in real time using the MIMIC-IV and validate using the MIMIC-IV and Kangwon National University Hospital (KNUH) dataset. To address the challenge of missing values in time series, we propose a novel recurrent neural network model, GRU-D++. This model outperformed other experimental models by achieving the area under the receiver operating characteristic curve (AUROC) of 0.945 for the on-time prediction and AUROC of 0.912 for 48h in-advance prediction. Furthermore, in the external validation with the KNUH dataset, the fine-tuned GRU-D++ model demonstrated superior performances, achieving an AUROC of 0.898 for on-time prediction and an AUROC of 0.897 for 48h in-advance prediction. The proposed GRU-D++, designed to consider temporal information and missing values, stands out for its predictive accuracy. Our findings suggest that this model can significantly alleviate the workload of medical staff and prevent the worsening of patient conditions by enabling timely interventions for PUs in the ICU.