Background: Despite excellent prediction performance, noninterpretability has undermined the value of applying deep-learning algorithms in clinical practice. To overcome this limitation, attention mechanism has been introduced to clinical research as an explanatory modeling method. However, potential limitations of using this attractive method have not been clarified to clinical researchers. Furthermore, there has been a lack of introductory information explaining attention mechanisms to clinical researchers.Objective: The aim of this study was to introduce the basic concepts and design approaches of attention mechanisms. In addition, we aimed to empirically assess the potential limitations of current attention mechanisms in terms of prediction and interpretability performance.Methods: First, the basic concepts and several key considerations regarding attention mechanisms were identified. Second, four approaches to attention mechanisms were suggested according to a two-dimensional framework based on the degrees of freedom and uncertainty awareness. Third, the prediction performance, probability reliability, concentration of variable importance, consistency of attention results, and generalizability of attention results to conventional statistics were assessed in the diabetic classification modeling setting. Fourth, the potential limitations of attention mechanisms were considered.Results: Prediction performance was very high for all models. Probability reliability was high in models with uncertainty awareness. Variable importance was concentrated in several variables when uncertainty awareness was not considered. The consistency of attention results was high when uncertainty awareness was considered. The generalizability of attention results to conventional statistics was poor regardless of the modeling approach.Conclusions: The attention mechanism is an attractive technique with potential to be very promising in the future. However, it may not yet be desirable to rely on this method to assess variable importance in clinical settings. Therefore, along with theoretical studies enhancing attention mechanisms, more empirical studies investigating potential limitations should be encouraged.

A screening model for undiagnosed diabetes mellitus (DM) is important for early medical care. Insufficient research has been carried out developing a screening model for undiagnosed DM using machine learning techniques. Thus, the primary objective of this study was to develop a screening model for patients with undiagnosed DM using a deep neural network. We conducted a cross-sectional study using data from the Korean National Health and Nutrition Examination Survey (KNHANES) 2013–2016. A total of 11,456 participants were selected, excluding those with diagnosed DM, an age < 20 years, or missing data. KNHANES 2013–2015 was used as a training dataset and analyzed to develop a deep learning model (DLM) for undiagnosed DM. The DLM was evaluated with 4444 participants who were surveyed in the 2016 KNHANES. The DLM was constructed using seven non-invasive variables (NIV): age, waist circumference, body mass index, gender, smoking status, hypertension, and family history of diabetes. The model showed an appropriate performance (area under curve (AUC): 80.11) compared with existing previous screening models. The DLM developed in this study for patients with undiagnosed diabetes could contribute to early medical care.

The common data model (CDM) has found widespread application in healthcare studies, but its utilization in cancer research has been limited. This article describes the development and implementation strategy for Cancer Clinical Library Databases (CCLDs), which are standardized cancer-specific databases established under the Korea-Clinical Data Utilization Network for Research Excellence (K-CURE) project by the Korean Ministry of Health and Welfare. Fifteen leading hospitals and fourteen academic associations in Korea are engaged in constructing CCLDs for 10 primary cancer types. For each cancer type-specific CCLD, cancer data experts determine key clinical data items essential for cancer research, standardize these items across cancer types, and create a standardized schema. Comprehensive clinical records covering diagnosis, treatment, and outcomes, with annual updates, are collected for each cancer patient in the target population, and quality control is based on six-sigma standards. To protect patient privacy, CCLDs follow stringent data security guidelines by pseudonymizing personal identification information and operating within a closed analysis environment. Researchers can apply for access to CCLD data through the K-CURE portal, which is subject to Institutional Review Board and Data Review Board approval. The CCLD is considered a pioneering standardized cancer-specific database, significantly representing Korea’s cancer data. It is expected to overcome limitations of previous CDMs and provide a valuable resource for multicenter cancer research in Korea.

This study aimed to evaluate the exposure levels and variation in airborne endotoxin and heavy metals such as aluminum, chromium (Cr), iron (Fe), manganese (Mn), nickel (Ni), zinc, and lead (Pb) in the three different South Korean subway environments (driver room, station office, and underground tunnel) and identify subway characteristics influencing endotoxin and heavy metals levels. Air samples were collected and analyzed using the kinetic Limulus Amebocyte Lysate assay and inductively coupled plasma mass spectrometers. The geometric mean was determined for endotoxin levels (0.693 EU/m³). It was also found that Fe (5.070 µg/m³) had the highest levels in subway environments while Pb (0.008 µg/m³) had the lowest levels. Endotoxin levels were higher in the underground tunnel and lower in the station office; the total heavy metal levels showed the same pattern with endotoxin levels. Endotoxins and total heavy metal levels were higher in the morning than at night. Positive correlations were found between endotoxin and Cr, Fe, Mn, and Ni levels. Given the correlation between airborne endotoxins and heavy metals, further studies with larger sample sizes are needed to identify the correlation between levels of airborne endotoxin and heavy metals.

We aimed to understand the decision-making process related to the willingness to undergo BRCA1/2 genetic testing, risk-reducing salpingo-oophorectomy (RRSO), or risk-reducing mastectomy (RRM) among the general public, cancer patients, and healthcare professionals in South Korea. In total, 3444 individuals (1496 from the general public, 1500 cancer patients, 108 clinicians, and 340 researchers) completed a survey addressing genetic testing and related risk management options in a hypothetical scenario. Differences in intent and associated factors for undergoing the above procedures or sharing test results were analyzed. Overall, 67% of participants were willing to undergo BRCA1/2 testing, with proportions of the general public (58%), cancer patients (70%), clinicians (88%), and researchers (90%). The willingness to undergo RRSO was highest among clinicians (58%), followed by among patients (38%), the general public (33%), and researchers (32%) (p < 0.001). Gender, age, education level, and household income were associated with willingness to undergo genetic testing, RRM, and RRSO (p < 0.05). The intent for undergo genetic testing, RRM, and RRSO were affected by many factors. Finally, 69% of the general public intended to share information with family, while this percentage was 92%, 91%, and 94% for patients, clinicians, and researchers, respectively (p < 0.05). These results highlight the requirement for developing targeted educational materials and counseling strategies for facilitating informed decision making.