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.

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.

Analysis of treatment pattern of anti-dementia medications in newly diagnosed Alzheimer's dementia using OMOP CDM.

Anti-dementia medications are widely prescribed to patients with Alzheimer's dementia (AD) in South Korea. This study investigated the pattern of medical management in newly diagnosed patients with AD using a standardized data format-the Observational Medical Outcome Partnership Common Data Model from five hospitals. We examined the anti-dementia treatment patterns from datasets that comprise > 5 million patients during 2009-2019. The medication utility information was analyzed with respect to treatment trends and persistence across 11 years. Among the 8653 patients with newly diagnosed AD, donepezil was the most commonly prescribed anti-dementia medication (4218; 48.75%), followed by memantine (1565; 18.09%), rivastigmine (1777; 8.98%), and galantamine (494; 5.71%). The rising prescription trend during observation period was found only with donepezil. The treatment pathways for the three cholinesterase inhibitors combined with N-methyl-D-aspartate receptor antagonist were different according to the drugs (19.6%; donepezil; 28.1%; rivastigmine, and 17.2%; galantamine). A 12-month persistence analysis showed values of approximately 50% for donepezil and memantine and approximately 40% for rivastigmine and galantamine. There were differences in the prescribing pattern and persistence among anti-dementia medications from database using the Observational Medical Outcome Partnership Common Data Model on the Federated E-health Big Data for Evidence Renovation Network platform in Korea.

Cancer risk in Korean patients with gout.

BACKGROUND/AIMS: Using a nationwide cohort, we investigated the cancer risk in Korean patients with gout. METHODS: Data were obtained from the Korean National Health Insurance Service Database. Patients with gout were defined as those aged ≥ 20 years who were diagnosed with gout and received anti-gout medication (allopurinol, colchicine, and benzbromarone) between 2008 and 2010. Patients with nail disorders were randomly assigned to a control group (1:1 ratio) after frequency matching for age and sex. Cancer incidence was then investigated between 2012 and 2018. Cox proportional hazard regression analysis was used to investigate the association between gout and cancer after adjusting for concomitant diseases. RESULTS: This study included 179,930 patients with gout and an equal number of matched controls. The incidence of overall cancer was higher in patients with gout than in controls (incidence rate ratio, 1.08). Cox proportional hazards regression analysis showed that gout was associated with a hazard ratio of 1.053 (95% confidence interval ,1.031 to 1.077) after adjusting for concomitant diseases. CONCLUSION: Gout was associated with a significantly high risk of cancer, especially esophageal, stomach, colon, liver, pancreatic, lung, ovarian, renal, and bladder cancers.

Word Embedding Reveals Cyfra 21-1 as a Biomarker for Chronic Obstructive Pulmonary Disease.

BACKGROUND: Although patients with chronic obstructive pulmonary disease (COPD) experience high morbidity and mortality worldwide, few biomarkers are available for COPD. Here, we analyzed potential biomarkers for the diagnosis of COPD by using word embedding. METHODS: To determine which biomarkers are likely to be associated with COPD, we selected respiratory disease-related biomarkers. Degrees of similarity between the 26 selected biomarkers and COPD were measured by word embedding. And we infer the similarity with COPD through the word embedding model trained in the large-capacity medical corpus, and search for biomarkers with high similarity among them. We used Word2Vec, Canonical Correlation Analysis, and Global Vector for word embedding. We evaluated the associations of selected biomarkers with COPD parameters in a cohort of patients with COPD. RESULTS: Cytokeratin 19 fragment (Cyfra 21-1) was selected because of its high similarity and its significant correlation with the COPD phenotype. Serum Cyfra 21-1 levels were determined in patients with COPD and controls (4.3 ± 5.9 vs. 3.9 ± 3.6 ng/mL, P = 0.611). The emphysema index was significantly correlated with the serum Cyfra 21-1 level (correlation coefficient = 0.219, P = 0.015). CONCLUSION: Word embedding may be used for the discovery of biomarkers for COPD and Cyfra 21-1 may be used as a biomarker for emphysema. Additional studies are needed to validate Cyfra 21-1 as a biomarker for COPD.

Tuberculosis-Infected Giant Bulla Treated by Percutaneous Drainage Followed by Obliteration of the Pulmonary Cavity Using Talc: Case Report.

Tuberculosis (TB)-infected giant bullae are rare. A 55-year-old man was referred when an infected bulla did not respond to empirical treatment. Computed tomography showed a giant bulla in the right upper lobe with an air-fluid level and surrounding infiltrate. Sputum culture, acid-fast bacilli (AFB) stain, and polymerase chain reaction (PCR) for TB were negative. Percutaneous drainage of the bullous fluid was performed. AFB stain and PCR were positive in the drained fluid. The patient was given anti-TB drugs and later underwent obliteration of the pulmonary cavity using talc. To summarize, we report a patient with a TB-infected giant bulla that was treated successfully with anti-TB drugs and obliteration of the pulmonary cavity using talc.

Difference of copy number variation in blood of patients with lung cancer.

BACKGROUND: Lung cancer is the leading cause of cancer-related deaths worldwide. Copy number variation (CNV) in several genetic regions correlate with cancer susceptibility. Hence, this study evaluated the association between CNV and non-small cell lung cancer (NSCLC) in the peripheral blood. METHODS: Blood samples of 150 patients with NSCLC and 150 normal controls were obtained from a bioresource center (Seoul, Korea). Through an epigenome-wide analysis using the MethylationEPIC BeadChip method, we extracted CNVs by using an SVS8 software-supplied multivariate method. We compared CNV frequencies between the NSCLC and controls, and then performed stratification analyses according to smoking status. RESULTS: We acquired 979 CNVs, with 582 and 967 copy-number gains and losses, respectively. We identified five nominally significant associations (ACOT1, NAA60, GSDMD, HLA-DPA1, and SLC35B3 genes). Among the current smokers, the NSCLC group had more CNV losses and gains at the GSDMD gene in chromosome 8 (P=0.02) and at the ACOT1 gene in chromosome 14 (P=0.03) than the control group. It also had more CNV losses at the NAA60 gene in chromosome 16 (P=0.03) among non-smokers. In the NSCLC group, current smokers had more CNV gains and losses at the ACOT1 gene in chromosome 14 (P=0.003) and at HLA-DPA1 gene in chromosome 6 (P=0.02), respectively, than non-smokers. CONCLUSION: Five nominally significant associations were found between the NSCLC and CNVs. CNVs are associated with the mechanism of lung cancer development. However, the role of CNVs in lung cancer development needs further investigation.

Normothermia in Patients With Sepsis Who Present to Emergency Departments Is Associated With Low Compliance With Sepsis Bundles and Increased In-Hospital Mortality Rate.

OBJECTIVES: To investigate the impact of normothermia on compliance with sepsis bundles and in-hospital mortality in patients with sepsis who present to emergency departments. DESIGN: Retrospective multicenter observational study. PATIENTS: Nineteen university-affiliated hospitals of the Korean Sepsis Alliance participated in this study. Data were collected regarding patients who visited emergency departments for sepsis during the 1-month period. The patients were divided into three groups based on their body temperature at the time of triage in the emergency department (i.e., hypothermia [< 36°C] vs normothermia [36-38°C] vs hyperthermia [> 38°C]). INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Of 64,021 patients who visited emergency departments, 689 with community-acquired sepsis were analyzed (182 hyperthermic, 420 normothermic, and 87 hypothermic patients). The rate of compliance with the total hour-1 bundle was lowest in the normothermia group (6.0% vs 9.3% in hyperthermia vs 13.8% in hypothermia group; p = 0.032), the rate for lactate measurement was lowest in the normothermia group (62.1% vs 73.1% vs 75.9%; p = 0.005), and the blood culture rate was significantly lower in the normothermia than in the hyperthermia group (p < 0.001). The in-hospital mortality rates in the hyperthermia, normothermia, and hypothermia groups were 8.5%, 20.6%, and 30.8%, respectively (p < 0.001), but there was no significant association between compliance with sepsis bundles and in-hospital mortality. However, in a multivariate analysis, compared with hyperthermia, normothermia was significantly associated with an increased in-hospital mortality (odds ratio, 2.472; 95% CI, 1.005-6.080). This association remained significant even after stratifying patients by median lactate level. CONCLUSIONS: Normothermia at emergency department triage was significantly associated with an increased risk of in-hospital mortality and a lower rate of compliance with the sepsis bundle. Despite several limitations, our findings suggest a need for new strategies to improve sepsis outcomes in this group of patients.

Effects of age on the association between pulmonary tuberculosis and lung cancer in a South Korean cohort.

BACKGROUND: Various studies have investigated the association between pulmonary tuberculosis (TB) and lung cancer However, how the relationship between TB and lung cancer may differ by age is not yet clear. This study investigated how risk for lung cancer after pulmonary TB may differ by age. METHODS: This study used the National Health Insurance Service-National Sample Cohort in South Korea. We compared 3,776 pulmonary TB patients with 18,880 controls matched for sex and age during the period from 2003 to 2013. We analyzed the incidence of lung cancer after diagnosis of active pulmonary TB. A multivariate Cox proportional hazard model was used to calculate the adjusted hazard ratio (HR) of lung cancer after adjusting for sex, age, house income, and smoking status. RESULTS: Among 3,776 pulmonary TB patients, 86 had lung cancer diagnoses, whereas there were 108 lung cancer patients among 18,880 controls. The incidence rate ratio in the pulmonary TB group was 12.26 within 1 year and 3.33 at 1-3.9 years after TB infection, compared to the control group. There was increased risk for lung cancer in pulmonary TB patients compared to controls (HR, 4.18; 95% CI, 3.15-5.56). Compared to patients <50 years of age, the risks for lung cancer were HR 9.85, 7.1, 3.32, and 2.57 in patients aged 50-59, 60-69, and ≥70 years, respectively. CONCLUSIONS: Pulmonary TB is a risk factor for lung cancer. Patients with pulmonary TB should be monitored for subsequent development of lung cancer, particularly in younger patients.

Characteristics, management and clinical outcomes of patients with sepsis: a multicenter cohort study in Korea.

BACKGROUND: Mortality rates associated with sepsis have increased progressively in Korea, but domestic epidemiologic data remain limited. The objective of this study was to investigate the characteristics, management and clinical outcomes of sepsis patients in Korea. METHODS: This study is a multicenter retrospective cohort study. A total of 64,021 adult patients who visited an emergency department (ED) within one of the 19 participating hospitals during a 1-month period were screened for eligibility. Among these, patients diagnosed with sepsis based on the third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) were included in the study. RESULTS: Using the Sepsis-3 criteria, 977 sepsis patients were identified, among which 36.5% presented with septic shock. The respiratory system (61.8%) was the most common site of infection. The pathogen involved was identified in 444 patients (45.5%) and multi-drug resistance (MDR) pathogens were isolated in 171 patients. Empiric antibiotic therapy was appropriate in 68.6% of patients, but the appropriateness was significantly reduced in infections associated with MDR pathogens as compared with non-MDR pathogens (58.8% vs. 76.0%, P<0.001). Hospital mortality was 43.2% and 18.5% in sepsis patients with and without shock, respectively. Of the 703 patients who survived to discharge, 61.5% were discharged to home and 38.6% were transferred to other hospitals or facilities. CONCLUSIONS: This study found the prevalence of sepsis in adult patients visiting an ED in Korea was 1.5% (15.2/1,000 patients). Patients with sepsis, especially septic shock, had a high mortality and were often referred to step-down centers after acute and critical care.

Levels of vitamin D-associated cytokines distinguish between active and latent tuberculosis following a tuberculosis outbreak.

BACKGROUND: Vitamin D levels are associated with the extent of mycobactericidal activity. Interleukin (IL)-15 and IL-32 play roles in the vitamin D-mediated tuberculosis (TB) defense mechanism. Vitamin D induces IL-1ß, which plays an important role in terms of resistance to TB. We evaluated whether the levels of vitamin D-related cytokines distinguished between those with active TB and latent TB infection (LTBI). METHODS: In total, 50 TB-infected patients (25 with active TB and 25 with LTBI following a TB outbreak in a high school) were enrolled. Plasma 25-hydroxyvitamin D (25[OH]D), IL-15, IL-32, and IL-1ß levels were measured via enzyme-linked immunosorbent assays. Mycobacterium tuberculosis-specific antigen-induced and unstimulated cytokine levels were measured in the supernatants of the QuantiFERON TB Gold-In-Tube (QFT-GIT) assay. RESULTS: Plasma 25(OH)D and plasma IL-15 levels were lower in patients with active TB than in LTBI subjects (25(OH)D: 16.64 ng/mL vs. 21.6 ng/mL, P = 0.031; IL-15: 148.9 pg/mL vs. 189.8 pg/mL, P = 0.013). Plasma 25(OH)D levels correlated with the plasma levels of IL-15 and IL-1ß in TB-infected patients. In addition, the plasma 25(OH)D levels correlated positively with the level of unstimulated IL-15 (IL-15nil) and negatively with that of TB antigen-stimulated IL-32 (IL-32TB) in QFT-GIT supernatants. Although the IL-15nil and IL-15TB levels were higher in LTBI subjects than patients with active TB, the IL-32nil and IL-32TB levels were higher in the latter patients. A combination of the IL-15nil and IL-32TB levels accurately predicted 91.3% of active TB patients and latent subjects, with an area under the curve of 0.964. CONCLUSIONS: Our preliminary data showed that the levels of the vitamin D-related cytokines IL-15 and IL-32 differed between active TB patients and LTBI subjects. This result might be used as a basic data for developing biomarkers distinguishing between active TB and LTBI.

Inhaled Corticosteroid-Related Tuberculosis in the Real World Among Patients with Asthma and COPD: A 10-Year Nationwide Population-Based Study.

BACKGROUND: There have been concerns about the risk of inhaled corticosteroid (ICS)-related tuberculosis (TB) development. OBJECTIVE: We investigated the occurrence of TB among ICS users according to underlying respiratory diseases and type of ICS. METHODS: A 12-year population cohort comprising approximately 1 million subjects collected from the Korean claims database were used. Adult ICS users (budesonide or fluticasone) were enrolled. The temporal relationship between TB development and the last ICS prescription before TB development was evaluated. A nested case-control study was performed with 1:4 matching for age, sex, and the initiation date of the ICS. RESULTS: There were 17,991 ICS users, and 175 developed TB during the study period. Approximately 80% (140/175) of patients who developed TB were diagnosed within 3 years after the last ICS prescription. In the nested case-control study, the occurrence of TB was not related to the type of ICS, but was related to a higher annual admission rate and a higher comorbidity score. The risk of TB was higher in patients with chronic obstructive pulmonary disease (COPD) than in those with asthma (odds ratio: 2.31; CI 95%: 1.39-3.38; P = .0011) after adjusting for covariates. The subgroup analysis revealed no difference between budesonide and fluticasone with respect to the risk of developing TB in patients with asthma, COPD, or asthma-COPD overlap syndrome. CONCLUSION: An increased risk of TB development may persist for 3 years after stopping the ICS and the risk is higher in patients with COPD regardless of the type of ICS used.

Plasma CRABP2 as a Novel Biomarker in Patients with Non-Small Cell Lung Cancer.

BACKGROUND: Lung cancer is the most common cause of cancer-related mortality worldwide. We previously reported the identification of a new genetic marker, cellular retinoic acid binding protein 2 (CRABP2), in lung cancer tissues. The aim of this study was to assess plasma levels of CRABP2 from patients with non-small cell lung cancer (NSCLC). METHODS: Blood samples that were collected from 122 patients with NSCLC between September 2009 and September 2013 were selected for the analysis, along with samples from age- (± 5 years), sex-, and cigarette smoking history (± 10 pack-years [PY])-matched controls from the Korea Biobank Network. The control specimens were from patients who were without malignancies or pulmonary diseases. We measured plasma levels of CRABP2 using commercially available enzyme-linked immunosorbent assay kits. RESULTS: The mean age of the NSCLC patients was 71.8 ± 8.9 years, and the median cigarette smoking history was 32 PY (range, 0-150 PY). Plasma CRABP2 levels were significantly higher in patients with NSCLC than in the matched controls (37.63 ± 28.71 ng/mL vs. 24.09 ± 21.09 ng/mL, P < 0.001). Higher plasma CRABP2 levels were also correlated with lower survival rates in NSCLC patients (P = 0.014). CONCLUSION: Plasma CRABP2 levels might be a novel diagnostic and prognostic marker in NSCLC.

Quantitative computed tomography features and clinical manifestations associated with the extent of bronchiectasis in patients with moderate-to-severe COPD.

Background: Few studies have investigated the quantitative computed tomography (CT) features associated with the severity of bronchiectasis in COPD patients. The purpose of this study was to identify the quantitative CT features and clinical values to determine the extent of bronchiectasis in moderate-to-severe COPD patients. Methods: A total of 127 moderate-to-severe COPD patients were selected from the cohort of COPD in Dusty Areas (CODA). The study subjects were classified into three groups according to the extent of bronchiectasis on CT: no bronchiectasis, mild bronchiectasis, and moderate-to-severe bronchiectasis. The three groups were compared with respect to demographic data, symptoms, medical history, serum inflammatory markers, pulmonary function, and quantitative CT values. Results: Among 127 moderate-to-severe COPD subjects, 73 patients (57.5%) were detected to have bronchiectasis, 51 patients (40.2%) to have mild bronchiectasis, and 22 patients (17.3%) to have moderate-to-severe bronchiectasis. Compared with COPD patients without bronchiectasis, those with bronchiectasis were older and had higher frequency of prior tuberculosis, lower prevalence of bronchodilator reversibility (BDR), and more severe air trapping (P < 0.05). Moderate-to-severe bronchiectasis patients had lower body mass index (BMI), higher frequency of prior tuberculosis, lower prevalence of BDR, worse pulmonary function, and more severe air trapping (P < 0.05) than those in the mild bronchiectasis group. Conclusion: Moderate-to-severe bronchiectasis was associated with a history of pulmonary tuberculosis, lower BMI, severe airflow obstruction, and lower BDR in moderate-to-severe COPD patients. Quantitative analysis of CT showed that severe air trapping was associated with the extent of bronchiectasis in these patients.

Changes in the Characteristics and Long-term Mortality Rates of Intensive Care Unit Patients from 2003 to 2010: A Nationwide Population-Based Cohort Study Performed in the Republic of Korea.

BACKGROUND: There are few studies on intensive care unit (ICU) patients in the Republic of Korea. We analyzed the characteristics and mortality changes of all ICU patients over the last 8 years. METHODS: This study used the cohort of the National Health Insurance Corporation, which provides medical care to all residents of the Republic of Korea. The cohort consists of patients aged 20 years or older between 2003 and 2010 with a history of ICU admission. We analyzed changes in sex, age, household income, number of hospital beds, emergency admissions, and reasons for admission using the Cochran-Armitage trend test. The adjusted hazard ratios (HRs) of mortality according to these variables and year of admission were calculated by Cox proportional hazards regression. RESULTS: The proportion of patients aged ≥70 years increased over that period, as did their average age (by 3.6 years). During the 8-year study period, the 3-year mortality rate was 32.91%-35.83%. The overall mortality was higher in males and older patients, in those with a lower household income and higher Charlson Comorbidity Index (CCI) score, those admitted to a hospital with a smaller number of beds, and those admitted via the emergency room. There was no significant change in crude mortality rate over the 8-year study period; however, the adjusted HR showed a decreasing trend. CONCLUSIONS: Patients admitted to the ICU were older and had higher CCI score. Nevertheless, there was a temporal trend toward a decrease in the HR of long-term mortality.