Deep learning-based natural language processing for detecting medical symptoms and histories in emergency patient triage.

OBJECTIVE: The manual recording of electronic health records (EHRs) by clinicians in the emergency department (ED) is time-consuming and challenging. In light of recent advancements in large language models (LLMs) such as GPT and BERT, this study aimed to design and validate LLMs for automatic clinical diagnoses. The models were designed to identify 12 medical symptoms and 2 patient histories from simulated clinician-patient conversations within 6 primary symptom scenarios in emergency triage rooms. MATERIALS AND METHOD: We developed classification models by fine-tuning BERT, a transformer-based pre-trained model. We subsequently analyzed these models using eXplainable artificial intelligence (XAI) and the Shapley additive explanation (SHAP) method. A Turing test was conducted to ascertain the reliability of the XAI results by comparing them to the outcomes of tasks performed and explained by medical workers. An emergency medicine specialist assessed the results of both XAI and the medical workers. RESULTS: We fine-tuned four pre-trained LLMs and compared their classification performance. The KLUE-RoBERTa-based model demonstrated the highest performance (F1-score: 0.965, AUROC: 0.893) on human-transcribed script data. The XAI results using SHAP showed an average Jaccard similarity of 0.722 when compared with explanations of medical workers for 15 samples. The Turing test results revealed a small 6% gap, with XAI and medical workers receiving the mean scores of 3.327 and 3.52, respectively. CONCLUSION: This paper highlights the potential of LLMs for automatic EHR recording in Korean EDs. The KLUE-RoBERTa-based model demonstrated superior classification performance. Furthermore, XAI using SHAP provided reliable explanations for model outputs. The reliability of these explanations was confirmed by a Turing test.

The prognostic value of the phosphate-to-albumin ratio in patients with OHCA: A multicenter observational study.

PURPOSE: In patients with out-of-hospital cardiac arrest (OHCA), early and accurate outcome prediction is crucial for making treatment decisions and informing their relatives. A previous study reported an association between high phosphate levels and unfavorable neurological outcomes after return of spontaneous circulation (ROSC); however, its prognostic value was insufficient when used independently. Therefore, this study aimed to validate the usefulness of the phosphate-to-albumin ratio (PAR) in predicting neurological outcomes and in-hospital mortality by incorporating albumin, another known prognostic indicator. MATERIALS AND METHODS: This multicenter observational study included adult OHCA survivors from October 2015 to June 2021. The primary endpoint was an unfavorable neurological outcome at hospital discharge, defined as a cerebral performance category score of 3-5. The in-hospital mortality rates were also evaluated. RESULTS: Of the 2397 adult OHCA survivors, PAR differed significantly between the unfavorable and favorable neurological outcome groups, as well as between the non-survival and survival to hospital discharge groups (2.4 vs 1.4, 2.5 vs 1.6, respectively). The area under the receiver operating characteristic curve (AUROC) value of the PAR for predicting unfavorable neurological outcome was 0.81 (95% confidence interval [CI], 0.79-0.83), and the AUROC value for predicting in-hospital mortality was 0.76 (95% CI, 0.74-0.78). In multivariable analysis, the PAR was independently associated with unfavorable neurological outcome (odds ratio [OR] 1.30, 95% CI 1.15-1.37; p < 0.001) and in-hospital mortality (OR 1.24, 95% CI 1.12-1.38; p < 0.001). CONCLUSION: The PAR is a readily obtainable and independent prognostic indicator for patients with ROSC after OHCA, helping healthcare providers in predicting outcomes.

Diagnostic performance of rapid antigen tests for SARS-CoV-2 transmission risk based on cycle threshold values in the emergency department.

BACKGROUND: This study aimed to investigate the diagnostic performance of the rapid antigen test (RAT) for screening patients with cycle threshold (Ct) values of SARS-CoV-2 reverse transcription-polymerase chain reaction (RT-PCR) in the emergency department. Previous studies have shown that Ct values could be used as indicators of infectiousness. Therefore, we considered the Ct value an indicator of potential infectiousness. METHODS: This single-center retrospective observational study was conducted between January 1, 2020, and March 31, 2022. Patients who underwent both RT-PCR and RAT for the diagnosis of COVID-19 were included. Patients with negative RT-PCR results were excluded. Patients with Ct values lower than 26 and 30 were considered potentially infectious for COVID-19. RESULT: A total of 386 patients were analyzed. At Ct value cutoffs of 26 and 30, the result of the RAT showed a sensitivity of 82% and 74%, specificity of 84% and 89%, and area under the curve (AUC) of 0.829 and 0.813, respectively, in the receiver operating characteristic curve. However, the NPV was relatively low at 55% and 25%. CONCLUSION: The RAT might be a rapid screening tool for detecting patients with the infectiousness of SARS-CoV-2. However, considering the low NPV, it is challenging to depend only on a negative test result from an antigen test to terminate quarantine. Clinicians should consider additional factors, such as the duration of symptoms and the immunocompromised state, for SARS-CoV-2 transmission.

Effect of Contrast Level and Image Format on a Deep Learning Algorithm for the Detection of Pneumothorax with Chest Radiography.

Under the black-box nature in the deep learning model, it is uncertain how the change in contrast level and format affects the performance. We aimed to investigate the effect of contrast level and image format on the effectiveness of deep learning for diagnosing pneumothorax on chest radiographs. We collected 3316 images (1016 pneumothorax and 2300 normal images), and all images were set to the standard contrast level (100%) and stored in the Digital Imaging and Communication in Medicine and Joint Photographic Experts Group (JPEG) formats. Data were randomly separated into 80% of training and 20% of test sets, and the contrast of images in the test set was changed to 5 levels (50%, 75%, 100%, 125%, and 150%). We trained the model to detect pneumothorax using ResNet-50 with 100% level images and tested with 5-level images in the two formats. While comparing the overall performance between each contrast level in the two formats, the area under the receiver-operating characteristic curve (AUC) was significantly different (all p < 0.001) except between 125 and 150% in JPEG format (p = 0.382). When comparing the two formats at same contrast levels, AUC was significantly different (all p < 0.001) except 50% and 100% (p = 0.079 and p = 0.082, respectively). The contrast level and format of medical images could influence the performance of the deep learning model. It is required to train with various contrast levels and formats of image, and further image processing for improvement and maintenance of the performance.

Initial creatine kinase level as predictor for delayed neuropsychiatric sequelae associated with acute carbon monoxide poisoning.

INTRODUCTION: The primary goal of treating patients with carbon monoxide (CO) poisoning is preventing delayed neuropsychiatric sequelae (DNS). It is difficult to predict DNS because there is no precise diagnostic method in the early phase of CO poisoning. In this study, we aimed to investigate the optimal cut-off value for creatine kinase level to predict DNS. METHODS: This retrospective observational study included patients with CO poisoning visiting a single tertiary center from January to July 2018. They were divided into two groups according to the presence of DNS. We compared baseline characteristics with variables that could affect the presence of DNS. The optimal cut-off value of initial creatine kinase concentration for DNS was calculated. Additionally, multivariate analysis was performed to confirm whether creatine kinase could be an independent predictor of DNS. RESULTS: Of the 138 patients, 12 patients developed DNS. Univariate analysis showed significant differences in the Glasgow Coma Scale, duration of exposure, laboratory tests, abnormal finding on MRI in acute phase, the number of hyperbaric oxygen therapy sessions, and duration of hospitalization. Receiver operating characteristic analyses of creatine kinase were performed (AUC = 0.92; 95% CI, 0.86-0.96) with a cut-off value of 1603 U/L; DNS was predicted with a sensitivity of 91.7% and specificity of 88.1%. In multivariate analysis, the adjusted odds ratio of creatine kinase was 51.516. CONCLUSION: In patients with CO poisoning, initial creatine kinase concentrations of >1603 U/L can be used as an independent predictor of DNS.

Automatic Classification of the Korean Triage Acuity Scale in Simulated Emergency Rooms Using Speech Recognition and Natural Language Processing: a Proof of Concept Study.

BACKGROUND: Rapid triage reduces the patients' stay time at an emergency department (ED). The Korean Triage Acuity Scale (KTAS) is mandatorily applied at EDs in South Korea. For rapid triage, we studied machine learning-based triage systems composed of a speech recognition model and natural language processing-based classification. METHODS: We simulated 762 triage cases that consisted of 18 classes with six types of the main symptom (chest pain, dyspnea, fever, stroke, abdominal pain, and headache) and three levels of KTAS. In addition, we recorded conversations between emergency patients and clinicians during the simulation. We used speech recognition models to transcribe the conversation. Bidirectional Encoder Representation from Transformers (BERT), support vector machine (SVM), random forest (RF), and k-nearest neighbors (KNN) were used for KTAS and symptom classification. Additionally, we evaluated the Shapley Additive exPlanations (SHAP) values of features to interpret the classifiers. RESULTS: The character error rate of the speech recognition model was reduced to 25.21% through transfer learning. With auto-transcribed scripts, support vector machine (area under the receiver operating characteristic curve [AUROC], 0.86; 95% confidence interval [CI], 0.81-0.9), KNN (AUROC, 0.89; 95% CI, 0.85-0.93), RF (AUROC, 0.86; 95% CI, 0.82-0.9) and BERT (AUROC, 0.82; 95% CI, 0.75-0.87) achieved excellent classification performance. Based on SHAP, we found "stress", "pain score point", "fever", "breath", "head" and "chest" were the important vocabularies for determining KTAS and symptoms. CONCLUSION: We demonstrated the potential of an automatic KTAS classification system using speech recognition models, machine learning and BERT-based classifiers.

Surge Capacity and Mass Casualty Incidents Preparedness of Emergency Departments in a Metropolitan City: a Regional Survey Study.

BACKGROUND: Emergency departments (EDs) generally receive many casualties in disaster or mass casualty incidents (MCI). Some studies have conceptually suggested the surge capacity that ED should have; however, only few studies have investigated measurable numbers in one community. This study investigated the surge capacity of the specific number of accommodatable patients and overall preparedness at EDs in a metropolitan city. METHODS: This cross-sectional study officially surveyed surge capacity and disaster preparedness for all regional and local emergency medical centers (EMC) in Seoul with the Seoul Metropolitan Government's public health division. This study developed survey items on space, staff, stuff, and systems, which are essential elements of surge capacity. The number of patients acceptable for each ED was investigated by triage level in ordinary and crisis situations. Multivariate linear regression analysis was performed on hospital resource variables related to surge capacity. RESULTS: In the second half of 2018, a survey was conducted targeting 31 EMC directors in Seoul. It was found that all regional and local EMCs in Seoul can accommodate 848 emergency patients and 537 non-emergency patients in crisis conditions. In ordinary situations, one EMC could accommodate an average of 1.3 patients with Korean Triage and Acuity Scale (KTAS) level 1, 3.1 patients with KTAS level 2, and 5.7 patients with KTAS level 3. In situations of crisis, this number increased to 3.4, 7.8, and 16.2, respectively. There are significant differences in surge capacity between ordinary and crisis conditions. The difference in surge capacity between regional and local EMC was not significant. In both ordinary and crisis conditions, only the total number of hospital beds were significantly associated with surge capacity. CONCLUSION: If the hospital's emergency transport system is ideally accomplished, patients arising from average MCI can be accommodated in Seoul. However, in a huge disaster, it may be challenging to handle the current surge capacity. More detailed follow-up studies are needed to prepare a surge capacity protocol in the community.

External Validation of Deep Learning Algorithm for Detecting and Visualizing Femoral Neck Fracture Including Displaced and Non-displaced Fracture on Plain X-ray.

This study aimed to develop a method for detection of femoral neck fracture (FNF) including displaced and non-displaced fractures using convolutional neural network (CNN) with plain X-ray and to validate its use across hospitals through internal and external validation sets. This is a retrospective study using hip and pelvic anteroposterior films for training and detecting femoral neck fracture through residual neural network (ResNet) 18 with convolutional block attention module (CBAM) + + . The study was performed at two tertiary hospitals between February and May 2020 and used data from January 2005 to December 2018. Our primary outcome was favorable performance for diagnosis of femoral neck fracture from negative studies in our dataset. We described the outcomes as area under the receiver operating characteristic curve (AUC), accuracy, Youden index, sensitivity, and specificity. A total of 4,189 images that contained 1,109 positive images (332 non-displaced and 777 displaced) and 3,080 negative images were collected from two hospitals. The test values after training with one hospital dataset were 0.999 AUC, 0.986 accuracy, 0.960 Youden index, and 0.966 sensitivity, and 0.993 specificity. Values of external validation with the other hospital dataset were 0.977, 0.971, 0.920, 0.939, and 0.982, respectively. Values of merged hospital datasets were 0.987, 0.983, 0.960, 0.973, and 0.987, respectively. A CNN algorithm for FNF detection in both displaced and non-displaced fractures using plain X-rays could be used in other hospitals to screen for FNF after training with images from the hospital of interest.

Risk of Venous Thromboembolism After Carbon Monoxide Poisoning: A Nationwide Population-Based Study.

STUDY OBJECTIVE: Few studies have investigated the association between carbon monoxide (CO) poisoning and risk of venous thromboembolism. We aim to identify the risk of pulmonary embolism and deep venous thrombosis after CO poisoning. METHODS: We conducted a nationwide cohort-crossover study using administrative claims data in Korea. We compared the risk of venous thromboembolism (pulmonary embolism or deep venous thrombosis) in the cohort period after CO poisoning to that of the same period 1 year later (crossover period), using conditional logistic regression analysis. RESULTS: We included 22,699 patients with a diagnosis of CO poisoning during the study period between 2004 and 2015. The risk of venous thromboembolism was significantly elevated during days 0 to 90 after CO poisoning (odds ratio 3.96; 95% confidence interval 2.50 to 6.25). However, this risk was not significantly elevated during subsequent postexposure periods through 360 days. During days 0 to 30 after CO poisoning, the risks of pulmonary embolism (odds ratio 22.00; 95% confidence interval 5.33 to 90.75) and deep venous thrombosis (odds ratio 10.33; 95% confidence interval 3.16 to 33.80) were significantly elevated. CONCLUSION: We found that the risk of venous thromboembolism persisted for up to 90 days after CO poisoning. The risk was increased 22-fold for pulmonary embolism and 10-fold for deep venous thrombosis, especially in the first month after CO poisoning. Patients should be monitored for venous thromboembolism risk after CO poisoning.

Pulmonary thromboembolism after carbon monoxide poisoning.

OBJECTIVE: Carbon monoxide (CO) poisoning is known to increase thrombotic tendency, and the risk of deep vein thrombosis in individuals who have experienced CO poisoning is higher than in the general population. However, there are a few reports describing cases of pulmonary thromboembolisms (PTE) secondary to CO poisoning. DATA SOURCES: Retrospective data analysis. STUDY SELECTION: Seven hundred fifty bed tertiary university affiliated hospital. DATA EXTRACTION AND SYNTHESIS: Five patients with PTE after CO poisoning were observed. Two patients experienced cardiac arrest; they were treated successfully with tissue plasminogen activators and targeted temperature management. Their cerebral performance scores at discharge were both 1. Three patients had PTE and were treated with anticoagulation. CONCLUSIONS: To date, the causal relationship between PTE and CO poisoning is unclear. However, PTE should be considered in patients with CO poisoning as a differential diagnosis when unexplained hypoxemia or shock are observed. Further studies on the association between CO poisoning and PTE are warranted.

A Feasibility Study of Telementoring for Identifying the Appendix Using Smartphone-Based Telesonography.

We investigated the feasibility of the clinical application of novice-practitioner-performed/offsite-mentor-guided ultrasonography for identifying the appendix. A randomized crossover study was conducted using a telesonography system that can transmit the ultrasound images displayed on the ultrasound monitor (ultrasound sequence video) and images showing the practitioner's operations (background video) to a smartphone without any interruption in motion over a Long-Term Evolution (LTE) network. Thirty novice practitioners were randomly assigned to two groups. The subjects in group A (n = 15) performed ultrasonography for the identification of the appendix under mentoring by an onsite expert, whereas those in group B (n = 15) performed the same procedure under mentoring by an offsite expert. Each subject performed the procedure on three simulated patients. After a 4-week interval, they performed the procedure again under the other type of mentoring. A total of 90 ultrasound examinations were performed in each scenario. The primary outcomes were the success rate for identifying the appendix and the time required to identify the appendix. The success rates for identifying the appendix were 91.1 % (82/90) in onsite-mentored ultrasonography and 87.8 % (79/90) in offsite-mentored ultrasonography; both rates were high, and there was no significant difference (p = 0.468) between them. The time required in the case of offsite mentoring (median, 242.9 s; interquartile range (IQR), 238.2) was longer than that for onsite mentoring (median, 291.4 s; IQR, 200.9); however, the difference was not significant (p = 0.051). It appears that offsite mentoring can allow novice onsite practitioners to perform ultrasonography as effectively as they can under onsite mentoring, even for examinations that require proficiency in rather complex practices, such as identifying the appendix.

A Feasibility Study of Smartphone-Based Telesonography for Evaluating Cardiac Dynamic Function and Diagnosing Acute Appendicitis with Control of the Image Quality of the Transmitted Videos.

Our aim was to prove the feasibility of the remote interpretation of real-time transmitted ultrasound videos of dynamic and static organs using a smartphone with control of the image quality given a limited internet connection speed. For this study, 100 cases of echocardiography videos (dynamic organ)-50 with an ejection fraction (EF) of ≥50 s and 50 with EF <50 %-and 100 cases of suspected pediatric appendicitis (static organ)-50 with signs of acute appendicitis and 50 with no findings of appendicitis-were consecutively selected. Twelve reviewers reviewed the original videos using the liquid crystal display (LCD) monitor of an ultrasound machine and using a smartphone, to which the images were transmitted from the ultrasound machine. The resolution of the transmitted echocardiography videos was reduced by approximately 20 % to increase the frame rate of transmission given the limited internet speed. The differences in diagnostic performance between the two devices when evaluating left ventricular (LV) systolic function by measuring the EF and when evaluating the presence of acute appendicitis were investigated using a five-point Likert scale. The average areas under the receiver operating characteristic curves for each reviewer's interpretations using the LCD monitor and smartphone were respectively 0.968 (0.949-0.986) and 0.963 (0.945-0.982) (P = 0.548) for echocardiography and 0.972 (0.954-0.989) and 0.966 (0.947-0.984) (P = 0.175) for abdominal ultrasonography. We confirmed the feasibility of remotely interpreting ultrasound images using smartphones, specifically for evaluating LV function and diagnosing pediatric acute appendicitis; the images were transferred from the ultrasound machine using image quality-controlled telesonography.

Hierarchically porous polymers from hyper-cross-linked block polymer precursors.

We report synthesis of hierarchically porous polymers (HPPs) consisting of micropores and well-defined 3D continuous mesopores by combination of hyper-cross-linking and block polymer self-assembly. Copolymerization of 4-vinylbenzyl chloride (VBzCl) with divinylbenzene (DVB) in the presence of polylactide (PLA) macro-chain-transfer agent produced a cross-linked block polymer precursor PLA-b-P(VBzCl-co-DVB) via reversible addition-fragmentation chain transfer polymerization. A nanoscopic bicontinuous morphology containing PLA and P(VBzCl-co-DVB) microdomains was obtained as a result of polymerization-induced microphase separation. While a basic treatment of the precursor selectively removed PLA to yield a reticulated mesoporous polymer, hyper-cross-linking of the precursor by FeCl3 generated micropores in the P(VBzCl-co-DVB) microdomain via Friedel-Crafts alkylation and simultaneously degraded PLA to produce the HPP containing micropores in the mesoporous framework. The mesopore size of the HPP could be precisely controlled from 6 to 15 nm by controlling the molar mass of PLA. We demonstrate acceleration in adsorption rate in the HPP compared to a hyper-cross-linked microporous polymer.

Clinical application of real-time tele-ultrasonography in diagnosing pediatric acute appendicitis in the ED.

PURPOSE: We investigated the effectiveness of tele-mentored ultrasonography between emergency medicine (EM) residents and remote experts in diagnosing acute appendicitis. METHODS: This prospective observational study was performed in an academic emergency department. Beginning in June 2014, the EM residents performed the initial ultrasonography for suspected pediatric acute appendicitis; then, the remote experts observed/mentored the residents' practice using the tele-ultrasonography system; and finally, an onsite expert verified the diagnosis. The diagnostic confidence of each examiner (resident, remote expert, and onsite expert) was rated on a 5-point Likert scale. The appendix identification rate and the diagnostic values; sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for each type of examination were calculated. RESULTS: A total of 115 pediatric cases with suspected acute appendicitis, including 36 with pathology-confirmed acute appendicitis, between June 2014 and February 2015 were enrolled in this study. In 9 of the 115 (7.8%) cases, a non-appendicitis diagnosis was determined in the absence of the successful identification of a normal appendix upon resident examination. Of these, seven appendices were identified upon expert tele-ultrasonography. The diagnostic values for expert tele-ultrasonography were higher (sensitivity: 1.000, specificity: 0.975, PPV: 0.947, NPV: 1.000) than those for resident-performed ultrasonography (sensitivity: 0.917, specificity: 0.899, PPV: 0.805, NPV: 0.959) and similar to those for onsite expert-performed ultrasonography (sensitivity: 1.000, specificity: 0.987, PPV: 0.973, NPV: 1.000). CONCLUSION: Tele-ultrasonography with tele-mentoring between EM residents and experienced mentors was effectively applied in diagnosing pediatric acute appendicitis in an emergency clinical setting.

Proper target depth of an accelerometer-based feedback device during CPR performed on a hospital bed: a randomized simulation study.

PURPOSE: Feedback devices are used to improve chest compression (CC) quality related to survival rates in cardiac arrest. However, several studies have shown that feedback devices are not sufficiently reliable to ensure adequate CC depth on soft surfaces. Here, we determined the proper target depth of feedback (TDF) using an accelerometer during cardiopulmonary resuscitation in hospital beds. METHODS: In prospective randomized crossover study, 19 emergency physicians performed CCs for 2 minutes continuously on a manikin in 2 different beds with 3 TDFs (5, 6, and 7 cm). We measured CC depth, the proportion of accurate compression depths, CC rate, the proportion of incomplete chest decompressions, the velocity of CC (CC velocity), the proportion of time spent in CC relative to compression plus decompression (duty cycle), and the time spent in CC (CC time). RESULTS: Mean (SD) CC depths at TDF 5, 6, and 7 were 45.42 (5.79), 52.68 (4.18), and 58.47 (2.48) on one bed and 46.26 (4.49), 53.58 (3.15), and 58.74 (2.10) mm on the other bed (all P<.001), respectively. The proportions of accurate compression depths and CC velocity at TDF 5, 6, and 7 differed significantly according to TDF on both beds (all P<.001).The CC rate, CC time, and proportion of incomplete chest decompression did not differ on both beds (all P>.05). The duty cycle differed significantly on only B2. CONCLUSIONS: The target depth of the real-time feedback device should be at least 6 cm but should not exceed 7 cm for optimal CC on patients on hospital beds.

A new chest compression depth feedback algorithm for high-quality CPR based on smartphone.

BACKGROUND: Although many smartphone application (app) programs provide education and guidance for basic life support, they do not commonly provide feedback on the chest compression depth (CCD) and rate. The validation of its accuracy has not been reported to date. This study was a feasibility assessment of use of the smartphone as a CCD feedback device. In this study, we proposed the concept of a new real-time CCD estimation algorithm using a smartphone and evaluated the accuracy of the algorithm. MATERIALS AND METHODS: Using the double integration of the acceleration signal, which was obtained from the accelerometer in the smartphone, we estimated the CCD in real time. Based on its periodicity, we removed the bias error from the accelerometer. To evaluate this instrument's accuracy, we used a potentiometer as the reference depth measurement. The evaluation experiments included three levels of CCD (insufficient, adequate, and excessive) and four types of grasping orientations with various compression directions. We used the difference between the reference measurement and the estimated depth as the error. The error was calculated for each compression. RESULTS: When chest compressions were performed with adequate depth for the patient who was lying on a flat floor, the mean (standard deviation) of the errors was 1.43 (1.00) mm. When the patient was lying on an oblique floor, the mean (standard deviation) of the errors was 3.13 (1.88) mm. CONCLUSIONS: The error of the CCD estimation was tolerable for the algorithm to be used in the smartphone-based CCD feedback app to compress more than 51 mm, which is the 2010 American Heart Association guideline.

The Incidence and Outcomes of Out-of-Hospital Cardiac Arrest During the COVID-19 Pandemic in South Korea: Multicenter Registry Study.

BACKGROUND: The COVID-19 pandemic has profoundly affected out-of-hospital cardiac arrest (OHCA) and disrupted the chain of survival. Even after the end of the pandemic, the risk of new variants and surges persists. Analyzing the characteristics of OHCA during the pandemic is important to prepare for the next pandemic and to avoid repeated negative outcomes. However, previous studies have yielded somewhat varied results, depending on the health care system or the specific characteristics of social structures. OBJECTIVE: We aimed to investigate and compare the incidence, outcomes, and characteristics of OHCA during the prepandemic and pandemic periods using data from a nationwide multicenter OHCA registry. METHODS: We conducted a multicenter, retrospective, observational study using data from the Korean Cardiac Arrest Resuscitation Consortium (KoCARC) registry. This study included adult patients with OHCA in South Korea across 3 distinct 1-year periods: the prepandemic period (from January to December 2019), early phase pandemic period (from July 2020 to June 2021), and late phase pandemic period (from July 2021 to June 2022). We extracted and contrasted the characteristics of patients with OHCA, prehospital time factors, and outcomes for the patients across these 3 periods. The primary outcomes were survival to hospital admission and survival to hospital discharge. The secondary outcome was good neurological outcome. RESULTS: From the 3 designated periods, a total of 9031 adult patients with OHCA were eligible for analysis (prepandemic: n=2728; early pandemic: n=2954; and late pandemic: n=3349). Witnessed arrest (P<.001) and arrest at home or residence (P=.001) were significantly more frequent during the pandemic period than during the prepandemic period, and automated external defibrillator use by bystanders was lower in the early phase of the pandemic than during other periods. As the pandemic advanced, the rates of the first monitored shockable rhythm (P=.10) and prehospital endotracheal intubation (P<.001) decreased significantly. Time from cardiac arrest cognition to emergency department arrival increased sequentially (prepandemic: 33 min; early pandemic: 35 min; and late pandemic: 36 min; P<.001). Both survival and neurological outcomes worsened as the pandemic progressed, with survival to discharge showing the largest statistical difference (prepandemic: 385/2728, 14.1%; early pandemic: 355/2954, 12%; and late pandemic: 392/3349, 11.7%; P=.01). Additionally, none of the outcomes differed significantly between the early and late phase pandemic periods (all P>.05). CONCLUSIONS: During the pandemic, especially amid community COVID-19 surges, the incidence of OHCA increased while survival rates and good neurological outcome at discharge decreased. Prehospital OHCA factors, which are directly related to OHCA prognosis, were adversely affected by the pandemic. Ongoing discussions are needed to maintain the chain of survival in the event of a new pandemic. TRIAL REGISTRATION: ClinicalTrials.gov NCT03222999; https://classic.clinicaltrials.gov/ct2/show/NCT03222999.

Correlation between Carboxyhemoglobin Levels Measured by Blood Gas Analysis and by Multiwave Pulse Oximetry.

Carbon monoxide (CO) poisoning is difficult to diagnose owing to its nonspecific symptoms. Multiwave pulse oximetry can be used to quickly screen patients for CO poisoning. However, few studies have analyzed patients with CO poisoning who presented to the emergency department (ED). The primary aim of our study was to determine the correlation between COHb levels measured in blood gas analysis and COHb levels measured in multiwave pulse oximetry. Secondary aims were the sensitivity and specificity of the COHb level cutoff value using multiwave pulse oximetry to predict a 25% COHb level in blood gas analysis. This single-center retrospective observational study included patients with CO poisoning who visited the ED of a university-affiliated hospital in Seoul, Korea between July 2021 and June 2023. COHb poisoning was determined using blood gas analysis and multiwave pulse oximetry. The correlation of COHb levels between the two tests was evaluated using correlation analysis. The area under the receiver operating characteristic curve (AUC) of multiwave pulse oximetry was calculated to predict COHb levels from the blood gas analysis. The optimal cutoff values, sensitivity, and specificity of COHb were determined. A total of 224 patients who had COHb levels measured using both multiwave pulse oximetry and blood gas analysis were included in the analysis. In the correlation analysis, COHb showed a high positive correlation with COHb measured using blood gas analysis (Spearman correlation coefficient = 0.86, p < 0.001). The AUC of COHb measured by multiwave pulse oximetry to predict 25% of the COHb level (which can be an indication of hyperbaric oxygen treatment) measured by blood gas analysis was 0.916. When the COHb levels measured with multiwave pulse oximetry were 20% the sensitivity was 81% and the specificity was 83%, and when the COHb levels were 25% the sensitivity was 50% and the specificity was 95%. The COHb value measured using multiwave pulse oximetry blood gas analysis showed a high correlation. However, additional research using large-scale studies is required for validation.

Kidney, ureter, and urinary bladder segmentation based on non-contrast enhanced computed tomography images using modified U-Net.

This study was performed to segment the urinary system as the basis for diagnosing urinary system diseases on non-contrast computed tomography (CT). This study was conducted with images obtained between January 2016 and December 2020. During the study period, non-contrast abdominopelvic CT scans of patients and diagnosed and treated with urinary stones at the emergency departments of two institutions were collected. Region of interest extraction was first performed, and urinary system segmentation was performed using a modified U-Net. Thereafter, fivefold cross-validation was performed to evaluate the robustness of the model performance. In fivefold cross-validation results of the segmentation of the urinary system, the average dice coefficient was 0.8673, and the dice coefficients for each class (kidney, ureter, and urinary bladder) were 0.9651, 0.7172, and 0.9196, respectively. In the test dataset, the average dice coefficient of best performing model in fivefold cross validation for whole urinary system was 0.8623, and the dice coefficients for each class (kidney, ureter, and urinary bladder) were 0.9613, 0.7225, and 0.9032, respectively. The segmentation of the urinary system using the modified U-Net proposed in this study could be the basis for the detection of kidney, ureter, and urinary bladder lesions, such as stones and tumours, through machine learning.