Association between microcirculation in spontaneous breathing trial and extubation success.

PURPOSE: This study assessed the association between changes in sublingual microcirculation after a spontaneous breathing trial (SBT) and successful extubation. MATERIALS AND METHODS: Sublingual microcirculation was assessed using an incident dark-field video microscope before and after each SBT and before extubation. Microcirculatory parameters before the SBT, at the end of the SBT, and before extubation were compared between the successful and failed extubation groups. RESULTS: Forty-seven patients were enrolled and analysed in this study (34 patients in the successful extubation group and 13 patients in the failed extubation group). At the end of the SBT, the weaning parameters did not differ between the two groups. However, the total small vessel density (21.2 [20.4-23.7] versus 24.9 [22.6-26.5] mm/mm2), perfused small vessel density (20.6 [18.5-21.8] versus 23.1 [20.9-25] mm/mm2), proportion of perfused small vessels (91 [87-96] versus 95 [93-98] %), and microvascular flow index (2.8 [2.7-2.9] versus 2.9 [2.9-3]) were significantly lower in the failed extubation group than in the successful extubation group. The weaning and microcirculatory parameters did not differ significantly between the two groups before the SBT. CONCLUSIONS: More patients are required to investigate the difference between baseline microcirculation before a successful SBT and the change in microcirculation at the end of the SBT between the successful and failed extubation groups. Better sublingual microcirculatory parameters at the end of SBT and before extubation are associated with successful extubation.

Etiology-associated heterogeneity in acute respiratory distress syndrome: a retrospective cohort study.

BACKGROUND: Heterogeneity in acute respiratory distress syndrome (ARDS) has led to many statistically negative clinical trials. Etiology is considered an important source of pathogenesis heterogeneity in ARDS but previous studies have usually adopted a dichotomous classification, such as pulmonary versus extrapulmonary ARDS, to evaluate it. Etiology-associated heterogeneity in ARDS remains poorly described. METHODS: In this retrospective cohort study, we described etiology-associated heterogeneity in gas exchange abnormality (PaO2/FiO2 [P/F] and ventilatory ratios), hemodynamic instability, non-pulmonary organ dysfunction as measured by the Sequential Organ Failure Assessment (SOFA) score, biomarkers of inflammation and coagulation, and 30-day mortality. Linear regression was used to model the trajectory of P/F ratios over time. Wilcoxon rank-sum tests, Kruskal-Wallis rank tests and Chi-squared tests were used to compare between-etiology differences. RESULTS: From 1725 mechanically ventilated patients in the ICU, we identified 258 (15%) with ARDS. Pneumonia (48.4%) and non-pulmonary sepsis (11.6%) were the two leading causes of ARDS. Compared with pneumonia associated ARDS, extra-pulmonary sepsis associated ARDS had a greater P/F ratio recovery rate (difference = 13 mmHg/day, p = 0.01), more shock (48% versus 73%, p = 0.01), higher non-pulmonary SOFA scores (6 versus 9 points, p < 0.001), higher d-dimer levels (4.2 versus 9.7 mg/L, p = 0.02) and higher mortality (43% versus 67%, p = 0.02). In pneumonia associated ARDS, there was significant difference in proportion of shock (p = 0.005) between bacterial and non-bacterial pneumonia. CONCLUSION: This study showed that there was remarkable etiology-associated heterogeneity in ARDS. Heterogeneity was also observed within pneumonia associated ARDS when bacterial pneumonia was compared with other non-bacterial pneumonia. Future studies on ARDS should consider reporting etiology-specific data and exploring possible effect modification associated with etiology.

Evaluation of the Bruker Biotyper matrix-assisted laser desorption ionization-time of flight mass spectrometry system for identification of blood isolates of Acinetobacter species.

Matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) (Bruker Biotyper) was able to accurately identify 98.6% (142/144) of Acinetobacter baumannii isolates, 72.4% (63/87) of A. nosocomialis isolates, and 97.6% (41/42) of A. pittii isolates. All Acinetobacter junii, A. ursingii, A. johnsonii, and A. radioresistens isolates (n = 28) could also be identified correctly by Bruker Biotyper.

Increase in the rate of azithromycin-resistant Streptococcus pneumoniae isolates carrying the erm(B) and mef(A) genes in Taiwan, 2006-2010.

BACKGROUND: This study investigated the molecular characteristics of azithromycin-resistant Streptococcus pneumoniae in Taiwan. METHODS: A total of 486 non-duplicate isolates of azithromycin-resistant S. pneumoniae recovered from various clinical sources of patients treated at 22 different hospitals in Taiwan from 2006 to 2010. The presence of erm(B) and mef(A) genes using duplex PCR, multilocus sequence typing (MLST), and pulsed-field gel electrophoresis of these isolates were studied. RESULTS: Of the isolates tested, 59% carried the erm(B) gene, 22% carried the mef(A) gene, and 19% carried both genes. The prevalence of isolates carrying the erm(B) and mef(A) genes increased from 10% (11/110) in 2006 to 25% (15/60) in 2010 (p-value = 0.0136). The majority of isolates carrying both erm(B) and mef(A) genes belonged to serotypes 19 F (64%) followed by 19 F A (24%). Of these isolates, 33% were sequence type 320 (ST320), 32% were ST236, and 12% were ST271. CONCLUSIONS: The increase in incidence of mef(A)/erm(B)-positive azithromycin-resistant S. pneumoniae isolates during the study period was primarily due to serotypes 19 F and 19A and ST236 and ST320.

The impact of nontuberculous mycobacterial lung disease in critically ill patients: Significance for survival and ventilator use.

BACKGROUND: This study investigates the impact of nontuberculous mycobacterial lung disease (NTM-LD) on mortality and mechanical ventilation use in critically ill patients. METHODS: We enrolled patients with NTM-LD or tuberculosis (TB) in intensive care units (ICU) and analysed their association with 30-day mortality and with mechanical ventilator-free survival (VFS) at 30 days after ICU admission. RESULTS: A total of 5996 ICU-admitted patients were included, of which 541 (9.0 %) had TB and 173 (2.9 %) had NTM-LD. The overall 30-day mortality was 22.2 %. The patients with NTM-LD had an adjusted hazard ratio (aHR) of 1.49 (95 % CI, 1.06-2.05), and TB patients had an aHR of 2.33 (95 % CI, 1.68-3.24), compared to ICU patients with negative sputum mycobacterial culture by multivariable Cox proportional hazard (PH) regression. The aHR of age<65 years, obesity, idiopathic pulmonary fibrosis, end-stage kidney disease, active cancer and autoimmune disease and diagnosis of respiratory failure were also significantly positively associated with ICU 30-day mortality. In multivariable Cox PH regression for VFS at 30 days in patients requiring invasive mechanical ventilation, NTM-LD was negatively associated with VFS (aHR 0.71, 95 % CI: 0.56-0.92, p = 0.009), while TB showed no significant association. The diagnosis of respiratory failure itself predicted unfavourable outcome for 30-day mortality and a negative impact on VFS at 30 days. CONCLUSIONS: NTM-LD and TB were not uncommon in ICU and both were correlated with increasing 30-day mortality in ICU patients. NTM-LD was associated with a poorer outcome in terms of VFS at 30 days.

Deep learning with test-time augmentation for radial endobronchial ultrasound image differentiation: a multicentre verification study.

PURPOSE: Despite the importance of radial endobronchial ultrasound (rEBUS) in transbronchial biopsy, researchers have yet to apply artificial intelligence to the analysis of rEBUS images. MATERIALS AND METHODS: This study developed a convolutional neural network (CNN) to differentiate between malignant and benign tumours in rEBUS images. This study retrospectively collected rEBUS images from medical centres in Taiwan, including 769 from National Taiwan University Hospital Hsin-Chu Branch, Hsinchu Hospital for model training (615 images) and internal validation (154 images) as well as 300 from National Taiwan University Hospital (NTUH-TPE) and 92 images were obtained from National Taiwan University Hospital Hsin-Chu Branch, Biomedical Park Hospital (NTUH-BIO) for external validation. Further assessments of the model were performed using image augmentation in the training phase and test-time augmentation (TTA). RESULTS: Using the internal validation dataset, the results were as follows: area under the curve (AUC) (0.88 (95% CI 0.83 to 0.92)), sensitivity (0.80 (95% CI 0.73 to 0.88)), specificity (0.75 (95% CI 0.66 to 0.83)). Using the NTUH-TPE external validation dataset, the results were as follows: AUC (0.76 (95% CI 0.71 to 0.80)), sensitivity (0.58 (95% CI 0.50 to 0.65)), specificity (0.92 (95% CI 0.88 to 0.97)). Using the NTUH-BIO external validation dataset, the results were as follows: AUC (0.72 (95% CI 0.64 to 0.82)), sensitivity (0.71 (95% CI 0.55 to 0.86)), specificity (0.76 (95% CI 0.64 to 0.87)). After fine-tuning, the AUC values for the external validation cohorts were as follows: NTUH-TPE (0.78) and NTUH-BIO (0.82). Our findings also demonstrated the feasibility of the model in differentiating between lung cancer subtypes, as indicated by the following AUC values: adenocarcinoma (0.70; 95% CI 0.64 to 0.76), squamous cell carcinoma (0.64; 95% CI 0.54 to 0.74) and small cell lung cancer (0.52; 95% CI 0.32 to 0.72). CONCLUSIONS: Our results demonstrate the feasibility of the proposed CNN-based algorithm in differentiating between malignant and benign lesions in rEBUS images.

A deep learning model using chest X-ray for identifying TB and NTM-LD patients: a cross-sectional study.

BACKGROUND: Timely differentiating between pulmonary tuberculosis (TB) and nontuberculous mycobacterial lung disease (NTM-LD), which are radiographically similar, is important because infectiousness and treatment differ. This study aimed to evaluate whether artificial intelligence could distinguish between TB or NTM-LD patients by chest X-rays (CXRs) from suspects of mycobacterial lung disease. METHODS: A total of 1500 CXRs, including 500 each from patients with pulmonary TB, NTM-LD, and patients with clinical suspicion but negative mycobacterial culture (Imitator) from two hospitals, were retrospectively collected and evaluated in this study. We developed a deep neural network (DNN) and evaluated model performance using the area under the receiver operating characteristic curves (AUC) in both internal and external test sets. Furthermore, we conducted a reader study and tested our model under three scenarios of different mycobacteria prevalence. RESULTS: Among the internal and external test sets, the AUCs of our DNN model were 0.83 ± 0.005 and 0.76 ± 0.006 for pulmonary TB, 0.86 ± 0.006 and 0.64 ± 0.017 for NTM-LD, and 0.77 ± 0.007 and 0.74 ± 0.005 for Imitator. The DNN model showed higher performance on the internal test set in classification accuracy (66.5 ± 2.5%) than senior (50.8 ± 3.0%, p < 0.001) and junior pulmonologists (47.5 ± 2.8%, p < 0.001). Among different prevalence scenarios, the DNN model has stable performance in terms of AUC to detect TB and mycobacterial lung disease. CONCLUSION: DNN model had satisfactory performance and a higher accuracy than pulmonologists on classifying patients with presumptive mycobacterial lung diseases. DNN model could be a complementary first-line screening tool.

Early prediction of delirium upon intensive care unit admission: Model development, validation, and deployment.

STUDY OBJECTIVE: To develop, validate, and deploy models for predicting delirium in critically ill adult patients as early as upon intensive care unit (ICU) admission. DESIGN: Retrospective cohort study. SETTING: Single university teaching hospital in Taipei, Taiwan. PATIENTS: 6238 critically ill patients from August 2020 to August 2021. MEASUREMENTS: Data were extracted, pre-processed, and split into training and testing datasets based on the time period. Eligible variables included demographic characteristics, Glasgow Coma Scale, vital signs parameters, treatments, and laboratory data. The predicted outcome was delirium, defined as any positive result (a score ≥ 4) of the Intensive Care Delirium Screening Checklist that was assessed by primary care nurses in each 8-h shift within 48 h after ICU admission. We trained models to predict delirium upon ICU admission (ADM) and at 24 h (24H) after ICU admission by using logistic regression (LR), gradient boosted trees (GBT), and deep learning (DL) algorithms and compared the models' performance. MAIN RESULTS: Eight features were extracted from the eligible features to train the ADM models, including age, body mass index, medical history of dementia, postoperative intensive monitoring, elective surgery, pre-ICU hospital stays, and GCS score and initial respiratory rate upon ICU admission. In the ADM testing dataset, the incidence of ICU delirium occurred within 24 h and 48 h was 32.9% and 36.2%, respectively. The area under the receiver operating characteristic curve (AUROC) (0.858, 95% CI 0.835-0.879) and area under the precision-recall curve (AUPRC) (0.814, 95% CI 0.780-0.844) for the ADM GBT model were the highest. The Brier scores of the ADM LR, GBT, and DL models were 0.149, 0.140, and 0.145, respectively. The AUROC (0.931, 95% CI 0.911-0.949) was the highest for the 24H DL model and the AUPRC (0.842, 95% CI 0.792-0.886) was the highest for the 24H LR model. CONCLUSION: Our early prediction models based on data obtained upon ICU admission could achieve good performance in predicting delirium occurred within 48 h after ICU admission. Our 24-h models can improve delirium prediction for patients discharged >1 day after ICU admission.

Improving the Process of Shared Decision-Making by Integrating Online Structured Information and Self-Assessment Tools.

The integration of face-to-face communication and online processes to provide access to information and self-assessment tools may improve shared decision-making (SDM) processes. We aimed to assess the effectiveness of implementing an online SDM process with topics and content developed through a participatory design approach. We analyzed the triggered and completed SDM cases with responses from participants at a medical center in Taiwan. Data were retrieved from the Research Electronic Data Capture (REDCap) database of the hospital for analysis. Each team developed web-based patient decision aids (PDA) with empirical evidence in a multi-digitized manner, allowing patients to scan QR codes on a leaflet using their mobile phones and then read the PDA content online. From July 2019 to December 2020, 48 web-based SDM topics were implemented in the 24 clinical departments of this hospital. The results showed that using the REDCap system improved SDM efficiency and quality. Implementing an online SDM process integrated with face-to-face communication enhanced the practice and effectiveness of SDM, possibly through the flexibility of accessing information, self-assessment, and feedback evaluation.

A Precision Health Service for Chronic Diseases: Development and Cohort Study Using Wearable Device, Machine Learning, and Deep Learning.

This paper presents an integrated and scalable precision health service for health promotion and chronic disease prevention. Continuous real-time monitoring of lifestyle and environmental factors is implemented by integrating wearable devices, open environmental data, indoor air quality sensing devices, a location-based smartphone app, and an AI-assisted telecare platform. The AI-assisted telecare platform provided comprehensive insight into patients' clinical, lifestyle, and environmental data, and generated reliable predictions of future acute exacerbation events. All data from 1,667 patients were collected prospectively during a 24-month follow-up period, resulting in the detection of 386 abnormal episodes. Machine learning algorithms and deep learning algorithms were used to train modular chronic disease models. The modular chronic disease prediction models that have passed external validation include obesity, panic disorder, and chronic obstructive pulmonary disease, with an average accuracy of 88.46%, a sensitivity of 75.6%, a specificity of 93.0%, and an F1 score of 79.8%. Compared with previous studies, we establish an effective way to collect lifestyle, life trajectory, and symptom records, as well as environmental factors, and improve the performance of the prediction model by adding objective comprehensive data and feature selection. Our results also demonstrate that lifestyle and environmental factors are highly correlated with patient health and have the potential to predict future abnormal events better than using only questionnaire data. Furthermore, we have constructed a cost-effective model that needs only a few features to support the prediction task, which is helpful for deploying real-world modular prediction models.

Effect of implementing decision support to activate a rapid response system by automated screening of verified vital sign data: A retrospective database study.

AIM: Activating a rapid response system (RRS) at general wards requires memorizing trigger criteria, identifying deterioration, and timely notification of abnormalities. We aimed to assess the effect of decision support (DS)-linked RRS activation on management and outcomes. METHODS: We retrospectively analyzed general ward RRS activation cases from 2013 to 2017 and the incidence of cardiopulmonary resuscitations (CPR) from 2013 to 2020. A DS-alerting mechanism was added to the conventional RRS activation process in 2017, with an alert window appearing whenever the system automatically detected any verified abnormal vital sign entry, alerting the nurse to take further action. Logistic and linear regression analyses were used to compare outcomes. RESULTS: We analyzed 27,747 activations and 64,592 DS alerts. RRS activations increased from 3.5 to 30.3 per 1,000 patient-days (P < 0.001) after DS implementation. The first DS activations occurred earlier than conventional ones (-2.9 days, 95% confidence interval = -3.6 to -2.1 days). After adjustment with inverse probability of treatment weighting, main (conventional vs DS-linked activations after implementation) and sensitivity analyses showed that DS activation cases had a lower risk of CPR and in-hospital mortality. Cases with more DS alerts before RRS activation had a higher risk of CPR (P trend = 0.017) and in-hospital mortality (P trend < 0.001). The incidence of CPR at the general ward decreased. CONCLUSION: Implementing a DS mechanism with an automated screening of verified abnormal vital signs linked to RRS activations at general wards was associated with improved practice and timeliness of hospital-wide RRS activations and reduced in-hospital resuscitations and mortality.

Training a Deep Contextualized Language Model for International Classification of Diseases, 10th Revision Classification via Federated Learning: Model Development and Validation Study.

BACKGROUND: The automatic coding of clinical text documents by using the International Classification of Diseases, 10th Revision (ICD-10) can be performed for statistical analyses and reimbursements. With the development of natural language processing models, new transformer architectures with attention mechanisms have outperformed previous models. Although multicenter training may increase a model's performance and external validity, the privacy of clinical documents should be protected. We used federated learning to train a model with multicenter data, without sharing data per se. OBJECTIVE: This study aims to train a classification model via federated learning for ICD-10 multilabel classification. METHODS: Text data from discharge notes in electronic medical records were collected from the following three medical centers: Far Eastern Memorial Hospital, National Taiwan University Hospital, and Taipei Veterans General Hospital. After comparing the performance of different variants of bidirectional encoder representations from transformers (BERT), PubMedBERT was chosen for the word embeddings. With regard to preprocessing, the nonalphanumeric characters were retained because the model's performance decreased after the removal of these characters. To explain the outputs of our model, we added a label attention mechanism to the model architecture. The model was trained with data from each of the three hospitals separately and via federated learning. The models trained via federated learning and the models trained with local data were compared on a testing set that was composed of data from the three hospitals. The micro F1 score was used to evaluate model performance across all 3 centers. RESULTS: The F1 scores of PubMedBERT, RoBERTa (Robustly Optimized BERT Pretraining Approach), ClinicalBERT, and BioBERT (BERT for Biomedical Text Mining) were 0.735, 0.692, 0.711, and 0.721, respectively. The F1 score of the model that retained nonalphanumeric characters was 0.8120, whereas the F1 score after removing these characters was 0.7875-a decrease of 0.0245 (3.11%). The F1 scores on the testing set were 0.6142, 0.4472, 0.5353, and 0.2522 for the federated learning, Far Eastern Memorial Hospital, National Taiwan University Hospital, and Taipei Veterans General Hospital models, respectively. The explainable predictions were displayed with highlighted input words via the label attention architecture. CONCLUSIONS: Federated learning was used to train the ICD-10 classification model on multicenter clinical text while protecting data privacy. The model's performance was better than that of models that were trained locally.

Estimation of quality-adjusted life expectancy in patients under prolonged mechanical ventilation.

OBJECTIVES: The purpose of this study was to estimate the quality-adjusted life expectancy (QALE) and the expected lifetime utility loss of patients with prolonged mechanical ventilation (PMV). METHODS: PMV was defined as more than 21 days of mechanical ventilation. A total of 633 patients fulfilled this definition and were followed for 9 years (1998-2007) to obtain their survival status. Quality of life of 142 patients was measured with the EuroQol five-dimensional (EQ-5D) questionnaire during the period 2008 to 2009. The survival probabilities for each time point were adjusted with a utility measurement of quality of life and then extrapolated to 300 months to obtain the QALE. We compared the age-, gender-matched reference populations to calculate the expected lifetime utility loss. RESULTS: The average age of subjects was 76 years old. The life expectancy and loss of life expectancy were 1.95 years and 8.48 years, respectively. The QALE of 55 patients with partial cognitive ability and the ability to respond was 0.58 quality-adjusted life years (QALY), whereas the QALEs of 87 patients with poor consciousness were 0.28 and 0.29 QALY for the EQ-5D measured by family caregivers and nurses, respectively. The loss of QALE for PMV patients was 9.87 to 10.17 QALY, corresponding to a health gap of 94% to 97%. CONCLUSIONS: Theses results of poor prognosis would provide stakeholders evidence for communication to facilitate clinical decisions. The estimation may be used in future studies to facilitate the cost-effectiveness and reduction of the health gap.

Life expectancies and incidence rates of patients under prolonged mechanical ventilation: a population-based study during 1998 to 2007 in Taiwan.

INTRODUCTION: The present study examined the median survival, life expectancies, and cumulative incidence rate (CIR) of patients undergoing prolonged mechanical ventilation (PMV) stratified by different underlying diseases. METHODS: According to the National Health Insurance Research Database of Taiwan, there were 8,906,406 individuals who obtained respiratory care during the period from 1997 to 2007. A random sample of this population was performed, and subjects who had continuously undergone mechanical ventilation for longer than 21 days were enrolled in the current study. Annual incidence rates and the CIR were calculated. After stratifying the patients according to their specific diagnoses, latent class analysis was performed to categorise PMV patients with multiple co-morbidities into several groups. The life expectancies of different groups were estimated using a semiparametric method with a hazard function based on the vital statistics of Taiwan. RESULTS: The analysis of 50,481 PMV patients revealed that incidence rates increased as patients grew older and that the CIR (17 to 85 years old) increased from 0.103 in 1998 to 0.183 in 2004 before stabilising thereafter. The life expectancies of PMV patients suffering from degenerative neurological diseases, stroke, or injuries tended to be longer than those with chronic renal failure or cancer. Patients with chronic obstructive pulmonary disease survived longer than did those co-morbid with other underlying diseases, especially septicaemia/shock. CONCLUSIONS: PMV provides a direct means to treat respiratory tract diseases and to sustain respiration in individuals suffering from degenerative neurological diseases, and individuals with either of these types of conditions respond better to PMV than do those with other co-morbidities. Future research is required to determine the cost-effectiveness of this treatment paradigm.

Factors Influencing the Use of Health Information Exchange by Physicians-Using the National Health Insurance PharmaCloud System in Taiwan.

This study aimed to investigate the factors influencing physicians use of the PharmaCloud system in Taiwan through Technology Continuance Theory (TCT) and to construct a TCT-based structured questionnaire to demonstrate the attitude and behavior of physicians in the Taiwanese medical system. It focused on investigating "confirmation", "perceived usefulness", "perceived ease of use", "attitude", "satisfaction", and "continuance intention" towards the preload-based comparison and manual search in PharmaCloud by attending physicians during their outpatient clinics. Path analysis was used to analyze the cause and effect relationship between variables. This study collected 528 valid questionnaires and the results of path analysis found that factors affecting physicians' continued use of preload-based comparison in PharmaCloud included "perceived usefulness", "satisfaction", and "attitude" (all p < 0.001); however, factors that influenced physicians' continued use of manual search in PharmaCloud were only "satisfaction" and "attitude" (all p < 0.001). Additionally, the effects of "perceived usefulness" and "perceived ease of use" on "satisfaction" could only be seen in preload-based comparison in PharmaCloud. In conclusion, when physicians' actual use of PharmaCloud met their expectations, physicians had higher levels of confirmation and better perceived usefulness, which naturally increased their satisfaction and attitude towards PharmaCloud and positively prompted them to continue using it.

Automatic ICD-10 Coding and Training System: Deep Neural Network Based on Supervised Learning.

BACKGROUND: The International Classification of Diseases (ICD) code is widely used as the reference in medical system and billing purposes. However, classifying diseases into ICD codes still mainly relies on humans reading a large amount of written material as the basis for coding. Coding is both laborious and time-consuming. Since the conversion of ICD-9 to ICD-10, the coding task became much more complicated, and deep learning- and natural language processing-related approaches have been studied to assist disease coders. OBJECTIVE: This paper aims at constructing a deep learning model for ICD-10 coding, where the model is meant to automatically determine the corresponding diagnosis and procedure codes based solely on free-text medical notes to improve accuracy and reduce human effort. METHODS: We used diagnosis records of the National Taiwan University Hospital as resources and apply natural language processing techniques, including global vectors, word to vectors, embeddings from language models, bidirectional encoder representations from transformers, and single head attention recurrent neural network, on the deep neural network architecture to implement ICD-10 auto-coding. Besides, we introduced the attention mechanism into the classification model to extract the keywords from diagnoses and visualize the coding reference for training freshmen in ICD-10. Sixty discharge notes were randomly selected to examine the change in the F1-score and the coding time by coders before and after using our model. RESULTS: In experiments on the medical data set of National Taiwan University Hospital, our prediction results revealed F1-scores of 0.715 and 0.618 for the ICD-10 Clinical Modification code and Procedure Coding System code, respectively, with a bidirectional encoder representations from transformers embedding approach in the Gated Recurrent Unit classification model. The well-trained models were applied on the ICD-10 web service for coding and training to ICD-10 users. With this service, coders can code with the F1-score significantly increased from a median of 0.832 to 0.922 (P<.05), but not in a reduced interval. CONCLUSIONS: The proposed model significantly improved the F1-score but did not decrease the time consumed in coding by disease coders.

Acute Exacerbation of a Chronic Obstructive Pulmonary Disease Prediction System Using Wearable Device Data, Machine Learning, and Deep Learning: Development and Cohort Study.

BACKGROUND: The World Health Organization has projected that by 2030, chronic obstructive pulmonary disease (COPD) will be the third-leading cause of mortality and the seventh-leading cause of morbidity worldwide. Acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are associated with an accelerated decline in lung function, diminished quality of life, and higher mortality. Accurate early detection of acute exacerbations will enable early management and reduce mortality. OBJECTIVE: The aim of this study was to develop a prediction system using lifestyle data, environmental factors, and patient symptoms for the early detection of AECOPD in the upcoming 7 days. METHODS: This prospective study was performed at National Taiwan University Hospital. Patients with COPD that did not have a pacemaker and were not pregnant were invited for enrollment. Data on lifestyle, temperature, humidity, and fine particulate matter were collected using wearable devices (Fitbit Versa), a home air quality-sensing device (EDIMAX Airbox), and a smartphone app. AECOPD episodes were evaluated via standardized questionnaires. With these input features, we evaluated the prediction performance of machine learning models, including random forest, decision trees, k-nearest neighbor, linear discriminant analysis, and adaptive boosting, and a deep neural network model. RESULTS: The continuous real-time monitoring of lifestyle and indoor environment factors was implemented by integrating home air quality-sensing devices, a smartphone app, and wearable devices. All data from 67 COPD patients were collected prospectively during a mean 4-month follow-up period, resulting in the detection of 25 AECOPD episodes. For 7-day AECOPD prediction, the proposed AECOPD predictive model achieved an accuracy of 92.1%, sensitivity of 94%, and specificity of 90.4%. Receiver operating characteristic curve analysis showed that the area under the curve of the model in predicting AECOPD was greater than 0.9. The most important variables in the model were daily steps walked, stairs climbed, and daily distance moved. CONCLUSIONS: Using wearable devices, home air quality-sensing devices, a smartphone app, and supervised prediction algorithms, we achieved excellent power to predict whether a patient would experience AECOPD within the upcoming 7 days. The AECOPD prediction system provided an effective way to collect lifestyle and environmental data, and yielded reliable predictions of future AECOPD events. Compared with previous studies, we have comprehensively improved the performance of the AECOPD prediction model by adding objective lifestyle and environmental data. This model could yield more accurate prediction results for COPD patients than using only questionnaire data.

CD28 engagement inhibits CD73-mediated regulatory activity of CD8+ T cells.

CD28 is required for T cell activation as well as the generation of CD4+Foxp3+ Treg. It is unclear, however, how CD28 costimulation affects the development of CD8+ T cell suppressive function. Here, by use of Hepa1.6.gp33 in vitro killing assay and B16.gp33 tumor mouse model we demonstrate that CD28 engagement during TCR ligation prevents CD8+ T cells from becoming suppressive. Interestingly, our results showed that ectonucleotidase CD73 expression on CD8+ T cells is upregulated in the absence of CD28 costimulation. In both murine and human tumor-bearing hosts, CD73 is upregulated on CD28-CD8+ T cells that infiltrate the solid tumor. UPLC-MS/MS analysis revealed that CD8+ T cells activation without CD28 costimulation produces elevated levels of adenosine and that CD73 mediates its production. Adenosine receptor antagonists block CD73-mediated suppression. Our data support the notion that CD28 costimulation inhibits CD73 upregulation and thereby prevents CD8+ T cells from becoming suppressive. This study uncovers a previously unidentified role for CD28 costimulation in CD8+ T cell activation and suggests that the CD28 costimulatory pathway can be a potential target for cancer immunotherapy.

Measurement of quality of life using EQ-5D in patients on prolonged mechanical ventilation: comparison of patients, family caregivers, and nurses.

PURPOSE: This study reports how QOL (quality of life) assessments differ between patients on prolonged mechanical ventilation (PMV) and their proxies (family caregivers and nurses). METHODS: We enrolled consecutive subjects on PMV for more than 21 days from five institutions. We conducted QOL assessments using the Taiwanese version of the EQ-5D in face-to-face interviews. Direct caregivers (family members and nurses) also completed the EQ-5D from the patient's point of view. RESULTS: For 55 of the 142 enrolled patients who were able to assess their QOL, we recruited 44 patient-family caregiver pairs, 53 patient-nurse pairs, and 42 family caregiver-nurse pairs. There were 81 family caregiver-nurse pairs out of 87 patients with poor cognition. The agreement between patient-family caregiver pairs was generally higher than that of patient-nurse pairs. As the proportions of exact agreement between family caregivers and nurses for patients with poor cognition were 98-99% for observable dimensions of mobility, self-care, and usual activities, they lead to a minimal difference in the final values. CONCLUSIONS: QOL assessments from family caregivers agreed more closely with patients than did those from nurses using EQ-5D evaluations for patients with clear cognition, but either proxy was acceptable for rating PMV patients with poor cognition.