Search

Abstract BackgroundElectronic medical records store extensive patient data and serve as a comprehensive repository, including textual medical records like surgical and imaging reports. Their utility in clinical decision support systems is substantial, but the widespread use of ambiguous and unstandardized abbreviations in clinical documents poses challenges for natural language processing in clinical decision support systems. Efficient abbreviation disambiguation methods are needed for effective information extraction. ObjectiveThis study aims to enhance the one-to-all (OTA) framework for clinical abbreviation expansion, which uses a single model to predict multiple abbreviation meanings. The objective is to improve OTA by developing context-candidate pairs and optimizing word embeddings in Bidirectional Encoder Representations From Transformers (BERT), evaluating the model’s efficacy in expanding clinical abbreviations using real data. MethodsThree datasets were used: Medical Subject Headings Word Sense Disambiguation, University of Minnesota, and Chia-Yi Christian Hospital from Ditmanson Medical Foundation Chia-Yi Christian Hospital. Texts containing polysemous abbreviations were preprocessed and formatted for BERT. The study involved fine-tuning pretrained models, ClinicalBERT and BlueBERT, generating dataset pairs for training and testing based on Huang et al’s method. ResultsBlueBERT achieved macro- and microaccuracies of 95.41% and 95.16%, respectively, on the Medical Subject Headings Word Sense Disambiguation dataset. It improved macroaccuracy by 0.54%‐1.53% compared to two baselines, long short-term memory and deepBioWSD with random embedding. On the University of Minnesota dataset, BlueBERT recorded macro- and microaccuracies of 98.40% and 98.22%, respectively. Against the baselines of Word2Vec + support vector machine and BioWordVec + support vector machine, BlueBERT demonstrated a macroaccuracy improvement of 2.61%‐4.13%. ConclusionsThis research preliminarily validated the effectiveness of the OTA method for abbreviation disambiguation in medical texts, demonstrating the potential to enhance both clinical staff efficiency and research effectiveness.

Studies show that statins users are at reduced risk of fracture and improved bone mineral density. However, the clinical effectiveness of statin use in patients with gout has not been investigated. This retrospective cohort study used data from Taiwan's National Health Insurance Research Database, consisting of 3443 patients with gout using statins aged 50 years and above and 6886 gout patients of non-statin users matched by sex, age and propensity score. The Cox proportional hazards regression analysis showed that statin use was associated with a reduced risk of hip fracture (adjusted hazard ratio [aHR] = 0.78, 95 % confidence interval [CI] = 0.64–0.94) after controlling for potential confounding factors. The association was significant in both genders aged 50–64 years, with aHRs of near 0.35, but not in the elderly. In addition, women aged 50–64 years who used statins also exhibited a lower risk of vertebral fracture (aHR = 0.70, 95 % CI = 0.50–0.99), but not men. In conclusion, the stating use in gout patients could reduce fracture risk for younger patients. Further research is warranted to confirm these findings.

Parkinson’s disease has been extensively studied using artificial intelligence for severity assessment. Previous research has achieved excellent results using wearable solutions. However, this approach has significant cost disadvantages, especially in the post-COVID-19 era. Recently, researchers are investigating vision-based alternatives for severity assessment. Unfortunately, the outcomes of these attempts often fall short when compared to wearable solutions. In this paper, we propose a novel feature extraction method suitable for temporal models, called ‘Multitemporal Feature Average Pooling.’ This method effectively captures crucial features from temporal data. Additionally, we employ interpretable artificial intelligence techniques to analyze the models, demonstrating their importance in severity assessment. We trained four severity classification models based on MDS-UPDRS. In the multi-class task of finger tapping, we achieved accuracies of 77% for the left hand and 85% for the right hand. These results represent an improvement of 4.6% and 12.6% compared to Shao et al.’s Three-Stream CNN model. In the binary classification, both left- and right-hand tasks achieved 92% accuracy, showing an increase of 11.4% and 14% compared to Chang et al. Furthermore, we achieved accuracies ranging from 60% to 80% for multi-class tasks involving Hand Movements, Rapid Alternating Movements of Hands, and Leg Agility. In binary classification, these tasks reached an even higher accuracy of 80% to 90%. This research is crucial for advancing vision-based severity assessment in clinical settings, providing more accurate evaluations for Parkinson’s disease patients. Through the application of artificial intelligence technology, we anticipate achieving substantial breakthroughs and advancements in clinical practice.

BackgroundHyperglycemia affects the outcomes of endovascular therapy (EVT) for acute ischemic stroke (AIS). This study compares the predictive ability of diabetes status and glucose measures on EVT outcomes using nationwide registry data.MethodsThe study included 1,097 AIS patients who underwent EVT from the Taiwan Registry of Endovascular Thrombectomy for Acute Ischemic Stroke. The variables analyzed included diabetes status, admission glucose, glycated hemoglobin (HbA1c), admission glucose-to-HbA1c ratio (GAR), and outcomes such as 90-day poor functional outcome (modified Rankin Scale score ≥ 2) and symptomatic intracranial hemorrhage (SICH). Multivariable analyses investigated the independent effects of diabetes status and glucose measures on outcomes. A receiver operating characteristic (ROC) analysis was performed to compare their predictive abilities.ResultsThe multivariable analysis showed that individuals with known diabetes had a higher likelihood of poor functional outcomes (odds ratios [ORs] 2.10 to 2.58) and SICH (ORs 3.28 to 4.30) compared to those without diabetes. Higher quartiles of admission glucose and GAR were associated with poor functional outcomes and SICH. Higher quartiles of HbA1c were significantly associated with poor functional outcomes. However, patients in the second HbA1c quartile (5.6–5.8%) showed a non-significant tendency toward good functional outcomes compared to those in the lowest quartile (

Background: Acute dizziness is a common symptom in the emergency department (ED), with strokes accounting for 3 %–5 % of cases. We investigated the risk of stroke in ED patients with acute dizziness and compared stroke characteristics diagnosed during and after the ED visit. Methods: We identified adult patients with acute dizziness, vertigo, or imbalance using a hospital research-based database. Patients with abdominal or flank pain were used as the comparison group. Patients with dizziness were 1:1 matched to comparison patients. Each patient was traced for up to one year until being hospitalized for a stroke. Results: Out of the 24,266 eligible patients, 589 (2.4 %) were hospitalized for stroke during the ED visit. For the remaining 23,677 patients, the risk of stroke at 7, 30, 90, and 365 days after ED discharge was 0.40 %, 0.52 %, 0.71 %, and 1.25 % respectively. Patients with dizziness had a higher risk of stroke compared to the comparison group at 7, 30, 90, and 365 days. The risk ratios decreased from 5.69 (95 % confidence interval [CI], 3.34–9.68) to 2.03 (95 % CI, 1.65–2.49). Compared to patients hospitalized for stroke during the ED visit, those hospitalized for stroke after the ED visit had greater stroke severity despite a lower initial triage acuity. Patients with early stroke (≤7 days) after ED discharge were less likely to have hypertension, diabetes, hyperlipidemia, and atrial fibrillation. They mostly experienced posterior circulation stroke. Patients with late stroke (>7 days) were older and less likely to have hypertension and hyperlipidemia but more likely to have a history of prior stroke and ischemic heart disease. Their strokes were mainly located in the anterior circulation territory. Conclusions: The risk of stroke after ED discharge was higher in patients with dizziness than in the comparison group, with gradually decreasing risk ratios in the following year. Patients hospitalized for stroke during and after the ED visit had different profiles of vascular risk factors and clinical characteristics.

Background Endovascular thrombectomy (EVT) is the standard therapy for patients with acute ischemic stroke secondary to large‐artery occlusion. In January 2019, the Taiwan Stroke Society established a nationwide TREAT‐AIS (Taiwan Registry of Endovascular Thrombectomy for Acute Ischemic Stroke). Here, we provide the study design, current progress, and baseline data of TREAT‐AIS. Methods TREAT‐AIS is a multicenter prospective registration program in Taiwan. Patients aged ≥20 years who underwent EVT for acute ischemic stroke were recruited. The key items on the registration form were divided into general stroke demographics and EVT‐related sections. The main outcome of effectiveness was functional independence (modified Rankin Scale score, 0–2) at 3 months. The influence of sex on post‐EVT outcomes was also analyzed in the presented study. Results By the end of June 2022, there were 10 medical centers and 9 community hospitals participating in the TREAT‐AIS and a total of 1522 patients (mean±SD age, 71.2±13.6 years; men, 55.6%) being enrolled. The median National Institutes of Health Stroke Scale score on admission was 18 (interquartile range, 12–23). The major cause of stroke was cardioembolism (43.6%), followed by large‐artery atherosclerosis (36.8%) and an undetermined cause (15.4%). Functional independence at 3 months poststroke was achieved in 36.2% of the patients. Male patients were more likely to have functional independence at 3 months compared with female patients (40.4% versus 30.8%; P

BackgroundStroke survivors are prone to osteoporosis and fractures. However, bone mineral density (BMD) testing and osteoporosis treatment were underutilized in patients with recent stroke. We aimed to examine whether stroke has an impact on the utilization of BMD testing and osteoporosis treatment as well as the determinants of their utilization in stroke patients using nationwide population-based data in Taiwan.MethodsWe identified patients aged 55 years and older who were hospitalized for hemorrhagic or ischemic stroke as the stroke cohort, and age- and sex-matched patients hospitalized for reasons other than stroke, fracture, or fall as the non-stroke cohort. We used the Fine-Gray sub-distribution hazard competing risk regression model to determine the predictors for BMD testing and osteoporosis treatment.ResultsA total of 32997 stroke patients and 32997 age- and sex-matched controls comprised the stroke and non-stroke cohorts, respectively. BMD testing and osteoporosis treatment were performed in 1.0% and 5.2% of the stroke patients, respectively, within one year after hospitalization while these measures were performed in 0.8% and 4.7% of the controls. Stroke patients were more likely to receive BMD testing (adjusted hazard ratio [HR] 1.33; 95% confidence interval [CI] 1.11–1.58) and osteoporosis treatment (adjusted HR 1.19; 95% CI 1.11–1.29). Female sex, osteoporosis, prior BMD testing, and low-trauma fractures after stroke increased the likelihood of using BMD testing and osteoporosis treatment whereas greater stroke severity reduced the likelihood of receiving both measures.ConclusionsBoth BMD testing and osteoporosis treatment were underutilized among stroke survivors even though they had a higher chance of receiving both measures than non-stroke patients.

BackgroundIdentifying patients at high risk of stroke-associated pneumonia (SAP) may permit targeting potential interventions to reduce its incidence. We aimed to explore the functionality of machine learning (ML) and natural language processing techniques on structured data and unstructured clinical text to predict SAP by comparing it to conventional risk scores.MethodsLinked data between a hospital stroke registry and a deidentified research-based database including electronic health records and administrative claims data was used. Natural language processing was applied to extract textual features from clinical notes. The random forest algorithm was used to build ML models. The predictive performance of ML models was compared with the A2DS2, ISAN, PNA, and ACDD4 scores using the area under the receiver operating characteristic curve (AUC).ResultsAmong 5,913 acute stroke patients hospitalized between Oct 2010 and Sep 2021, 450 (7.6%) developed SAP within the first 7 days after stroke onset. The ML model based on both textual features and structured variables had the highest AUC [0.840, 95% confidence interval (CI) 0.806–0.875], significantly higher than those of the ML model based on structured variables alone (0.828, 95% CI 0.793–0.863, P = 0.040), ACDD4 (0.807, 95% CI 0.766–0.849, P = 0.041), A2DS2 (0.803, 95% CI 0.762–0.845, P = 0.013), ISAN (0.795, 95% CI 0.752–0.837, P = 0.009), and PNA (0.778, 95% CI 0.735–0.822, P < 0.001). All models demonstrated adequate calibration except for the A2DS2 score.ConclusionsThe ML model based on both textural features and structured variables performed better than conventional risk scores in predicting SAP. The workflow used to generate ML prediction models can be disseminated for local adaptation by individual healthcare organizations.

BackgroundTimely detection of atrial fibrillation (AF) after stroke is highly clinically relevant, aiding decisions on the optimal strategies for secondary prevention of stroke. In the context of limited medical resources, it is crucial to set the right priorities of extended heart rhythm monitoring by stratifying patients into different risk groups likely to have newly detected AF (NDAF). This study aimed to develop an electronic health record (EHR)-based machine learning model to assess the risk of NDAF in an early stage after stroke.MethodsLinked data between a hospital stroke registry and a deidentified research-based database including EHRs and administrative claims data was used. Demographic features, physiological measurements, routine laboratory results, and clinical free text were extracted from EHRs. The extreme gradient boosting algorithm was used to build the prediction model. The prediction performance was evaluated by the C-index and was compared to that of the AS5F and CHASE-LESS scores.ResultsThe study population consisted of a training set of 4,064 and a temporal test set of 1,492 patients. During a median follow-up of 10.2 months, the incidence rate of NDAF was 87.0 per 1,000 person-year in the test set. On the test set, the model based on both structured and unstructured data achieved a C-index of 0.840, which was significantly higher than those of the AS5F (0.779, p = 0.023) and CHASE-LESS (0.768, p = 0.005) scores.ConclusionsIt is feasible to build a machine learning model to assess the risk of NDAF based on EHR data available at the time of hospital admission. Inclusion of information derived from clinical free text can significantly improve the model performance and may outperform risk scores developed using traditional statistical methods. Further studies are needed to assess the clinical usefulness of the prediction model.

BackgroundPoststroke atrial fibrillation (AF) screening aids decisions regarding the optimal secondary prevention strategies in patients with acute ischemic stroke (AIS). We used an electronic medical record (EMR) algorithm to identify AF in a cohort of AIS patients, which were used to validate eight risk scores for predicting AF detected after stroke (AFDAS).MethodsWe used linked data between a hospital stroke registry and a deidentified database including EMRs and administrative claims data. EMR algorithms were constructed to identify AF using diagnostic and medication codes as well as free clinical text. Based on the optimal EMR algorithm, the incidence rate of AFDAS was estimated. The predictive performance of 8 risk scores including AS5F, C2HEST, CHADS2, CHA2DS2-VASc, CHASE-LESS, HATCH, HAVOC, and Re-CHARGE-AF scores, were compared using the C-index, net reclassification improvement, integrated discrimination improvement, calibration curve, and decision curve analysis.ResultsThe algorithm that defines AF as any positive mention of AF-related keywords in electrocardiography or echocardiography reports, or presence of diagnostic codes of AF was used to identify AF. Among the 5,412 AIS patients without known AF at stroke admission, the incidence rate of AFDAS was 84.5 per 1,000 person-year. The CHASE-LESS and AS5F scores were well calibrated and showed comparable C-indices (0.741 versus 0.730, p = 0.223), which were significantly higher than the other risk scores.ConclusionThe CHASE-LESS and AS5F scores demonstrated adequate discrimination and calibration for predicting AFDAS. Both simple risk scores may help select patients for intensive AF monitoring.

BackgroundAmong poststroke morbidities, poststroke epilepsy (PSE) has been identified as a significant clinical issue. Although middle cerebral artery (MCA) infarct is the most common type of stroke among all vascular territories, very few studies specifically focused on the risk factors leading to PSE in patients with MCA infarct.MethodsA population study in Taiwan has been conducted, linking the National Health Insurance Research Database and Hospital Stroke Registry, from 2001 to 2015 and 2006 to 2010, respectively. Patients were divided into MCA and non-MCA groups, and the diagnosis of incident epilepsy between the groups has been compared. The multivariable Cox proportional hazard model was used to identify the risk factors for developing PSE. The distribution of time to PSE was estimated using the Kaplan–Meier method.ResultsIn total, 1,838 patients were recruited, with 774 and 1,064 in the MCA and non-MCA groups, respectively. PSE incidence in the MCA group was 15.5% vs. 6.2% in the non-MCA group, with a hazard ratio of (95% CI) 2.06 (1.33–3.19). Factors significantly associated with PSE included atrial fibrillation, depression, National Institutes of Health Stroke Scale (NIHSS) scores of ≥ 16, and alert on arrival. For patients with MCA infarct, higher NIHSS and Glasgow coma scale scores, the presence of visual field defects and weakness, urination control impairment, and complications during hospitalization were associated with a higher risk for PSE development.ConclusionsThis study established the conditions leading to a higher risk of PSE and identified the important clinical risk factors in patients experiencing MCA infarct. Efforts to manage these risk factors may be important in preventing PSE in patients with MCA infarct.

BackgroundSeveral prognostic scores have been proposed to predict functional outcomes after an acute ischemic stroke (AIS). Most of these scores are based on structured information and have been used to develop prediction models via the logistic regression method. With the increased use of electronic health records and the progress in computational power, data-driven predictive modeling by using machine learning techniques is gaining popularity in clinical decision-making. ObjectiveWe aimed to investigate whether machine learning models created by using unstructured text could improve the prediction of functional outcomes at an early stage after AIS. MethodsWe identified all consecutive patients who were hospitalized for the first time for AIS from October 2007 to December 2019 by using a hospital stroke registry. The study population was randomly split into a training (n=2885) and test set (n=962). Free text in histories of present illness and computed tomography reports was transformed into input variables via natural language processing. Models were trained by using the extreme gradient boosting technique to predict a poor functional outcome at 90 days poststroke. Model performance on the test set was evaluated by using the area under the receiver operating characteristic curve (AUC). ResultsThe AUCs of text-only models ranged from 0.768 to 0.807 and were comparable to that of the model using National Institutes of Health Stroke Scale (NIHSS) scores (0.811). Models using both patient age and text achieved AUCs of 0.823 and 0.825, which were similar to those of the model containing age and NIHSS scores (0.841); the model containing preadmission comorbidities, level of consciousness, age, and neurological deficit (PLAN) scores (0.837); and the model containing Acute Stroke Registry and Analysis of Lausanne (ASTRAL) scores (0.840). Adding variables from clinical text improved the predictive performance of the model containing age and NIHSS scores, the model containing PLAN scores, and the model containing ASTRAL scores (the AUC increased from 0.841 to 0.861, from 0.837 to 0.856, and from 0.840 to 0.860, respectively). ConclusionsUnstructured clinical text can be used to improve the performance of existing models for predicting poststroke functional outcomes. However, considering the different terminologies that are used across health systems, each individual health system may consider using the proposed methods to develop and validate its own models.

BackgroundThe efficacy and safety of intravenous alteplase administered 3–4.5 h after acute ischemic stroke have been demonstrated. However, whether responses differ between low-dose and standard-dose alteplase during this time window and whether certain subgroups benefit more remain unknown.Patients and MethodsThe current analysis was based on a multicenter matched-cohort study conducted in Taiwan. The treatment group comprised 378 patients receiving intravenous alteplase 3–4.5 h after stroke onset, and the control group comprised 378 age- and sex-matched patients who did not receive alteplase treatment during the same period. Standard- and low-dose alteplase was administered to patients at the physician's discretion.ResultsOverall, patients receiving alteplase exhibited more favorable outcomes than did controls [34.0 vs. 22.7%; odds ratio (OR): 1.75, 95% confidence interval (CI): 1.27–1.42], and the effectiveness was consistent in all subgroups. Although patients in the standard-dose group (n = 182) were younger than those in the low-dose (n = 192) group, the proportions of patients with favorable outcomes (36.3 vs. 31.8%; OR: 1.22, 95% CI: 0.80–1.88) and symptomatic hemorrhage (2.8 vs 4.2%; OR: 0.65, 95% CI: 0.21–2.02) were consistently comparable in a covariate-adjusted model and an age-matched cohort. In the subgroup analysis, patients with cardioembolism, atrial fibrillation, and hypercholesterolemia were more likely to achieve favorable outcomes after receiving standard-dose than low-dose alteplase.ConclusionIn the 3–4.5 h time window, the effectiveness and safety of standard-dose and low-dose alteplase may be comparable. A standard dose may be selected for patients with cardioembolism, atrial fibrillation, or hypercholesterolemia.

Objective: Lower serum low-density lipoprotein cholesterol (LDL-C) levels are associated with increased intracerebral hemorrhage (ICH) risk. However, reverse causality and residual confounding has not attracted public attention. Therefore, we assessed whether people with LDL-C have increased risk of mortality adjusting for potential confounders using two large Taiwan cohorts.Methods: The Mei-Jhao (MJ) cohort has 414,372 adults participating in a medical screening program with 378 ICH deaths within 15 years of follow-up (1994–2008). Cox proportional hazards regressions estimated hazard death ratios according to LDL-C levels. We identified 4,606 ICH patients from the Taiwan Stroke Registry (TSR) and analyzed the impact of LDL-C on 3-month mortality.Results: Low cholesterol (LDL-C

Background Conventional prognostic scores usually require predefined clinical variables to predict outcome. The advancement of natural language processing has made it feasible to derive meaning from unstructured data. We aimed to test whether using unstructured text in electronic health records can improve the prediction of functional outcome after acute ischemic stroke. Methods and Results Patients hospitalized for acute ischemic stroke were identified from 2 hospital stroke registries (3847 and 2668 patients, respectively). Prediction models developed using the first cohort were externally validated using the second cohort, and vice versa. Free text in the history of present illness and computed tomography reports was used to build machine learning models using natural language processing to predict poor functional outcome at 90 days poststroke. Four conventional prognostic models were used as baseline models. The area under the receiver operating characteristic curves of the model using history of present illness in the internal and external validation sets were 0.820 and 0.792, respectively, which were comparable to the National Institutes of Health Stroke Scale score (0.811 and 0.807). The model using computed tomography reports achieved area under the receiver operating characteristic curves of 0.758 and 0.658. Adding information from clinical text significantly improved the predictive performance of each baseline model in terms of area under the receiver operating characteristic curves, net reclassification improvement, and integrated discrimination improvement indices (all P

Background: Enhancing detection of undiagnosed atrial fibrillation (AF) in hospitalized patients with a recent ischemic stroke is important because of the treatment implications; especially since presence of paroxysmal AF may not be picked up in a single 12-lead electrocardiogram (ECG) test. While several trials have shown improved detection of AF with prolonged ECG monitoring, this strategy is associated with relatively high cost, labor intensity, and patient inconvenience, thereby making it challenging to routinely implement in all hospitals. Fortunately, conventional 24-h Holter monitoring and repeated 12-lead ECGs are readily available to detect paroxysmal AF in all hospitals, but is unclear which is the better strategy for evaluating undiagnosed AF. The objective of his study is to conduct a randomized trial of serial 12-lead ECGs vs. 24-hour Holter monitoring in the detection of AF in ischemic stroke patients without known AF. Methods and analysis: We plan to enroll 1200 participants from six hospitals in Taiwan. Patients will be eligible for enrollment if they are admitted for an acute ischemic stroke within 2 days, are ≥65 years of age, and have no known AF by history or on baseline ECG at admission. We will randomly assign participants in a 1:1 ratio to undergo daily 12-lead ECG once daily for 5 days (intervention group) or 24-h Holter monitoring (control group). Primary outcome is newly detected AF on a 12-lead ECG or AF lasting ≥30 s on Holter monitoring. Trial registration number: ClinicalTrials.gov Identifier: NCT02578979.

Background and Objectives: Intravenous recombinant tissue plasminogen activator (rt-PA) has been approved for acute ischemic stroke (AIS) within 3 h after onset and the treatment was then extended to 4.5 h. However, the Food and Drug Administration did not approve the indication in the expanded time window. This retrospective, matched cohort study aims to investigate the effectiveness and safety of rt-PA in AIS at 3–4.5 h after onset.Materials and Methods: The treatment group included AIS patients receiving rt-PA at 3–4.5 h after onset, otherwise complying with the regulation, in the stroke registries in 16 hospitals between 2008 and 2017. The control group included age- and sex-matched patients not receiving intravenous thrombolysis from the same registries, excluding those with contraindications. The primary outcome was modified Rankin Scale (mRS) 0–1 at day 90. The safety outcomes were any intracerebral hemorrhage (ICH), early neurological deterioration and 3-month mortality.Results: Each group had 374 patients. There were 34.0% of patients with 3-month mRS 0-1 in the treatment group vs. 22.7% in the control group with an odds ratio of 1.75 (95% confidence intervals, 1.27 to 2.42, P = 0.001). There was no difference in symptomatic ICH, early neurological deterioration and 3-month mortality rates between two groups. The 3-month mRS and symptomatic ICH did not differ significantly in patients receiving standard dose or low dose of rt-PA.Conclusions: Our results support the prescription of rt-PA in AIS patients 3–4.5 h after onset as an effective and tolerable treatment in their functional recovery.

Background/purpose: The in-hospital mortality rates of patients with acute ischemic stroke remain between 3% and 18%. For improving the quality of stroke care, we investigated the factors that contribute to an increased risk of in-hospital mortality in patients with acute ischemic stroke. Methods: A total of 2556 patients with acute ischemic stroke, who were admitted to a stroke unit between January 1, 2007 and December 31, 2011, were included in this study. Factors such as demographic and clinical characteristics, comorbidities, and complications related to in-hospital mortality were assessed. Results: Of the 2556 ischemic stroke patients, 157 received thrombolytic therapy. Eighty of the 2556 patients (3.1%) died during hospitalization, which included 14 (8.9%) patients who received thrombolytic therapy. A history of heart disease (p < 0.01) and stroke severity (p < 0.01) were independent risk factors, whereas herniation, sepsis, and basilar artery occlusion were the most common causes of in-hospital mortality. Approximately 70% of in-hospital mortality was found to be related to stroke severity (total middle cerebral artery occlusion with herniation, basilar artery occlusion, and hemorrhagic transformation). The remaining 30% of in-hospital mortality is attributable to sepsis, heart disease, and other complications. Conclusion: Nearly 30% of in-hospital mortality is associated with preventable factors. The prognosis of acute stroke can be improved by increased focus on reducing serious complications after stroke, particularly on the prevention of infection, heart disease, and increased intracranial pressure.

Readmissions after stroke are not only associated with greater levels of disability and a higher risk of mortality but also increase overall medical costs. Predicting readmission risk and understanding its causes are thus essential for healthcare resource allocation and quality improvement planning. By using machine learning techniques on initial admission data, this study aimed to develop prediction models for readmission or mortality after stroke. During model development, resampling methods were implemented to balance the class distribution. Two-layer nested cross-validation was used to build and evaluate the prediction models. A total of 3422 patients were included for analysis. The 90-day rate of readmission or mortality was 17.6%. This study identified several important predictive factors, including age, prior emergency department visits, pre-stroke functional status, stroke severity, body mass index, consciousness level, and use of a nasogastric tube. The Naïve Bayes model with class weighting to compensate for class imbalance achieved the highest discriminatory capacity in terms of the area under the receiver operating characteristic curve (0.661). Despite having room for improvement, the prediction models could be used for early risk assessment of patients with stroke. Identification of patients at high risk for readmission or mortality immediately after admission has the potential of enabling early discharge planning and transitional care interventions.

Electronic medical records (EMRs) have been used extensively in most medical institutions for more than a decade in Taiwan. However, information overload associated with rapid accumulation of large amounts of clinical narratives has threatened the effective use of EMRs. This situation is further worsened by the use of “copying and pasting”, leading to lots of redundant information in clinical notes. This study aimed to apply natural language processing techniques to address this problem. New information in longitudinal clinical notes was identified based on a bigram language model. The accuracy of automated identification of new information was evaluated using expert annotations as the reference standard. A two-stage cross-over user experiment was conducted to evaluate the impact of highlighting of new information on task demands, task performance, and perceived workload. The automated method identified new information with an F1 score of 0.833. The user experiment found a significant decrease in perceived workload associated with a significantly higher task performance. In conclusion, automated identification of new information in clinical notes is feasible and practical. Highlighting of new information enables healthcare professionals to grasp key information from clinical notes with less perceived workload.

Background and purpose: Severe stenosis in the internal carotid artery may increase the risk of ischemic stroke. The factors that affect the progression of carotid artery stenosis in patients with ischemic stroke are poorly studied. No guidelines for the duration of follow-up of patients with ischemic stroke through carotid ultrasonography exist.Methods: In this retrospective study, 179 patients (108 men; mean age, 68 years) with ischemic stroke and mild to moderate stenosis in the internal carotid artery (ICA) were recruited. Carotid artery ultrasonography was performed over the period of January 2013 to June 2016 with a median follow-up of 36 months (mean 36.5 ± 3.5 months). The severity of carotid artery stenosis was estimated with the following equation: 1– (narrowest ICA diameter/total lumen diameter at the narrowest site). The severity of stenosis was categorized into grades I (0–29%), II (30–49%), III (50–59%), and IV (60–69%). The patient's stenosis grade was defined on the basis of the stenosis rate of the ICA side with most severe stenosis.Results: Stenosis progressed in 17.9% (64/358) of the vessels in 30.7% (55/179) of patients. The risk of stenosis progression increased as the severity of ICA stenosis increased. Patients with stenosis rates of above 50% are at a higher risk of stenosis progression than those with stenosis rate of < 50%. Relative to the patient group with an ICA stenosis rate of 0–29%, the adjusted odds ratios of stenosis progression were 2.33 (p = 0.03; 95% CI: 1.05~5.17), 3.50 (p = 0.09; 95% CI: 0.81~15.84), and 6.61 (p = 0.03; 95% CI: 1.01~39.61) in patient groups with ICA stenosis rates of 30–49%, 50–59%, and 60–69%, respectively. Hyper-LDL-cholesterolemia (Hyper-LDL-c) also increased the risk of stenosis progression, with an adjusted odds ratio of 2.22 (p = 0.03; 95% CI: 1.05~4.71).Conclusion: The rate of ICA stenosis progression increases with stenosis grade. Patients with ICA stenosis severity >50% and Hyper-LDL-c have high rates of stenosis progression. For the patients with stroke and ICA stenosis severity >50%, annual follow up through carotid artery ultrasonography may be necessary.

Stroke registries are observational databases focusing on the clinical information and outcomes of stroke patients. They play an important role in the cycle of quality improvement. Registry data are collected from real-world experiences of stroke care and are suitable for measuring quality of care. By exposing inadequacies in performance measures of stroke care, research from stroke registries has changed how we manage stroke patients in Taiwan. With the success of various quality improvement campaigns, mortality from stroke and recurrence of stroke have decreased in the past decade. After the implementation of a nationwide stroke registry, researchers have been creatively expanding how they use and collect registry data for research. Through the use of the nationwide stroke registry as a common data model, researchers from many hospitals have built their own stroke registries with extended data elements to meet the needs of research. In collaboration with information technology professionals, stroke registry systems have changed from web-based, manual submission systems to automated fill-in systems in some hospitals. Furthermore, record linkage between stroke registries and administrative claims databases or other existing databases has widened the utility of registry data in research. Using stroke registry data as the reference standard, researchers have validated several algorithms for ascertaining the diagnosis of stroke and its risk factors from claims data, and have also developed a claims-based index to estimate stroke severity. By making better use of registry data, we believe that we will provide better care to patients with stroke.

OBJECTIVE:Stroke is a well-known and devastating complication during the perioperative period. However, detailed stroke risk profiles within 90 days in patients discharged without stroke after inpatient surgery are not fully understood. Using the case-crossover design, we aimed to evaluate the risk of ischemic stroke in these patients. METHODS:We included adult patients with the first hospitalization for ischemic stroke between 2011 and 2012 from 23 million enrollees in the National Health Insurance Research Database. Admission date of the hospitalization was defined as the case day and exactly 365 days before the admission date as the control day. The exposure was the last hospitalization for surgery within 1-30, 31-60, or 61-90 days (case period) before the case day or similar time intervals (control period) before the control day. Surgical types were grouped based on the International Classification of Diseases procedure codes. We performed conditional logistic regression adjusting for time-varying variables to determine the relationship between surgery and subsequent stroke, and case-time-control analyses to examine whether the results were confounded by the time-trend in surgery. RESULTS:A total of 56596 adult patients (41% female, mean age 69 years) comprised the study population. After adjustment was made for confounding variables, an association between stroke and prior inpatient surgery within 30 days was observed (adjusted odds ratio 1.44; 95% confidence interval 1.29-1.61). Cardiothoracic, vascular, digestive surgery, and musculoskeletal surgery within 30 days independently predicted ischemic stroke in the case-crossover analysis. In the case-time-control analysis, inpatient surgery remained an independent risk factor for ischemic stroke, whereas only cardiothoracic, vascular, and digestive surgery independently predicted ischemic stroke. CONCLUSIONS:Surgery as a whole independently increased the risk of ischemic stroke within 30 days. Among various types of surgery, cardiothoracic, vascular, and digestive surgery significantly increased the risk of ischemic stroke.

Sex-related differences in the clinical presentation and outcomes of stroke patients are issues that have attracted increased interest from the scientific community. The present study aimed to investigate sex-related differences in the risk factors for in-hospital mortality and outcome in ischemic stroke patients.A total of 4278 acute ischemic stroke patients admitted to a stroke unit between January 1, 2007 and December 31, 2014 were included in the study. We considered demographic characteristics, clinical characteristics, co-morbidities, and complications, among others, as factors that may affect clinical presentation and in-hospital mortality. Good and poor outcomes were defined as modified Ranking Score (mRS)≦2 and mRS>2. Neurological deterioration (ND) was defined as an increase of National Institutes of Health Stroke Score (NIHSS) ≥ 4 points. Hemorrhagic transformation (HT) was defined as signs of hemorrhage in cranial CT or MRI scans. Transtentorial herniation was defined by brain edema, as seen in cranial CT or MRI scans, associated with the onset of acute unilateral or bilateral papillary dilation, loss of reactivity to light, and decline of ≥ 2 points in the Glasgow coma scale score.Of 4278 ischemic stroke patients (women 1757, 41.1%), 269 (6.3%) received thrombolytic therapy. The in hospital mortality rate was 3.35% (139/4278) [4.45% (80/1757) for women and 2.34% (59/2521) for men, p < 0.01]. At discharge, 41.2% (1761/4278) of the patients showed good outcomes [35.4% (622/1757) for women and 45.2% (1139/2521) for men]. Six months after stroke, 56.1% (1813/3231) showed good outcomes [47.4% (629/1328) for women and 62.2% (1184/1903) for men, p < 0.01]. Atrial fibrillation (AF), diabetes mellitus, stroke history, and old age were factors contributing to poor outcomes in men and women. Hypertension was associated with poor outcomes in women but not in men in comparison with patients without hypertension. Stroke severity and increased intracranial pressure were associated with increased in-hospital mortality in men and women. AF was associated with increased in-hospital mortality in women but not in men compared with patients without AF.The in-hospital mortality rate was not significantly different between women and men. Functional outcomes at discharge and six months after stroke were poorer in women than in men. Hypertension is an independent factor causing poorer outcomes in women than in men. AF is an independent factor affecting sex differences in hospital mortality in women.

The relationship between cholesterol level and hemorrhagic stroke is inconclusive. We hypothesized that low cholesterol levels may have association with intracerebral hemorrhage (ICH) severity at admission and 3-month outcomes. This study used data obtained from a multi-center stroke registry program in Taiwan. We categorized acute spontaneous ICH patients, based on their baseline levels of total cholesterol (TC) measured at admission, into 3 groups with 200 mg/dL of TC. We evaluated risk of having initial stroke severity, with National Institutes of Health Stroke Scale (NIHSS) >15 and unfavorable outcomes (modified Rankin Scale [mRS] score >2, 3-month mortality) after ICH by the TC group. A total of 2444 ICH patients (mean age 62.5±14.2 years; 64.2% men) were included in this study and 854 (34.9%) of them had baseline TC 15), with an adjusted odds ratio [aOR] of 1.80; 95% confidence interval [CI], 1.41-2.30), and 3-month mRS >2 (aOR, 1.41; 95% CI, 1.11-1.78) using patients with TC >200 mg/dL as reference. Those with TC >160 mg/dL and body mass index (BMI)

Code stroke systems are widely implemented to expedite acute stroke treatment. Although this system requires considerable resources, so far no reimbursement has been provided by the Bureau of National Health Insurance (BNHI) in Taiwan. We investigated how often a code stroke was initiated and the percentage of patients treated with intravenous (IV) tissue plasminogen activator, and draw attention to the resulting mismatch. Methods: From January 2010 to September 2011, we prospectively registered all consecutive code stroke patients. Patient characteristics, including demographic data, medical history, comorbidity conditions, treatments, and discharge diagnosis were collected, together with the exact time of onset (or last known normal time) and management. The eligibility of thrombolysis for each patient recorded originally on the chart was reviewed retrospectively on the basis of two sets of criteria, namely, the BNHI reimbursement criteria and the Taiwan Stroke Society (TSS) guideline. Results: During the study period, code strokes were activated for 419 patients at an average of around 20 patients per month. About 57% of code strokes were initiated outside of office hours. Stroke was diagnosed in 377 (90%) patients and 304 (73%) patients had ischemic stroke or transient ischemic attack. A total of 42 (10%) patients according to the BNHI reimbursement criteria and 101 (24%) patients by the TSS guideline were eligible for IV thrombolytic therapy. Of all the code stroke patients, only 49 (12%) were actually treated. Before each additional patient was thrombolysed, about eight patients had been evaluated and excluded from treatment. Conclusion: The majority of code stroke patients were stroke patients; however, most of them could not be treated with thrombolytic therapy. These findings underscore the need for further support from the BNHI in order for health-care providers to implement the code stroke systems successfully.

Silent infarcts (SIs) are commonly found on brain computed tomography (CT) or magnetic resonance imaging (MRI) among elderly subjects, but their risk factors and impact on outcome in stroke patients are unknown. We evaluated the prevalence, risk factors and impact of SIs on the outcome of patients admitted with first-ever ischemic stroke or transient ischemic attack (TIA). Methods: A prospective study of 446 patients admitted consecutively to the neurology service with a diagnosis of TIA or stroke between July 2003 and June 2005, including 226 without any history of prior TIA or stroke. All patients underwent brain CT on the day of admission to the hospital. Risk factors analyzed included age, history of hypertension, diabetes mellitus, cardiovascular disease or stroke, smoking habit and alcohol use. Cholesterol and triglyceride levels were measured on the second day of admission. We monitored these patients for 24 months after stroke onset. Results: The frequency of SIs among the 226 patients with first-ever stroke or TIA was 20%. Most of the SIs were small and deep. Small-artery disease was more frequently observed in patients with SIs. Age, hypertension, diabetes mellitus, hypercholesterolemia, hypertriglyceridemia, alcohol use, smoking habits and atrial fibrillation did not significantly differ between patients with SIs and those without SIs. During the 24-month follow-up period, the frequency of recurrent stroke was higher in patients with SIs than those without SIs. The mortality rate was higher in patients without SIs than those with SIs. The interval from stroke onset to rehospitalization was shorter in patients without SIs than in those with SIs. Conclusion: The study showed a higher frequency of small artery disease in patients with SIs. First-ever stroke patients with SIs should be considered at high risk for recurrent stroke.

Purpose. Status epilepticus (SE) is an important neurological emergency. Early diagnosis could improve outcomes. Traditionally, SE is defined as seizures lasting at least 30 min or repeated seizures over 30 min without recovery of consciousness. Some specialists argued that the duration of seizures qualifying as SE should be shorter and the operational definition of SE was suggested. It is unclear whether physicians follow the operational definition. The objective of this study was to investigate whether the incidence of SE was underestimated and to investigate the underestimate rate. Methods. This retrospective study evaluates the difference in diagnosis of SE between operational definition and traditional definition of status epilepticus. Between July 1, 2012, and June 30, 2014, patients discharged with ICD-9 codes for epilepsy (345.X) in Chia-Yi Christian Hospital were included in the study. A seizure lasting at least 30 min or repeated seizures over 30 min without recovery of consciousness were considered SE according to the traditional definition of SE (TDSE). A seizure lasting between 5 and 30 min was considered SE according to the operational definition of SE (ODSE); it was defined as underestimated status epilepticus (UESE). Results. During a 2-year period, there were 256 episodes of seizures requiring hospital admission. Among the 256 episodes, 99 episodes lasted longer than 5 min, out of which 61 (61.6%) episodes persisted over 30 min (TDSE) and 38 (38.4%) episodes continued between 5 and 30 min (UESE). In the 38 episodes of seizure lasting 5 to 30 minutes, only one episode was previously discharged as SE (ICD-9-CM 345.3). Conclusion. We underestimated 37.4% of SE. Continuing education regarding the diagnosis and treatment of epilepsy is important for physicians.

White matter hyperintensities (WMHs) of presumed vascular origin are common in ageing population, especially in patients with acute cerebral infarction and the volume has been reported to be associated with mental impairment and the risk of hemorrhage from antithrombotic agents. WMHs delineation can be computerized to minimize human bias. However, the presence of cerebral infarcts greatly degrades the accuracy of WMHs detection and thus limits the application of computerized delineation to patients with acute cerebral infarction. We propose a computer-assisted segmentation method to depict WMHs in the presence of cerebral infarcts in combined T1-weighted, fluid attenuation inversion recovery, and diffusion-weighted magnetic resonance imaging (MRI). The proposed method detects WMHs by empirical threshold and atlas information, with subtraction of white matter voxels affected by acute infarction. The method was derived using MRI from 25 hemispheres with WMHs only and 13 hemispheres with both WMHs and cerebral infarcts. Similarity index (SI) and correlation were utilized to assess the agreement between the new automated method and a gold standard visually guided semi-automated method done by an expert rater. The proposed WMHs segmentation approach produced average SI, sensitivity and specificity of 83.142±11.742, 84.154±16.086 and 99.988±0.029% with WMHs only and of 68.826±14.036, 74.381±18.473 and 99.956±0.054% with both WMHs and cerebral infarcts in the derivation cohort. The performance of the proposed method with an external validation cohort was also highly consistent with that of the experienced rater.

Introduction. Quick thrombolysis after stroke improved clinical outcomes. The study objective was to shorten door-to-needle time for thrombolysis. Methods. After identifying the sources of in-hospital delays, we developed a protocol with a parallel algorithm and recruited nurse practitioners into the acute stroke team. We applied the new protocol on stroke patients from October 2009 to September 2010. Patients from the previous two years were used for comparison. Results. For ischemic stroke patients within 3 hours of onset, the median time from arrival to computed tomography scanning was reduced from 29 to 20 minutes (𝑃