Machine learning models for predicting in-hospital mortality in patient with sepsis: Analysis of vital sign dynamics.

Purpose: To build machine learning models for predicting the risk of in-hospital death in patients with sepsis within 48 h, using only dynamic changes in the patient's vital signs. Methods: This retrospective observational cohort study enrolled septic patients from five emergency departments (ED) in Taiwan. We adopted seven variables, i.e., age, sex, systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, and body temperature. Results: Among all 353,253 visits, after excluding 159,607 visits (45%), the study group consisted of 193,646 ED visits. With a leading time of 6 h, the convolutional neural networks (CNNs), long short-term memory (LSTM), and random forest (RF) had accuracy rates of 0.905, 0.817, and 0.835, respectively, and the area under the receiver operating characteristic curve (AUC) was 0.840, 0.761, and 0.770, respectively. With a leading time of 48 h, the CNN, LSTM, and RF achieved accuracy rates of 0.828, 0759, and 0.805, respectively, and an AUC of 0.811, 0.734, and 0.776, respectively. Conclusion: By analyzing dynamic vital sign data, machine learning models can predict mortality in septic patients within 6 to 48 h of admission. The performance of the testing models is more accurate if the lead time is closer to the event.

The Prognostic Value of Serum Soluble TREM-1 on Outcome in Adult Patients with Sepsis.

Increased soluble triggering receptor expressed on myeloid cells 1 (sTREM-1) levels have been reported in patients with sepsis. We tested the hypotheses that serum sTREM-1 levels increase in the early phase of sepsis and decrease after sepsis under appropriate treatment and that sTREM-1 levels can predict therapeutic outcomes. One hundred and fifty-five patients prospectively underwent blood samples including biochemical data, sTREM-1, and biomarkers on endothelial dysfunction as well as clinical severity index examinations. Blood samples from Days 1, 4, and 7 after admission were checked. For comparison, 50 healthy subjects were selected as healthy control. Those patients who had sepsis had significantly higher sTREM-1 levels than those of healthy control. sTREM-1 levels positively correlated with biomarkers for endothelial dysfunction (ICAM-1, VCAM-1, and E-selectin) and lactate level as well as clinical severity index (maximum 24 h APACHE score and Sequential Organ Failure Assessment (SOFA) score) upon admission. sTREM-1 concentrations were significantly higher from Day 1 to Day 7 in the non-survivors than in the survivors. A stepwise logistic regression analysis showed only sTREM-1 level and maximum 24 h SOFA score upon admission were significantly associated with fatality. Area under the receiver operating characteristic curve analysis for the diagnostic accuracy of sTREM-1 in sepsis-related fatality gave a value of 0.726, with a cutoff value of 384.6 pg/mL (sensitivity = 80.8% and specificity = 61.5%). sTREM-1 level may be valuable in auxiliary diagnosis, and it can serve as a useful biomarker as a screening service and follow-up therapeutic outcomes in sepsis.

Effects of Hyperbaric Oxygen Therapy on Serum Adhesion Molecules, and Serum Oxidative Stress in Patients with Acute Traumatic Brain Injury.

BACKGROUND: Serum concentrations of adhesion molecules and oxidative stress is thought to participate in the pathobiology of secondary brain injury after acute traumatic brain injury (TBI). We aimed to study the hypothesis that hyperbaric oxygen therapy (HBOT) both improves the adhesion molecules levels and antioxidant capacity. METHODS: Thirty blood samples from ten patients after acute TBI were obtained after injury and before and after HBOT. Four patients received early HBOT started two weeks after injury, four patients received late HBOT started ten weeks after injury and two patients did not receive HBOT and served as control in this study. The HBOT patients received total 30 times HBOT in six weeks period. RESULTS: Those serum biomarkers in patients with TBI had not significantly difference in glutathione (GSH), thiobarbituric acid reactive substances (TBARS), soluble intercellular cell adhesion-molecule-1 (sICAM-1) and soluble vascular cell adhesion molecule-1 (sVCAM-1) concentrations on admission between early HBOT, late HBOT, and control group (p = 0.916, p = 0.98, p = 0.306, and p = 0.548, respectively). Serum GSH levels were higher at 10 weeks after injury in the early HBOT group than in the late HBOT group and control group (mean, 1.40 µmol/L, 1.16 µmol/L, and 1.05 µmol/L, respectively). Then the serum GSH level was increased at 18 weeks after injury in the late HBOT group (mean, 1.49 µmol/L). However, there was only statistically significant difference at Weeks 18 (p = 0.916, p = 0.463, and p = 0.006, at Week 2, Week 10, and Week 18, respectively). Serum TBARS levels were decreased at 10 weeks after injury in the early HBOT group than in the late HBOT group and control group (mean, 11.21 µmol/L, 17.23 µmol/L, and 17.14 µmol/L, respectively). Then the serum TBARS level was decreased at 18 weeks after injury in the late HBOT group (mean, 12.06 µmol/L). There was statistically significant difference after HBOT (p = 0.98, p = 0.007, and p = 0.018, at Week 2, Week 10, and Week 18, respectively). There was no statistically significant difference between the three groups on sICAM-1 and sVCAM-1 levels from Week 2 to Week 18. CONCLUSIONS: HBOT can improve serum oxidative stress in patients after TBI. These molecules may be added as evaluation markers in clinical practice. Perhaps in the future it may also become part of the treatment of patients after acute traumatic brain injury. Further large-scale study may be warrant.

Predicting in-hospital mortality in adult non-traumatic emergency department patients: a retrospective comparison of the Modified Early Warning Score (MEWS) and machine learning approach.

BACKGROUND: A feasible and accurate risk prediction systems for emergency department (ED) patients is urgently required. The Modified Early Warning Score (MEWS) is a wide-used tool to predict clinical outcomes in ED. Literatures showed that machine learning (ML) had better predictability in specific patient population than traditional scoring system. By analyzing a large multicenter dataset, we aim to develop a ML model to predict in-hospital morality of the adult non traumatic ED patients for different time stages, and comparing performance with other ML models and MEWS. METHODS: A retrospective observational cohort study was conducted in five Taiwan EDs including two tertiary medical centers and three regional hospitals. All consecutively adult (>17 years old) non-traumatic patients admit to ED during a 9-year period (January first, 2008 to December 31th, 2016) were included. Exclusion criteria including patients with (1) out-of-hospital cardiac arrest and (2) discharge against medical advice and transferred to other hospital (3) missing collect variables. The primary outcome was in-hospital mortality and were categorized into 6, 24, 72, 168 hours mortality. MEWS was calculated by systolic blood pressure, pulse rate, respiratory rate, body temperature, and level of consciousness. An ensemble supervised stacking ML model was developed and compared to sensitive and unsensitive Xgboost, Random Forest, and Adaboost. We conducted a performance test and examine both the area under the receiver operating characteristic (AUROC) and the area under the precision and recall curve (AUPRC) as the comparative measures. RESULT: After excluding 182,001 visits (7.46%), study group was consisted of 24,37,326 ED visits. The dataset was split into 67% training data and 33% test data for ML model development. There was no statistically difference found in the characteristics between two groups. For the prediction of 6, 24, 72, 168 hours in-hospital mortality, the AUROC of MEW and ML mode was 0.897, 0.865, 0.841, 0.816 and 0.939, 0.928, 0.913, 0.902 respectively. The stacking ML model outperform other ML model as well. For the prediction of in-hospital mortality over 48-hours, AUPRC performance of MEWS drop below 0.1, while the AUPRC of ML mode was 0.317 in 6 hours and 0.2150 in 168 hours. For each time frame, ML model achieved statistically significant higher AUROC and AUPRC than MEWS (all P < 0.001). Both models showed decreasing prediction ability as time elapse, but there was a trend that the gap of AUROC values between two model increases gradually (P < 0.001). Three MEWS thresholds (score >3, >4, and >5) were determined as baselines for comparison, ML mode consistently showed improved or equally performance in sensitivity, PPV, NPV, but not in specific. CONCLUSION: Stacking ML methods improve predicted in-hospital mortality than MEWS in adult non-traumatic ED patients, especially in the prediction of delayed mortality.

Influence of the Coronavirus Disease 2019 Pandemic on Patients with ST-Segment Elevation Myocardial Infarction in Taiwan.

The outbreak of the new coronavirus disease 2019 (COVID-19) has notably affected the medical system worldwide and influenced the health-seeking behavior of people while depleting medical resources, causing a delay in ST-elevation myocardial infarction (STEMI) management. In this single-center, retrospective cohort study, we compared the clinical pictures of nontransfer patients who presented to the emergency department directly and received primary percutaneous cardiovascular intervention (PPCI) from February 1 to April 30, 2020 (group 2, N = 28), with patients who received PPCI from February 1 to April 30, 2016-2019 (group 1, N = 130). A total of 158 patients with STEMI who received PPCI were included in the study. A decrease in the percentage of patients with door-to-balloon time <90 minutes was found in group 2 (64.3% vs. 81.5%, p = 0.044). The adjusted odds ratio was calculated using logistic regression, according to potential confounding factors such as age, sex, off-hours, and Killip class. An adjusted odds ratio of 2.45 (95% confidence interval, 1.1-6.0, p = 0.048) was reported for group 2. A decrease in the percentage of patients meeting the criteria of door-to-balloon time <90 minutes was demonstrated, and differences were revealed in the clinical pictures of patients with STEMI after the pandemic. While systemic factors contributed the most, improvements and adjustments in the protocols for managing patients with STEMI for better outcomes in the COVID-19 era have yet to be studied.

The timing of last hemodialysis influences the prognostic value of serum lactate levels in predicting mortality of end-stage renal disease patients with sepsis in the emergency department.

ABSTRACT: Sepsis is a life-threatening condition, and serum lactate levels have been used to predict patient prognosis. Studies on serum lactate levels in patients undergoing regular hemodialysis who have sepsis are limited. This study aimed to determine the predictive value of serum lactate levels for sepsis-related mortality among patients who underwent last hemodialysis at three different times before admission to the emergency department (ED).This retrospective cohort study was conducted from January 2007 to December 2013 in southern Taiwan. All hemodialysis patients with sepsis, receiving antibiotics within 24 hours of sepsis confirmation, admitted for at least 3 days, and whose serum lactate levels were known were examined to determine the difference in the serum lactate levels of patients who underwent last hemodialysis within 4 hours (Groups A), in 4-12 hours (Group B), and beyond 12 hours (Group C) before visited to the ED. All the continuous variables, categorical variables and mortality were compared by using Kruskal-Wallis test or Mann-Whitney test, the χ2 or Fisher exact tests, and multiple logistic regression model, respectively.A total of 490 patients were enrolled in the study, and 8.0% (39), 21.5% (84), and 74.9% (367) of the patients were in Group A, Group B and Group C, respectively; the serum lactate levels (2.91 vs 2.13 vs 2.79 mmol/L, respectively; P = .175) and 28-day in-hospital mortality (17.9% vs 14.6% vs 22.9%) showed no statistically significant difference between 3 groups. The association between serum lactate levels and 28-day in-hospital mortality was reliable in Group B (P = .002) and Group C (P < .001), but it was unreliable in Group A (P = .629).Serum lactate level has acceptable sensitivity in predicting 28-day in-hospital mortality among patients with sepsis who undergo last hemodialysis after 4 hours, but is not reliable when the last hemodialysis takes place within 4 hours.

Liver cirrhosis affects serum lactate level measurement while assessing disease severity in patients with sepsis.

BACKGROUND: Elevated serum lactate is associated with higher mortality in sepsis, whereas liver dysfunction is associated with higher serum lactate levels. We assessed the predictive ability of serum lactate in patients with liver cirrhosis and sepsis. METHODS: This retrospective study included 12 281 cases of suspected infection with initial serum blood lactate drawn during January 2007-December 2013. RESULTS: Using one-to-two propensity score matching analysis, 1053 and 2106 septic patients with and without underlying liver cirrhosis, respectively, were successfully matched. Lactate levels of survivors and nonsurvivors were 2.58 and 5.93 mmol/L, respectively, in patients without liver cirrhosis (WLC), 2.96 and 7.29 mmol/L, respectively, in patients with nondecompensated liver cirrhosis (NDLC), and 4.08 and 7.16 mmol/L, respectively, in patients with decompensated liver cirrhosis (DLC). In receiver operating characteristic curve analysis, the sensitivity and specificity for predicting mortality were 0.81 and 0.55, respectively, in the WLC group, 0.85 and 0.45, respectively, in the NDLC group, and 0.86 and 0.33, respectively, in the DLC group, using serum lactate levels >2.0 mmol/L. CONCLUSIONS: The serum lactate level can be used to predict the severity of sepsis in patients with liver cirrhosis; however, its specificity would be lower at a cutoff of 2.0 mmol/L.

Does inappropriate initial antibiotic therapy affect in-hospital mortality of patients in the emergency department with Escherichia coli and Klebsiella pneumoniae bloodstream infections?

Extended-spectrum ß-lactamase (ESBL)-positive bloodstream infection (BSI) is on the rise worldwide. The purpose of this study is to evaluate the impact of inappropriate initial antibiotic therapy (IIAT) on in-hospital mortality of patients in the emergency department (ED) with Escherichia coli and Klebsiella pneumoniae BSIs. This retrospective single-center cohort study included all adult patients with E. coli and K. pneumoniae BSIs between January 2007 and December 2013, who had undergone a blood culture test and initiation of antibiotics within 6 h of ED registration time. Multiple logistic regression was used to adjust for bacterial species, IIAT, time to antibiotics, age, sex, quick Sepsis Related Organ Failure Assessment (qSOFA) score ⩾ 2, and comorbidities. A total of 3533 patients were enrolled (2967 alive and 566 deceased, in-hospital mortality rate 16%). The patients with K. pneumoniae ESBL-positive BSI had the highest mortality rate. Non-survivors had qSOFA scores ⩾ 2 (33.6% vs 9.5%, P < 0.001), more IIAT (15.0% vs 10.7%, P = 0.004), but shorter mean time to antibiotics (1.70 vs 1.84 h, P < 0.001). A qSOFA score ⩾ 2 is the most significant predictor for in-hospital mortality; however, IIAT and time to antibiotics were not significant predictors in multiple logistic regression analysis. In subgroup analysis divided by qSOFA scores, IIAT was still not a significant predictor. Severity of the disease (qSOFA score ⩾ 2) is the key factor influencing in-hospital mortality of patients with E. coli and K. pneumoniae BSIs. The time to antibiotics and IIAT were not significant predictors because they in turn were affected by disease severity.

Medical Liability of Residents in Taiwan Criminal Court: An Analysis of Closed Malpractice Cases.

OBJECTIVE: By analyzing closed criminal malpractice claims involving resident physicians, we aimed to clarify the characteristics of litigations and examine the litigious errors leading to guilty verdicts. DESIGN: A retrospective descriptive study. Setting/Study Participants. The verdicts pertaining to physicians recorded on the national database of the Taiwan justice system were reviewed. Main Outcome Measures. The characteristics of litigations were documented. Negligence and guilty verdicts were further analyzed to identify litigious errors. RESULTS: Between January 1, 2000, and December 31, 2014, from a total of 436 closed criminal malpractice cases, 40 included resident physicians. Five (12.5%) cases received guilty verdicts with mean imprisonment sentences of 5.4 ± 4.1 months. An average of 77.2 months was required for the final adjudication, and surgery residents were involved most frequently (38.9%). Attending physicians were codefendants in 82.5% of cases and were declared guilty in 60% of them. Sepsis (37.5%) was the most common disease in the 40 cases examined, followed by operation/procedure complications (25%). Performance errors (70%) were more than twice as common than diagnostic errors (30%), but the percentage of guilty verdicts in performance error cases was much lower (7.1% vs. 25%). Four negligence cases received nonguilty verdicts, which were mostly due to lack of causation. CONCLUSION: Closed criminal malpractice cases involving residents took on average 6.22 years to conclude. Performance errors accounted for 70% of cases, with treatment of sepsis and operation/procedure complications predominant. To reduce medicolegal risk, residents should learn experiences from analyzing malpractice cases to avoid similar litigious pitfalls.

Impact of oxidative stress on treatment outcomes in adult patients with sepsis: A prospective study.

PURPOSE: Oxidative stress has been shown to reflect on the development of sepsis and disease severity. In the present study, we evaluated the effects of increased levels of oxidative stress and decreased antioxidant coactivity in patients with sepsis, and the importance of oxidative stress on treatment outcomes. METHODS: Biomarkers of oxidative stress (thiobarbituric acid-reactive substances [TBARS]) and antioxidant capacity (glutathione peroxidase [GPx] and glutathione content [thiol]) were prospectively evaluated along with biochemical and clinical data in 100 patients with sepsis on days 1, 4, and 7 after admission. RESULTS: The TBARS level of the non-survivor group was significantly higher than that of the survivor group on day 1 and day 4 and negatively correlated with thiol upon admission. However, thiol was positively correlated with lactate concentration. The TBARS and lactate levels upon admission were independent predictors of fatality. CONCLUSIONS: We conclude that a TBARS cut-off value of 18.30 µM can be used to predict fatality, and an increase in the TBARS concentration by 1 µM will increase the fatality rate by 0.94%. In the panel of biomarkers, the TBARS assay can be considered as a prognostic biomarker for the treatment of patients with sepsis.

Early antibiotics use in young infants with invasive bacterial infection visiting emergency department, a single medical center's experience.

BACKGROUND: The Surviving Sepsis Campaign (SSC) recommends that intravenous antimicrobials be administered within one hour to treat sepsis in both adult and pediatric patients. However, most studies on the timing of antibiotics have been based on adults so far. Therefore, the goal of this study was to clarify the clinical outcomes of the timing of antibiotics use in young infants diagnosed with invasive bacterial infections. METHODS: We conducted this retrospective, cohort study at a single tertiary medical center during the period of 2007-2013 and included young infants visiting the PED diagnosed with an invasive bacterial infection. Early antibiotics use is defined as broad-spectrum intravenous antibiotics given within one hour of a patient's triage at PED. In this study, we have analyzed the factors associated with early antibiotics use and measured such primary outcomes as mortality, ICU admission, ventilation support, and length of hospital stay. RESULTS: A total of 1231 visits of young infants with fever occurred during our study period, and we included 41 patients in our study; 29 of them were diagnosed with bacteremia and the remaining 12 with bacterial meningitis. Early antibiotics use was also associated with a shortened hospital stay (14 ± 8.5 vs. 24 ± 19.6; p = 0.043). However, no statistical differences were observed with regard to mortality, ICU admission, or ventilation support. CONCLUSION: Among young infants that visited the pediatric department with an invasive bacterial infection, early broad-spectrum antibiotics use was not associated with reduced mortality rate, ICU admission, or ventilator support. However, it may have advantages regarding hospital length of stay.

The clinical features and therapy of community-acquired gram negative bacteremia in children less than three years old.

BACKGROUND: Community-acquired Gram-Negative (GN) bacteremia caused more morbidity and mortality recently in children. The increasing drug resistance was also an important issue. However, published reference was few about children. METHODS: We conducted a retrospective study to collect febrile patients with blood culture from a pediatric emergency department during 2007∼2013, and exclude cases ever admitted to hospital within 14 days. These blood cultures all showed single GN organism. The demographic characteristics of enrolled patients and the antibiogram of pathogens were recorded, and then were compared statistically to find out the immediate and appropriate antibiotics. RESULTS: Total 143 sets of blood culture were GN bacilli and the median age of cases was 2 (IQR, 1-5) years old. Male gender was predominant. Non-fermenting Gram-negative bacilli group (NFGNB spp.), Salmonella spp. and Escherichia coli were first three common pathogens respectively. However, total 37 cases of NFGNB spp. other than Pseudomonas aeruginosa were the possible pathogens. By multiple logistic regression analysis, lower hemoglobin and higher alanine aminotransferase were significant difference between common pathogens and possible ones. Besides, the prevalent age regarding resistant strains of Escherichia coli and Pseudomonas aeruginosa were both focused on less than 1 year old. However, Salmonella spp. were prevalent in the age from 1 to 3 years old. CONCLUSIONS: For different age groups, Salmonella spp. and Escherchia coli were the most common pathogens of community-acquired GN bacteremia. For infants, Pseudomonas aeruginosa sepsis and resistant strain of Escherchia coli should be alert, and broader antibiotics should be considered.

C-Reactive Protein Concentration Can Help to Identify Bacteremia in Children Visiting the Emergency Department: A Single Medical Center Experience.

BACKGROUND: For febrile children who are evaluated in a pediatric emergency department (PED), blood culture can be considered the laboratory criterion standard to detect bacteremia. However, high rates of negative, false-positive, or contaminated blood cultures in children often result in this testing being noncontributory. This study determined the factors associated with true-positive blood cultures in children. METHODS: This retrospective study was conducted at a tertiary medical center's PED. The blood culture use reports were prepared by an infectious disease specialist and were classified as bacteremia, nonbacteremia, and contamination. RESULTS: We registered a total of 239,459 PED visits during the 8-year period, and 21,841 blood culture samples were taken. Of the laboratory test studies, higher C-reactive protein (CRP) levels and lower hemoglobin levels were observed in the bacteremia group compared with other groups (all P < 0.001). The cut-off value calculated for each age group was adjusted for better clinical usage and significantly improved the blood culture clinical utility documented in the following age groups: 0 to 1 years (CRP level = 30 mg/L, odds ratio [OR] = 5.4, P < 0.001), 1 to 3 years (CRP level = 45 mg/L, OR = 3.7, P < 0.001), and 12 to 18 years (CRP level = 50 mg/L, OR = 6.3, P = 0.006). Using the CRP cut-off value established in this study, we could reduce the blood culture samples in the PED by 14,108 (64.6%). CONCLUSIONS: This study provides new evidence that CRP may be a useful indicator for blood culture sampling in certain age groups and may help improve the efficiency of blood culture in the PED.

The role of blink reflex R1 latency as an electrophysiological marker in diabetic distal symmetrical polyneuropathy.

OBJECTIVE: Studies showed a relatively prolonged blink R1 latency in patients with diabetic distal symmetrical polyneuropathy (DSPN) compared to that without DSPN. We tested the hypothesis that blink R1 latency would provide a diagnostic alternative to nerve conduction studies (NCS) in DSPN and act as a marker of the severity of NCS abnormalities in DSPN. METHOD: A total of 109 patients with type 2 diabetes underwent blink reflex studies and NCS. We used the composite amplitude scores of nerve conductions (CAS), which consisted of motor (tibial, peroneal and ulnar) and sensory (sural and ulnar) amplitudes for estimating the severity of NCS. RESULTS: Patients with DSPN had longer blink R1, R2, and contralateral R2 latencies (P < 0.0001, P = 0.001, and P = 0.031, respectively) and higher CAS (P < 0.0001). Area under curve on receiver operating characteristic curve analysis in diagnosing occurrence of DSPN in blink R1 latency was 0.772 (P < 0.0001). Multiple linear regression analysis showed that blink R1 latency was independently associated with CAS. CONCLUSION: Blink R1 latency may be valuable in auxiliary diagnosis and in determining the severity of NCS abnormalities in DSPN. SIGNIFICANCE: Blink R1 latency can be added as a supplemental marker of severity of NCS in DSPN, especially if the patient's sural amplitudes has a floor effect.

Sural nerve sensory response in diabetic distal symmetrical polyneuropathy.

INTRODUCTION: The sural sensory nerve action potential (SNAP) amplitude is a measure of the number of axons. We tested the hypothesis that sural SNAP amplitude can be used as a marker in screening, severity evaluation, and follow-up of diabetic distal symmetrical polyneuropathy (DSPN). METHODS: Patients with type 2 diabetes underwent nerve conduction studies and were followed for 6 years. Composite amplitude scores (CASs) were determined to evaluate DSPN severity. RESULTS: Sural SNAP amplitudes were negatively correlated with CAS (r = -.790, P < .0001), and changes in sural SNAP amplitudes were negatively correlated with those of CAS after controlling for follow-up duration (r = -.531, P = .028). DISCUSSION: When a patient's baseline sural SNAP amplitude is above zero, it can be used as one measure of DSPN in screening, severity evaluation, and follow-up. However, if the patient's sural SNAP value is zero, CAS can be used as a follow-up measure.

Mortality Prediction of Septic Patients in the Emergency Department Based on Machine Learning.

In emergency departments, the most common cause of death associated with suspected infected patients is sepsis. In this study, deep learning algorithms were used to predict the mortality of suspected infected patients in a hospital emergency department. During January 2007 and December 2013, 42,220 patients considered in this study were admitted to the emergency department due to suspected infection. In the present study, a deep learning structure for mortality prediction of septic patients was developed and compared with several machine learning methods as well as two sepsis screening tools: the systemic inflammatory response syndrome (SIRS) and quick sepsis-related organ failure assessment (qSOFA). The mortality predictions were explored for septic patients who died within 72 h and 28 days. Results demonstrated that the accuracy rate of deep learning methods, especially Convolutional Neural Network plus SoftMax (87.01% in 72 h and 81.59% in 28 d), exceeds that of the other machine learning methods, SIRS, and qSOFA. We expect that deep learning can effectively assist medical staff in early identification of critical patients.

Adjunctive statin therapy reduces intracranial hemorrhage and 1-year mortality in patients with atrial fibrillation after acute ischemic stroke: A population-based epidemiological study from Taiwan.

BACKGROUND: Atrial fibrillation (AF)-related stroke causes severe disability and poor prognosis. Adjunctive statin therapy has been recommended for atherosclerotic-related stroke but not AF-related stroke. This study investigated the effects of statin in AF patients who experienced acute ischemic stroke. METHODS: Data from patients with AF experiencing first-ever ischemic stroke between 2001 and 2010 were collected from the Taiwan National Health Insurance Research Database and categorized into non-statin and statin groups. The statin group was further divided into pre-stroke statin (those who began statin therapy before stroke) and post-stroke statin (those who began statin therapy after stroke) groups. The risks for recurrent ischemic stroke, coronary artery disease (CAD), intracranial hemorrhage (ICH), and 1-year mortality were compared among the groups. RESULTS: A total of 43,242 patients were in the non-statin, 2858 in the pre-stroke statin and 4640 in post-stroke statin groups. Comparing the risk for recurrent stroke and CAD among the three groups, the pre-stroke statin and post-stroke statin groups did not exhibit a significant difference compared with the non-statin group. In terms of ICH risk, the statin group had a lower risk for ICH (odds ratio [OR] 0.79, 95% confidence interval [CI] 0.68-0.90; p = 0.0007) compared with the non-statin group. The overall 1-year mortality in both statin subgroups was lower than that in the non-statin group (pre-stroke statin, OR 0.55 [95% CI 0.49-0.61]; p < 0.0001 versus post-stroke statin, OR 0.53 [95% CI 0.48-0.58]; p < 0.0001). CONCLUSIONS: Statin therapy reduced the risk of ICH and 1-year mortality in AF patients who experienced acute ischemic stroke.

α-1-Acid Glycoprotein Concentration as an Outcome Predictor in Adult Patients with Sepsis.

BACKGROUND: α-1-Acid glycoprotein (AGP) is an acute-phase protein that plays a role in first-line defense against infection and is therefore elevated in sepsis. We tested the hypothesis that AGP levels increase initially in sepsis and decrease after antimicrobial therapy and that these levels may predict treatment outcomes. METHODS: AGP, biomarkers widely used in clinical practice, and maximum 24-h acute physiology and chronic health evaluation (APACHE)-II scores upon emergency department (ED) admission were prospectively evaluated and compared. We further examined changes in AGP concentrations 1, 4, and 7 days after admission and determined the value of AGP that may be used to accurately and reliably predict the prognosis in patients with sepsis. RESULTS: Mechanical ventilation, white blood cell (WBC) counts, C-reactive protein (CRP) and lactate levels, maximum 24-h APACHE-II scores, and AGP concentrations were significantly higher upon admission in patients with sepsis who died. AGP and lactate concentrations were also significantly higher in non-survivors than in survivors on days 1, 4, and 7. As indicated by the stepwise logistic regression model analysis and area under the curve analysis, AGP was the best prognostic indicator, and the cut-off value for predicting fatality was 1307 µg/mL, and any increase 1-ng/mL in AGP concentration would increase the fatality rate by 0.5%. CONCLUSION: Based on our observations, AGP may be a good prognostic predictor in patients with sepsis. In addition, serial AGP levels meet the requirements for predicting outcomes in patients with sepsis.

Long-term effects of surgical treatment on baroreflex function in patients with obstructive sleep apnea: an 18-month follow-up.

OBJECTIVE: Depressed baroreflex sensitivity (BRS) is reported in obstructive sleep apnea (OSA). Improvement of BRS short-term after surgical treatment is also reported. We tested the hypothesis that surgical treatment not only improves clinical outcomes, but also improves BRS after 18 months. METHODS: Cardiovascular autonomic tests, polysomnography (PSG), and biochemical testing were prospectively evaluated in 54 OSA patients at three time points (preoperatively, 6 months and 18 months postoperatively) and compared with 20, age- and body mass index (BMI)-matched, healthy controls. RESULTS: The BRS increased after surgical treatment at 18-month follow-up, with results similar to the healthy control. Additionally, average O2, mO2 <90% (% per night), and lowest O2 showed an increase after surgical treatment at the 18-month follow-up. CONCLUSIONS: Besides improvement in clinical outcomes, depressed BRS in OSA patients is reversible and these patients have the potential for total recovery of baroreflex function after 18 months of treatment.