Gordon R. Bernard
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Gordon R. Bernard has a total of 111 co-authors. Here are the top co-authors by number of shared publications:
- Lorraine B. Ware
Vanderbilt University Medical Center
9 shared publications - Tatsuki Koyama
Department of Biostatistics, Vanderbilt University, Nashville, TN, USA
6 shared publications - Addison K. May
Department of Surgery, Atrium Health Carolinas Medical Center
5 shared publications - Arthur P. Wheeler
Division of Allergy, Pulmonary and Critical Care Medicine, Vanderbilt University Medical Center, Nashville, TN
4 shared publications - David R. Janz
Division of Allergy, Pulmonary, and Critical Care Medicine, Department of Medicine, Vanderbilt University School of Medicine, Nashville, TN Division of Allergy, Pulmonary, and Critical Care Medicine, Department of Medicine; and, Department of Pathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN
4 shared publications - Michael A. Matthay
Cardiovascular Research Institute, Departments of Medicine and Anesthesia, University of California, San Francisco, USA
4 shared publications
Latest Publications
- Open Access
- 13 June 2018
- Published by Springer Nature in BMC Bioinformatics
- Vol. 19 (S8) , 210-74
Identifying tweets of personal health experience through word embedding and LSTM neural network
As Twitter has become an active data source for health surveillance research, it is important that efficient and effective methods are developed to identify tweets related to personal health experience. Conventional classification algorithms rely on features engineered by human domain experts, and engineering such features is a challenging task and requires much human intelligence. The resultant features may not be optimal for the classification problem, and can make it challenging for conventional classifiers to correctly predict personal experience tweets (PETs) due to the various ways to express and/or describe personal experience in tweets. In this study, we developed a method that combines word embedding and long short-term memory (LSTM) model without the need to engineer any specific features. Through word embedding, tweet texts were represented as dense vectors which in turn were fed to the LSTM neural network as sequences. Statistical analyses of the results of 10-fold cross-validations of our method and conventional methods indicate that there exist significant differences (p < 0.01) in performance measures of accuracy, precision, recall, F1-score, and ROC/AUC, demonstrating that our approach outperforms the conventional methods in identifying PETs. We presented an efficient and effective method of identifying health-related personal experience tweets by combining word embedding and an LSTM neural network. It is conceivable that our method can help accelerate and scale up analyzing textual data of social media for health surveillance purposes, because of no need for the laborious and costly process of engineering features.
- 13 June 2018
- Published by Springer Nature in BMC Bioinformatics
- Vol. 19 (S8) , 210-74
Open AccessIdentifying tweets of personal health experience through word embedding and LSTM neural network
As Twitter has become an active data source for health surveillance research, it is important that efficient and effective methods are developed to identify tweets related to personal health experience. Conventional classification algorithms rely on features engineered by human domain experts, and engineering such features is a challenging task and requires much human intelligence. The resultant features may not be optimal for the classification problem, and can make it challenging for conventional classifiers to correctly predict personal experience tweets (PETs) due to the various ways to express and/or describe personal experience in tweets. In this study, we developed a method that combines word embedding and long short-term memory (LSTM) model without the need to engineer any specific features. Through word embedding, tweet texts were represented as dense vectors which in turn were fed to the LSTM neural network as sequences. Statistical analyses of the results of 10-fold cross-validations of our method and conventional methods indicate that there exist significant differences (p < 0.01) in performance measures of accuracy, precision, recall, F1-score, and ROC/AUC, demonstrating that our approach outperforms the conventional methods in identifying PETs. We presented an efficient and effective method of identifying health-related personal experience tweets by combining word embedding and an LSTM neural network. It is conceivable that our method can help accelerate and scale up analyzing textual data of social media for health surveillance purposes, because of no need for the laborious and costly process of engineering features.
- 13 February 2018
- Published by American Medical Association (AMA) in JAMA
- Vol. 319 (6) , 614-615
Early Resuscitation for Adults With Sepsis in a Low-income Country—Reply
In Reply Drs Nedel and Deutschendorf highlight 2 key considerations for the interpretation of our trial and the management of patients with sepsis across settin
- 13 February 2018
- Published by American Medical Association (AMA) in JAMA
- Vol. 319 (6) , 614-615
Early Resuscitation for Adults With Sepsis in a Low-income Country—Reply
In Reply Drs Nedel and Deutschendorf highlight 2 key considerations for the interpretation of our trial and the management of patients with sepsis across settin
- Top Cited PapersOpen Access
- 3 October 2017
- Published by American Medical Association (AMA) in JAMA
- Vol. 318 (13) , 1233-1240
Effect of an Early Resuscitation Protocol on In-hospital Mortality Among Adults With Sepsis and Hypotension
Importance The effect of an early resuscitation protocol on sepsis outcomes in developing countries remains unknown. Objective To determine whether an early resuscitation protocol with administration of intravenous fluids, vasopressors, and blood transfusion decreases mortality among Zambian adults with sepsis and hypotension compared with usual care. Design, Setting, and Participants Randomized clinical trial of 212 adults with sepsis (suspected infection plus ≥2 systemic inflammatory response syndrome criteria) and hypotension (systolic blood pressure ≤90 mm Hg or mean arterial pressure ≤65 mm Hg) presenting to the emergency department at a 1500-bed referral hospital in Zambia between October 22, 2012, and November 11, 2013. Data collection concluded December 9, 2013. Interventions Patients were randomized 1:1 to either (1) an early resuscitation protocol for sepsis (n = 107) that included intravenous fluid bolus administration with monitoring of jugular venous pressure, respiratory rate, and arterial oxygen saturation and treatment with vasopressors targeting mean arterial pressure (≥65 mm Hg) and blood transfusion (for patients with a hemoglobin level P = .03). In the 6 hours after presentation to the emergency department, patients in the sepsis protocol group received a median of 3.5 L (interquartile range, 2.7-4.0 L) of intravenous fluid compared with 2.0 L (interquartile range, 1.0-2.5 L) in the usual care group (mean difference, 1.2 L [95% CI, 1.0-1.5 L];P < .001). Fifteen patients (14.2%) in the sepsis protocol group and 2 patients (1.9%) in the usual care group received vasopressors (between-group difference, 12.3% [95% CI, 5.1%-19.4%];P < .001). Conclusions and Relevance Among adults with sepsis and hypotension, most of whom were positive for HIV, in a resource-limited setting, a protocol for early resuscitation with administration of intravenous fluids and vasopressors increased in-hospital mortality compared with usual care. Further studies are needed to understand the effects of administration of intravenous fluid boluses and vasopressors in patients with sepsis across different low- and middle-income clinical settings and patient populations. Trial Registration clinicaltrials.gov Identifier:NCT01663701
- 3 October 2017
- Published by American Medical Association (AMA) in JAMA
- Vol. 318 (13) , 1233-1240
Top Cited PapersOpen AccessEffect of an Early Resuscitation Protocol on In-hospital Mortality Among Adults With Sepsis and Hypotension
Importance The effect of an early resuscitation protocol on sepsis outcomes in developing countries remains unknown. Objective To determine whether an early resuscitation protocol with administration of intravenous fluids, vasopressors, and blood transfusion decreases mortality among Zambian adults with sepsis and hypotension compared with usual care. Design, Setting, and Participants Randomized clinical trial of 212 adults with sepsis (suspected infection plus ≥2 systemic inflammatory response syndrome criteria) and hypotension (systolic blood pressure ≤90 mm Hg or mean arterial pressure ≤65 mm Hg) presenting to the emergency department at a 1500-bed referral hospital in Zambia between October 22, 2012, and November 11, 2013. Data collection concluded December 9, 2013. Interventions Patients were randomized 1:1 to either (1) an early resuscitation protocol for sepsis (n = 107) that included intravenous fluid bolus administration with monitoring of jugular venous pressure, respiratory rate, and arterial oxygen saturation and treatment with vasopressors targeting mean arterial pressure (≥65 mm Hg) and blood transfusion (for patients with a hemoglobin level P = .03). In the 6 hours after presentation to the emergency department, patients in the sepsis protocol group received a median of 3.5 L (interquartile range, 2.7-4.0 L) of intravenous fluid compared with 2.0 L (interquartile range, 1.0-2.5 L) in the usual care group (mean difference, 1.2 L [95% CI, 1.0-1.5 L];P < .001). Fifteen patients (14.2%) in the sepsis protocol group and 2 patients (1.9%) in the usual care group received vasopressors (between-group difference, 12.3% [95% CI, 5.1%-19.4%];P < .001). Conclusions and Relevance Among adults with sepsis and hypotension, most of whom were positive for HIV, in a resource-limited setting, a protocol for early resuscitation with administration of intravenous fluids and vasopressors increased in-hospital mortality compared with usual care. Further studies are needed to understand the effects of administration of intravenous fluid boluses and vasopressors in patients with sepsis across different low- and middle-income clinical settings and patient populations. Trial Registration clinicaltrials.gov Identifier:NCT01663701
- 1 July 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2017 (1557170X) , 1174-1177
Identifying personal health experience tweets with deep neural networks
Twitter, as a social media platform, has become an increasingly useful data source for health surveillance studies, and personal health experiences shared on Twitter provide valuable information to the surveillance. Twitter data are known for their irregular usages of languages and informal short texts due to the 140 character limit, and for their noisiness such that majority of the posts are irrelevant to any particular health surveillance. These factors pose challenges in identifying personal health experience tweets from the Twitter data. In this study, we designed deep neural networks with 3 different architectural configurations, and after training them with a corpus of 8,770 annotated tweets, we used them to predict personal experience tweets from a set of 821 annotate tweets. Our results demonstrated a significant amount of improvement in predicting personal health experience tweets by deep neural networks over that by conventional classifiers: 37.5% in accuracy, 31.1% in precision, and 53.6% in recall. We believe that our method can be utilized in various health surveillance studies using Twitter as a data source.
- 1 July 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2017 (1557170X) , 1174-1177
Identifying personal health experience tweets with deep neural networks
Twitter, as a social media platform, has become an increasingly useful data source for health surveillance studies, and personal health experiences shared on Twitter provide valuable information to the surveillance. Twitter data are known for their irregular usages of languages and informal short texts due to the 140 character limit, and for their noisiness such that majority of the posts are irrelevant to any particular health surveillance. These factors pose challenges in identifying personal health experience tweets from the Twitter data. In this study, we designed deep neural networks with 3 different architectural configurations, and after training them with a corpus of 8,770 annotated tweets, we used them to predict personal experience tweets from a set of 821 annotate tweets. Our results demonstrated a significant amount of improvement in predicting personal health experience tweets by deep neural networks over that by conventional classifiers: 37.5% in accuracy, 31.1% in precision, and 53.6% in recall. We believe that our method can be utilized in various health surveillance studies using Twitter as a data source.
- Open Access
- 13 April 2017
- Published by Springer Nature in Trials
- Vol. 18 (1) , 1-8
Saline versus balanced crystalloids for intravenous fluid therapy in the emergency department: study protocol for a cluster-randomized, multiple-crossover trial
Prior studies in critically ill patients suggest the supra-physiologic chloride concentration of 0.9% (“normal”) saline may be associated with higher risk of renal failure and death compared to physiologically balanced crystalloids. However, the comparative effects of 0.9% saline and balanced fluids are largely unexamined among patients outside the intensive care unit, who represent the vast majority of patients treated with intravenous fluids. This study, entitled Saline Against Lactated Ringer’s or Plasma-Lyte in the Emergency Department (SALT-ED), is a pragmatic, cluster, multiple-crossover trial at a single institution evaluating clinical outcomes of adults treated with 0.9% saline versus balanced crystalloids for intravenous fluid resuscitation in the emergency department. All adults treated in the study emergency department receiving at least 500 mL of isotonic crystalloid solution during usual clinical care and subsequently hospitalized outside the intensive care unit are included. Treatment allocation of 0.9% saline versus balanced crystalloids is assigned by calendar month, with study patients treated during the same month assigned to the same fluid type. The first month (January 2016) was randomly assigned to balanced crystalloids, with each subsequent month alternating between 0.9% saline and balanced crystalloids. For balanced crystalloid treatment, clinicians can choose either Lactated Ringer’s or Plasma-Lyte A©. The study period is set at 16 months, which will result in an anticipated estimated sample size of 15,000 patients. The primary outcome is hospital-free days to day 28, defined as the number of days alive and out of the hospital from the index emergency department visit until 28 days later. Major secondary outcomes include proportion of patients who develop acute kidney injury by creatinine measurements; major adverse kidney events by hospital discharge or day 30 (MAKE30), which is a composite outcome of death, new renal replacement therapy, and persistent creatinine elevation >200% of baseline; and in-hospital mortality. This ongoing pragmatic trial will provide the most comprehensive evaluation to date of clinical outcomes associated with 0.9% saline compared to physiologically balanced fluids in patients outside the intensive care unit. ClinicalTrials.gov, NCT02614040 . Registered on 18 November 2015.
- 13 April 2017
- Published by Springer Nature in Trials
- Vol. 18 (1) , 1-8
Open AccessSaline versus balanced crystalloids for intravenous fluid therapy in the emergency department: study protocol for a cluster-randomized, multiple-crossover trial
Prior studies in critically ill patients suggest the supra-physiologic chloride concentration of 0.9% (“normal”) saline may be associated with higher risk of renal failure and death compared to physiologically balanced crystalloids. However, the comparative effects of 0.9% saline and balanced fluids are largely unexamined among patients outside the intensive care unit, who represent the vast majority of patients treated with intravenous fluids. This study, entitled Saline Against Lactated Ringer’s or Plasma-Lyte in the Emergency Department (SALT-ED), is a pragmatic, cluster, multiple-crossover trial at a single institution evaluating clinical outcomes of adults treated with 0.9% saline versus balanced crystalloids for intravenous fluid resuscitation in the emergency department. All adults treated in the study emergency department receiving at least 500 mL of isotonic crystalloid solution during usual clinical care and subsequently hospitalized outside the intensive care unit are included. Treatment allocation of 0.9% saline versus balanced crystalloids is assigned by calendar month, with study patients treated during the same month assigned to the same fluid type. The first month (January 2016) was randomly assigned to balanced crystalloids, with each subsequent month alternating between 0.9% saline and balanced crystalloids. For balanced crystalloid treatment, clinicians can choose either Lactated Ringer’s or Plasma-Lyte A©. The study period is set at 16 months, which will result in an anticipated estimated sample size of 15,000 patients. The primary outcome is hospital-free days to day 28, defined as the number of days alive and out of the hospital from the index emergency department visit until 28 days later. Major secondary outcomes include proportion of patients who develop acute kidney injury by creatinine measurements; major adverse kidney events by hospital discharge or day 30 (MAKE30), which is a composite outcome of death, new renal replacement therapy, and persistent creatinine elevation >200% of baseline; and in-hospital mortality. This ongoing pragmatic trial will provide the most comprehensive evaluation to date of clinical outcomes associated with 0.9% saline compared to physiologically balanced fluids in patients outside the intensive care unit. ClinicalTrials.gov, NCT02614040 . Registered on 18 November 2015.
- Open Access
- 3 February 2017
- Published by Springer Nature in BMC Pulmonary Medicine
- Vol. 17 (1) , 1-7
Early exposure to hyperoxia and mortality in critically ill patients with severe traumatic injuries
Hyperoxia is common early in the course of resuscitation of critically ill patients. It has been associated with mortality in some, but not all, studies of cardiac arrest patients and other critically ill cohorts. Reasons for the inconsistency are unclear and may depend on unmeasured patient confounders, the timing and duration of hyperoxia, population characteristics, or the way that hyperoxia is defined and measured. We sought to determine whether, in a prospectively collected cohort of mechanically ventilated patients with traumatic injuries with and without head trauma, higher maximum partial pressure of arterial oxygen (PaO2) within 24 hours of admission would be associated with increased risk of in-hospital mortality. Critically ill patients with traumatic injuries undergoing invasive mechanical ventilation enrolled in the Validating Acute Lung Injury biomarkers for Diagnosis (VALID) study were included in this study. All arterial blood gases (ABGs) from the first 24 hours of admission were recorded. Primary analysis was comparison of the highest PaO2 between hospital survivors and non-survivors. A total of 653 patients were evaluated for inclusion. Of these, 182 were not mechanically ventilated or did not have an ABG measured in the first 24 hours, leaving 471 patients in the primary analysis. In survivors, the maximum PaO2 was 141 mmHg (median, interquartile range 103 - 212) compared to 148 mmHg (IQR 105 - 209) in non-survivors (p = 0.82). In the subgroup with head trauma (n = 266), the maximum PaO2 was 133 mmHg (IQR 97 - 187) among survivors and 152 mmHg (108 - 229) among nonsurvivors (p = 0.19). After controlling for age, injury severity score, number of arterial blood gases, and fraction of inspired oxygen, maximum PaO2 was not associated with increased mortality (OR 1.27 for every fold increase of PaO2 (95% CI 0.72 - 2.25). In mechanically ventilated patients with severe traumatic injuries, hyperoxia in the first 24 hours of admission was not associated with increased risk of death or worsened neurological outcomes in a setting without brain tissue oxygenation monitoring.
- 3 February 2017
- Published by Springer Nature in BMC Pulmonary Medicine
- Vol. 17 (1) , 1-7
Open AccessEarly exposure to hyperoxia and mortality in critically ill patients with severe traumatic injuries
Hyperoxia is common early in the course of resuscitation of critically ill patients. It has been associated with mortality in some, but not all, studies of cardiac arrest patients and other critically ill cohorts. Reasons for the inconsistency are unclear and may depend on unmeasured patient confounders, the timing and duration of hyperoxia, population characteristics, or the way that hyperoxia is defined and measured. We sought to determine whether, in a prospectively collected cohort of mechanically ventilated patients with traumatic injuries with and without head trauma, higher maximum partial pressure of arterial oxygen (PaO2) within 24 hours of admission would be associated with increased risk of in-hospital mortality. Critically ill patients with traumatic injuries undergoing invasive mechanical ventilation enrolled in the Validating Acute Lung Injury biomarkers for Diagnosis (VALID) study were included in this study. All arterial blood gases (ABGs) from the first 24 hours of admission were recorded. Primary analysis was comparison of the highest PaO2 between hospital survivors and non-survivors. A total of 653 patients were evaluated for inclusion. Of these, 182 were not mechanically ventilated or did not have an ABG measured in the first 24 hours, leaving 471 patients in the primary analysis. In survivors, the maximum PaO2 was 141 mmHg (median, interquartile range 103 - 212) compared to 148 mmHg (IQR 105 - 209) in non-survivors (p = 0.82). In the subgroup with head trauma (n = 266), the maximum PaO2 was 133 mmHg (IQR 97 - 187) among survivors and 152 mmHg (108 - 229) among nonsurvivors (p = 0.19). After controlling for age, injury severity score, number of arterial blood gases, and fraction of inspired oxygen, maximum PaO2 was not associated with increased mortality (OR 1.27 for every fold increase of PaO2 (95% CI 0.72 - 2.25). In mechanically ventilated patients with severe traumatic injuries, hyperoxia in the first 24 hours of admission was not associated with increased risk of death or worsened neurological outcomes in a setting without brain tissue oxygenation monitoring.
- 4 March 2016
- Published by Springer Nature in Intensive Care Medicine
- Vol. 42 (5) , 640-642
The definition of ARDS revisited: 20 years later
- 4 March 2016
- Published by Springer Nature in Intensive Care Medicine
- Vol. 42 (5) , 640-642
The definition of ARDS revisited: 20 years later
- 1 October 2014
- Published by Wolters Kluwer Health in Survey of Anesthesiology
- Vol. 58 (5) , 212-213
Long-Term Cognitive Impairment After Critical Illness
- 1 October 2014
- Published by Wolters Kluwer Health in Survey of Anesthesiology
- Vol. 58 (5) , 212-213
Long-Term Cognitive Impairment After Critical Illness
- Open Access
- 1 January 2014
- Published by Elsevier in Annals of Intensive Care
- Vol. 4 (1) , 5
Vitamin D deficiency and risk of acute lung injury in severe sepsis and severe trauma: a case-control study
- 1 January 2014
- Published by Elsevier in Annals of Intensive Care
- Vol. 4 (1) , 5
Open AccessVitamin D deficiency and risk of acute lung injury in severe sepsis and severe trauma: a case-control study