scholarly journals Multicenter validation of a sepsis prediction algorithm using only vital sign data in the emergency department, general ward and ICU

2018 ◽  
Author(s):  
Qingqing Mao ◽  
Melissa Jay ◽  
Jana Hoffman ◽  
Jacob Calvert ◽  
Christopher Barton ◽  
...  
Keyword(s):  
Machine Learning ◽  
Septic Shock ◽  
Severe Sepsis ◽  
San Francisco ◽  
Vital Sign ◽  
Operating Characteristic ◽  
Medical Center ◽  
Prediction Algorithm ◽  
Data Set ◽  
Sepsis And Septic Shock

Objectives: We validate a machine learning-based sepsis prediction algorithm (InSight) for detection and prediction of three sepsis-related gold standards, using only six vital signs. We evaluate robustness to missing data, customization to site-specific data using transfer learning, and generalizability to new settings. Design: A machine learning algorithm with gradient tree boosting. Features for prediction were created from combinations of only six vital sign measurements and their changes over time. Setting: A mixed-ward retrospective data set from the University of California, San Francisco (UCSF) Medical Center (San Francisco, CA) as the primary source, an intensive care unit data set from the Beth Israel Deaconess Medical Center (Boston, MA) as a transfer learning source, and four additional institutions' datasets to evaluate generalizability. Participants: 684,443 total encounters, with 90,353 encounters from June 2011 to March 2016 at UCSF. Interventions: none Primary and secondary outcome measures: Area under the receiver operating characteristic curve (AUROC) for detection and prediction of sepsis, severe sepsis, and septic shock. Results: For detection of sepsis and severe sepsis, InSight achieves an area under the receiver operating characteristic (AUROC) curve of 0.92 (95% CI 0.90 - 0.93) and 0.87 (95% CI 0.86 - 0.88), respectively. Four hours before onset, InSight predicts septic shock with an AUROC of 0.96 (95% CI 0.94 - 0.98), and severe sepsis with an AUROC of 0.85 (95% CI 0.79 - 0.91). Conclusions: InSight outperforms existing sepsis scoring systems in identifying and predicting sepsis, severe sepsis, and septic shock. This is the first sepsis screening system to exceed an AUROC of 0.90 using only vital sign inputs. InSight is robust to missing data, can be customized to novel hospital data using a small fraction of site data, and retained strong discrimination across all institutions.

scholarly journals Pediatric Severe Sepsis Prediction Using Machine Learning

2017 ◽  
Author(s):  
Thomas Desautels ◽  
Jana Hoffman ◽  
Christopher Barton ◽  
Qingqing Mao ◽  
Melissa Jay ◽  
...  
Keyword(s):  
Machine Learning ◽  
Severe Sepsis ◽  
San Francisco ◽  
High Performance ◽  
Cross Validation ◽  
Learning Algorithm ◽  
Medical Center ◽  
Prediction Algorithm ◽  
Machine Learning Algorithm ◽  
Logistic Organ Dysfunction

Early detection of pediatric severe sepsis is necessary in order to administer effective treatment. In this study, we assessed the efficacy of a machine-learning-based prediction algorithm applied to electronic healthcare record (EHR) data for the prediction of severe sepsis onset. The resulting prediction performance was compared with the Pediatric Logistic Organ Dysfunction score (PELOD-2) and pediatric Systemic Inflammatory Response Syndrome score (SIRS) using cross-validation and pairwise t-tests. EHR data were collected from a retrospective set of de-identified pediatric inpatient and emergency encounters drawn from the University of California San Francisco (UCSF) Medical Center, with encounter dates between June 2011 and March 2016. Patients (n = 11,127) were 2-17 years of age and 103 [0.93%] were labeled severely septic. In four-fold cross-validation evaluations, the machine learning algorithm achieved an AUROC of 0.912 for discrimination between severely septic and control pediatric patients at onset and AUROC of 0.727 four hours before onset. Under the same measure, the prediction algorithm also significantly outperformed PELOD-2 (p < 0.05) and SIRS (p < 0.05) in the prediction of severe sepsis four hours before onset. This machine learning algorithm has the potential to deliver high-performance severe sepsis detection and prediction for pediatric inpatients.

scholarly journals Clinician Perception of a Machine Learning–Based Early Warning System Designed to Predict Severe Sepsis and Septic Shock*

2019 ◽  
Vol 47 (11) ◽  
pp. 1477-1484 ◽  
Author(s):  
Jennifer C. Ginestra ◽  
Heather M. Giannini ◽  
William D. Schweickert ◽  
Laurie Meadows ◽  
Michael J. Lynch ◽  
...  
Keyword(s):  
Machine Learning ◽  
Septic Shock ◽  
Severe Sepsis ◽  
Early Warning ◽  
Early Warning System ◽  
Warning System ◽  
Sepsis And Septic Shock

scholarly journals Clinical and Demographic Factors Associated with Antipyretic Use in Gram-Negative Severe Sepsis and Septic Shock

2011 ◽  
Vol 45 (10) ◽  
pp. 1207-1216 ◽  
Author(s):  
Nicholas M Mohr ◽  
Brian M Fuller ◽  
Lee P Skrupky ◽  
Hawnwan Moy ◽  
Robert Alunday ◽  
...  
Keyword(s):  
Septic Shock ◽  
Severe Sepsis ◽  
Medical Center ◽  
Academic Medical Center ◽  
Demographic Factors ◽  
Medication Administration ◽  
Patient Treatment ◽  
Gram Negative ◽  
Factors Associated ◽  
Sepsis And Septic Shock

Background: Antipyretic therapy is commonly prescribed for patients with Infection, but studies of its impact on clinical outcomes have yielded mixed results. No data exist to characterize the use of antipyretic medications in patients with severe sepsis or septic shock. Objective: To identify clinical and demographic factors associated with antipyretic medication administration in severe sepsis and septic shock. Methods: This single-center, retrospective, cohort study assessed febrile patients [temperature ≥38.3°C) with gram-negative severe sepsis or septic shock at an 1111-bed academic medical center between January 2002 and February 2008, Patients were excluded if they had liver disease, acute brain injury, or allergy to acetaminophen. Generalized estimating equations were used to estimate the effect of clinical factors on treatment of patients with antipyretic medications, Results: Although 76% of patients in this febrile cohort (n = 241) were prescribed an antipyretic agent, only 42% received antipyretic therapy; 95% of the doses were acetaminophen. Variables associated with antipyretic treatment were maximum body temperature (OR 2.11, 95% CI 1.53 to 2.89), time after sepsis diagnosis (OR 0.88, 95% CI 0.82 to 0.95), surgery during hospitalization (OR 0.49, 95% CI 0.31 to 0.80), death within 36 hours (OR 0.35, 95% CI 0.15 to 0.85). and mechanical ventilation (OR 0.58, 95% CI 0.34 to 0.98). Severity of illness factors, demographic factors, and patient treatment location did not predict who would receive antipyretic therapy. Conclusions: Most febrile episodes in patients with gram-negative severe sepsis or septic shock were not treated with antipyretic medications. Further studies are needed to demonstrate the effect of antipyretics on clinically relevant outcomes in severe sepsis and septic shock.

scholarly journals A Machine Learning Algorithm to Predict Severe Sepsis and Septic Shock

2019 ◽  
Vol 47 (11) ◽  
pp. 1485-1492 ◽  
Author(s):  
Heather M. Giannini ◽  
Jennifer C. Ginestra ◽  
Corey Chivers ◽  
Michael Draugelis ◽  
Asaf Hanish ◽  
...  
Keyword(s):  
Machine Learning ◽  
Septic Shock ◽  
Severe Sepsis ◽  
Learning Algorithm ◽  
Machine Learning Algorithm ◽  
Sepsis And Septic Shock

Effect of fluid bolus triggers and their combination on fluid responsiveness in optimization phase of severe sepsis and septic shock resuscitation

MedPharmRes ◽  
2018 ◽  
Vol 2 (3) ◽  
pp. 27-32
Author(s):  
Bien Le ◽  
Dai Huynh ◽  
Mai Tuan ◽  
Minh Phan ◽  
Thao Pham ◽  
...  
Keyword(s):  
Blood Pressure ◽  
Septic Shock ◽  
Severe Sepsis ◽  
Fluid Responsiveness ◽  
Statistical Significance ◽  
Venous Pressure ◽  
Fluid Bolus ◽  
Low Blood Pressure ◽  
Low Central Venous Pressure ◽  
Sepsis And Septic Shock

Objectives: to evaluate the fluid responsiveness according to fluid bolus triggers and their combination in severe sepsis and septic shock. Design: observational study. Patients and Methods: patients with severe sepsis and septic shock who already received fluid after rescue phase of resuscitation. Fluid bolus (FB) was prescribed upon perceived hypovolemic manifestations: low central venous pressure (CVP), low blood pressure, tachycardia, low urine output (UOP), hyperlactatemia. FB was performed by Ringer lactate 500 ml/30 min and responsiveness was defined by increasing in stroke volume (SV) ≥15%. Results: 84 patients were enrolled, among them 30 responded to FB (35.7%). Demographic and hemodynamic profile before fluid bolus were similar between responders and non-responders, except CVP was lower in responders (7.3 ± 3.4 mmHg vs 9.2 ± 3.6 mmHg) (p 0.018). Fluid response in low CVP, low blood pressure, tachycardia, low UOP, hyperlactatemia were 48.6%, 47.4%, 38.5%, 37.0%, 36.8% making the odd ratio (OR) of these triggers were 2.81 (1.09-7.27), 1.60 (0.54-4.78), 1.89 (0.58-6.18), 1.15 (0.41-3.27) and 1.27 (0.46-3.53) respectively. Although CVP < 8 mmHg had a higher response rate, the association was not consistent at lower cut-offs. The combination of these triggers appeared to raise fluid response but did not reach statistical significance: 26.7% (1 trigger), 31.0% (2 triggers), 35.7% (3 triggers), 55.6% (4 triggers), 100% (5 triggers). Conclusions: fluid responsiveness was low in optimization phase of resuscitation. No fluid bolus trigger was superior to the others in term of providing a higher responsiveness, their combination did not improve fluid responsiveness as well.

Sign in / Sign up

Export Citation Format

Share Document