The Experts below are selected from a list of 4413 Experts worldwide ranked by ideXlab platform

Leo Anthony Celi - One of the best experts on this subject based on the ideXlab platform.

  • Evaluating Progress on Machine Learning for Longitudinal Electronic Healthcare Data
    arXiv: Learning, 2020
    Co-Authors: David Bellamy, Leo Anthony Celi, Andrew L. Beam
    Abstract:

    The Large Scale Visual Recognition Challenge based on the well-known Imagenet dataset catalyzed an intense flurry of progress in computer vision. Benchmark tasks have propelled other sub-fields of machine learning forward at an equally impressive pace, but in healthcare it has primarily been image processing tasks, such as in dermatology and radiology, that have experienced similar benchmark-driven progress. In the present study, we performed a comprehensive review of benchmarks in medical machine learning for structured data, identifying one based on the Medical Information Mart for Intensive Care (MIMIC-III) that allows the first direct comparison of predictive performance and thus the evaluation of progress on four clinical prediction tasks: mortality, length of stay, phenotyping, and patient decompensation. We find that little meaningful progress has been made over a 3 year period on these tasks, despite significant community engagement. Through our meta-analysis, we find that the performance of deep recurrent models is only superior to logistic regression on certain tasks. We conclude with a synthesis of these results, possible explanations, and a list of desirable qualities for future benchmarks in medical machine learning.

  • Real-world characterization of blood glucose control and insulin use in the intensive care unit
    Scientific Reports, 2020
    Co-Authors: Lawrence Baker, Jason H. Maley, Aldo Arévalo, Francis Demichele, Roselyn Mateo-collado, Stan Finkelstein, Leo Anthony Celi
    Abstract:

    The heterogeneity of critical illness complicates both clinical trial design and real-world management. This complexity has resulted in conflicting evidence and opinion regarding the optimal management in many intensive care scenarios. Understanding this heterogeneity is essential to tailoring management to individual patients. Hyperglycaemia is one such complication in the intensive care unit (ICU), accompanied by decades of conflicting evidence around management strategies. We hypothesized that analysis of highly-detailed electronic medical record (EMR) data would demonstrate that patients vary widely in their glycaemic response to critical illness and response to insulin therapy. Due to this variability, we believed that hyper- and hypoglycaemia would remain common in ICU care despite standardised approaches to management. We utilized the Medical Information Mart for Intensive Care III v1.4 (MIMIC) database. We identified 19,694 admissions between 2008 and 2012 with available glucose results and insulin administration data. We demonstrate that hyper- and hypoglycaemia are common at the time of admission and remain so 1 week into an ICU admission. Insulin treatment strategies vary significantly, irrespective of blood glucose level or diabetic status. We reveal a tremendous opportunity for EMR data to guide tailored management. Through this work, we have made available a highly-detailed data source for future investigation.

  • A machine learning-based model for 1-year mortality prediction in patients admitted to an Intensive Care Unit with a diagnosis of sepsis.
    Medicina Intensiva, 2020
    Co-Authors: Javier E. García-gallo, Leo Anthony Celi, N.j. Fonseca-ruiz, John F. Duitama-muñoz
    Abstract:

    Abstract Introduction Sepsis is associated to a high mortality rate, and its severity must be evaluated quickly. The severity of illness scores used are intended to be applicable to all patient populations, and generally evaluate in-hospital mortality. However, patients with sepsis continue to be at risk of death after hospital discharge. Objective To develop a model for predicting 1-year mortality in critical patients diagnosed with sepsis. Patients The data corresponding to 5650 admissions of patients with sepsis from the Medical Information Mart for Intensive Care (MIMIC-III) database were evaluated, randomly divided as follows: 70% for training and 30% for validation. Design A retrospective register-based cohort study was carried out. The clinical Information of the first 24 h after admission was used to develop a 1-year mortality prediction model based on Stochastic Gradient Boosting (SGB) methodology. Variable selection was addressed using Least Absolute Shrinkage and Selection Operator (LASSO) and SGB variable importance methodologies. The predictive power was evaluated using the area under the ROC curve (AUROC). Results An AUROC of 0.8039 (95% confidence interval (CI): [0.8033 0.8045]) was obtained in the validation subset. The model exceeded the predictive performances obtained with traditional severity of disease scores in the same subset. Conclusion The use of assembly algorithms, such as SGB, for the generation of a customized model for sepsis yields more accurate 1-year mortality prediction than the traditional scoring systems such as SAPS II, SOFA or OASIS.

  • BIBE - One-Year Mortality Prediction in ICU Patients with Diagnosis of Sepsis Driven by Population Similarities
    2019 IEEE 19th International Conference on Bioinformatics and Bioengineering (BIBE), 2019
    Co-Authors: Javier E. García-gallo, Leo Anthony Celi, N.j. Fonseca-ruiz, John F. Duitama-muñoz
    Abstract:

    Traditional Intensive Care Unit (ICU) outcome prediction models are based on the analysis of large populations, and often provide statistically rigorous results for an average patient but are also expensive, time-consuming, and prone to selection bias; moreover, these indicators usually lack the precision required for use at the individual level, since they present significant errors at patient data away from the average. In this work we present an evaluation of the impact and relevance of three different patient similarity metrics on the one-year mortality prediction when patients are related by the same diagnosis, sepsis. We use data of 16.000 admissions of patients with sepsis from the Medical Information Mart for Intensive Care (MIMIC-III) database.

  • The MIMIC Code Repository: enabling reproducibility in critical care research.
    Journal of the American Medical Informatics Association, 2017
    Co-Authors: Alistair E W Johnson, Leo Anthony Celi, David J. Stone, Tom J Pollard
    Abstract:

    Objective Lack of reproducibility in medical studies is a barrier to the generation of a robust knowledge base to support clinical decision-making. In this paper we outline the Medical Information Mart for Intensive Care (MIMIC) Code Repository, a centralized code base for generating reproducible studies on an openly available critical care dataset. Materials and Methods Code is provided to load the data into a relational structure, create extractions of the data, and reproduce entire analysis plans including research studies. Results Concepts extracted include severity of illness scores, comorbid status, administrative definitions of sepsis, physiologic criteria for sepsis, organ failure scores, treatment administration, and more. Executable documents are used for tutorials and reproduce published studies end-to-end, providing a template for future researchers to replicate. The repository's issue tracker enables community discussion about the data and concepts, allowing users to collaboratively improve the resource. Discussion The centralized repository provides a platform for users of the data to interact directly with the data generators, facilitating greater understanding of the data. It also provides a location for the community to collaborate on necessary concepts for research progress and share them with a larger audience. Consistent application of the same code for underlying concepts is a key step in ensuring that research studies on the MIMIC database are comparable and reproducible. Conclusion By providing open source code alongside the freely accessible MIMIC-III database, we enable end-to-end reproducible analysis of electronic health records.

James A. Russell - One of the best experts on this subject based on the ideXlab platform.

  • Mean arterial pressure and mortality in patients with distributive shock: a retrospective analysis of the MIMIC-III database
    Annals of Intensive Care, 2018
    Co-Authors: Jean-louis Vincent, Nathan D. Nielsen, Nathan I. Shapiro, Margaret E. Gerbasi, Aaron Grossman, Robin Doroff, Feng Zeng, Paul J. Young, James A. Russell
    Abstract:

    Background Maintenance of mean arterial pressure (MAP) at levels sufficient to avoid tissue hypoperfusion is a key tenet in the management of distributive shock. We hypothesized that patients with distributive shock sometimes have a MAP below that typically recommended and that such hypotension is associated with increased mortality. Methods In this retrospective analysis of the Medical Information Mart for Intensive Care (MIMIC-III) database from Beth Israel Deaconess Medical Center, Boston, USA, we included all intensive care unit (ICU) admissions between 2001 and 2012 with distributive shock, defined as continuous vasopressor support for ≥ 6 h and no evidence of low cardiac output shock. Hypotension was evaluated using five MAP thresholds: 80, 75, 65, 60 and 55 mmHg. We evaluated the longest continuous episode below each threshold during vasopressor therapy. The primary outcome was ICU mortality. Results Of 5347 patients with distributive shock, 95.7%, 91.0%, 62.0%, 36.0% and 17.2%, respectively, had MAP 

  • Mean arterial pressure and mortality in patients with distributive shock: a retrospective analysis of the MIMIC-III database
    Annals of Intensive Care, 2018
    Co-Authors: Jean-louis Vincent, Nathan D. Nielsen, Nathan I. Shapiro, Margaret E. Gerbasi, Aaron Grossman, Robin Doroff, Feng Zeng, Paul M. Young, James A. Russell
    Abstract:

    Maintenance of mean arterial pressure (MAP) at levels sufficient to avoid tissue hypoperfusion is a key tenet in the management of distributive shock. We hypothesized that patients with distributive shock sometimes have a MAP below that typically recommended and that such hypotension is associated with increased mortality. In this retrospective analysis of the Medical Information Mart for Intensive Care (MIMIC-III) database from Beth Israel Deaconess Medical Center, Boston, USA, we included all intensive care unit (ICU) admissions between 2001 and 2012 with distributive shock, defined as continuous vasopressor support for ≥ 6 h and no evidence of low cardiac output shock. Hypotension was evaluated using five MAP thresholds: 80, 75, 65, 60 and 55 mmHg. We evaluated the longest continuous episode below each threshold during vasopressor therapy. The primary outcome was ICU mortality. Of 5347 patients with distributive shock, 95.7%, 91.0%, 62.0%, 36.0% and 17.2%, respectively, had MAP   0 to  2 h. Episodes of prolonged hypotension were associated with higher mortality.

Jingye Pan - One of the best experts on this subject based on the ideXlab platform.

  • Elevated serum iron level is a predictor of prognosis in ICU patients with acute kidney injury.
    BMC Nephrology, 2020
    Co-Authors: Jie Shu, Jia-xiu Chen, Jingye Pan
    Abstract:

    Accumulation of iron is associated with oxidative stress, inflammation, and regulated cell death processes that contribute to the development of acute kidney injury (AKI). We aimed to investigate the association between serum iron levels and prognosis in intensive care unit (ICU) patients with AKI. A total of 483 patients with AKI defined as per the Kidney Disease: Improving Global Guidelines were included in this retrospective study. The data was extracted from the single-centre Medical Information Mart for Intensive Care III database. AKI patients with serum iron parameters measured upon ICU admission were included and divided into two groups (low group and high group). The prognostic value of serum iron was analysed using univariate and multivariate Cox regression analysis. The optimal cut-off value for serum iron was calculated to be 60 μg/dl. Univariable Cox regression analysis showed that serum iron levels were significantly correlated with prognosis of AKI patients. After adjusting for possible confounding variables, serum iron levels higher than 60 μg/dl were associated with increases in 28-day (hazard [HR] 1.832; P 

  • Elevated serum iron level is a predictor of prognosis in severe patients with acute kidney injury
    2020
    Co-Authors: Jie Shu, Jia-xiu Chen, Jingye Pan
    Abstract:

    Abstract Background: Accumulation of iron is associated with oxidative stress (OS), inflammation and regulated cell death. The above three reactions contribute to the development of acute kidney injury (AKI). Here we aimed to investigate the association between the serum iron level and prognosis in severe patients with AKI.Methods: A total of 483 patients with AKI defined by Kidney Disease: Improving Global Guidelines (KIDGO) were included in this retrospective study. The data was extracted from the single-center Medical Information Mart for Intensive Care Ⅲ (MIMIC-Ⅲ) database. The max serum iron concentration measured after Intensive Care Unit (ICU) admission was defined as the serum iron in the study and divided into three groups (Low group, Middle group, High group). We plotted boxplots and Kaplan–Meier curves and used cox regression analysis to analyze data.Results: In univariable Cox regression analysis, serum iron levels were significantly correlated to the prognosis of AKI patients. After adjusting for confounding variables, higher serum iron level was remained to associate with the increase in 90-day mortality in the multivariable Cox regression analysis. Moreover, the risk of 90-day mortality stepwise increased as the groups of serum iron levels increased in AKI patients.Conclusions: From our study, we investigated that high serum iron level was associated with the increased mortality in severe patients with AKI. Serum iron levels on admission can be a predictor for predicting the prognosis of AKI patients.

  • Red blood cell distribution width is associated with mortality risk in patients with acute respiratory distress syndrome based on the Berlin definition: A propensity score matched cohort study
    Heart & Lung, 2020
    Co-Authors: Zhi-qiang Chen, Jie Shu, Jia-xiu Chen, Jingye Pan
    Abstract:

    Abstract Background Acute respiratory distress syndrome (ARDS) is a severe inflammatory disorder of the lungs and is associated with oxidative damage. However, red blood cell distribution width (RDW), as an indicator of body response to inflammation and oxidative stress, has not been studied for its relationship with ARDS as diagnosed by the Berlin definition. Objectives To examine the value of RDW in predicting the prognosis of in patients with ARDS. Methods This is a retrospective study based on the Medical Information Mart for Intensive Care III (MIMIC-III) database. Berlin-defined ARDS patients using mechanical ventilation for more than 48 hours were selected using structured query language. The primary statistical methods were propensity score matching and sensitivity analysis, including an inverse probability weighting model to ensure the robustness of our findings. Results A total of 529 intensive care unit (ICU) patients with ARDS according to the Berlin definition were enrolled in the study. The adjusted OR showed an adverse effect between the higher RDW group and 30-day mortality [OR 2.33, 95% CI (1.15–4.75), P=0.019]. However, we found that length of ICU stay was not related to RDW (P=0.167), and in the anaemia group, RDW was poorly predictive of 30-day mortality (P=0.307). Conclusion In unselected ARDS patients, higher RDW was associated with higher 30-day mortality rate. Further investigation is required to validate this relationship with prospectively collected data.

Michelle S. Chew - One of the best experts on this subject based on the ideXlab platform.

  • LiSep LSTM: A Machine Learning Algorithm for Early Detection of Septic Shock
    Scientific Reports, 2019
    Co-Authors: Josef Fagerström, Magnus Bång, Daniel Wilhelms, Michelle S. Chew
    Abstract:

    Sepsis is a major health concern with global estimates of 31.5 million cases per year. Case fatality rates are still unacceptably high, and early detection and treatment is vital since it significantly reduces mortality rates for this condition. Appropriately designed automated detection tools have the potential to reduce the morbidity and mortality of sepsis by providing early and accurate identification of patients who are at risk of developing sepsis. In this paper, we present “LiSep LSTM”; a Long Short-Term Memory neural network designed for early identification of septic shock. LSTM networks are typically well-suited for detecting long-term dependencies in time series data. LiSep LSTM was developed using the machine learning framework Keras with a Google TensorFlow back end. The model was trained with data from the Medical Information Mart for Intensive Care database which contains vital signs, laboratory data, and journal entries from approximately 59,000 ICU patients. We show that LiSep LSTM can outperform a less complex model, using the same features and targets, with an AUROC 0.8306 (95% confidence interval: 0.8236, 0.8376) and median offsets between prediction and septic shock onset up to 40 hours (interquartile range, 20 to 135 hours). Moreover, we discuss how our classifier performs at specific offsets before septic shock onset, and compare it with five state-of-the-art machine learning algorithms for early detection of sepsis.

Ritankar Das - One of the best experts on this subject based on the ideXlab platform.

  • Supervised Machine Learning for the Early Prediction of Acute Respiratory Distress Syndrome (ARDS)
    2020
    Co-Authors: Emily Pellegrini, Abigail Green-saxena, Charlotte Summers, Jana Hoffman, Jacob Calvert, Ritankar Das
    Abstract:

    Purpose: Acute respiratory distress syndrome (ARDS) is a serious respiratory condition with high mortality and associated morbidity. The objective of this study is to develop and evaluate a novel application of gradient boosted tree models trained on patient health record data for the early prediction of ARDS. Materials and Methods: 9919 patient encounters were retrospectively analyzed from the Medical Information Mart for Intensive Care III (MIMIC-III) data base. XGBoost gradient boosted tree models for early ARDS prediction were created using routinely collected clinical variables and numerical representations of radiology reports as inputs. XGBoost models were iteratively trained and validated using 10-fold cross validation. Results: On a hold-out test set, algorithm classifiers attained area under the receiver operating characteristic curve (AUROC) values of 0.905, 0.827, 0.810, and 0.790 when tested for the prediction of ARDS at 0-, 12-, 24-, and 48-hour windows prior to onset, respectively. Conclusion: Supervised machine learning predictions may help predict patients with ARDS up to 48 hours prior to onset.

  • Supervised machine learning for the early prediction of acute respiratory distress syndrome (ARDS).
    Journal of Critical Care, 1
    Co-Authors: Emily Pellegrini, Abigail Green-saxena, Charlotte Summers, Jana Hoffman, Jacob Calvert, Ritankar Das
    Abstract:

    Abstract Purpose Acute respiratory distress syndrome (ARDS) is a serious respiratory condition with high mortality and associated morbidity. The objective of this study is to develop and evaluate a novel application of gradient boosted tree models trained on patient health record data for the early prediction of ARDS. Materials and methods 9919 patient encounters were retrospectively analyzed from the Medical Information Mart for Intensive Care III (MIMIC-III) data base. XGBoost gradient boosted tree models for early ARDS prediction were created using routinely collected clinical variables and numerical representations of radiology reports as inputs. XGBoost models were iteratively trained and validated using 10-fold cross validation. Results On a hold-out test set, algorithm classifiers attained area under the receiver operating characteristic curve (AUROC) values of 0.905 when tested for the detection of ARDS at onset and 0.827, 0.810, and 0.790 for the prediction of ARDS at 12-, 24-, and 48-h windows prior to onset, respectively. Conclusion Supervised machine learning predictions may help predict patients with ARDS up to 48 h prior to onset.