David Castiñeira1,2, Katherine R Schlosser3,4,5, Alon Geva2,3,4, Amir R Rahmani6, Gaston Fiore2, Brian K Walsh3,4,7, Craig D Smallwood3,4, John H Arnold3,4, Mauricio Santillana8. 1. Massachusetts Institute of Technology, Cambridge, Massachusetts. davidcastineira@outlook.com msantill@g.harvard.edu. 2. Computational Health Informatics Program, Boston Children's Hospital, Boston, Massachusetts. 3. Department of Anesthesiology, Critical Care and Pain Medicine, Boston Children's Hospital, Boston, Massachusetts. 4. Department of Anaesthesia, Harvard Medical School, Boston, Massachusetts. 5. Department of Pediatrics, Division of Pediatric Critical Care, Columbia University Irving Medical Center, New York, New York. 6. Data Science Institute, Columbia University at the time the research was conducted. 7. Department of Allied Health Professions, School of Health Sciences, Liberty University, Lynchburg, Virginia. 8. Department of Pediatrics, Harvard Medical School, Boston, Massachusetts. davidcastineira@outlook.com msantill@g.harvard.edu.
Abstract
BACKGROUND: Bedside monitors in the ICU routinely measure and collect patients' physiologic data in real time to continuously assess the health status of patients who are critically ill. With the advent of increased computational power and the ability to store and rapidly process big data sets in recent years, these physiologic data show promise in identifying specific outcomes and/or events during patients' ICU hospitalization. METHODS: We introduced a methodology designed to automatically extract information from continuous-in-time vital sign data collected from bedside monitors to predict if a patient will experience a prolonged stay (length of stay) on mechanical ventilation, defined as >4 d, in a pediatric ICU. RESULTS: Continuous-in-time vital signs information and clinical history data were retrospectively collected for 284 ICU subjects from their first 24 h on mechanical ventilation from a medical-surgical pediatric ICU at Boston Children's Hospital. Multiple machine learning models were trained on multiple subsets of these subjects to predict the likelihood that each of these subjects would experience a long stay. We evaluated the predictive power of our models strictly on unseen hold-out validation sets of subjects. Our methodology achieved model performance of >83% (area under the curve) by using only vital sign information as input, and performances of 90% (area under the curve) by combining vital sign information with subjects' static clinical data readily available in electronic health records. We implemented this approach on 300 independently trained experiments with different choices of training and hold-out validation sets to ensure the consistency and robustness of our results in our study sample. The predictive power of our approach outperformed recent efforts that used deep learning to predict a similar task. CONCLUSIONS: Our proposed workflow may prove useful in the design of scalable approaches for real-time predictive systems in ICU environments, exploiting real-time vital sign information from bedside monitors. (ClinicalTrials.gov registration NCT02184208.).
BACKGROUND: Bedside monitors in the ICU routinely measure and collect patients' physiologic data in real time to continuously assess the health status of patients who are critically ill. With the advent of increased computational power and the ability to store and rapidly process big data sets in recent years, these physiologic data show promise in identifying specific outcomes and/or events during patients' ICU hospitalization. METHODS: We introduced a methodology designed to automatically extract information from continuous-in-time vital sign data collected from bedside monitors to predict if a patient will experience a prolonged stay (length of stay) on mechanical ventilation, defined as >4 d, in a pediatric ICU. RESULTS: Continuous-in-time vital signs information and clinical history data were retrospectively collected for 284 ICU subjects from their first 24 h on mechanical ventilation from a medical-surgical pediatric ICU at Boston Children's Hospital. Multiple machine learning models were trained on multiple subsets of these subjects to predict the likelihood that each of these subjects would experience a long stay. We evaluated the predictive power of our models strictly on unseen hold-out validation sets of subjects. Our methodology achieved model performance of >83% (area under the curve) by using only vital sign information as input, and performances of 90% (area under the curve) by combining vital sign information with subjects' static clinical data readily available in electronic health records. We implemented this approach on 300 independently trained experiments with different choices of training and hold-out validation sets to ensure the consistency and robustness of our results in our study sample. The predictive power of our approach outperformed recent efforts that used deep learning to predict a similar task. CONCLUSIONS: Our proposed workflow may prove useful in the design of scalable approaches for real-time predictive systems in ICU environments, exploiting real-time vital sign information from bedside monitors. (ClinicalTrials.gov registration NCT02184208.).
Keywords:
big data in medicine; biomedical and health data science; clinical decision making; critical care; data driven machine learning; decision support systems; intensive care; length of stay; length of stay estimation; machine learning; mechanical ventilation; pediatrics; precision medicine; prediction; predictive analytics
Authors: David A Gruenberg; Wayne Shelton; Susannah L Rose; Ann E Rutter; Sophia Socaris; Glenn McGee Journal: Am J Crit Care Date: 2006-09 Impact factor: 2.228
Authors: Scott R Levin; Eric T Harley; James C Fackler; Christoph U Lehmann; Jason W Custer; Daniel France; Scott L Zeger Journal: Crit Care Med Date: 2012-11 Impact factor: 7.598
Authors: Brian J Douthit; Rachel L Walden; Kenrick Cato; Cynthia P Coviak; Christopher Cruz; Fabio D'Agostino; Thompson Forbes; Grace Gao; Theresa A Kapetanovic; Mikyoung A Lee; Lisiane Pruinelli; Mary A Schultz; Ann Wieben; Alvin D Jeffery Journal: Appl Clin Inform Date: 2022-02-09 Impact factor: 2.342