Dynamic multi-outcome prediction after injury: Applying adaptive machine learning for precision medicine in trauma

Objective Machine learning techniques have demonstrated superior discrimination compared to conventional statistical approaches in predicting trauma death. The objective of this study is to evaluate whether machine learning algorithms can be used to assess risk and dynamically identify patient-specific modifiable factors critical to patient trajectory for multiple key outcomes after severe injury. Methods SuperLearner, an ensemble machine-learning algorithm, was applied to prospective observational cohort data from 1494 critically-injured patients. Over 1000 agnostic predictors were used to generate prediction models from multiple candidate learners for outcomes of interest at serial time points post-injury. Model accuracy was estimated using cross-validation and area under the curve was compared to select among predictors. Clinical variables responsible for driving outcomes were estimated at each time point. Results SuperLearner fits demonstrated excellent cross-validated prediction of death (overall AUC 0.94–0.97), multi-organ failure (overall AUC 0.84–0.90), and transfusion (overall AUC 0.87–0.9) across multiple post-injury time points, and good prediction of Acute Respiratory Distress Syndrome (overall AUC 0.84–0.89) and venous thromboembolism (overall AUC 0.73–0.83). Outcomes with inferior data quality included coagulopathic trajectory (AUC 0.48–0.88). Key clinical predictors evolved over the post-injury timecourse and included both anticipated and unexpected variables. Non-random missingness of data was identified as a predictor of multiple outcomes over time. Conclusions Machine learning algorithms can be used to generate dynamic prediction after injury while avoiding the risk of over- and under-fitting inherent in ad hoc statistical approaches. SuperLearner prediction after injury demonstrates promise as an adaptable means of helping clinicians integrate voluminous, evolving data on severely-injured patients into real-time, dynamic decision-making support.

[1]  Trevor Hastie,et al.  Regularization Paths for Generalized Linear Models via Coordinate Descent. , 2010, Journal of statistical software.

[2]  M. Bracken.  CRASH (Corticosteroid Randomization after Significant Head Injury Trial): landmark and storm warning. , 2005, Neurosurgery.

[3]  M. J. van der Laan,et al.  Statistical Applications in Genetics and Molecular Biology Super Learner , 2010 .

[4]  Sharon R. Weeks,et al.  Is the Kampala Trauma Score an Effective Predictor of Mortality in Low-Resource Settings? A Comparison of Multiple Trauma Severity Scores , 2012, World Journal of Surgery.

[5]  M. N. Chawdaa,et al.  Predicting outcome after multiple trauma : which scoring system ? , 2004 .

[6]  Alan Y. Chiang,et al.  Generalized Additive Models: An Introduction With R , 2007, Technometrics.

[7]  B. Jennett,et al.  Assessment of coma and impaired consciousness. A practical scale. , 1974, Lancet.

[8]  R. Goris,et al.  [Revision of the trauma score]. , 1992, Nederlands tijdschrift voor geneeskunde.

[9]  Arthur S Slutsky,et al.  Acute Respiratory Distress Syndrome The Berlin Definition , 2012 .

[10]  S. Bruijns,et al.  The Kampala Trauma Score has poor diagnostic accuracy for most emergency presentations. , 2017, Injury.

[11]  J. Ioannidis,et al.  Comparison of effect sizes associated with biomarkers reported in highly cited individual articles and in subsequent meta-analyses. , 2011, JAMA.

[12]  Mohammad H Rahbar,et al.  Time-dependent prediction and evaluation of variable importance using superlearning in high-dimensional clinical data , 2013, The journal of trauma and acute care surgery.

[13]  M. Simard,et al.  Performance of International Classification of Diseases–based injury severity measures used to predict in-hospital mortality and intensive care admission among traumatic brain-injured patients , 2017, The journal of trauma and acute care surgery.

[14]  M. Soares,et al.  ICU severity of illness scores: APACHE, SAPS and MPM , 2014, Current opinion in critical care.

[15]  L. Breiman.  Random Forests--random Features , 1999 .

[16]  Radford M. Neal.  Pattern Recognition and Machine Learning , 2007, Technometrics.

[17]  Antonio Paulo Nassar,et al.  Caution when using prognostic models: a prospective comparison of 3 recent prognostic models. , 2012, Journal of critical care.

[18]  W. Haddon,et al.  The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. , 1974, The Journal of trauma.

[19]  Foster J. Provost,et al.  Predictive Modeling With Big Data: Is Bigger Really Better? , 2013, Big Data.

[20]  S. Dudoit,et al.  Asymptotics of cross-validated risk estimation in estimator selection and performance assessment , 2005 .

[21]  T Gennarelli,et al.  Progress in characterizing anatomic injury. , 1990, The Journal of trauma.

[22]  Mohammad M. Ghassemi,et al.  A data-driven approach to optimized medication dosing: a focus on heparin , 2014, Intensive Care Medicine.

[23]  Lesly A. Dossett,et al.  Early prediction of massive transfusion in trauma: simple as ABC (assessment of blood consumption)? , 2009, The Journal of trauma.

[24]  B. Cairns,et al.  The Utility of the Kampala Trauma Score as a Triage Tool in a Sub-Saharan African Trauma Cohort , 2014, World Journal of Surgery.

[25]  H. Xiang,et al.  Trauma with Injury Severity Score of 75: Are These Unsurvivable Injuries? , 2015, PloS one.

[26]  S. Katsaragakis,et al.  Comparison of Acute Physiology and Chronic Health Evaluation II (APACHE II) and Simplified Acute Physiology Score II (SAPS II) scoring systems in a single Greek intensive care unit , 2018 .

[27]  M. Cohen,et al.  Protein C depletion early after trauma increases the risk of ventilator-associated pneumonia. , 2009, The Journal of trauma.

[28]  M. J. Laan,et al.  Targeted Learning: Causal Inference for Observational and Experimental Data , 2011 .

[29]  P. Giannoudis,et al.  Predicting outcome after multiple trauma: which scoring system? , 2004, Injury.

[30]  H. Champion,et al.  Organ injury scaling: spleen, liver, and kidney. , 1989, The Journal of trauma.

[31]  T. H. Kyaw,et al.  Multiparameter Intelligent Monitoring in Intensive Care II: A public-access intensive care unit database* , 2011, Critical care medicine.

[32]  E. Draper,et al.  APACHE II: A severity of disease classification system , 1985, Critical care medicine.

[33]  David B Hoyt,et al.  Transfusion of plasma, platelets, and red blood cells in a 1:1:1 vs a 1:1:2 ratio and mortality in patients with severe trauma: the PROPPR randomized clinical trial. , 2015, JAMA.

[34]  Alan Hubbard,et al.  Variable Importance and Prediction Methods for Longitudinal Problems with Missing Variables , 2015, PloS one.

[35]  H. Ishwaran.  Variable importance in binary regression trees and forests , 2007, 0711.2434.

[36]  E. Cook,et al.  Revised trauma scoring system to predict in-hospital mortality in the emergency department: Glasgow Coma Scale, Age, and Systolic Blood Pressure score , 2011, Critical care.

[37]  Leo Anthony Celi,et al.  Dynamic Clinical Data Mining: Search Engine-Based Decision Support , 2014, JMIR medical informatics.

[38]  Goris Rj,et al.  Revision of the trauma score , 1992 .

[39]  M. Cohen,et al.  Early release of high mobility group box nuclear protein 1 after severe trauma in humans: role of injury severity and tissue hypoperfusion , 2009, Critical care.

[40]  Anna Rumshisky,et al.  Unfolding physiological state: mortality modelling in intensive care units , 2014, KDD.

[41]  H. Champion,et al.  Trauma score , 1981, Critical care medicine.

[42]  T. Osler,et al.  Injury Severity Scoring , 1999 .

[43]  W. Copes,et al.  Evaluating trauma care: the TRISS method. Trauma Score and the Injury Severity Score. , 1987, The Journal of trauma.

[44]  A. Sauaia,et al.  VALIDATION OF POSTINJURY MULTIPLE ORGAN FAILURE SCORES , 2009, Shock.