TY - JOUR T1 - Supervised Machine Learning for the Early Prediction of Acute Respiratory Distress Syndrome (ARDS) JF - medRxiv DO - 10.1101/2020.03.19.20038364 SP - 2020.03.19.20038364 AU - Sidney Le AU - Emily Pellegrini AU - Abigail Green-Saxena AU - Charlotte Summers AU - Jana Hoffman AU - Jacob Calvert AU - Ritankar Das Y1 - 2020/01/01 UR - http://medrxiv.org/content/early/2020/03/23/2020.03.19.20038364.abstract N2 - 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.Competing Interest StatementSL, EP, AGS, JH, JC, and RD are employees of Dascena.Funding StatementThis research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.Author DeclarationsAll relevant ethical guidelines have been followed; any necessary IRB and/or ethics committee approvals have been obtained and details of the IRB/oversight body are included in the manuscript.YesAll necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived.YesI understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).YesI have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable.YesData are publicly available.AECCAmerican-European Consensus ConferenceAPACHEAcute Physiologic Assessment and Chronic Health EvaluationARDSAcute Respiratory Distress SyndromeAUROCArea Under Receiver Operating CharacteristicCDSClinical Decision SupportDORDiagnostic Odds RatioEDEmergency DepartmentEHRElectronic Health RecordsGCSGlasgow Coma ScaleICUIntensive Care UnitICD-9International Statistical Classification of Diseases version 9INRInternational Normalised RatioLIPSLung Injury Prediction ScoreLR+Positive Likelihood RatioLR-Negative Likelihood RatioMAPMean Arterial PressureMIMIC IIIMedical Information Mart for Intensive Care version IIIMLAMachine Learning AlgorithmPEEPPositive End Expiratory PressureP/F ratioPaO2/FiO2 ratioPPPulse PressureROCReceiver Operating CharacteristicSAPSSimplified Acute Physiology ScoreSOFASequential Organ Failure AssessmentWBCWhite Blood Cell Count ER -