Risk factors for adverse outcomes during mechanical ventilation of 1152 COVID-19 patients: a multicenter machine learning study with highly granular data from the Dutch Data Warehouse

Lucas M Fleuren,Michael De Neree Tot Babberich,Annemieke Dijkstra,Harald J Faber,Peter Koetsier,Dennis Geutjes,Judith Lens,Fuda Van Diggelen,Arend Jan Meinders,Remko Van Den Akker,Remko De Jong,Klaas Sierk Arnold,Olaf L Cremer,Hans Van Der Spoel,Aletta A De Beer,M C Reuland,H J De Grooth,Ellen G M Smit,Albertus Beishuizen,Roberto Noorduijn Londono,Marco Peters,Dave A Dongelmans,Daan P De Bruin,Lot Wagemakers,Tom Hendriks,Robbert C A Lalisang,Peter Pickkers,Sefanja Achterberg,Sesmu Arbous,Leo Heunks,G J Zijlstra,Barbara Festen-Spanjer,Taco Houwert,Mark Hoogendoorn,Martijn G Scholtemeijer,Tomas Machado,Tim Frenzel,Wouter De Ruijter,Marlijn J A Kamps,David Romero Guzman,Jacomar J M Van Koesveld,Thomas Klausch,Johan Lutisan,Jan Jaap Spijkstra,Sander Rigter,Arjan Slooter,Laura Van Manen,Bart Van De Gaauw,Hilde G A Van Der Pol,Mattia Fornasa,Alexander D Cornet,Michele Tonutti,Dharmanand Ramnarain,Evelien Oostdijk,Arne Jon Van Tienhoven ,Ronald H Driessen ,Frits Van Osch ,Angélique De Man ,Attila Karakus ,Bart Grady ,Nicolas F Schroten ,P De Jong ,Paul Elbers ,Cornelis P.c De Jager ,Thijs C.d Rettig ,Wim Boersma ,Nicole P Juffermans ,J Berend ,D.j Mehagnoul-Schipper ,Giovanni Cinà ,Louise C Urlings‐Strop ,B J Kullberg ,Joost A M Labout ,Evert‐Jan Wils ,Leon J Montenij ,Gert B Brunnekreef ,Patrick Thoral ,N F De Keizer ,Dagmar M Ouweneel ,Bas Van Den Bogaard ,Luca F Roggeveen ,Martijn D De Kruif ,Virginia Ceni Silva,Nardo J M Van Der Meer,Martijn Van Tellingen,Rutger Van Raalte,Sebastiaan J J Vonk,Niels C Gritters Van Den Oever,Pim Van Der Heiden,Stefaan H A Hendriks,Davide Quintarelli,Tariq A Dam,Martijn Beudel,Claudia C W Van Gemeren,Evelien Van Der Heiden,Willem E Herter,A A Rijkeboer,Anisa Hana,Martha De Bruin,Ellen Van Geest,Ali El Hassouni,Diederik Gommers,Bram Simons,Walter Van Den Tempel,Auke C Reidinga,Roy Van Den Berg,Michael Kuiper,Sandjai Bhulai,Heder De Vries,J W Fijen,Hidde Hovenkamp,Emma Rademaker,Ralph Nowitzky,Jasper Van Bommel,Olivier Thijssens,Rob J Bosman,Marcel Aries,Tom Dormans,Fleur G C A Nooteboom,Jan Peppink,Menno Beukema,Armand R J Girbes,Robert Entjes,Koen S Simons,Age D Boelens

doi:10.1186/s40635-021-00397-5

Abstract

BackgroundThe identification of risk factors for adverse outcomes and prolonged intensive care unit (ICU) stay in COVID-19 patients is essential for prognostication, determining treatment intensity, and resource allocation. Previous studies have determined risk factors on admission only, and included a limited number of predictors. Therefore, using data from the highly granular and multicenter Dutch Data Warehouse, we developed machine learning models to identify risk factors for ICU mortality, ventilator-free days and ICU-free days during the course of invasive mechanical ventilation (IMV) in COVID-19 patients.MethodsThe DDW is a growing electronic health record database of critically ill COVID-19 patients in the Netherlands. All adult ICU patients on IMV were eligible for inclusion. Transfers, patients admitted for less than 24 h, and patients still admitted at time of data extraction were excluded. Predictors were selected based on the literature, and included medication dosage and fluid balance. Multiple algorithms were trained and validated on up to three sets of observations per patient on day 1, 7, and 14 using fivefold nested cross-validation, keeping observations from an individual patient in the same split.ResultsA total of 1152 patients were included in the model. XGBoost models performed best for all outcomes and were used to calculate predictor importance. Using Shapley additive explanations (SHAP), age was the most important demographic risk factor for the outcomes upon start of IMV and throughout its course. The relative probability of death across age values is visualized in Partial Dependence Plots (PDPs), with an increase starting at 54 years. Besides age, acidaemia, low P/F-ratios and high driving pressures demonstrated a higher probability of death. The PDP for driving pressure showed a relative probability increase starting at 12 cmH2O.ConclusionAge is the most important demographic risk factor of ICU mortality, ICU-free days and ventilator-free days throughout the course of invasive mechanical ventilation in critically ill COVID-19 patients. pH, P/F ratio, and driving pressure should be monitored closely over the course of mechanical ventilation as risk factors predictive of these outcomes.

Highlights

The identification of risk factors for adverse outcomes and prolonged intensive care unit (ICU) stay in COVID-19 patients is essential for prognostication, determining treatment intensity, and resource allocation
Cohort description A total of 1152 patients were on invasive mechanical ventilation and included in the modeling. 883 of these patients were admitted before the 1st of September 2020 during the first wave in the Netherlands, 269 patients were admitted after this date during the second wave
Mortality during ICU admission occurred in 28.8% of patients that survived at least 24 h on invasive mechanical ventila‐ tion (IMV) and only slightly increased throughout the course of mechanical ventilation; 32.4% of patients that survived up until day 14 on IMV still died afterwards

Summary

Introduction

The identification of risk factors for adverse outcomes and prolonged intensive care unit (ICU) stay in COVID-19 patients is essential for prognostication, determining treatment intensity, and resource allocation. A better understanding of these predictors could aid clinicians in the prognosis of critically ill patients and may aid policy-makers and medical professionals in optimizing resource allocation This is of pivotal importance at the time of possible ICU admission, and throughout the entire course of ICU treatment. ICU-specific models often fail to incorporate the wide variety of dedicated ICU therapies such as mechanical ventilation or high-risk medication. Many of these models are single center and are frequently limited to risk factors at ICU admission, while COVID-19 often requires lengthy intensive care stays. We identified a gap for reproducible, multicenter predictive models in the ICU that include ICU-specific predictors over time

Objectives

Methods

Results

Discussion

Conclusion