Background
The COVID-19 pandemic hit Germany in spring 2020 and since then intensive care resources were heavily utilized up to now [
1]. Although large numbers of SARS-CoV-2 patients required intensive care unit (ICU) admission, ICU capacity in Germany was not exceeded. However, risk stratification and prediction of outcomes continues to be challenging. Several investigators have reported their ICU COVID-19 experience during this time period, yet these data show great variability in the number of cases and outcomes reported [
2‐
16].
Few of these reports attempted to identify risk factors predicting morbidity, mortality and overall clinical outcome. This may be the result of the reporting of (1) incomplete data sets earlier in the pandemic as many patient were still undergoing ICU care for SARS-CoV-2 infection [
10,
13,
15,
16], and/or (2) data sets biased by the need to triage ICU care to patients in the face of the exhaustion of local/regional ICU capacity [
7,
10,
14,
15]. Nonetheless, there was consensus that SARS-CoV-2 ICU patients experienced lengthy ICU stays with ICU mortality in the range of 25 to 41% [
14,
17]. Classical statistical analysis identified risk factors in these patient populations including age, renal function, the degree of pulmonary compromise and severity of acute respiratory distress syndrome (ARDS). But standard statistical techniques are limited in their ability to integrate diverse data types such as past medical history, therapeutic ICU interventions and many more in relation to clinical outcome variables [
18].
To overcome these limitations, we employed machine learning methods to optimize risk stratification and prediction of overall outcomes for individual COVID-19 ICU patients. It has been recently shown that machine learning (ML) algorithms in combination with numerous, multidimensional variables with non-linear relationships may have advantages in clinical outcome prediction. Machine learning strategies were found to be superior to classical methods of outcome prediction typically used in cardiovascular pathologies [
18,
19]. To take advantage of this superior technique for outcome prediction, we investigated 1186 PCR-confirmed COVID-19 patients receiving ICU care at 27 German hospitals that were enrolled retrospectively and prospectively. The aim of this study is to investigate whether ML can provide additional and interpretable insights for outcome prediction and weigh the identified outcome factors in COVID-19 ICU patients.
Discussion
In this multi-center retrospective—prospective cohort study we identified and weighed possible predictive factors on COVID-19 outcome using a machine learning approach on 49 variables. Using the present ML approach, we confirmed previously reported factors and extend knowledge to novel factors and factor combinations likely predicting outcome in COVID-19 patients. Shape functions for each of these variables show the individual influence of the variable for the prediction of the outcome. For ICU survival these include age, platelet/neutrophil ratio, D-dimers, and ARDS severity. The most important factors for the prediction of RRT need include the combination of Age and D-Dimers, Creatinine levels and SOFA score without GCS.
Previous studies have shown that older age, obesity, diabetes, being immunocompromised, lower P
aO
2/F
iO
2, higher hemodynamic and renal SOFA score at ICU admission were independently associated with 90-day mortality in COVID-19 [
14]. This has also been reported by other investigators, yet they did not show individual cutoff values nor weigh the individual importance for the identified factors [
30,
31]. To exclude an early effect or a late effect as seen when logistic regression is performed, we included almost all admission variables collected for our cohort. Variable selection influencing outcome can be performed in ML models but is less crucial than for logistic regression. We refrained from such a variable selection in our EBM model’s decision process. In our analysis we were able to confirm that age and pulmonary function on admission are important predictors in COVID-19 ICU patients. The present shape functions clearly show a non-linear association between the predictive factors and the outcome variable. Patient’s age, for instance, as the most important predictive factor, shows a higher chance for ICU survival below 61 years. Additionally, the ML approach identified the D-dimer level and platelet/neutrophil ratio at ICU admission as important factors. This is especially interesting in the context of reported thrombotic complications of COVID-19 patients [
32,
33]. When activated, neutrophils complex with platelets to form platelet-neutrophil complexes (PNCs) activating both cell types. These PNCs enhance inflammation, increases neutrophil extracellular trap formation, and result in micro-thrombosis [
34,
35]. The same is applicable when looking at D-dimer levels. High D-dimer levels reflect an activation of inflammation and the formation of micro-thrombi with neutrophil extracellular trap formation. We can therefore say that our data reflects the inflammatory markers known from translational science and confirm their relevance to outcome [
35].
In everyday clinical practice, it is of great interest to assess the further course of patients in intensive care, such as a necessity for renal replacement or ECMO therapy. The present ML model predicting the need for ECMO therapy identified age and pulmonary compromise (Murray lung injury score) as important factors. Admission both from an external hospital and already in an intubated state are associated with the need for ECMO therapy. This result is not surprising, as both younger and more severely pulmonary compromised patients were typically transferred for ECMO therapy to our participating centers [
36]. Our ML models assessing the need for RRT include age as an important factor as well as variables quantifying disease severity (SOFA score) or inflammatory and thrombotic activity (D-dimers and Platelet/neutrophil ratio). Our models do not only permit the identification of risk factors in COVID-19 patients, they also provide insights to the weight of each individual variable for the selected ICU outcome of the individual patient [
18,
37]. The ML models chosen allow for transparent assessment of various variables in a non-linear fashion which overcomes limitations of currently employed regression models. The use of shape functions in GAMs for each variable allows for complex relationships (even non-linear) between the variable and the outcome prediction. Therefore, EBMs can be significantly more accurate than simple linear models [
27]. Interactions of different variables extend the analyzing capabilities of the ML approach. Overall, the results from the EBM offer a greater degree of interpretability than a p-value of a linear regression, or an odds ratio analysis. As shown in Figs.
2 and
3 the visualizations offer insight into transition values from positive to negative impact, plateaus, as well as confidence intervals as a certainty measure.
A limitation of the present study is that we were not able to include even more patients into the analysis. This is of course a valid point of criticism, yet the data used for our analyses were manually collected and curated. The data was not simply exported from an electronic medical record where missing data are prevalent and validity of the information has not been confirmed. Missing data often needs to be imputed prior to analysis. As a result of the design of our study, we were largely able to reduce imputation of missing data, again adding to the significance of our findings. The predictiveness of the models presented here differed for the three outcomes (survival, ECMO, RRT). This is likely due to the underlying dataset containing more information for predicting e.g. survival compared to ECMO. Since the study was designed with a focus on predicting survival, some variables which might better predict ECMO or RRT might not have been included in this study (for details see Additional file
1: Table E9). Furthermore, whereas the validation of survival prediction was largely consistent between the retrospective and prospective datasets, there was more variability with regard to ECMO and RRT. A possible reason for this might be structural differences between the retro- and prospective datasets, e.g. changes in treatment or age cohort over time. However, the moderate predictive capabilities of the variables used in these ML models leave open the opportunity to add further, even translational technologies for risk prediction in future. A strength of our approach is the ability to determine a weight for individual patient factors with respect to an individual prediction. Additionally, risk factors are presented with a shape function. This allows for a more detailed interpretation and segmentation of risk factors than a simple linear incrementation, as it is the case for the linear regression. Finally, due to the imbalanced dataset (more patients survived ICU therapy, more patients did not need ECMO or RRT), our model is more reliable for predicting “survival” than “mortality”. Nonetheless, the strength of these clinical data is the generalizability across institutions and even other similarly resourced countries.
Acknowledgements
The authors and collaborators thank the employees of the participating intensive care units for their immeasurable efforts in the care of COVID-19 patients. The authors thank Nico Pfeifer, Medical Informatics, University of Tübingen, Germany, for the numerous fruitful discussions and valuable input on the application of ML methods. The authors thank Sascha Rehm and Christian Erhardt (meDIC, University Hospital of Tübingen, Germany) for server and REDCap administration and support. The authors also greatly appreciate the contributions of the assistant scientists, data scientists, study nurses and physicians towards all aspects of data collection, data analysis and the general support of this study, namely: H. Gloeckner & M. Keim (Department of Anaesthesiology and Intensive Care Medicine, University Hospital Tübingen, Eberhard-Karls-University Tübingen, Tübingen, Germany), T. Simon (Department of Intensive Care Medicine, University Hospital RWTH Aachen, Aachen, Germany), S. Winands & N. Andrees & G. Vorderwülbecke & K. Steinecke (Department of Anesthesiology and Operative Intensive Care Medicine (CCM, CVK), Charité—Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Germany), K. Willemsen & B. Rahmel (Department of Anesthesiology/Intensive Care Medicine/Pain Therapy, Knappschaftskrankenhaus Bochum, Bochum, Germany), N. Theuerkauf & C. Bode & C.J. Schewe & U. Lohmer (Department of Anaesthesiology and Intensive Care Medicine, University Hospital Bonn, Bonn, Germany), S. May & P. Spieth & A. Gueldner (Department of Anesthesiology and Intensive Care Medicine, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany), T. Dimski & C.Pelletier (Department of Anaesthesiology, University Hospital Düsseldorf, Düsseldorf, Germany), A. Skarabis & M. Overbeck & B. Breuing & S. Kattner (Department of Anesthesiology and Intensive Care Medicine, University Hospital Essen, Essen, Germany), T. Bui (Department of Anaesthesiology, Intensive Care Medicine and Pain Therapy, University Hospital Frankfurt, Goethe University, Frankfurt, Germany), J. Kalbhenn (Department of Anesthesiology and Critical Care, Medical Center—University of Freiburg, Freiburg, Germany), A. Kernchen (Center for Anesthesiology, Emergency and Intensive Care Medicine, University of Göttingen, Göttingen, Germany), N. Piasta & M. Gerber (Department of Anesthesiology, University Medicine Greifswald, Greifswald, Germany), K. Ludwig & J. Mühlhaus (Department Cardiology, Angiology and Intensive Care Medicine, University Hospital Halle (Saale), Halle (Saale), Germany), M. Fiedler (Department of Anesthesiology, Heidelberg University Hospital, Heidelberg, Germany), C.Schlegel & A. Seidel (Department of Anesthesiology and Intensive Care, Leipzig University Hospital, Leipzig, Germany), H. Linnen & G. Bergmann (Department of Anesthesiology and Intensive Care, University Medical Center Schleswig-Holstein, Campus Lübeck and University of Lübeck, Lübeck, Germany), L. Bartning & K. Albizky (Department of Anaesthesiology and Intensive Care Therapy, Otto-von-Guericke-University Magdeburg, Magdeburg, Germany), C. Arndt & T. Koch & R. Axt & G. Kix ( University Hospital Marburg, UKGM, Philipps University Marburg, Marburg, Germany), U. Flechtner (Department of Anesthesiology, Intensive Care Medicine and Pain Therapy, University Hospital Giessen and Marburg, and Justus-Liebig University Giessen, Giessen, Germany), N. Moßbacher & S. Maluche & P. Feddersen (Departemnt of Anesthesiology and Intensive Care, Technical University of Munich, School of Medicine, Klinikum rechts der Isar, Munich, Germany), S. Kaiser (Department for Anesthesiology, Intensive Care Medicine, Emergency Medicine and Pain Medicine, St. Elisabethen Klinikum, Ravensburg, Germany), K. Meiers (Department of Anesthesiology, Intensive Care Medicine and Pain Medicine, Saarland University Hospital Medical Center, Homburg/Saar, Germany), F. Branz & S. Hoheisen & S. Appel (Department of Anaesthesiology and Intensive Care Medicine, Ulm University, Ulm, Germany), J.M. Defosse & F. Strobel (Department of Anaesthesiology and Intensive Care Medicine, Cologne-Merheim Medical Centre, Witten/Herdecke University, Germany), P. Kranke & C. Lotz (Department of Anaesthesiology, Intensive Care, Emergency and Pain Medicine, University Hospital Wuerzburg, University Wuerzburg, Wuerzburg, Germany).
Declarations
Competing interests
Harry Magunia: HM received a speaker’s honorarium from CSL Behring (Germany) outside the submitted work. Simone Lederer: SL has no competing interests to report. Raphael Verbuecheln: RV has no competing interests to report. Bryant Joseph Gilot: BJG has no competing interests to report. Michael Koeppen: MK has no competing interests to report. Helene A Haeberle: HH has no competing interests to report. Valbona Mirakaj: VM has no competing interests to report. Pascal Hoffmann: PH has no competing interests to report. Gernot Marx: GM reports personal fees from Philips Health Care, personal fees from B. Braun, during the conduct of the study; and GM is co-founder of Clinomics GmbH, Germany. Johannes Bickenbach: JB reports personal fees from Biotest AG, Germany, during the conduct of the study. Boris Nohe: BN has no competing interests to report. Michael Lay: ML has no competing interests to report. Claudia Spies: CS reports no conflicts of interests within the submitted manuscript, she reports grants from outside the submitted manuscript from Drägerwerk AG & Co. KGaA, grants from Deutsche Forschungsgemeinschaft / German Research Society, grants from Deutsches Zentrum für Luft- und Raumfahrt e. V. (DLR) /German Aerospace Center, grants from Einstein Stiftung Berlin/ Einstein Foundation Berlin, grants from Gemeinsamer Bundesausschuss / Federal Joint Committee (G-BA), grants from Inneruniversitäre Forschungsförderung / Inner University Grants, grants from Projektträger im DLR / Project Management Agency, grants from Stifterverband/Non-Profit Society Promoting Science and Education, grants from WHOCC, grants from Baxter Deutschland GmbH, grants from Cytosorbents Europe GmbH, grants from Edwards Lifesciences Germany GmbH, grants from Fresenius Medical Care, grants from Grünenthal GmbH, grants from Masimo Europe Ltd., grants from Pfizer Pharma PFE GmbH, personal fees from Georg Thieme Verlag, grants from Dr. F. Köhler Chemie GmbH, grants from Sintetica GmbH, grants from Stifterverband für die deutsche Wissenschaft e.V. / Philips, grants from Stiftung Charité, grants from AGUETTANT Deutschland GmbH, grants from AbbVie Deutschland GmbH & Co. KG, grants from Amomed Pharma GmbH, grants from InTouch Health, grants from Copra System GmbH, grants from Correvio GmbH, grants from Max-Planck-Gesellschaft zur Förderung der Wissenschaften e.V., grants from Deutsche Gesellschaft für Anästhesiologie & Intensivmedizin (DGAI, grants from Stifterverband für die deutsche Wissenschaft e.V. / Metronic, grants from Philips ElectronicsNederland BV, grants from BMG, grants from BMBF, grants from BMBF, grants from Deutsche Forschungsgemeinschaft / German Research Society, outside the submitted work; In addition, Dr. Spies has a patent 10 2014 215 211.9 licensed, a patent 10 2018 114 364.8 licensed, a patent 10 2018 110 275.5 licensed, a patent 50 2015 010 534.8 licensed, a patent 50 2015 010 347.7 licensed, and a patent 10 2014 215 212.7 licensed. Andreas Edel: AE has no competing interests to report. Fridtjof Schiefenhövel: FS has no competing interests to report. Tim Rahmel: TR has no competing interests to report. Christian Putensen: CP has no competing interests to report. Timur Sellmann: TS has no competing interests to report. Thea Koch: TK has no competing interests to report. Timo Brandenburger: TBra has no competing interests to report. Detlef Kindgen-Milles: DKM has co competing interests to report. Thorsten Brenner: TBre reports grants from Deutsche Forschungsgemeinschaft (DFG), Dietmar Hopp Stiftung and Stiftung Universitätsmedizin Essen, personal fees from CSL Behring GmbH, Schöchl medical education GmbH, Boehringer Ingelheim Pharma GmbH, Biotest AG, Baxter Deutschland GmbH, Astellas Pharma GmbH, B. Braun Melsungen AG and MSD Sharp & Dohme GmbH outside the submitted work. Marc Berger: MB has no competing interests to report. Kai Zacharowski: KZ has no competing interests to report. Elisabeth Adam: EA has no competing interests to report. Matthias Posch: MP has no competing interests to report. Onnen Moerer: OM has no competing interests to report. Christian S. Scheer: CS has no competing interests to report. Daniel Sedding: DS has no competing interests to report. Markus A. Weigand: MW has no competing interests to report. Falk Fichtner: FF has no competing interests to report. Carla Nau: CN has no competing interests to report. Florian Prätsch: FP has no competing interests to report. Thomas Wiesmann: TW reports personal fees from Pajunk, Germany, personal fees from Vygon, Germany, outside the submitted work. Christian Koch: CK has no competing interests to report. Gerhard Schneider: GS has no competing interests to report. Tobias Lahmer: TL has no competing interests to report. Andreas Straub: AS reports personal fees from CSL Behring GmbH (Munich, Germany), personal fees from Schöchl Medical Education GmbH (Mattsee, Austria), personal fees from Aspen Germany GmbH (Munich, Germany), outside the submitted work. Andreas Meiser: AM reports personal fees from Sedana Medical, Danderyd, Sweden, outside the submitted work. Manfred Weiss: MW has no competing interests to report. Bettina Jungwirth: BJ has no competing interests to report. Frank Wappler: FW has no competing interests to report. Patrick Meybohm: PM has no competing interests to report. Johannes Herrmann: JH has no competing interests to report. Nisar Malek: NM has no competing interests to report. Oliver Kohlbacher: OK has no competing interests to report. Stephanie Biergans: SB has no competing interests to report. Peter Rosenberger: PR has no competing interests to report.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.