Theme 1 | (COVID-19 related) mental health care
Predicting mortality of individual COVID-19 patients: A multicenter Dutch cohort.
Maarten C Ottenhoff1, Lucas L Ramos2,3, Wouter Potters4, Marcus LF Janssen5, Deborah Hubers6, 7, Shi Hu12, Egill A Fridgeirsson8, Dan Piña-Fuentes4, Rajat Thomas8, Iwan CC van der Horst6, Christian Herff1, Pieter Kubben9, Paul WG Elbers10, Henk A Marquering2,11, Max Welling12, Suat Simsek13,14, Martijn D de Kruif15, Tom Dormans16, Lucas M Fleuren18, Michiel Schinkel17, Peter G Noordzij18, Joop P van den Bergh19, Caroline E Wyers19, David TB Buis20, Joost Wiersinga21, Ella HC van den Hout13, Auke C Reidinga22, Daisy Rusch23, Kim CE Sigaloff24, Renée A Douma25, Lianne de Haan25, Niels C Gritters van den Oever26, Roger JMW Rennenberg27, Guido A van Wingen8, Marcel JH Aries6*, Martijn Beudel4*,, on behalf of The Dutch COVID-PREDICT research group
1Department of Neurosurgery, School for Mental Health and Neuroscience, Maastricht University, the Netherlands
2Department of Biomedical Engineering and Physics, Amsterdam University Medical Centers, Amsterdam, the Netherlands
3Department of Epidemiology & Data science, Amsterdam University Medical Centers, Amsterdam, the Netherlands
4Department of Neurology, Amsterdam Neuroscience Institute, Amsterdam University Medical Center, Amsterdam, the Netherlands
5Department of Clinical Neurophysiology, Maastricht University Medical Center, School for Mental Health and Neuroscience
6Department of Intensive Care, Maastricht University Medical Center, Maastricht, the Netherlands
7Technical University Twente, Enschede, the Netherlands
8Department of Psychiatry, Amsterdam Neuroscience, Amsterdam University Medical Centers, University of Amsterdam, Amsterdam, the Netherlands
9Department of Neurosurgery, Maastricht University Medical Center, the Netherlands
10EDIC, Department of Intensive Care Medicine, Research VUmc Intensive Care (REVIVE), Amsterdam Medical Data science (AMDS), Amsterdam Cardiovascular Sciences (ACS), Amsterdam Infection and Immunity Institute (AI&II), Amsterdam UMC, location VUmc, VU Amsterdam, the Netherlands
11Department of Radiology & Nuclear Medicine, Amsterdam University Medical Centers, Amsterdam, the Netherlands
12Informatics Institute, University of Amsterdam, Amsterdam, the Netherlands.
13Department of Internal Medicine, Northwest Clinics, Alkmaar, the Netherlands
14Department of Internal Medicine & Endocrinology, Amsterdam University Medical Center, Amsterdam, the Netherlands
15Department of Pulmonary Medicine, Zuyderland Medical Center, Heerlen, the Netherlands
16Department of Intensive Care, Zuyderland Medical Center, Heerlen, the Netherlands
17Center for Experimental and Molecular Medicine (C.E.M.M.), Amsterdam UMC, location Academic Medical Center, Amsterdam, the Netherlands
18Department of Anesthesiology and Intensive Care, St Antonius Hospital, Nieuwegein, the Netherlands
19Department of Internal Medicine, VieCuri Medical Centre, Venlo, the Netherlands
20Department of Internal Medicine, Amsterdam institute for Infection and Immunity, Amsterdam University Medical Center, the Netherlands
21Department of Internal Medicine, Division of Infectious Diseases, Amsterdam Institute for Infection and Immunity, Amsterdam UMC, the Netherlands
22Department of Intensive Care, Martiniziekenhuis, Groningen, the Netherlands
23Research, Martiniziekenhuis, Groningen, the Netherlands
24Department of Internal Medicine, Division of Infectious Diseases, Amsterdam Institute for Infection and Immunity, Amsterdam UMC, the Netherlands
25Department of Internal Medicine, Flevoziekenhuis, Almere, the Netherlands
26Department of Intensive Care, Treant Zorggroep, Emmen, the Netherlands
27Department of Internal Medicine, Maastricht University, Maastricht University Medical Center, Maastricht, the Netherlands
*Both authors contributed equally
Several waves of the COVID-19 pandemic had a dramatic effect on society and severely disrupted our daily lives, economies and healthcare systems. During the peak of these waves, hospitals and intensive care units (ICU) throughout Europe were (nearly) overwhelmed and resources were exhausted. Given the novelty of the virus, accurate information about the clinical course and prognosis of individual patients is largely unknown, which led to the use of crude limits to unilaterally withhold advanced life support measures to face the large numbers of pulmonary insufficient patients. Although criticized, several hospitals in Europe already solely used age as triage criterion. To prevent these forced socio-ethical debatable choices, triage should be made based on medical criteria with an evidence base. Given the largely unknown prognosis of the virus, we proposed and developed a predictive model that uses medical records to predict mortality of COVID-19 patients admitted to the hospital, aiming to support clinical decision making during scarcity of hospital or ICU beds.
COVID-19 patients (≥18 y/o) admitted to 10 different hospitals across the Netherlands were included between February 27th to June 8th 2020. We modelled the outcome prediction as the 21-day mortality from the day of hospital admission. Five feature sets, including premorbid, clinical presentation and laboratory & radiology values, were derived from a total of 80 features. Additionally, the 10-most predictive features were selected. Two models (logistic regression, LR and tree-based gradient boosting, XGB) were trained using leave-one-hospital-out cross validation and evaluated using the area under the receiver operator curve (AUC). Furthermore, the results were compared with decision rules used in practice.
2273 patients were included, of whom 516 died or were discharged to palliative care within 21 days after admission. The resulting cohort represents about 16% of all hospitalized COVID-19 patients from the first wave in the Netherlands. A linear logistic regression (LR) and non-linear tree-based gradient boosting (XGB) algorithm fitted the data with an AUC of 0.81 (95% confidence interval 0.77 to 0.85) and 0.82 (0.79 to 0.85), respectively, using the ten selected features: age, number of home medications, urea nitrogen, lactate dehydrogenase, albumin, oxygen saturation (%), oxygen saturation is measured on room air, oxygen saturation is measured on oxygen therapy, blood gas pH and history of chronic cardiac disease (in order of highest to lowest F-value). Both models outperformed age-based decision rules used in practice (AUC of 0.69, 0.65 to 0.74 for age > 70). Furthermore, performance remained stable when excluding age as predictor (AUC of 0.78, 0.75 to 0.81)
Both models showed excellent performance and had better test characteristics than age-based decision rules. Additionally, outcome predictions can be made using ten admission features readily available in Dutch hospitals. Furthermore, we demonstrated that age does not have to be included as feature to achieve high performance, contributing to the ongoing social-ethical debate in the Netherlands. We conclude that both models hold promise to aid decision making during a hospital bed shortage, but acknowledge that the models should be thoroughly validated in practice before being applied.
The School for Mental Health and Neuroscience (MHeNs) strives to advance our understanding of brain-behaviour relationships by using an approach integrating various disciplines in neuro- and behavioural science, medicine, and the life sciences more widely. MHeNs performs high-impact mental health and neuroscience research and educates master's students and PhD researchers. MHeNs performs translational research, meaning practical collaboration between researchers in the lab and in the hospital.
MHeNs is one of six graduate schools of the Faculty of Health, Medicine and Life Sciences (FHML) aligned to the Maastricht University Medical Centre+ (MUMC+).