Development and Internal Validation of a Gradient-Boosted Trees Model for Prediction of Delirium after Surgery and Anesthesia (the BioCog Study)

IMPORTANCE: Postoperative delirium (POD) is a multietiological condition and affects 20% of older surgical patients. It is associated with poor clinical outcome and increased mortality. OBJECTIVE: We aimed to develop and validate a risk prediction algorithm for POD based on a multimodal biomarker database exploiting preoperative data (predisposing factors) and procedural factors as well as perioperative molecular changes associated with POD (precipitating factors). DESIGN: BioCog is a prospective cohort study conducted from November 2014 to April 2017. Patients were followed up for seven postoperative days after surgery for POD. Gradient-boosted trees (GBT) with nested cross-validation was used for POD prediction. SETTING: Patients aged ≥65 years were enrolled at the anesthesiologic departments of two tertiary care centers. EXPOSURE: All patients underwent surgery with an expected duration of at least 60min. Clinical, neuropsychological, neuroimaging data and blood were collected and clinically well established as well as non-established biomarkers (e.g., gene expression profiling) were measured pre- and postoperatively. MAIN OUTCOME: POD according to DSM 5 until the seventh postoperative day RESULTS: 184 of 929 (20%) patients experienced POD. A GBT algorithm using both preoperative data, characteristics of the intervention and postoperative changes in laboratory parameters achieved the highest area under the curve (0.83, [0.79; 0.86]) with a Brier score of 0.12 (0.12; 0.13). CONCLUSIONS AND RELEVANCE: Models combining predisposing factors with precipitating factors predict POD best. Non-routine laboratory data provide useful information for POD risk prediction, providing relevant results for future studies on the molecular factors of POD. In addition, possibly relevant molecular mechanisms contributing to the development of POD were identified, mostly indicating a dysregulated postoperative immune response. This study constitutes the basis for future hypothesis-driven analyses or implementation of prediction expert system for clinical practice. ### Competing Interest Statement Georg Winterer is currently licensing a Class IIa medical device (web-based software tool for risk prediction of POD and POCD in clinical practice). Dr. Winterer is CEO of PharmaImage Biomarker Solutions GmbH Berlin (Germany) and President of its subsidiary Pharmaimage Biomarkers Incl. (Cambridge, MA, USA). Dr. Spies, Dr. Winterer, Dr. Boraschi, Dr. Dschietzig, Dr. Kuehn, Dr. Nuernberg, Dr. Pischon, Dr. Pietzsch, Dr. Slooter, Dr. Stamatakis, Dr. Weber, report grants from the European Commission during the conduct of the study. Dr. Winterer reports grants from the Deutsche Forschungsgemeinschaft (DFG)/German Research Society and from the German Ministry of Health. Dr. Spies reports grants from DFG/German Research Society, Einstein Foundation Berlin, Deutsches Zentrum fuer Luft- und Raumfahrt e.V. (DLR)/German Aerospace Center, Projekttraeger im DLR/Project Management Agency, Gemeinsamer Bundesausschuss (GBA)/Federal Joint Committee, inneruniversity grants, Stifterverband/Non-Profit Society Promoting Science and Education, European Society of Anesthesiology and Intensive Care, BMWI - Federal Ministry of Economic Affairs and Climate Action, Dr. F. Koehler Chemie GmbH, Sintetica GmbH, Max-Planck-Gesellschaft zur Foerderung der Wissenschaft e.V., Metronic, BMBF - Federal Ministry of Education and Research, Robert Koch Institute and payments by Georg Thieme Verlag, board activity for Prothor, Takeda Pharmaceutical Company Ltd., Lynx Health Science GmbH, AWMF (Association of the Scientific Medical Societies in Germany), DFG, Deutsche Akademie der Naturforscher Leopoldina e.V. (German National Academy of Sciences Leopoldina), Berliner Medizinische Gesellschaft, European Society of Intensive Care Medicine (ESICM), European Society of Anaesthesiology and Intensive Care (ESAIC), Deutsche Gesellschaft fuer Anaesthesiologie und Intensivmedizin (DGAI)/German Society of Anaesthesiology and Intensive Care Medicine, German Interdisciplinary Association for Intensive Care and Emergency Medicine (DIVI) as well as patents 15753 627.7, PCT/EP 2015/067731, 3 174 588, 10 2014 215 211.9, 10 2018 114 364.8, 10 2018 110 275.5, 50 2015 010 534.8, 50 2015 010 347.7, 10 2014 215 212.7. Gunnar Lachmann and Maria Heinrich report grants from the BIH Charite Clinician Scientist Program during conduct of the study. Dr. Dschietzig reports personal fees from Immundiagnostik AG during the conduct of the study. Dr. Lammers-Lietz and Anton Wiehe report personal fees from Pharmaimage Biomarker Solutions GmbH during the conduct of the study. Dr. Lachmann reports personal fees from Sobi, the University of Zurich and Thieme outside the submitted work. Dr. Wolf receives fees from the Kompetenz-Centrum Qualitaetssicherung. Dr. Stamatakis reports funding from Stephen Erskine Fellowship from Queens' College of the University of Cambridge, UK outside the BioCog study. Dr. Bresser reports funding from Alzheimer Nederland outside of the study. Dr. Gallinat received funding from the German Research Foundation (DFG), Federal Ministry of Education and Research (BMBF) and received payment for five lectures and presentations with about 1.500 euro per presentation sponsored by Lundbeck, Janssen-Cilag and Boehringer. Dr. Heilmann-Heimbach receives personal fees from Life&Brain GmbH. None of the other authors have a conflict of interest to declare. ### Funding Statement The research leading to these results has received funding from the European Union Seventh Framework Program [FP7/2007-2013] under grant agreement no. 602461. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Ethics committees of the Charite-Universitaetsmedizin Berlin, Germany, (EA2/092/14) and University Medical Center Utrecht (UMC), Utrecht University, Netherlands, (14-469) gave ethical approval of this work. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I 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). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Due to the protection of intellectual property, machine learning algorithms will not be made publicly available, but can be obtained from Dr. Winterer (georg.winterer@pi-pharmaimage.com) after signing a confidentiality agreement. Participant data may be made available upon request following publication to researchers who provide a methodologically sound proposal in accordance with applicable legal and regulatory restrictions. Proposals for data analysis must be directed to both claudia.spies@charite.de and georg.winterer@pi-pharmaimage.com. To gain access, requesting researchers will need to sign a data access agreement. Analyses will be limited to those approved in appropriate ethics and governance arrangements. All study documents which do not identify individuals (e.g. study protocol, informed consent form) will be freely available on request.