gms | German Medical Science

73. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie (DGNC)
Joint Meeting mit der Griechischen Gesellschaft für Neurochirurgie

Deutsche Gesellschaft für Neurochirurgie (DGNC) e. V.

29.05. - 01.06.2022, Köln

ICU after elective craniotomies – Can a score help identify those who need it?

Intensivmedizinische Überwachung nach elektiven Kraniotomien – Kann ein Score unterscheiden, welche Patienten dies brauchen?

Meeting Abstract

  • presenting/speaker Stephanie Schmidt - Universitätsklinikum Heidelberg, Neurochirurgische Klinik, Heidelberg, Deutschland
  • Amin Nohman - Universitätsklinikum Heidelberg, Neurochirurgische Klinik, Heidelberg, Deutschland
  • Andreas W. Unterberg - Universitätsklinikum Heidelberg, Neurochirurgische Klinik, Heidelberg, Deutschland
  • Jan-Oliver Neumann - Universitätsklinikum Heidelberg, Neurochirurgische Klinik, Heidelberg, Deutschland

Deutsche Gesellschaft für Neurochirurgie. 73. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie (DGNC), Joint Meeting mit der Griechischen Gesellschaft für Neurochirurgie. Köln, 29.05.-01.06.2022. Düsseldorf: German Medical Science GMS Publishing House; 2022. DocV255

doi: 10.3205/22dgnc247, urn:nbn:de:0183-22dgnc2470

Published: May 25, 2022

© 2022 Schmidt et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License. See license information at http://creativecommons.org/licenses/by/4.0/.


Outline

Text

Objective: Following elective craniotomy, patients are usually monitored at the ICU for 24 hours. However, studies have shown that most do not need any specific intervention. While certain risk factors for the need of intensive care intervention have been identified, we lack the tools to evaluate each patient to stratify the need for postoperative ICU monitoring.

Methods: 1000 patients were evaluated retrospectively after receiving an elective craniotomy. Data concerning demography, patient history, diagnosis, radiological findings as well as the intraoperative and postoperative course was collected. We defined several therapeutic measures requiring either ICU or IMC monitoring, such as new or worsening neurological deficits, ICP-monitoring, reanimation, reintubation, intravenous blood pressure therapy or revision surgery. Two common machine learning methods, neuronal networks and boosted decision trees, were trained using 800 patients from our data set. Relevant risk factors were identified and the clinical effect of risk stratification using machine learning methods were evaluated by simulation in a validation cohort of 200 patients.

Results: 90% of patients were monitored in an ICU/IMC setting, of which less than half received any of the aforementioned interventions. Only 1% of the patients had to be transferred to the ICU after being transferred directly to the surgical ward. Using machine learning methods, a AUC-ROC of over 80% could be achieved. Choosing a sensible threshold, a level of roughly 1% false-negative predictions could be maintained, while significantly reducing the planned ICU admission rate to about 60%.

Conclusion: Our data show that using machine learning methods for risk stratification, a larger number of patients with a low risk for postoperative ICU intervention can be identified and transferred directly to the surgical ward while maintaining the safety level previously achieved with conventional decision-making.