Art

Reichweite

Bereiche

Status

Start

Ende

Finanzierungsart

Kurzbeschreibung/Zielsetzung

Background: In patients suffering from aneurysmal subarachnoid hemorrhage (aSAH) devastating and largely unpredictable sequels such as re-bleeding, cerebral vasospasm and delayed cerebral ischemia have a large impact on patient outcome. However, until now no adaptive prediction model for aSAH exists that reflects the clinical course and incorporates time sensitive information.

Aim: Within the framework of an established and successful Swiss national multicenter collaboration we aim to develop and validate a dynamic outcome prediction tool in aSAH patients based on machine learning algorithms.

Design and Methods: For the first part of this project an prospectively collected dataset of anonymized aSAH patients from the Swiss SOS database will be used to develop a dynamic outcome model. Based on a computed core model at admission we will integrate available time dependent parameters influencing patient outcome to build a dynamic model. This model will show diurnal variation to estimate the chance of specific events and their consequences on the clinical course and outcome using state of the art machine learning algorithms. In a second step, we aim to implement the established outcome model prospectively in daily clinical practice at the Kantonsspital St. Gallen. Important parameters will be highlighted, and missing variables linked to changes in patient outcome should be detected and integrated into the algorithm. Intuitive ways to display outcome prognosis such as “heat maps” will be developed and information regarding specific events and treatment pathways may be displayed in relation to predicted patient outcome. As a final step, the goal is to roll out a prospective dynamic outcome analysis and extended data collection to other neurovascular centers in Switzerland. An application programming interface will be developed to assist flow of information to and from hospital information systems and an app will be programmed that displays patient outcome prediction and certainty.