ICU Cockpit Apps: Interventional Study With First ICU Cockpit Software Applications

Background:

Since 2014, the project "ICU-Cockpit" has been set up at the Neurointensive Unit of the University Hospital Zurich in collaboration with the University of Zurich and ETH Zurich. State-of-the-art information technology collects, synchronizes and encrypts data from numerous medical devices in real time. The project is aimed at initiating a fundamental development in emergency and intensive medicine - and bringing about a substantial improvement in the way diagnostics, treatment and risk management are handled in everyday clinical practice.

Based on video monitoring, algorithms for early alarm systems and therapy support have been already developed. In order to make the algorithms usable for clinical practice, a user interface (GUI) is currently being developed in cooperation with the University of Applied Sciences Northwestern Switzerland (FHNW), see appendix 15. The GUI allows to visualize and interpret the multimodal patient data in a comprehensive manner.

In 2016, a research project started at the NICU of the University Hospital Zurich started with the title "ICU Cockpit: Integrated platform for multimodal patient monitoring and therapy support in neurocritical care" (BASEC No. 2016-01101). The purpose of the single site project is to set up an integrated platform (scientific prototype of ICU Cockpit software platform) for multimodal patient monitoring and therapy support in neurocritical care, and as a second step to develop and validate algorithms for some first clinical applications. Furthermore, automatic artefacts detections (i.e., number of false and true alarms in ICU Cockpit in comparison to standard monitoring system) are assessed and algorithms are developed for the early detection of epileptic seizures and secondary impairment of cerebral perfusion.

The objective of this current clinical investigation is the verification and validation of the ICU Cockpit software platform and the following three different applications for prognostication and prediction of complications:

1. ICU Cockpit COVID-19 for remote monitoring of isolated patients, b) ICU Cockpit Stable State for a comprehensive visualization of vital parameters and as additional aid in early detection of imminent critical complications c) ICU Cockpit Cerebral Ischemia for the prediction of delayed cerebral ischemia in patients with subarachnoid hemorrhage (SAH).
2. ICU Cockpit COVID-19 for remote monitoring of isolated patients Having infected more than 100 000 people in 100 countries, the coronavirus disease 2019 (COVID-19) has become a pandemic. In China, up to 10% of patients with COVID-19 became severely ill requiring intensive care treatment mostly due to pneumonia, acute respiratory distress syndrome (ARDS) and multiorgan failure. In Italy, having about 5200 beds in intensive care units (ICUs), already 1028 were already devoted to patients with COVID-19 by March 11. Furthermore, 20% of health-care professionals, at this time, have become already infected, and some have died. For the United States, without self-isolation and for R0 = 2 and R0 = 2.5, respectively, ICU admissions are projected to become 17.4 to 25.7 per 1,000 population, which would significantly exceed the capacity of the existing 100,000 ICU beds in the US. Also, in Switzerland the health care system and especially ICUs are facing extraordinary challenges as human resources for ICU professionals are limited. Advances in medical informatics have the potential to facilitate medical care and save personnel resources not only in the ambulatory sector but in the ICUs especially during epidemics. In intensive and emergency medicine, the situation is compounded by real-time signals from multiple sensors on, as well as inside the human body. In an emergency situation, in particular, it is not possible to integrate this flood of information rapidly into the decision-making process. Remote monitoring of strictly isolated patients especially reduces routes and walking of the staff, the number of close contacts with infected patients and therefore, virus transmission. The safety of patients will be increased by a reduction of stress and burnouts in nurses and medical doctors, which often work in crisis over time.
3. ICU Cockpit Stable State As a first application a software called "stable state" has been developed recently. Vital parameters as arterial blood pressure, intracranial pressure and many others are hard to read in a glance due to their high information density. Since the human body is a complexly inter-connected system, these biosignals are also heavily correlated. Experienced clinicians can recognize deviations in the correlational structure of several signals, but it is impossible to grasp the full picture for the huge amount of available data. However, subtle correlational changes may indicate an instable state of a patient before these changes are visible in a specific signal or its compound measure. An algorithm which computes the inter-dependencies in an arbitrary set of signals during a stable phase of the patient was recently developed. This stable state is compared to an on-line estimate of the current state. Correlational changes are didactically reduced and visualized in a Mondrian-like display for quick comparisons of current and past states. The application "stable state" as part of the GUI ICU Cockpit is planned to be implemented as an additional aid in early detection of imminent critical complications.
4. ICU Cockpit Cerebral Ischemia Cerebral vasospasm (CVS) is a delayed morphologic narrowing of cerebral arteries, occurring 4 to 10 days after aneurysmal subarachnoid hemorrhage. Classically, CVS has been associated with delayed cerebral ischemia (DCI), which ultimately leads to cerebral infarction associated with morbidity and mortality. This has been sort of a paradigm, and the term CVS has been often misused to describe clinical signs of DCI. The majority of research was focused on the assumption CVS to be solely responsible for DCI. Today, it is well accepted that not all patients with CVS develop DCI. Inversely, DCI can occur in the absence of CVS. Recent review of the literature indicates that CVS is not the only cause of DCI and that the entire picture of DCI is multifactorial. There is an ongoing debate about this issue, however cerebral infarction on imaging studies might be the most appropriate measure of DCI beside functional outcome. Therefore, an algorithm to predict DCI based on a random forest model has been developed, which allows to identify biomarkers for the occurrence 24h to 48 hours before.