Project / product name: zCase Plus
Link to the project: http://
Team leader: Martin Ron
Challenge: no. 7: zCase – forecasting probability of ambulance arrival
Problem: The delay between patient calling an ambulance and patient getting treatment is currently still affected by the Emergency Admission readiness. There are surgical procedures occupying the operating room that could have been replanned, specialists to be alerted to adjust their workday plans. A predictive model supporting such a decision making process is missing.
Solution: Our ambulance arrival forecasting system zCase Plus provides prediction of arrival rate based on the online weather data, season, time, mobility data from Google, Waze and Apple and even public events (football matches). Our system provides forecasts for up to 24 hours, it's updated online, and learns every day. It is accessible through a web interface, and it's lightweight - it does not need a powerful server. We bring patient’s historic data from FHIR to zCase to notify the doctor sooner.
Impact: We expect to save on average 50 minutes per day on delays resulting in preventing 5 unnecessary deaths per 30 unstable patients (based on IKEM provided analysis). The engine can be easily deployed to any other hospital in a few days by deploying the web application locally. The engine is implemented so that additional online data resources can be quickly incorporated into live forecasts. Users can access forecasts from their phone at home. No such forecast system is in use so far.
Feasibility & financials: Expert mode is Dockerized and ready to use. Model used in this application could be tuned up by additional testing and hyperparameters selection. Integration directly in the zCase dashboard has to be integrated in cooperation with IKEM IT department in the future. Code itself needs some refactoring for future sustainability and extensibility. In the end the application itself has to be deployed on a publicly available server instead of a local network.
What you have built at the hackathon - text explanation + code (e.g. GitHub link): - public databases scrapers - scrapers directly from web pages - unique forecast model based on Gaussian processes - web interface in Streamlit - interconnection with FHIR cloud database - https://github.com/factorio-solutions/ehh2021-challange-7
What you had before the hackathon, please mention open source as well: - great team - IKEM already has a zCase application - open source frameworks PyTorch, GPytorch, Pyro, Streamlit, Scikit, etc.
What comes next and what you wish to achieve: Wishes: - more saved lifes by optimal medical stuff shifts planning - zCase Plus deployed in every Czech hospital - extend application with additional domains data What comes next: - cooperation between our team and IKEM IT department to deploy our application into the production environment - integration to zCase dashboard - testing and thorough forecasts validation - many hours of sleep :)