Tivi - Tracking Ventilators
Inspiration
Currently decision makers and news outlets are worrying about the supply of ventilators and speculating about their availability in Switzerland and beyond. There are many questions surrounding this issue.
The federal and cantonal level
- need to know how many ventilators exist and where
- if and where resources are not optimally used
- which manufacturers produce the spare parts
- where should they deploy spare ventilators by the army
- have to consider machine hacks and DIY solutions
...
The hospital and medical staff level
- wants to have and overview over their devices, ask which ones are in use or defect?
- where can we place the next patient?
- how does the machine work? how do I clean it?
...
What it does
Tivi provides a real-time tracking and visualization system for crisis inventory management.
It tracks medical devices via a webapp and visualizes machine statuses, information about each machine and groups of machines in Switzerland can be updated and shared at the hospital, cantonal and federal level.
How we built it
To gain information, we implemented a React based webapp to ensure availability on any device (phone, laptop, ...). Machines can be registered or looked up by scanning their medical barcode (data matrix and qr code scanner by Scandit) that each existing machine possesses in CH. Once registered, medical staff can scan the machine and set its current state with a click. The machine view also offers quick manuals for usage and cleaning, because ventilators differ from producer to producer and consist of a wide array of parts. We also added the option for medical staff to change the capacity of 1 person (default), such that the count of available ventilator slots remains accurate. We also added a QR code (Scandit) generator for DIY ventilators for the same reason.
Status, Location data (from input and IP location) and IDs are transferred via a REST API (FastAPI) to a database (SQL Server database, hosted on Azure). The raw data is enriched (geolocations are mapped with hospitals) and transformed before being delivered to Tableau for visualisation. For different levels of detail, e.g. the cantonal, and federal level, we implemented Tableau dashboards. For the demo, we populated the database with (researched) simulated data.
The implementation does not include any information about patients. In terms of security, we enriched our system with keycloak.
Challenges we ran into
Initially we wanted to build a rasperryPI based setup that could monitor vibration or airflow to automatically assess the state of each machine. Due to quarantine limitations of some group members, we thus decided to build a webapp interface via barcode identification. We discuss this in the _ what's next _ section
Accomplishments that we are proud of
- We digitalized our hardware approach (using disused mobile phones, raspberryPI, ..)
- We built our barcode scanner on the medical data matrix standard, which makes this system interoperable
- We also considered the ways in which hacked or DIY machines might come into the system
- We applied a clustering algorithm to match IP locations of ventilators to the nearest hospital (all Swiss hospitals included)
- We were able to technically implement the system (webapp, barcode scanning/generating, database, & dashboards)
- We like that our system is pragmatic and extendable. The system can be replicated to other medical devices that may represent a bottleneck in future crises; where the registration of the devices could become a preparation step in action plans by the federal ministry of health (BAG).
What I learned and next steps
- Testing the POC in a hospital environment
- Automate status updates by connecting the devices or adding sensors that track vibration, airflows
- Extend system for scalability of data transactions & storage
- Build recommender system specifying where machines should be deployed and for how long
- Extend Predictive Maintenance by tracking bottleneck spare parts (tubes, connectors, mouth part, ...)
- Extend to other devices that could be critical in a pandemic and integrate prep into BAG action plans
Acknowledgements
Many thanks go to:
Philipp Bogner (Public Health Professional, DEZA)
Pablo-David Rojas (PhD in Infectious diseases)
Corry Curtis (Public Health Professional)
Rob Scott (Healthcare Experience)
Anthony Ajose (Health Tech and Clinical Pharmacy Specialist)
Franziska Smely, Theresa Ling (Medical Doctors)
Log in or sign up for Devpost to join the conversation.