Corona Score

Medical WebApp

The corona score algorithms use standard demographic, laboratory and imaging data to compute a score that expresses a prediction on the likelihood of a patient, reporting to the emergency department with COVID-19-like symptoms, to test positive with RT-PCR. The aim of the algorithm is to help doctors with rapid triaging of patients and was the developed in the early stage of the Covid-19 pandemic in collaboration with a groupd of talented people from four Dutch hospitals.

A Corona score can be calculated in two ways:

Machine Learning: This algorithm was trained using the Random Forest method and a dataset of over 500 patients reporting to the emergency departments of Dutch hospitals with respiratory symptoms .

Human readable: The second is by implementing a different version of the algorithm which was created by sampling for optimal thresholds and scores per parameter on a dataset with 375 patients and validated with 592 patients. This second version is less sophisticated and accurate but allows doctors to see how different parameters contribute to the score. This is essential for gaining insight and validating the scores generated in the clinical context.

The Corona score was rapidly implemented at four Dutch hospitals.

I was personally responsible for the computational sampling of the human readable algorithm (using Python, Pandas, skLearn) as well as both the front and backend web development (using ReactJs, Flask, Python, Heroku).

Links:
Peer-reviewed article
Free pre-print article
Public website