Our team is looking at track 2 challenge 1 COVID-19 data COVID-19 continues to cause difficulties for hospitals and clinical staff. Predicting the outcome of patients will enable clinical staff to make better informed resource management decisions, improving patient care.

What it does;

Using previous COVID-19 patient data to create a model that predicts whether a patient is likely to be a short or long term patient in the hospital based on readings of vital signs taken in the first few days of hospital. The dashboard assesses patient data and is able to show the user the details about median results informing clinical decision making and resource allocation.

How we built it;

Individual patient results across 6 tests were plotted against the longevity of their hospital stay, in order to reveal a correlation in their first few days of vital sign readings and the length of stay.

Challenges we ran into;

The data came from 974 patients who were hospitalised over a 9 month period in 2020. Not least has medical treatment for COVID-19 patients developed since that time, but the treatment for those patients during these 9 months would have developed rapidly too. This would have affected the response and test results of those patients, of which we remained conscious of.

Accomplishments that we're proud of;

As a developer, we get to decide on how best to visualise and examine the data, but it is an achievement to be able to develop a dashboard that allows the medical professional to interact with their patient data and access useful insights quickly to assist decision making. We developed a dashboard with interactive usability so that the consumer could drill down and choose which metrics they want to view therefore, answering their own questions related to the data.

What we learned;

Improving our understanding of the capabilities of machine learning and SAS VA predictive modelling. Time management and collaborating with team members effectively who are based in 3 different time zones and countries and improving our video editing & presentation skills.

What's next for Butterfly Effect;

This dashboard uses data from March to November 2020 however, we are looking to build a program that can absorb up to date data and review the previous 3 months of patients. This will give a more accurate view of the new patients as COVID-19 medication and patient care develops over the pandemic, therefore ensuring a more reliable result.

Built With

  • azure
  • clustering
  • forest
  • linearregression
  • logisticregression
  • random
  • sas
  • viya
Share this project: