Case Studies

PrioVax: Using Machine Learning to Improve Distribution of COVID-19 Vaccines with Dropbase

Learn how the PrioVax team was able to build machine learning models to better distribute vaccines in just 36 hours, with the help of Dropbase for preparing their datasets.

This article is an ongoing series showcasing some of the best hacks we encountered through the Hack the North competition. To view more of these projects, please checkout the full list of hacks using the Dropbase API

Despite only forming their team hours before the event started, the PrioVax team was able to build an impressive project, managing to win the Vonage API prize, and narrowly missing out on winning our Dropbase API Prize as well. The team was comprised of three members from the University of Toronto, Hilary Wang, a third year Industrial Engineering student, Louis Liu, a third year Mechanical Engineering student, and Malhar Shah, a third year Computer Engineering student alongside Braden Collingwood, who is a third year Software Engineering student at the University of Ottawa.

The three University of Toronto engineers came across Braden somewhat by chance, but it turned out to be a great fit for the team. Hilary commented that "We wanted someone to help us with the frontend, and he was a perfect fit, he actually came up with the idea for [PrioVax]". Braden's initial Hack the North team had to drop out due to scheduling conflicts, so the Software Engineering student found the rest of the team through the Hack the North discord, and fit in perfectly with the strengths of the other team members.

How can we improve the distribution of the COVID-19 Vaccine?

After nearly a year in partial and full lockdowns due to Coronavirus, for many of us the COVID vaccine is the light at the end of the tunnel. In Canada, the current plan is to have a phased rollout of the vaccine for different segments of the population. The challenge for provincial governments and local health is determining who within their populations is at greatest risk of contracting the disease, who should be given priority in the vaccination efforts, and how to effectively communicate this to their population.

The PrioVax team attempted to build a tool to answer all these questions, and reduce the load on municipal health organizations, allowing them to make better, data-driven decisions surrounding vaccine rollout.

Using Machine Learning to Identify and Prioritize High Risk Individuals

The PrioVax team realized that the best way to identify the most vulnerable individuals is by taking past COVID patient medical histories and mortality rates, and training a machine learning model to assign priorities to individuals based on their medical histories. Using factors like age, gender and pre-existing health conditions, the model is able to give healthcare providers a mortality prediction if the patient was to contract COVID, allowing them to make informed decisions about when the patient should be vaccinated.

Screenshots of the patient SMS and Batch CSV upload functionalities

The team also created the ability for hospitals to input new data either through a web portal or through batch CSV uploads, further allowing for refinement in the machine learning algorithm. As well, once priorities are assigned the application will be able to alert patients via SMS of when they are eligibly for vaccination, and allow them to book an appointment with a healthcare provider. To see a more in depth picture of the capabilities of PrioVax, checkout the product demo created by the team:

Using Dropbase to clean and prepare data for machine learning models

When the team was looking for datasets to run their machine learning algorithms against, they struggled to find a clean dataset that was large enough to have predictive validity, and not too simplistic in nature. When they did come across a dataset that fit their criteria, they realized that it still had certain columns with errors and certain columns that weren't relevant to the work they wanted to do with the dataset. So they used Dropbase to quickly and programatically clean their datasets, by removing unnecessary columns and building custom data cleaning functions to fix the columns with errors.

Fining errors in the data and filtering out those results so it wouldn't confuse our data algorithm was key for us, which is why we decided to use Dropbase

Next Steps for PrioVax

The team spoke of a number of new features that they hope to add to PrioVax in the near future. One of the largest ones is further refinement of the machine learning algorithm by feeding it larger datasets through Dropbase. They see the future use case of the platform being the aggregation of COVID data from individual hospitals, allowing for a ML model that is more robust, with far more data points.

Another feature they hope to implement is scheduling appointments for vaccinations. Given access to patient data and the patient notification system they built through Vonage, the PrioVax application will be able to help hospitals optimize their vaccination schedules, and communicate these schedules to the patients.

Find out how Dropbase can help your team clean datasets to create more robust machine learning models! Sign up for free today, or contact us to schedule a demo

Insights and updates from the Dropbase team.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
By signing up you agree to our Terms of Service