Disease Detective

Click! At the touch of the button, you find the location of every deadly disease outbreak instantly. No flu, COVID, or measles. The best part? No outbreak has to occur for these disease to be found.

Comment

Inspiration

Ever since the introduction of COVID-19, our team has noticed a conspicuous flaw in the healthcare system. Outbreaks of COVID-19 are widely unpredictable to healthcare providers, with propagated waves of infection coupled with varying incubation periods and mutated strains with a diverse set of common symptoms changing the case definition for an outbreak. These delays in outbreak recognition lead to interruptions in the medical supply system, leading to extreme shortages of personal protective equipment (PPE) and medical equipment to safely treat an influx of patients during these times of demand. This results in a greater mortality rates, longer periods of sickness, and the incidence of cases rising significantly. In response to this issue, our team aims to innovate a solution to accurately and effectively predict future outbreaks.

What it does

DiseaseDetector analyzes current hospitalization data provided by the CDC, WHO, and local hospitals to anticipate rising trends of disease incidence, so that hospitals may be better equipped to host more people. Our current model focuses on Virginia based data specifically, and primarily targets infectious diseases, but is being developed to accommodate food-borne illnesses

How we built it

We leveraged google trends of symptoms pertaining towards common food-borne illnesses. We used these trends to anticipate an outbreak of non-communicable disease. These trends can be used to identify the origin of the outbreak. For communicable diseases, we used hospitalization data and mapped it over time to identify whether cases were increasing drastically, which would imply that there was an additional wave of infection.

Challenges we ran into

We had a hard time manipulating the google API to be compatible with our program and work at the specificity that we needed it to. We also struggled with pandas on replit, as the two features were supposedly incompatible. We struggled with simple Django tasks as well, as we were quite new to the framework, and had to do quite a bit of researching in order to figure out even how to iterate through a list in HTML.

Accomplishments that we're proud of

We were able to effectively use the google trends API to track the search of several key words and see spiking cases over specific points of time. We were able to create graphs from copious amounts of data, combing through the CSVs with pandas and matplotlib to create easy-to-understand graphs and plots. Additionally, we were able to learn Django in a relatively short time, something we're quite proud of.

What we learned

We learnt how to scale a larger level coding project. Beyond that we also learned how to use the Django framework when paired with replit, as well as how to use many data manipulation libraries, including pandas, matplotlib, and more. We learned how to work as a team, tackling individual components of the project before assembling it together into one cohesive product.

What's next for DiseaseDetector

We are now working on expanding our model to include foodborne illnesses on a large scale - i.e. finding adequate amounts of data, rather than just relying on google trends. Additionally, we are working to include a function SIR model into our code, which utilizes artificial intelligence to predict the various trends of a certain disease, including susceptibility, infectious percentage, and more.

Built With

Share this project:

Updates