Inspiration

What it does

Utilizes a simplified medication regimen protocol paired with an operant conditioning paradigm to significantly improve medication adherence, as well as establish a metric for medication adherence, create a database to protect providers against medical malpractice, and live assessment of whether a treatment plan is working or not.

In order to simplify the medication regimen, we use an evidence-based approach that utilizes a tool called a medication grid. This grid works with a notification system that not only reminds the patient when to takes their medication, but makes it easier to understand what they have to take.

Operant conditioning is achieved based on machine learning forecasting of risk of coronary events (primarily myocardial infarction and CVD-related death), as determined by latent variable analysis. Patient systolic and diastolic blood pressure data is pooled into risk categories, who's bounds are defined by a Bayesian learning model initialized to the current literature. As we obtain our user population's demographics and mortality data, this data serves as posterior evidence into our model and update bounds accordingly. Thus, risk is individualized based on patient's personal characteristics and health related behaviors. Once forecasted risk is determined, it is communicated to the patient in simple, effective visuals that will reinforce medication adherence. In addition, forecasted risk is assessed next to a null hypothesis of risk decrease with 100% adherence to treatment. This serves to determine whether or not the degree of medication adherence corresponds with degree of improvement in forecasted risk, as well as if the current treatment plan is effective.

How we built it

Python (Numpy, Scikit-learn), D3.js for visualizations, Swift for iOS app that was developed.

Challenges we ran into

Learning D3.js and database integration

Accomplishments that we're proud of

Implementing D3.js and Neural Network

What we learned

D3.js

What's next for Heart attack forecasting

Scalability to other disease types, integration with wearable health monitors and Alexa, implementation in different languages, introduction of more dynamic visuals, moving toward an continuous assessment of cardiovascular disease risk, and integration with electronic health report data from provider (de-identification and encryption measures).

Built With

Share this project:

Updates