Helping people decide which drugs are best for them.
What it does
We are hoping to crowd-source data to train a machine learning algorithm to help patients make an informed decision for a drug based on both side effects for their demographic and for other drugs in that class, with full transparency and data contributed by patients and available for viewing.
For our weekend prototype, we chose to focus on birth control options for women. Currently, women have a plethora of options for birth control without a centralized way to figure out what is best for them, leading to a trial-and-error process to figure out which side effects they can tolerate, which is often emotionally, physically, and financially taxing. This program aspires to help them choose a form of birth control that works for them the FIRST TIME based on their demographic/clinical data and goals while minimizing side effects. This is just a proof-of-concept prototype and still has many improvements in store!
How I built it
We spent the first night brainstorming different project ideas and their feasibility. We were fortunate to have an incredibly diverse team with an MD/PhD student, pharmacy student, and three experienced computer scientists and web developers, all of whom contributed valuable insights. We went through many iterations of the website design and data collection and analysis, and our joint perspectives helped create the best product possible.
We cared deeply about the set of questions we were asking, and wanted to be able to iterate quickly as we refined this list. Therefore, we created a json format for our questions and designed the whole project to pull from this json and build itself on this list, for maximum flexibility.
Challenges I ran into
Our original vision was too big for a weekend hackathon and we had to pair it down. Originally, we wanted something that could be integrated electronic medical records and seamlessly take into account patient clinical data, but did not have the data available, so we decided to crowdsource the data instead and focus on a broad population-- birth control in women.
Accomplishments that I'm proud of
Building a web app that addresses a niche in health care that people really want to be available.
What I learned
While we were originally hesitant to crowdsource the data, by sending out the survey to people we were also able to receive feedback from them and we learned that many people felt like this was something that urgently needed in the healthcare field, and should be do-able given the wealth of information stored in electronic medical records.
What's next for Our Experience (OE)
Cleaning up the interface, making it more user friendly, improving the machine learning aspect, and integrating it with the healthcare system.
For the github code, please see: https://github.com/HastingsGreer/PharmaceuticalThing and https://github.com/mishuvs/pittchallenge20-dbpopulater