IntelliTrack

Featured on the 2021 HackOHI/O winners blog

Idea

The pharmaceutical industry uses prescriber data to target doctors with life-saving medications and therapies. Companies can use this prescriber data to help understand how doctors prescribe their product, as well as to help them track total prescriptions (TRx) and new prescriptions (NRx) for a product in a given market.

IntelliTrack is a full-stack web app that takes prescriber data for doctors and leverages that to build out trending and targeting reports in an easily comprehensible dashboard.

Junbo Chen, Nathan Rogers and I built this in November, 2021 at HackOHI/O. Source code is available on GitHub.

Features Added

1. Report top doctors based on total prescriptions (TRx) by product -- We determined that more prescriptions equates to a more successful doctor and filtered through the data accordingly. The final result was the sorting of each drug in descending order with respect to TRx_Month_i, for each i in {1, 2, 3, 4, 5, 6}.
2. Report trends of total prescriptions (TRx) based on product -- Created a line graph with four distinct lines correlating to each product. These lines vary over the course of the time period under evaluation, allowing the audience to draw conclusions based on the ebb and flow of each trendline.
3. Predict future top targets based on new prescriptions (NRx) -- The trickiest problem to solve–first, we find statistical descriptive counts such as the min, max, and mean. Then, if a doctor's total NRx count meets or exceeds the mean, he or she advances to the "next round" of candidacy. If not, they are excluded from consideration of being a future top target in accordance with this process. After narrowing the field using the average, we next use the cumulative delta change between NRx counts for the remaining doctors. For instance, consider an example doctor named "Nathan Chen" who has NRx counts of [1,3,14,11,12,19] over the past six months. In this case, to compute the delta change value, d, perform the following calculations: d = |(3 - 1)| + |(14 - 3)| + |(11 - 14)| + |(12 - 11)| + |(19 - 12)|. We then took the top 6 delta change values and output them in a line graph. The takeaway here is that the consistent improvement of these doctors makes them optimal candidates for future targeting, while the methodology used allows for users to pinpoint doctors who may not have top tier statistics right now but are trending in an upward trajectory. Compared to top performing doctors who are stagnant or even trending down, these perpetually improving performers are candidates to overtake current top doctors moving forward.
4. Database functionality -- Used for storing training data for the machine learning algorithms and for storing user information.

Future Improvements

Currently, IntelliTrack is only able to recognize limited data types from a span of around six months, but we'd prefer to improve it to be able to recognize any enumerable data types over n month timeframes.

The "dashboard" aspect of it was mostly devised as a fun way to give the web application a social media type feel. It turns out that this feature was legitimately useful and set us apart from competing web apps. If we could build out the interactivity components of the social media dashboard, it would increase its utility and usability for doctors and medical centers alike.