TL-DR Health

Insurance companies and insurance plans have frequently appeared in the news throughout the past few months, frequently appearing with negative headlines headlined by data showing astonishingly high denial rates and concerted efforts to continuously limit the responsibilities of coverage. The black box surrounding insurance coverage has only continued to grow and there seems to be no solution to truly help consumers understand the companies that they are entrusting their livelihoods and even their lives to. The growing uncertainty led to increasingly more polarity, especially in terms of civil discourse, with these emotions erupting in the public assassination of a healthcare CEO that was captured on camera and seen my tens of millions of Americans.

This is where the idea of an AI model that could accurately and honestly inform people about their choice in health insurer. What is now often seen as a shot in the dark could become a decision made upon incredibly relevant and accurate data. We wanted to build a model that could give the power back to the people, as the truth is insurance companies have become extremely capitalistic and it is only through the influence of the educated consumer that true change can be made within the luxurious conference rooms of these extremely profitable corporations that are supposed to help those when they are at their weakest or most vulnerable.

We wanted our model to be able to provide users the most up-to-date and relevant data on the health insurer choices they have access to. Because of this our first goal was to be able to find and properly integrate data that gave valuable insights into these insurance companies. We were able to reach this goal and this resulted in us having a dataset containing the top insurance companies in New York City as well as data on the premiums, deductibles, max out-of-pocket costs, policy claim rate, and policy denial rates by plan level for these companies. From this we were able to train a classifier that rated each company based on a variety of feature weights and provide a general overall rating for each dataset.

From here, we realized that we wanted to provide comprehensive medical information for our consumers and as such developed multiple models in order to predict annual medical costs based on specific individual metrics. This actually turned out to be extremely helpful as it helped us to continue to make adjustments on our dataset of New York City insurance companies. With this additional information about a consumer’s estimated annual medical spending, we could add or subtract feature importance from critical factors such as deductible amount or max out-of-pocket spending. All of this resulted in us being to confidently recommend the best insurance company and policy level for an user based on all of the information provided to us.

TL-DR started off as an ambitious project intended to rebalance the scales for the consumer against for-profit conglomerations that claim to serve as a safety net. And yet here we are, less than 24 hours later, and our model is no longer just an ambition but a reality in being able to give accurate policy recommendations to users based on a small number of individual features. We are so excited to be able to share TL-DR, even after such a short time working on it, as we truly believe it has the ability to revolutionize the way insurance policies are not only bought, but also marketed and honored first in New York and then throughout the rest of the country.

Built With

• css
• fastapi
• google-cloud
• groq
• jupyternotebook
• next.js
• python
• tailwindcss
• typesscript