Prescription for Safety

Inspiration

As a team, we were inspired by the reality of the situation we ourselves are in. Products being used daily by thousands of Americans are going unregulated, which really hit close to home. While the FDA does work to collect adverse event data and identify any dangerous trends, we currently feel that there is a lot of room for improvements in the FDA guidelines and problems they have identified, and in highlighting concerning trends we hope to contribute to the process of strengthening the guidelines to help keep everyone safe. We feel that the biggest concern to us was the power of large corporations lobbying to prevent safety information from being made available to the public as well as a low amount of regulation and oversight in certain industries like the vitamin and supplement industry.

What it does

Our project both analyzed the given data on a high level as well as on a lower level, taking both different categories and subsets into account. One major part of our analysis focuses on symptoms correlated with the adverse events, and we developed code to identify the most frequently reported symptoms and group and analyze them based on a number of categories, including product category, product name, sex, age, etc. We also broke down the adverse events into outcomes and were able to identify that a statistically significant proportion of hospitalizations were caused by vitamins and supplements. We analyzed specific cases we found in the data as well, including the large number of cosmetic-related ovarian cancer cases with the product name excluded under Exemption 4.

How we built it

We built our project in both Python and RStudio, using the pandas library in Python to help with analyzing the data and using R to create workable subsets of data. We also used Excel to search through the data and give us an initial understanding of the dataset to set goals in our data analyses.

Challenges we ran into

We ran into challenges with generating proper subsets, data cleaning, and identification of our end goals for the project, as well as addressing new issues that came up while programming and working with the data. With the very large size of the dataset, the variety of different ways information was listed or spelled, and the open-endedness of the project goal, going through the data and identifying a few key areas to narrow our focus for our final project proved to be a challenge.

Accomplishments that we're proud of

We were able to create a program that grouped symptoms with products used so we could do a high level analysis of frequency of symptoms. We are proud of identifying a potential dangerous product, Super Beta Prostate, through our high-level analysis of symptom frequency categorized by product name. We are also proud of being able to identify discrepancies in the numbers of ovarian cancer cases that show a statistically significant difference from the national average, which indicates a statistically significant likelihood that the cosmetic ovarian cancer cases are the result of an adverse effect from the product.

What we learned

We learned a lot about data set manipulation, especially with large data sets. We didn't have much experience with data science, so we learned a lot about how to use tools like pandas for working with large datasets and how to analyze and draw conclusions from large, open-ended data sets. We also learned a lot about how the FDA's adverse reporting system works, a number of past and current incidents involving adverse effects from various products, about FDA legal policies, and about potentially dangerous products and trends in the data.

What's next for Prescription for Safety

We hope to learn more about FDA regulations and the process for changing them. We also plan to conduct more research and analysis regarding the products and trends we focused on in this project. In addition, limited regulations in the supplement industry leads to misinformation that leads to a tarnishing of trust between consumers and companies that we hope to bring awareness to.