Inspiration
Our project was inspired by the need to address the issue of adverse food events and provide valuable insights to regulatory bodies like the FDA. Adverse event data can help regulatory agencies like the FDA identify and address potential safety concerns early on, before they lead to widespread harm. Analysis of adverse events allows the FDA to evaluate the safety of products and make informed decisions about their regulation,
What it does
We identified patterns and trends in the data to understand the relationship between products, adverse events, and outcomes. We developed a scoring system to quantify the impact of different products on consumers. We conducted a case study on Kratom and Johnson and Johnson products to investigate the underlying patterns in adverse events, and the contexts surrounding the product.
How we built it
We used data cleaning and processing techniques to prepare the dataset for analysis, employed various visualization techniques to explore the data and extract meaningful insights, applied statistical methods to analyze the data and identify correlations between variables, and reviewed outside literature to inform our analysis.
Data Wrangling
We cleaned the data using Pandas by removing null values and other anomalies, Exemption 4 products (due to their lack of specificity), as well as standardizing the ages for all patients. We also separated all case outcomes according to individual cases to look at the distinctive impacts of products.
Visualization
We used a variety of visualizations, including bar plots, heat maps, violin plots, histograms, pie charts, word cloud, and more. We also used a Latent Dirichlet Allocation model to analyze the association between sets of symptoms and find the top 30 most informative adverse events.
Findings and Analysis
We created a scaling system that assigned points to specific outcomes based on their severity, and averaged the point values across all product descriptions to identify the mean outcome for every product. We report that Alcoholic Beverages, Powder Formula, Medical Foods, Vitamins/Supplemtns, and Meal Replacements achieved the highest overall outcomes scores overall, indicative of relative severity across all reports. We looked at deviation within this scaling system to identify what products did not show regular outcomes. The FDA should employ better strategies when regulating these products. We also analyzed two case studies that exhibited notable trends in our analyses and incorporated outside socio political data to obtain a better understanding of the trends in cases for Kratom, as well as Johnson and Johnson products.
We request the FDA to implement better regulation on supplementary products and cosmetics across the board. When looking at specific products, Kratam and Johnson and Johnson and any affiliated products ought to be banned immediately.
Technical Skill
This project involved us going beyond our comfort zone through exploring new libraries, visualization techniques, and methods to parse through data. We also created a model scoring system based on severity of outcome in order to use our given data to predict products that should be removed. This scoring also predicts the severity of any incoming reports based on previous data.
Social Impact
In addition to thoroughly analyzing our categorical and quantitative variables, we also employed an interdisciplinary approach by extending beyond what we were given, and finding socio political evidence to support the trends in our case studies. Not only did this make our observations more sound, it also increased the scope of our project to involve a holistic approach.
Log in or sign up for Devpost to join the conversation.