FDA Drug Recalls

If you or anyone you know has ever gotten sick ever, then this article is for you. 

As allergy/flu/general sickness season starts to set in, people may be picking up some more medicine soon. While we can typically trust what we get, every now and then, a drug gets recalled by the FDA. A recall is when some issue occurs with the quality or production of any specific drug and therefore its usage has to be stopped. To better understand the cause of recalls and if we should fear them, we did a deep dive into historical FDA Recall Data. 

TLDR: maybe don’t buy pharmaceuticals in the summer

Our data was the publicly available FDA recall database, which initially had the following columns:

We filtered only for the rows with Product-Types including “Drugs” as we wanted to only focus on pharmaceutical recalls in this analysis.

Unfortunately, most of this data is text data and doesn’t immediately give us much usage. Not to worry; we asked our best friend (ChatGPT) to label the Product Descriptions and Recall Reasons to categorize each recall with (1) a product family and (2) a specific cause of recalls. 

Here are the categories of product families:

And here are the causes of recalls:

Now we have our data. Let’s get into it!

Before we explore the causes or product families, let’s take a look at when recalls occur the most. We group by month and plot, and see something slightly interesting…

Why is there a huge spike in July and August? (if you look at the axis, you’ll notice that over 3 years, July had 23 recalls whereas May only had 5 - a 400%+ difference)

I did some research and found a major reason - high temperatures causing quality issues. A study in JSQH (Journal on Study of Quality in Healthcare) states that “The study showed a simple implementation of the combination of SPC tools, which demonstrated that the major contributors to recalls are microbiological quality issues, problems with product compositions, and packaging defects. Months that contributed by more than 60% of the total recalls were from May to August, November, and December. “.

First looking at proportions alone, we see that chemical impurities and ingredient issues make up the highest proportion (around 30% and 20% respectively). Contamination issues also are significant, whereas the other causes are less likely. 

But we also have a column that determines if a recall was terminated (i.e. if the FDA decided that the recall wasn’t needed anymore). Let’s see if there are any significant relationships here. 

If the light blue and dark blue bars are fairly similar, there isn’t any big difference. The two instances that are somewhat different:

Note: The percentages should add up to 100% but don’t. This is due to the labelling mechanism -- if ChatGPT didn’t know what to do, it would err on labelling recalls as nothing as opposed to guessing. This is why the numbers don’t add up. 

Sanitizers and injectables are the most common product to be recalled, but the distribution is still fairly even, especially when compared to the causes up above. 

Now a similar analysis as above:

Three standouts here:

These insights are important no doubt, but given that drug recalls don’t happen too frequently, it’s important to know that if you don’t act on these, you may still be okay. But they’re easy enough to just do them.