It is common for the FDA and others to show a map of the United States with states color-coded by intensity to show the total number of inspections performed in that state. Indeed, the FDA includes such a map in its new publication dashboard for FDA inspections. Looking at this map with the United States map color-coded to reflect the location of medical device establishments, do you notice anything? Not to destroy the suspense for you, but it turns out the FDA tends to inspect where the medical device inspection facilities are. Really.
We wanted to get below those numbers in two ways. First, it’s much more informative to look at county-level data because there’s actually quite a bit of variation from county to county. Second, and most importantly, we wanted to normalize inspection data based on the number of installations. In other words, by looking at inspections by facility, we can get a better idea of the frequency of inspections in each county.
This is an interactive chart where you can move your cursor over the counties to see the particular percentage. Please do not read the percentages in the legend too literally. For example, 100% means that the number of inspections was equal to or greater than the number of businesses registered there. But as we all know, that probably means some companies have never been inspected and others have been inspected more than once in the 14 years. Thus, it is only a comparative measure that allows you to see the differences between counties.
It takes a little more work than you think and the product here should be considered an estimate. There are several places where we had to make assumptions and gloss over minor discrepancies. Here’s how we did it.
The FDA inspection database, at the time we performed this study, included dates from October 1, 2008 through September 15, 2021. We chose to leave out this entire period as it gives a smoother assessment on these 13 years approximately. In some years, the FDA may perform many inspections; other years not so much. We don’t care so much about year-to-year variations as about establishing a typical inspection level.
There were just over a quarter of a million FDA inspections during this time. We filtered the data to focus on national inspections, and carried out under the auspices of the Center for Devices and Radiological Health. There were about 33,000 left. The FDA conducts inspections for many different reasons, and we’ve chosen to focus on those that involve device compliance. That brought us down to just over 18,000. That’s the inspection pool we’ve chosen to use.
Then we went to the registration database. Obviously, the records change from year to year, so the fact that we’re using 13 years of inspection data adds some complexity. We chose to use the number of registered installs in fiscal year 2021. We believe that the geographic dispersion of registered installs has not changed so much over the past 13 years that it makes the results too imprecise. Also, the inspection database published by the FDA only included the name of the company, not the type of establishment. But looking at the range of establishments, it is possible that any of the different types of registered establishments will be inspected for device compliance, from manufacturers of medical devices to reprocessors of single-use devices, to exporters. We have therefore chosen to leave all types of establishments in the data. Obviously it would have been better to be able to filter both the inspection database and the registration database by type of establishment, but since that was not available on the public inspection database , we could not refine the analysis in this way.
We have merged the two lists using postal codes as the common geographic denominator. In the end, because of database imperfections and because of changes over time, there were counties where there were inspections, but there were no establishments registered in 2021. Fortunately, as you can guess, in those cases there was only usually one county-wide inspection, probably because there was a facility there that no longer exists. But the numbers were very low. We treated these counties as zeros.
Producing the visualization at the county level is a bit tricky. As you may know, postal codes do not correspond to counties. Many postal codes span multiple counties. So it took some work to create a crosswalk between the zip codes and the county data (using a database from the Department of Housing and Urban Development which includes the estimated location of businesses in the county) and we had to make some assumptions along the way. But, again, we think it produced a reasonable estimate.
As we suspected, there is a lot more nuance to the data than simply concluding that inspections are most prevalent where registered businesses are most prevalent. There are counties where the inspection rate is very high and others where it is considerably lower. Where you operate has a big impact on the likelihood of an inspection.
What interests us is that it does not necessarily appear to correspond to district offices or regions of the FDA. In other words, we do not notice any particular region of the United States having a higher prevalence of FDA medical device inspections than other regions of the United States. Indeed, at the state level, we do not really discern any trends. There are definitely states like North Dakota that don’t have hotspots. But honestly, that’s because North Dakota has few medical device establishments to begin with.
Ultimately, it looks like there are county-level hotspots, but not state or regional hotspots.
What does this map tell you?
©2022 Epstein Becker & Green, PC All rights reserved.National Law Review, Volume XII, Number 4