Day 11 – what have we learned (part 1)

This post is a part of a series on drug shortages.

Today I’d like to take a step back and attempt to weave together some of the threads I’ve pulled on over the last few weeks, and perhaps begin to identify some larger takeaways, gaps that need filling, and follow-on questions.

In this post, we will review measures of drug shortages that give an overview of the problem, and then in later posts look the impact of shortages (e.g. essential medicines list), and data quality.

Drug shortages: the 5 Ws

This section reviews findings that locate drug shortages in Canada in aggregate to help us understand the scale and breadth of shortages. I have intentionally been descriptive in my exploration by relying on very simple measures and charts to display trends and patterns. Perhaps with more specific questions or hypotheses, I will turn to statistical tests.

Who: which manufacturers are involved in shortages?

The main bit of work I did to answer this question was to provide an updated figure comparing of shortage rates for the top 10 manufacturers, but I find those particular figures confusing, so I won’t repeat them. The basic question they are meant to answer is: does the size of manufacturer relate to the number of shortages they have? So, here is just a straight up plot of # new shortages per year vs. # of marketed products per manufacturer:

Shortages per manufacturer size

Are they correlated? I haven’t checked, but my guess is not very well. I suppose that’s a takeway itself because you might have expected to find larger companies either better or worse at managing shortages, but instead nothing in the data jumps out to push us one way or another. However, you can see I’ve included a few company names, which lets me gesture at the outliers: e.g. Teva seems to have a low shortage rate for the number of drugs it produces when you compare it to, say, Apotex.

A more fulsome analysis is needed to say much more but I could see this being a useful, although complex, line of investigation to dig into. With some work we could begin to identify companies that are performing better or worse than their competitors but to do that you’d want to make sure you were comparing manufacturers in relevant ways – i.e. by more than just the absolute number of drugs they produce.

What: which drugs are involved in shortages?

The bulk of the aggregate analysis I did on this question was related to the essential medicines list which I discuss in the next part. However, I did update that monster of a chart from our paper which shows medications with the most shortages lumped together by ingredient group. (If I was to do this again, I would group by ATC code – roughly the same thing, if I recall correctly.) I’m not sure if I can summarize the data much more than pointing to the chart itself. It’s not pretty.

Top 10 lists are fine, I suppose, but they seem a rather narrow window into the understanding which drugs are involved in shortages. I want more rich ways of digging into the data here, but I’m not sure what the question really is. Perhaps its that I want to know which kinds of medications are being affected by shortages, over time, and I especially want to know about those classes of medications affected out of proportion compared to other classes of varying size. Perhaps what’s needed is a look at the data via ATC classification at different levels of granularity to see the shifts in shortages over time and be able to zoom in to where the action is.

The takeaway: on the scale of ingredient groups there does not appear to be any trends in which drugs go into shortage.

When: the timing of shortages

Here are the essential figures from my look into the flow of shortages over time:

New shortages per month
Drugs in shortage over time
Shortage duration

The first chart shows us that the rate of new shortages has been roughly constant or slightly decreasing in rate over time. The linear regression line in the chart takes into account the initial spike in shortages when the DSC database first came online, and I’d bet that if you excluded the first few months the slope would be shallower still.

The second chart shows that in the last two years the number of drugs in shortage in any given month has been, again, roughly stable (between 1700 and 2000 per month). The sharp increase in drugs in shortage in 2017 is likely an artifact of poor capture of drugs shortages in the voluntary drug shortages database that preceded the DSC. In other words, voluntary reporting didn’t work.

Lastly, the third chart shows what looks to be an average decrease in the duration of shortages from 2017-2019 within the expected power law distribution.

The takeaway: voluntary reporting didn’t work, drug shortages appear to continue to happen at a roughly constant rate, and the average duration of shortages has been decreasing.

Advanced warning

Apparently manufacturers are supposed to give 6 months of warning when possible and if unable to do so, at least file shortage reports with a few days of their start. I looked into that:

By shortage reason, the above chart shows a histogram of the time between filing the report and the shortage starting for 2019. The grey box shows the 6 month period before the start date. If all reporting was going to plan we should see all of the reports happening to the right of the grey box (i.e. with 6 months of warning), and certainly none of the reports several months after the shortage start. Clearly things aren’t going to plan. You might rightly say that 6 months is perhaps too optimistic a time period for advanced warning. Offhand, I don’t see any good reason for the late filings. I’d be curious to know how many shortages reports are filed after the shortage itself is over… My guess is the number is greater than 0.

One thought I had to extend this analysis is to start looking at particular manufacturers (blame game!) or drug classes to see if we could identify factors driving delayed the reports.

The takeaway: reporting has not been timely, and reports can be months late.

Why: reasons for drug shortages

This question has tendrils that run deep, and I have only taken a very superficial look: the stated reason given in the shortage report. Here’s what I found:

I’m not sure what more to say. The shortage reports only use a very short list of possible reasons (that’s all 6 of them in the chart above), so it’s not possible to do much more of a fine-grained analysis. That said, there is a space for free form comments which could be mined, but only ~18% of reports have those filled in, and from a cursory look they are mostly garbage.

The takeaway: manufacturing disruption is reported to account for over half of the shortages in 2019.

5Ws in summary

What have we learned since our last look at Canadian drug shortages in 2018?

  • Shortages continue to happen at roughly the same rate (maaaaybe slowing slightly) but appear to not last as long. Nevertheless, there has been between 1700-2000 drugs in shortage in any given month over the last two years.
  • Over half of the shortages are caused by manufacturing disruptions in 2019.
  • There is _a lot _of variability between manufacturers in their rates of shortages. More investigation needs to be done to tease that apart.
  • Manufacturers almost never file shortage reports with 6 months of notice, and often are >1 month late in filing.
  • Voluntary reporting didn’t work which we know because we’re seeing the rate in new shortages jumping significantly with mandatory reporting.

There’s a lot of further exploration to be done. In particular, much of the above analysis did not factor in discontinuation reports. Additionally, I think there’s something interesting to be explored in the flow of shortages, perhaps with some more sophisticated event modelling.

1 Comment

Comments are closed.