Risk Severity Estimation for Medical Devices as per ISO 14971



Dear Everyone,

When estimating the severity of a risk I often find that there are multiple possible outcomes (severities). Obviously, the conservative thing to do would be to go with the worst case but what if the worst case is not very likely?

How do you guys handle this? Do you go with the average case? the worst case? do you split the risk into two different risks?

The reason that I am asking is that I am analyzing a risk where part of a device breaks off and hits the patient. In most cases this is unlikely to cause serious harm but if "the stars are right" permanent injury or even death could be a possibility. Assigning a severity of "catastrophic" to the risk is not without problems ...


Jen Kirley

Quality and Auditing Expert
Welcome to the Cove!

Not being a medical device person myself, I will suggest you try looking at the similar threads shown at the bottom of this page.

I hope this helps!


Fully vaccinated are you?
I'm not a medical device person myself either, and I do not have a copy of ISO 14971, but in automotive and aerospace I always go with "worst case" when assessing "severity". If you have multiple outcomes, split the risk and address individually. If you do not, you better have a statistically valid reason (among other things) why you didn't because if the impossible happens (struck by lightning comes to mind, as does the GM ignition switch fiasco) you'll have a lot of explaining to do. I think the old saying is "better safe than sorry".


I understand that when Disney do an annual service on a rollercoaster they take it completely apart, inspect and replace things as necessary, then reassemble it. The first people to test-ride it are the maintenance team including the manager.

So for me the tests are: would I allow my child to be the patient? And would the CEO allow her child to be the patient?

Just 2c


Thanks you for the answers!

Splitting a risk to address the different scenarios based on the severity of harm and analyzing them individually sounds like a promising idea.

I could conclude that for a given event a "serious" outcome is "unlikely" and that a "catastrophic" outcome is "improbable". Both risks would be acceptable according to my acceptance criteria. Fantastic :)

But is this approach acceptable ???

One could argue that because the difference in probability hinges only on issues related to exposure that the different outcomes for the event must be treated as one risk?

If this is not so, then it would be possible to postulate that because each actual incident will be different, the probability component should be considered in relation to a single interaction with the equipment rather than with total amount of interactions over the cause of the equipment's lifetime.

At the very least this would be something that should be considered during "residual risk assessment"?


Quite Involved in Discussions
I do not know how happy I would be to participate in an extremely rare case when a improbable catastrophic event does happen ... :cool:

Assessing probability is wordsmithing for me, it is subjective, up to the assessment team, up to cultural and educational background, etc. I see no real benefit of emphasising too much on probability to be honest.



If you calculate 'probability' in with the severity aren't you then "double dipping" on 'Likelihood' in your risk assessment?


Hello Nusse.

Score the worst case scenario against the likelihood of the event
This is all about reasonably foreseeable sequences of events following the identification of hazards and not unlikely sequences. :agree1:


Risk is made up of two factors - severity and probability. Both have to be considered when evaluating risk, so a risk with high severity, but extremely low probability might be considered acceptable. It is the responsibility of senior management to determine criteria for acceptability of risk. Also, in the EU, the Medical Devices Directive requires that all risks are reduced "as far as possible" taking account of the "generally recognised state of the art".

There is also the relationship between a hazard, a hazardous situation and harm. An example of this might be a shark in a swimming pool is a hazard; a person getting in the pool with the shark is a hazardous situation and the person being attacked by the shark is a harm. Each of those occurrences has a lower probability than the one preceding it. Someone might not get into the pool if they know a shark is in there and even if they do, the shark might not be hungry.
Top Bottom