Reading a number of posts above, I think misunderstanding persists about "safety". In the 2nd edition of IEC 60601-1, safety was defined as "direct" harm. This means energy (electrical, mechanical, thermal), flow, shock, sharp points, radiation etc. coming from the device itself.
Looking just at the function or performance of the device, many devices are unable to create this kind of direct harm, in which case a definition of "no EP" makes sense.
With the 3rd edition, the term "direct" was quietly removed from the standard, in order to align it with ISO 14971. This meant the inclusion of "indirect" harm as well, such as failure to diagnose, monitor or compensate, which are all considered potential risks in ISO 14971.
The revised definition means failure to perform becomes a risk. For low risk devices, this risk may be small, but it still exists.
Our gut reaction is to ignore these kind of risks because of a hidden assumption that in the normal course of events, a manufacturer is economically incentivised to ensure the device works "most" of the time. Thus, when we consider for example "failure of a hearing aid to amplify sound", we assume unconsciously the failure rate of <1%, in which case the risk is acceptable. Most users are OK for a short time without the aid and will have a backup device in the home (they are normally sold in pairs).
However, if the failure rate was >1% , the risk should be unacceptable, especially if the cause of the failure is easily preventable.
So in general, EP should exist, but with fairly relaxed expectations for low risk devices. They should work most of the time in normal, real world conditions, but not necessarily in the "worst case normal conditions" that we often use as "normal use" in handling high risk issues.
Unfortunately, standards often tend towards overkill. For example, the 4th edition of IEC 60601-1-2 changed the test levels for ESD air from 8kV to 15kV. Does 15kV represent the normal environment, or is it an outlier?
If 15kV is an outlier, then a "no EP" claim makes sense for a hearing aid.
But there's a problem: how do we know the device works OK at levels that aren't outliers?
Consider a hearing aid that gives up the ghost with a relatively low 2kV ESD pulse: is that acceptable? It's winter, the first one dies out of the box. Get the next one, it dies one the 3rd day of use. Go to the shop and get a replacement. Both of those die after a couple of days, all due to poor ESD. The user can't work, join in conversations, walk around without fear of not hearing cars etc.
This should be unacceptable, because immunity to 2kV is run of the mill, nothing special, designers should be able to do it. There's no excuse. So the "no EP" is not a reasonable approach, but we need standard writers to differentiate between "general normal use" and "worst case normal use". This is not only for EMC, but for other tests that are designed to represent normal use, for example, water ingress testing in IEC 60601-1-11 (home use), IPX2 makes sense for serious risks, but a 30s test makes sense for lower risk EP.
You might say: surely manufacturers would do the right thing anyway? Well I have three case studies in the past year of exactly this kind of situation, unreasonably hiding behind "no EP". In each case the manufacturer said they recognised the problem, agree it's unreasonable, plan to fix it, but they have deadlines for FDA 510(k) applications, NB review etc so they want to hide behind the "no EP" to get a report as soon as possible. Which is another story ... !
