Expected service life comes from IEC 60601-1.
In the 2005 (third) edition it is stated to be:
"maximum period of useful life as defined by the MANUFACTURER"
In its integral update (3.1 edition; currently valid in nearly all situations) it is stated to be:
"time period specified by the MANUFACTURER during which the ME EQUIPMENT or ME SYSTEM is expected to remain safe for use (i.e. maintain BASIC SAFETY and ESSENTIAL PERFORMANCE)"
With the note: "Maintenance can be necessary during the EXPECTED SERVICE LIFE."
Beware you note 'define' instead of specify. Are you following the appropriate edition?
Two key things then are:
Specified by the manufacturer, meaning you state it.
During which it is expected to remain BASIC SAFETY and ESSENTIAL PERFORMANCE.
These last two are risk management core principles.
In short: first do no harm, second do the (clinical) good you must do to avert (an increase of) risk.
Thus reliability of combination of components leading to either aspect remains important.
If you always fail-safe, you don't have basic safety concerns that affect expected life. [This is the main concern to prove for 'shorter than expected service life' replaceable parts; and you could make a case for preventive/predictive maintenance through replacement (think tires of your car) to also serve this].
If you have no risky absence of clinical performance you have no essential performance concern.
The twist comes when you do have some failure modes (both unintended/unexpected as well as intended/expected) that increase risk, and often a case can be made that such risks do exist (but might be accepted!).
Hence on it becomes a bit of a discussion and interpretation, where usually the reliably predictable absence & accepted presence of (unsafe) failure modes determines what is useful life. Needed is a mastery of the risk management report and where you decide what is and is not acceptable risk, and how you justify accepting unacceptable risk becomes key inputs in the cut-offs for this, taking note that given the 'right' state-of-the-art indicated by standards and competitors these need not be zero risk.
Widely understandable example: as long as your post-market surveillance observes the occurrence and severity of irreducible (if you want to maintain your main effect) side-effects for medication to not exceed the levels you set as acceptable (based also on your competitors/society requirements) for within-shelf life product, the occurrence of the side-effect does not mean you broke basic safety or essential performance.