There seems to be a range of opinions on how to associate clinical observations generated by devices with the settings in effect on the device at the time the clinical observation was made. Associating the device settings and clinical observations is important to provide context for the clinical observations, and to provide historical evidence. 

Some have suggested that the clinical observation should use hasMember relationships to the settings observations (either directly or indirectly through a "settings set" observation). This has the advantage of unambiguously tying the clinical observation to the settings in effect.

Others have suggested simply recording the settings observations either when they change, or at defined times (daily at noon, at the start of a procedure, etc.) and recording the clinical observations without regard for the settings; the settings in effect for any clinical observation can be determined by finding the most recent setting observations (for each of the device metrics) prior to the clinical observation. This approach doesn't laden the clinical observations with concern over the settings observations; but makes it more difficult to detect a missing/deleted setting.

There is also some confusion about the difference between device settings (recorded as Observations linked to DeviceMetrics linked to Devices) and device properties, and in general how to record device settings.

An example or guidance that ties these all together, perhaps starting with a clinical Observation linking to settings Observations, would help provide a common approach to these challenges.