The Oxygen saturation by Pulse oximetry profile should not require both the 59408-5, Oxygen Saturation by Pulse oximetry and 2708-6, Oxygen saturation in Arterial blood LOINC codes.  I realize that the less granular 2708-6 code is required by the base spec but, if both codes are used in the US Core profile then the only situation in which a system can be compliant to the profile is if all oxygen saturation observations are done via pulse oximetry.

The seeming intent behind FHIR core using the 2708-6 code is to act as a grouping code for all other oxygen saturation codes.  It is true that most of the more granular oxygen saturation codes can be inferred as being descendants of 2708-6.  But, not all of them can be inferred as being vital signs.  Some of the descendant codes would be included as a vital sign based on contextual use while others would not be included at all.  Also, the inference is just that, an inference.  2708-6 is not a LOINC group code and (as far as I know) there is no computable way to link the more granular codes to it in the LOINC terminology model.  There is a LOINC group code for oxygen saturation, LG33051-0 Oxygen Saturation and 2708-6 is listed as a member.

Our proposal is that the FHIR core Oxygen saturation profile should use a value set binding to Observation.code, limited to very few LOINC oxygen saturation codes that are widely accepted as part of vital signs.  By doing this confusion is removed from profiles that derive from it (US Core for example) by not forcing them to use multiple, different codes and also maintaining compliance to the core spec.

We also think that using the LOINC group code is insufficient due to the fact stated above, not all its members could be considered as a vital sign.