Do I need a LOINC code in my observation resource?

Hi everyone. I have a question related to this work I am doing (link to previous questions):

I am creating FHIR observation resources using data collected from point-of-care devices (e.g. bedside patient monitors), and I am trying to understand how to properly identify each observation. I have dug through the “vital signs profile” and a “PoCD IG” of the FHIR spec, and it appears the specs recommend (require?) the use of LOINC to identify each observation. It even refers to LOINC as a “magic value” which is kind of confusing me.

I feel compelled to use LOINC. But my challenge is that the PoC devices do not provide LOINC codes for each value. But they do provide the MDC code! My questions are:

Do all FHIR observations require a LOINC code? Or maybe just the subset listed in the vital signs profile? Or can I just use an MDC code?

It would obviously be so much easier if I could just use the MDC code because the manufacturer of the patient monitor provides it. Otherwise, I will need to map from MDC to LOINC.

It sort of depends on where you are and what data you’re representing. The core specification enforces that certain vital signs SHALL use specific LOINC codes. Other codes (from LOINC or elsewhere) can also be sent, but the expectation is that systems must map whatever codes they use internally to the appropriate LOINC code. This then allows clients to be confident that they can find key vital signs (pulse rate, body temperature, etc.) regardless of where or how they’re captured.

For other types of Observations, FHIR imposes no requirements, but countries might. For example, US-Core mandates the use of LOINC for lab (where codes exist) and the UK mandates SNOMED for the same.