I am responsible for writing the 2020 CTSA Common Metrics queries for OMOP/BigQuery along with @rtmill and @apeshansky for other DBMSs. One metric is the number of encounters with at least one vital sign, which I am interpreting as any form of BP, HR, RR, and/or Temperature measurement. My ask: Is there an already established concept set for these common measures and if not, any advice for how to get a “good enough” set.
I did not find any relevant concept sets defined in atlas.ohdsi.org. or atlas-covid19.ohdsi.org. Using ATHENA, I see that LOINC has a classification concept (concept_id: 1003157) that incorporates lower-level LOINC Vitals measurements. But the 1263 descendant concepts from concept_ancestor far exceeds BP/HR/RR/T (e.g. cardiac index, LVEDP, O2 sat), even after limiting to only standard concepts and measurement domain_id.
Seeking advice how to proceed in the most efficient way.
Hi, I’m using a few LOINC codes for a subset of vitals contained in the MIMIC-IV waveform database (WFDB) - BP, HR, etc. This is a pilot/proof of concept level project. I’d be interested in knowing if there are better codes to describe vitals obtained in the ICU setting.
We need to standardize this badly, @mgkahn. We started that with smoking, but not yet with these. @mkwong’s list is a good starting point, but right now you’d have to create a comprehensive list of all possible standard concepts, unfortunately.
Hello, @mkwong! Can you provide us with more information about these vitals that need to be mapped?
And I also want to add that we don’t use Meas Value domain as separate domain in mappings (your variant for ABPm - 46237881, RR [rpm] - 36310358), it is not equal to the domain Measurement.
Hi, The MIMIC-IV waveform database reports out the following vitals - some of which I have assigned a LOINC code. It would be great to hear suggestions from the community regarding proper mapping in our next iteration.
HR
Yes - Heart rate
bpm
LOINC.3027018
NBPd
No
mmHg
NBPm
No
mmHg
NBPs
No
mmHg
PPV
No - Pulse Pressure Variation
%
PVC
Yes - Premature ventricular contractions [#]
/min
LOINC.21490839
Pulse (ABP)
Yes - Ambulatory blood pressure monitor study report
Michael - are you working on those vitals that are reported out of the monitors or include derived vitals like heart rate variability, ST, activity, etc.
So let me alter my request to the community: Anybody have a list of codes for “vital signs” that they like? Even better, a list that has been used in a OHDSI Network study (less likely to be local mappings). I am very into reusing rather than opening up a blank USAGI or Athena screen…
If you are looking for concept sets, we have a list we used in All of Us. It’s not perfect but we leveraged a combination of Loinc groups and loinc hierarchy for a set of vitals and lab concept sets. We are in the middle of updating the concept sets after Jim Cimino pointed out that the list was outdated (ie contained deprecated, non-standard codes). I can share the list hopefully next week if that works for you, @mgkahn.
@cukarthik Yes, yes, yes! I would rather use a “good enough” list that others have given some consideration than me opening up a blank USAGI or ATHENA screen and start typing “B-L-O-O-D-sp-P-R-E-S-S…” The CTSA Common Metrics measures are intended to be high level swags, not research-quality definitions so I will take whatever you have that you are comfortable sharing as a swag…
Let’s say we identify some candidate concept sets for vital signs like the LOINC one above or one that crosses LOINC and SNOMED as sketched out a little earlier on. Would it be useful to create one or more classifications that could be used to define / circumscribe these concept sets? I guess my question goes to whether there is a mechanism we might use to identify one or more “standard” concept sets for standard vital signs.
Usually, if both LOINC and SNOMED with the same meaning concepts are present, we map to a LOINC one. The SNOMED is used when the LOINC concepts are too detailed and such details are absent in a source data.
I would suggest to run a collaborative study, if you meet the pair of similar SNOMED and LOINC concepts , store it, and then the vocabulary team will review it and leave only concept standard.
@Polina_Talapova do we have such cases when the SNOMED - LOINC hierarchy exists, but they mean the same?
I had an impression that the hierarchy exists only when LOINC concepts are more granular
I am working through an example right now – a section of the WHO COVID-19 CORE CRF. The section is Module 2 → Signs and Symptoms. Measurements here are repeated on admission, upon entry to the ICU and daily as available. Signs and Symptoms includes many vitals, various oxygenation and Co2 measurements as well as a few consciousness / state of mind measurements. Mostly, I can annotate using LOINC. But there are a couple of exceptions that may be instructive. In a few days I will try to summarize my results.