OHDSI Home | Forums | Wiki | Github

Collecting the degrees of prone positioning (135° vs 180°)

CDM Builders
#1

Hi everyone,

it’s me again :slight_smile:

When placing patients in a prone position techniques differ between countries. Some hospitals tend to put the patients at a 135° angle (good lung down/up) whereas others put the patients flat on the belly (180° prone position).

From what I can see there are only two SNOMED concepts that cover this topic:

  • 431182000 Placing subject in prone position

  • 514971000000103 Placing subject in semiprone position (this is not part of the newest iteration of SNOMED CT as it seems - the concept cannot be found via the SNOMED CT browser)

    image
    image932×288 22.2 KB

I can see that the PROCEDURE_OCCURRENCE table has a modifier_concept_id. According to the documentation, this column is not primarily suited for such information but the alternative would be to store a separate OBSERVATION with concepts containing the degrees (creating new ones if they do not exist, e.g. 135°).

As an interesting side note: You will find 134/180 values ranging between 1 and 180 in the OMOP CDM (as part of the Meas Value domain and Qualifier Value concept_class). Fortunately, typical values such as 90, 135, and 180 are found in there and could be subsequently used as modifier_concept_id.

Any thoughts :slight_smile:?

#2

I think we can guess technique in your hospital :slight_smile:

Why do you want to use procedures? Snomed has a good hierachy for body positions. We put them in Condition domain, shouldn’t have done that probably, but still you can use them.

There are good targets for Semiproneposition, Prone position, Lateral position in this hierarchy.

Measuring angles in clinical settings is not really feasible in my opinion. You cannot measure them accurate enough. And I think that positions are a good substitute to be used in analysis.

#3

…too obvious I guess :laughing:

I was thinking about the procedures table because it fitted the description the best.

…This table contains records of activities or processes ordered by, or carried out by, a healthcare provider on the patient with a diagnostic or therapeutic purpose.

I agree that we never measure angles and use 135° as a surrogate for a semi-prone position. How would you map which side is up/down with the “semi-prone” approach or would you say this is not relevant for analyses either? I guess they would only make sense if you have a correlation to chest X-rays, CT scans or EIT devices. But you never know…maybe one day those analyses are possible and I’d rather have this information than let it go.

#4

Two things to consider here. Indeed it is relevant for case management. But for the RWE world when you have a lot of patients I would argue that it matters.

If it is your usecase, you can use more than one concept to postcoordinate clinical entity, which is the patient lying left/right side up/down in semiprone position. Left lateral tilt and Right lateral tilt, same Snomed axis. The patient could lie simultaneously in a semi-prone and tilted position, right?

Of course, you want to keep everything, but you have to let something go, it is inevitable. And not necessarily evil. If you want to keep angles, a good approach would be to find a concept in the Measurement domain, use correct units and put the results in value_as_number field. It is possible. However, after a brief search, I was unable to find anything suitable. So maybe you need to create a 2bil+ one, it is also a valid approach, less elegant though.

The process of rotation is probably a procedure, but the result of the procedure is what you are interested in. Observation is much better.
Eg. Appendectomy is a procedure. No appendix (in OHDSI - History of + appendectomy) is an Observation.

#5

This is a feasible approach, I think, and I will try to incorporate it into our system.

About the location of this information. Do you think using the modifier_concept_id would be okay here? I see that OBSERVATION also has this column though I would stick to the procedure table because the observation table does not have an end_datetime column. Calculating the difference between Observation A (patient in prone) and Observation B (patient in supine position) makes less sense to me compared to having one PROCEDURE_OCCURRENCE entry with a start_datetime and end_datetime.

I think this makes sense in some but not all cases. If I’m interested in the number of appendectomies performed by our surgeons I would need the actual procedures. Appendix absent (300309008) would not be enough. Interestingly this is a Clinical Finding/Condition and no Observation in the OMOP CDM - to make it even more confusing to the layman that I am :rofl:

#6

I see where are you going.

All the concepts from the abovementioned hierarchy of Snomeds are Conditions (might not be the worst decision to put them there after all). Condition table has condition_start_datetime and condition_end_datetime. I believe these fields are not often used, but they may be used for your usecase. And the beauty of this decision lies in its simplicity. You don’t need to create anything custom. It may be a bit counterintuituve, but it should work.

It can be measurement, condition, procedure, observation. Whatever works best for you. I think with certain exercises in debating skills and techniques this information may fit almost anywhere.

Regarding modifyers (from official CDM 5.4 documentation)

image
image863×232 8.35 KB

They are under construction at the moment, as you can see here - the main thread for modifyers, and here, and there are even more topics.
So ideally you need to pick one concept to be a modifyer for a procedure. But thicky thing. You don’t have one concept for each position. So you may need to create new concepts.

Absolutely. But when the patient comes and says “I had an appendectomy 25 years ago”, the best you can do is History of event longer than 10 years ago + appendectomy + patient reported.

1 Like
t