OHDSI Home | Forums | Wiki | Github

Procedure. Anesthesia concept


Hi, I am anesthesiologist who want to map anesthesia procedure concepts correctly to CDM.
The problem is one case of general anesthesia consists of several concepts. I just wonder how all these attributes put on one procedure occurrence.

This is the typical case of general anesthesia, procedure.
On arrival to the operating room, patient received ✱ lidocaine 40 mg and 1% propofol 80 mg for anesthesia induction, and ① rapid sequence ② tracheal intubation was performed after ③ muscle relaxation with rocuronium 80 mg IV. ④ Sevoflurane 1 to 2 vol%:o2:air was administered for maintenance of anesthesia.

① Rapid sequence induction
② Endotracheal anesthesia
③ General anesthesia and muscle relaxant
④Inhalation general anesthesia

✱ Intravenous induction

(Qi Yang) #2

I just wonder how all these attributes put on one procedure occurrence.

They are not put in one procedure occurrence. They get put into multiple records in Procedure_Occurrence table. Each concept represents one record. If you want to link them together, you need to use Fact_Relationship table. However, Fact_Relationship table does not give you a high level view of the entire process. Right now, two new tables - Episode and Episode_Event tables are introduced in Oncology Extension model. There is thinking now to use these two tables to represent other area besides oncology such as this one

(Christian Reich) #3

Hang on, @QI_omop, let’s first understand what’s going on.

@pissces: We usually don’t record intention. So, things like “induction” are not explicitly stated, instead, the lidocaine 40mg iv etc. speaks for itself. So, in your case you would just create DRUG_EXPOSURE records of those drugs.

Would that work?


Thanks. @Christian_Reich

Of course the drug should go to DRUG_EXPOSURE records. But ‘induction’ is not an ‘intention’ but type of procedure. As you can see fig. 2., there are four types of induction of general anesthesia and intravenous induction is one of them.

Thanks @QI_omop.
Anesthesia procedures really need those tables in Oncology Extension.

(Christian Reich) #5


Look. You are saying “for anesthesia induction”, “for maintenance”. This is what I meant by “intention”. You could also call it “purpose” or “step in a process”. From an OMOP perspective, the procedure for all these is “Drug injection”, and the Drug Exposure is “Lidocain”, “Propofol” and all those. The fact that the anesthesiologist has something in mind when doing these injections is not the intent of data capture in the OMOP CDM. Because it happens in the mind of the anesthesiologist, not in body of the patient.

Of course, you could do what @QI_omop said, and that is a legitimate way of capturing this. Except no analytical tool will have a clue what it means and even look for it. So, you would solve the problem in your own data set, but you couldn’t network with anybody else.

Does that make sense? Can you share the analytical use case you want to investigate? That would help with giving you the right advice.


Thank you for your help.

Let’s say I want to know the outcome according to the induction type such as intravenous induction vs. inhalational induction. This is common subject of anesthesia research.

These are few examples

If there is a drug-exposure record of Propofol without a procedure record of induction type(id:4082996) on the same day when anesthesia was delivered, we could presume Propofol was injected for anesthesia induction, but the problem is Propol could be used at other instance not for anesthesia induction for example treatment for nausea or seizure.
For precision of information, we just can not assume Propofol was induction agent just because it was injected on the same day as anesthesia delivered. That why we need concept specific to induction procedure type.
I hope this helps.

(Chris Knoll) #7

I think inhaled vs. injection can be identified via the dose form of the drug in drug exposure.

I think what @Christian_Reich is saying is that the CDM focuses on what happens to the person, not what was in the head of the person doing the drug administration (were they thinking they were treating a seizure or administering an anesthesia? The CDM answer is: it doesn’t matter, only that the drug entered the body). The CDM concern is definitely ‘how’ the drug entered the body, but this can be captured at the drug exposure level, and not linked over to a procedure.


Well, then I just wonder what is the use of concept intravenous anesthesia induction?
Also, we could identify inhaled vs. injection but we need to differentiate iv drug is used for anesthesia induction or other treatment to see the outcome according to the induction method.
Both Inhaled anesthesia and iv drug can be used in one patient, then we don’t know which one is induction agent?
That’s why we need both procedure and drug exposure record, respectively.
Maybe I don’t understand CDM well and Anesthesia procedure may not be suitable for CDM.
Thank you for your help.

(Christian Reich) #9

As @Chris_Knoll says: It’s the fact that matters. Not the intention. The fact is that the patient got lidocain and propofol IV or inhalant.

Absolutely, and we want to support it. IV lidocain injection is done by using a drug product that is formulated for injection. These are all the descendants of RxNorm 35604127 Lidocaine Injection. The inhalants are children of 21113643 Lidocaine Inhalant Solution. For propofol it looks like there isn’t any inhalant on the market, only Propofol Injection. Is that correct?

@pissces: If you need to understand how the data are represented and used for analytics please look at the Book of OHDSI or the Tutorial. Both will help you to wrap you head around it. Or keep asking here, but that will be the slow route.

(Chris Knoll) #10

Just to answer this one: one might imagine a use case where they want to exclude people who experienced an intravenous anesthesia induction of any kind. The drug doesn’t matter, it’s just the scientific question says you don’t want people who underwent anesthesia. So, that concept will identify those people. it’s better than having to go through saying ‘not exposed to drug x for anesthesia induction, not exposed to drug y for anesthesia induction, not exposed to drug z for anesthesia induction’, etc etc, you can see the pain that would be.


Exactly, I am not saying propofol itself map to intravenous induction. What I am saying is the case itself map to intravenous induction to include those who received intravenous induction with any kind of drug.
Patient 1 received propofol for anesthesia induction.
Patient 2 received midazolam for anesthesia induction.
Patient 1. drug exposure record for propofol
Procedure record for intravenous anesthesia induction

Patient 2. Drug exposure record for midazolam
Procedure record for intravenous anesthesia induction

Then I can use concept intravenous anesthesia induction for cohort generation.

(Chris Knoll) #12

So, I think a procedure record and a drug record out of the single event of ‘received propofol for anesthesia induction’ is what I’d do.

This is similar to the case of ‘chemotherapy’…saying I am undergoing chemotherapy doesn’t say anything about the drugs involved in the chemo, but it’s definitely an important observation in a patients procedure history that they underwent chemo. So, if you received a source record saying ‘chemo with cytoxan’, I think it would be appropriate to record the start of chemo in the procedure_occurrence and the individual drugs used at the various times in drug_exposure’. Some sources, you don’t get the drugs, you just know they underwent chemo. So you couldn’t do a drug study on these people but you could potentially exclude them from another clinical question by nature that they did have chemo.

Note, I don’t want to start an argument here around ‘chemo is a drug exposure’, I was just explaining how I think of it and how I compare it to your case of anesthesia.

(Antoine Lamer) #13

Dear @pissces,

Do you plan to use a standard vocabulary (as SNOMED) to characterize these facts ?
I also work on the mapping of operating room data to OMOP. I finished the structural mapping and am currently dealing with semantic mapping (we used many local vocabularies in France).


Hi @AntoineLamer

Yes, I want. I visited your site. It was interesting. I will attend OHDSI Europe 2020 at Oxford. I hope to see you there and hear from you in person.