| OHDSI Home | Forums | Wiki | Github |
Hello,
Looking for some guideance on best practice for populating discharge_to_concept_id when a patient dies. Assuming this is just concept_id = 0 (unmapped).
Examples are:
Of each of these discharge dispositions, only the medical examiners office seems like a medical visiit…
Any insight appreciated.
Piper
Seems you are asking two questions. 1) What do use as a discharge_to_concept_id? 2) Should there be a visit_occurrence record?
I agree with your assumption to set discharge_to_concept_id = 0 because there are no disposition concepts for the morgue, autopsy or funeral home. But then your statement “only the medical examiners office seems like a medical visit…” seem like asking if transferred to [ morgue, autopsy, funeral home] constitutes a visit. The second question depend on whether the transfer part of an existing visit, or being treated as a stand alone event. Regardless, create a death record.
Thanks @DTorok this helps.
I wasn’t thinking in terms of addding a record to visit ocurrnce for an autopsy “visit” to the medical examiner. I was thinking more in terms of the appropriate concept_id for field discharge_to_concept_id.
The research use case would be identifying a cohort for whom autopsy reports exist.
Thanks again,
Piper
Visits are generally assumed to be encounters of the patient with the healthcare system. So, after death we wouldn’t expect any more visits, at least in this world. So, leave the discharge_to_concept_id empty.
Autopsy reports would create Condition records. The Type Concept would indicate where the information was obtained from. We currently don’t have “Autopsy report” in there, but could add that. Are you doing research with autopsy reports?
I don’t think that one additional short visit will interrupt something. Sounds like a proposal for convention changing.
As well as some Lab Measurements. Wouldn’t it be more consistent to use Visits rather than _type_concept_id while filtering them out?
I think the contrary is true. Visits are relationships between patients and the healthcare system, and their choice is causing outcomes. None of that obviously is the case any longer. All we do is to collect some final data, and the provenance is defined through Type Concepts.
If I was tasked with defining this use case, my cohort would consist of Person’s with an autopsy report in the Note table.
@Christian_Reich, to my mind, an encounter with a medical examiner for an autopsy (procedure) strikes me as important kind of encounter with the healthcare system – at least from a research pespective. The goal is for healthcare system to learn (post mortem) what we were not able to determine when the person was being treated.
The other examples (funderal home, morgue) strike me as potentially useful simply because they reflect a discharge disposition that I see in our data set - and I suspect many others have in theirs. As is, when a patient is released to a funderal home or the morgue, the discharge_to_concept_id can only be set to ‘0’.
@Alexdavv Visits strike me as the ideal way to catpure this too.
@Christian_Reich - At the individual patient level I agree - an autopsy does nothing to change outcomes for this particular patient. From the perspective of understanding a particular disease state (or adverse response to a drug or procedure) the autopsy provides crtical clinical information that is potentially significant for improving outcomes for folks with the same disease state (or being exposed to the same drug or procedure). The fact that an autopsy visit occurs also strkes me as an important event in itself.
Totally agree. Except, this is not a Visit. But we should add the autopsy report to the list of Type Concepts. So, insights derived from an autopsy report will be tagged that way.
Thoughts?
What is the use case? Why does it matter to large scale observational research where the corpse of the patient goes?
Thanks everybody for this important discussion/information. Where are the autopsy reports collected, and how can they be accessed?
I am too new to OMOP to appreciate the signifiance of this decidion. I derfer to you and others who have deeper experience and knowledge than I do!
That said, I do worry about cases where we have information that a person was dsicharged to the ME but don’t end up with the report itself. I also have a bias towards minimizating the number of times we map data to concept_id = 0 (unmappable). If we know where a patient goes upon discharge, it would be nice to be able to capture it.
Piper
Upon furhter relection, I agree with you.
This was a misdirected attempt to capture a person’s discharge disposition as completely and accurately as possible. Upon further reflection it seems to make sense to say their journey with the healthcare system ends here - just as we do for people who are discharged to home.
Nothing wrong with that. But if it is information we don’t need for our use case we drop it. I know it causes anxieties, but we are not running healthcare institutions. Our use cases are analyzing natural history and epidemiology of disease and the effect of healthcare interventions. What happens outside that realm, i.e. in people’s lives that is not affecting the above, or people’s afterlife - we can happily drop. Makes data clean.