OHDSI Home | Forums | Wiki | Github

Visit Occurrence - What is a Visit? Why is there so much confusion?

CDM Builders
#1

The Visit Occurrence table has been getting a lot of discussion lately. See here, here, here, here, here, here.

Right now some standard concept_ids in the Visit domain are combined with Care Sites, such as 38004307, Adult Care Home or 8782, Urgent Care Facility. Other standard concept_ids in the Visit domain are combined with clinical events, such as 38004443, Child Mental Retardation and / or Developmental Disability Respite Care or 44777691, Well babies (care given by the mother / substitute, with nursing advice if needed). And this is confusing for people building their CDMs. Why are some combos allowed and others are not? And I’m going to put it out there, and would like for others to chime in, but I think we made a mistake when we expanded the Visit domain. Years ago we had a handful of high level standard concepts in the visit domain. These included OP, IP, ER, ER to IP, and LTC. And those five fit Christian’s definition perfectly. Then the standard concepts in the domain were expanded and now it is confusing because some encounter types are a perfect match to a standard concept_id and other encounter types aren’t present. As a community, we need to:

  1. Define the use case for Visits. What questions do we need to answer with these data?
  2. Define what a Visit represents.
  3. Review the structure of the Visit Occurrence table to ensure it fits our use cases.
  4. Review the current list of standard concept_ids in the Visit domain and add or subtract as necessary with remapping on non-standard concepts to standard concepts.
  5. Document, document and more documentation on the use cases, decisions made, rationale behind the decision, changes to the CDM (if needed), changes to the Vocabs (if needed), ETL guidance, and end user (analyst/researcher) guidance.

I’m tagging this as a Themis issue, but I would like the discussion kicked off here.

@Christian_Reich, @DTorok, @Alexdavv, @MaximMoinat, @jmethot, @clairblacketer, @Gowtham_Rao

1 Like
Mapping Oncology Encounter data to VISIT_OCCURRENCE - visit concept granularity
#2

The only one we are using, not from you standard list is, 42898160, which we are using for long term care visit:

image
image1587×101 5.17 KB

ATHENA says LTCP, is this the same as LTC from your list?

EDIT: Thank you for your answer below, I am adding this as an edit instead of making the thread harder to read. I have taken this under advisement and asked for clarification on my side.

#3

No, the concept you are using is not the same as the one I mentioned in my post.

LTC = Long Term Care, concept_id = 38004277, Long Term Care Hospital

In the US, these facilities are generally called Skilled Nursing Facilities (SNF), but also can include long term psychiatric facilities and other facilities where trained medical professionals provide 24-hour care to persons with chronic illness or injuries. Patients in these facilities aren’t sick enough to need hospital care where a physician is always on duty, but do need care provided by trained medical professions, such as CNAs or LPNs. The physician, PA or NP oversees the care and visits the patients at the facility. Other services are provided, but the distinguishing factor between a LTC visit and other custodial care (group homes, assisted living facilities, home care) is the presence of trained medical professionals providing care. It’s a murky difference and the definitions are usually regulated by the government.

The slight difference in concept_name and the ambiguousness in the concepts subsumed between your LTCP example and my LTC example further illustrates the need to clean up this domain. One subsumes “nursing home” and the other subsumes “nursing facility”. One is a child of “inpatient hospital” and the other doesn’t have a parent.

1 Like
#4

@MPhilofsky:

Thanks for starting this. @mik and I had the plan to work on the definition of each Visit. But then it frizzled.

Visits are supposed to be about the configuration of how healthcare is provided. Not about what happens. That’s in the other tables. So far, the elements of this definition are:

  • Patient coming to provider (of any kind) or the other way around
  • Patient retains control of his/her life or is put to bed
  • Service level is 24 hours or on demand
  • Service by category of provider (nursing facility vs doctor’s office)
  • Specific aspects of the service (closed mental institution)

It is not about

  • Specialty of an institution (pediatric hospital)
  • What happens during the visit (infusion clinic)

Obviously, there is gray zone in here. We just need to sit down. We will need an international crowd of curators, because usually these things are heavily influenced by how healthcare is run and paid for.

#5

What is a visit?

  • “Visit” was probably originally introduced by OMOP, and it tried to be different from the source data construct of encounters/events.
  • Encounters/events refer to timestamped observations of a person’s healthcare experience, involving a person, a date/time, and the care received.
  • Historically: OMOP’s concept of a visit is based on the patient’s perspective, e.g., if a patient sleeps in their own bed, it’s considered an inpatient visit, otherwise, it’s an outpatient visit.
  • The visit domain was then expanded to include clinic visits and emergency room visits, which slightly blurred the patient-centric perspective. i.e. blurry conceptId’s were made standard concepts of the visit domain.
  • Further complexity was added to the visit domain to capture granular details, some of which may be more relevant to administrative/billing systems than to patients, such as the HIPAA place of service codes. But these now introduced semantic ambiguity. This has led to scenarios where administrative codes might influence the designation of a visit, for instance, a patient may be wheeled to an oncology clinic from an inpatient ward due to hospital regulations.

Overall, I think the challenge arises from the vocabulary, not the Common Data Model (CDM). The issue is mainly due to the addition of standard concepts to the visit domain without a clear semantic understanding.

#6

My apologies if this is the wrong place for my question, but after looking at the links @MPhilofsky posted at the beginning of the thread, I think I’m close.

We use Epic and would like to know what other institutions with Epic use to determine the visit_source_value and subsequently the concept id. We’ve been using adt_pat_class_c, which seems to more closely match available concepts than enc_type_c. However, neither map well and it sounds like that’s the norm. Still, if anyone has better fields to use from Epic, we’d love to know. Please let me know if there’s a better place to ask this question.

Also, has there been and change on this topic of What is a Visit?

#7

Hello @WustlRob,

This thread is a discussion on how we define and differentiate Visits from Care Sites in the OMOP CDM. You should read this thread where I discuss the difficulties in mapping EHR encounters to the OMOP CDM. And please ask any questions there. Also, the Epic User Web has some information and discussion on mapping to OMOP.

In summary: There isn’t a 1:1 mapping between EHR encounter records and OMOP CDM Visit Occurrence / Visit Detail records.

1 Like
t