OHDSI Home | Forums | Wiki | Github

Visit_occurrence, Outpatient and Long Term care visits

CDM Builders
#1

Hello!
I have a question about building Outpatient and Long Term Care visits. In the source Outpatient records, we have an intersection between service date start and service date end - for a person going to the same provider, at the same place of service. (please see all possible cases in the attached excel file). Outpatient data.xls (9.4 KB)

How can we build Outpatient visits for such data? For now, we have 3 ideas.

  1. Create Outpatient visits for each distinct combination of service start date and service end date (svcdate and tsvcdat).
  2. Create Outpatient visits for each distinct svcdate, ignoring service end dates. In this case, we assume that an outpatient visit couldn’t be longer than 1 day.
  3. If there is an intersection between service start and end dates, then to take min(svcdate) as visit_start_date and max(tsvcdat) as visit_end_date.

If any of these ideas is correct? How to identify Outpatient visit?

P.s. the same question we have for building Long Term Care visits

#2

@TBanokina:

Outpatient visits cannot be more than one day. Where does the patient sleep? Under the doctor’s desk?

That shouldn’t be a problem. You can go from one doctor to the other.

Same as the first.

What’s the question? Long term care can be more than one day (typically a lot more than one day), but only one at a time. if it overlaps in the source data you got a problem.

#3

So, to summarize topic on Outpatient visits.
If we have difference between service end date and service start date > 1 day. For example 32 days. We shall create just one Outpatient visit on the first day? or 32 visits on each of these days?

#4

In our case it overlaps. And the question that we have about Long Term care visit is “how to merge them correctly in one visit” ?

#5

No. You cannot have an outpatient visit of 32 days. So, it is either one day (the mistake is in the service end date), or you really have an inpatient visit. I can’t tell which one. You can try looking into the procedure data to find out if whatever happened in those 32 days really is inpation treatment or not.

#6

@Christian_Reich

Seems like some Outpatient visits could be longer than 1 day (please follow the link, examples are in the table)

Description of the procedure codes from the sample above, when a patient has Outpatient stay longer than 1 day:
E1390 - Oxygen concentrator, single delivery port, capable of delivering 85 percent or greater oxygen concentration at the prescribed flow rate (source domain Device)
E0601 - Continuous positive airway pressure (cpap) device (source domain Device’)
T2021 - Day habilitation, waiver; per 15 minutes (source domain Observation)

#7

@TBanokina:

Not sure. It could be the patient got an oxygen contentrator at the visit day (or the "first day of the 32-day vist). Same with the airway pressure cap.

The Day habilitation - this could be something that happens very day 32 days long. Who knows. I don’t know this very well. @jenniferduryea: Can you help?

#8

Thanks for tagging me @Christian_Reich. Yes, I do know what this is about, though not specifcally about the codes that @TBanokina has in her excel sheet. This is a limitation of the codes and billing practices. But I have some questions for @TBanokina:

  1. do you have access to a units variable that tells you how many units were billed, specifically for the T2021? I’m not sure why this is spread over 7 days (2/1-2/7). My impression is that there were 7 units billed (where the unit is every 15 minutes) and two things happened: 1) there was one 15 minute unit billed every day from 2/1 through 2/7 or 2) there was 7 units billed for one day (1 hour 45 min meeting) and it was a billing error to expand the days (I have seen this with billing systems).
  2. The E HCPCS codes you are billing look like DME (durable medical equipment) claims. DO you know if you are looking at DME claims? Or are these claims coming from physician services? If they are DME claims, the billing procedures are different (in some circumstances) in that the days and units billed represent the quantity of the supply billed. I would argue this code should not even get a visit_occurrence record assigned to it because the this is a DME supplier billing Medicare for equipment.

IF you can give us some background as to the data you are working with, the payers where these claims come from, billing practices could explain what you are seeing.

Just a note that we have had to deal with this when looking at dialysis claims. And there is no elegant way around it. So because of the limitation of the data, you could get Outpatient visits that are longer than one day. But, depending on your definition of “Outpatient”, you could also consider hospital observation stays (where the patient is in the hospital under observation so their bills are covered under their Outpatient coverage instead of their Inpatient coverage) as Outpatient visits - thus truly having Outpatient stays longer than a day.

#9

If the source data is a representation of billing, it may not have 1:1 correlation with patients physical experience.

The way i understand it, in OMOP - we represent a patients physical experience. So something like:
Was the patient sleeping in a hospital/providers bed overnight – inpatient
Was the patient sleeping in his own bed at home, hotel, inn - outpatient

Generally, if the source data says that the record is ‘inpatient’ then the patient most probably slept in a hospital/providers bed if from_date < to_date. If your source data does not have information about inpatient vs outpatient – then we need to best guess.

In one physical visit - there may be many billing/administrative records - that is probably what you are seeing.

#10

@jenniferduryea @Gowtham_Rao thank you for your engagement into the discussion.

We are working with Truven claims dataset.

@jenniferduryea answers on your questions

  1. No. we don’t have such variable, just financial variables with payment information. But there is the ‘quantity of services’, which equals to 1 for T2021. Maybe it will help?
  2. It is insurance claims, which are divided into 3 visit types: Inpatient, Outpatient and Long term care. That’s all what we know.
#11

@TBanokina well, I’m not a big fan of that dataset due to limited documentation. The “quantity of services” is equivalent to “units”. Do you know if the T2021 is billed on a physician claim or a facility claim? The dates could represent different things (either the date of service, the billing dates relevant to the claim, or the dates the waiver is eligible for, etc.) based on the claim type. Without getting into the weeds on this, there are different billing procedures for different types of HCPCS/CPT billed. The T2021 is a code for Medicaid billing (see documentation I found: https://magellanprovider.com/media/1816/habbilling_claims02182014.pdf). It looks like it’s an error because the similar code, T2020 can be billed with a date range as long as the units = number of days.

Basically, any time you see weird dates on claims, it generally revolves around a billing practice that is required of that code or provider. I have not found a good resource that lists all of the offending codes. You just pick this up as you keep doing ETLs under different datasets. Then you have to create ETL exceptions based on the codes that you find. I know this is not entirely solving your problem, but at least it gives you some comfort :smile:

#12

@jenniferduryea

for T2021 is billed on a professional claim.

I just read the documentation about the billing process that you mentioned. From the information in the article I see how it should be, but in my case I don’t see a correlation between the value of “units” and date difference. Do you think that date difference happens due to the billing process and it doesn’t mean that procedure was carried out during several days?

#13

My impression is that T2021 was billed incorrectly. How you want to interpret that is up to you. I can only guess what actually happened, but based on how T2020 is billed, my thought is that the biller got confused and used date spans for T2021, instead of units. I would probably just pick the start date as the only date of interest for this code. But, I’m not sure if you want to use that logic for all physician records billed with a date range (case in point, procedure code T2020 does justifiably bill with a date range). But, I could see the argument that you should retain the dates and not make too many assumptions during the ETL process and keep both start and end dates. I know other people have more strict guidelines about what to include in their CDMs. So for this specific instance, this is an error. But I’m not sure how to generalize this to the rest of your ETL.

#14

@TBanokina @jenniferduryea I think I may have an answer for you regarding the mysteries of TSVCDAT. Not sure if this will apply to your specific case but conceptually it may help you understand the data pattern. I recently was attempting to identify reoperations after an initial operation. In this case the operation was performed in an outpatient hospital (stdplac=22). There were two instances of the CPT code for the operation, one was a facility record (facprof=F, billed with SVCDAT = 3/6/2015) and one was a professional record (facprof=P, billed with SVCDAT = 3/24/2015) - excuse me that I am using the Truven database terminology, I am not sure if these fields are available within the CDM.

Anyhow, for an operation, the facility will bill (F) and the surgeon will bill §, thus to total surgical episode comprises both kinds of claims. However, the facility may submit one big outpatient “claim” for multiple services rendered over several days. In this case, the facility’s records had SVCDAT = 3/6/2015, but they had TSVCDAT = 3/25/2015 (the day after the surgeon’s [P] claim). What happened, when looking at the data, is that the outpatient hospital had provided a chest imaging diagnostic on 3/6/2015 as part of the pre-surgery workup, but submitted all records associated with that surgery into an episode for which the service dates lasted from 3/6/2015 to 3/25/2015. In reality, services at the outpatient hospital were rendered only on 3/6/2015, 3/24/2015, and 3/25/2015, as evidenced by the corresponding P records which had consistent SVCDAT and TSVCDAT.

In summary, in the case of an outpatient facility, multiple related services (pre-surgery workup, actual surgery, etc.) may be rendered over several days. The facility may submit all records as a single “claim” with SVCDAT and TSVCDAT that span the period of time over which these services were rendered. You can try to match physician claims that have single (matching) SVCDAT and TSVCDAT to try to determine when the actual target service was rendered.

Outpatient visits that last more than one day
#15

@Stephen_Johnston thank you for the response. And I do agree with your assessments in regards to facility claims. However, @TBanokina’s issue is with a CPT service = T2021 that was billed on a professional claim, which rarely has a difference between SVCDAT and TSVCDAT. When there is a difference, there is usually a very specific billing requirement the provider is trying to accomplish with that code. For example, T2020 is billed with a date range to indicate the number of days the service was provided. So the SVCDAT and TSVCDAT is different in this instance due to a billing requirement for that code.

t