We’re working with TRUVEN HPM data and found that there are Medical Payments associated with the worker’s compensation cases. The only event that is linked to these cases is a diagnosis of a patient. On the other hand, the specification for CDM COST table does not expect to incorporate the findings on condition`s payment amount. All we can do is to link this cost with condition indicated in compensation case, but the current CDM specs dо not allow. Another option, we may link cost with VISIT when the visit is assumed like a set of all received medical services including procedures, drug administrations, etc.
@Gowtham_Rao and others, which approach is seemed to be correct?
In general, I assume every clinical event is linked to through a FK to a record in visit_occurrence table. If that assumption is correct, then in your case, every worker’s compensation case visit should have a visit_occurrence record. Diagnosis/condition maybe linked to the visit. Procedure/services during the visit should be linked the visit. etc.
So, i would ETL in such a way that every workers compensation claim is a visit_occurrence record. Condition, procedure, drug, costs are all linked to this visit_occurrence_id record.
Is it correct that for OMOP version 6.0 cdm.cost table still does not capture records containing the cost of events recorded in cdm.condition_occurrence table?
I found 2 mentions about condition: in incurred_date description and in revenue_code_concept_id convention description. Is it typo? We can not populate cdm.cost table with records from cdm.condition_occurrence table?
@gowtham_rao, that assumption will not necessarily hold for all databases. The FK relationship to VISIT_OCCURRENCE is optional. You can use it when it’s there, but you can’t expect that it will always exist. There are many circumstances when clinical data is not linked to a visit record, such as when a patient completes a survey (like a health risk assessment) or has a mail-order prescription drug order, or self-reports OTC medication history. I don’t know how all workers comp claims are generated, but if they don’t have to involve an encounter with a clinician, then it seems very reasonable they could not have a visit tied to them.
Thanks, fine. But let’s think how the cost of condition event should be interpreted. Is it a cumulative cost of all medical events associated with entire condition (diagnostics, treatment procedures, drug administrations, etc.) or something different?
This source dataset of worker’s compensation cases includes indemnity payments, medical payments, short-term disability payments. What of them mat be recommended to store in the CDM and which concepts may be used for cost_concept_id & cost_type_concept_id? Or some new should be created here?
@Alexdavv
I have only one example. We make Dx claims conversion. Claims have services and each service (HCPCS / CPT4 / ICD9Proc / ICD10PCS) has a charge.
And only services with source code G8771 ‘Documentation of diagnosis of chronic kidney disease’ are mapped to Condition domain concept_id = 46271022 (SNOMED) ‘Chronic kidney disease’. So probably it is problem with mapping HCPCS to standard concept_id SNOMED
@Alexdavv There is two ways to think about it at a generic level
First:
Cost maybe accounted at the transaction level i.e. cost for a visit, drug, procedure etc.
Cost maybe accounted at the person level i.e. cumulative out of pocket expense till date for the calendar year.
The later is rare - and i think it is better to do that on the analytic side, rather than ETL side (opinion).
Second:
In the case of the former, i.e. cost accounted at the transaction level – in general transactions have a summary and detail, where detail are the components of the summary. e.g. if a person goes for hip replacement surgery and the total cost was 10,000$; the components may be surgeon fee for procedure $1,000, hospitals fee for organizational overhead 5,000$, anesthesia drug fee $500, anesthetist professional fee $500 etc. - these are details.
Specifically on workers compensation: in general these are time-limited episode e.g. backpain at work – 60 days of workers compensation episode. During that time-limited episode,the person may have had many visits, clinical events, financial transactions. Some of the financial transactions may be visit related, others may be not visit related.
Can you please post a (dummy data) table structure of how your data source captures this type of information for one person. Maybe we could help translate that to CDM parlance. Something like this
