Would someone with knowledge of the drug cost data be able to provide a brief summary of the data in CDM V4? Also, what changes are planned for CDM v5?
Thanks in advance,
-Rich Boyce
OHDSI Home | Forums | Wiki | Github |
Would someone with knowledge of the drug cost data be able to provide a brief summary of the data in CDM V4? Also, what changes are planned for CDM v5?
Thanks in advance,
-Rich Boyce
Hey Rich: No changes from CDMv5 to CDMv5 in the cost fields in the _COST tables. Only ancillary changes are: 1) support for CURRENCY_CONCEPT_ID, 2) expansion to allow costs specific to visits in VISIT_COST and to devices in DEVICE_COST, and 3) DRGs moved to OBSERVATION records from PROCEDURE_COST field.
I haven’t seen much activity within the community using these tables for analytical use cases, but the original intention was to enable capture of costs (not charges) from both payer and patient perspective. The primary fields were loosely modeled off what others have recieved in claims data (common convention I’ve seen in ETLs: costs associated with medical claims go into PROCEDURE_COST while costs for pharmacy claims go in DRUG_COST). The key insight I’ve gained from these tables is that it is critical to understand which perspective(s) the source data provides when thinking about cost, as many databases have incomplete capture of the total expenditure.
Cheers,Patrick
This answer is a little bit tangential, but we are looking at the cost tables and I thought I would pass along some top line thoughts.
The table names really are a misnomer – “costs”. Based on the CDM v5 document, they are generally “paid amounts” or “reimbursed amounts” as described. The only exception is the drug AWP which is generally not what is paid.
Some data sources that have both reimbursed amounts and charges will require the ETL process to make a decision about which should be included. Generally nobody works with charges unless that is all that is available. Even then there are cost-to-charge ratios that can be applied (for example, with Nationwide Inpatient Sample data). Again, something that can be addressed in the ETL process. Very few databases have actual costs in them (Geisinger/MedMining data is one exception that I know of).
Again, this is not a direct answer, but these forums are good for storing thoughts and ideas. It may be worthwhile to set up some best practices for the cost ETL process, or for others to share their experiences.