OHDSI Home | Forums | Wiki | Github

Meps etl

Hi all,

According to the data network page on ohdsi.org, the MEPS dataset from AHRQ is part of the OHDSI network. Does anyone have an ETL of MEPS that they could share?

Tagging @jenniferduryea as I saw some earlier post from you mentioning MEPS.

Thanks,
Ajit

We had plans to do this, but we have not done the ETL itself. We may have some old v4 specs that we can share.

Keep in mind that this data has survey weights. In fact, I believe it has 2 sets, one for the cross-sectional and one for the longitudinal.

I believe I started spec’ing out the ETL but then ran into roadblocks as to how to store the survey weights in the CDM. So I abandoned the project. Plus, I started to run into other issues on responses. Most notably, how to represent the following responses; “no”, “missing value”, “refused to respond”. They are all negative values, but statistically could be used differently for analysis. I would love to review anyone’s ETL specs of MEPS to see how they overcame these obstacles in the CDM.

Ah okay, thanks Mark and Jen. My study colleagues have decided to use a different dataset, so unfortunately at this time I won’t be adding anything new to the MEPS ETL knowledge. But I’m curious: is this a good value-add ETL for us to have? Are there a lot of people interested in having a CDM MEPS (assuming roadblocks are cleared)?

I think it would be a great benefit since MEPS is one of the only datasets users can download from the internet for free. So new adopters of OMOP will have easily available raw data and that can then start ETL’ing data immediately. If the roadblocks are cleared, I think it would be a great test case for researchers to explore the CDM. But the roadblocks are quite extreme, so I’m not sure how feasible it is.

t