OHDSI Home | Forums | Wiki | Github

Concept ID set for ASCVD diagnoses

Researchers
#1

Hello everyone,

I have been searching for projects which define a cohort of ASCVD patients using the OMOP CDM. I am working with a large hospital system in the US to design data gathering processes for a proposed study and if at all possible, I would like to build it using the OMOP CDM so that it is reproducible.

I haven’t found anything by searching ASCVD in this forum other than risk calculators employing non-diagnosis data. I also didn’t see evidence of a working group that seeks to define this. Does anyone know of any efforts toward this end or is there any interest?

Andrew

#2

@AndrewNute:

You either use the Atherosclerotic cardiovascular disease risk assessment score if it is in the data, or you calculate it from the EHR data. Claims data won’t work.

#3

As Christian said, ASCVD is calculated from clinical data (e.g., the numerical value of lab results), not insurance claims data.

As a mathematical calculation, the OHDSI tools currently do not provide means to compute it.

If you want to calculate ASCVD scores on an OMOP CDM dataset, and you can work with a non-OHDSI tool, then you can see my vignette on how to do it using R package Phea: Calculating a risk score, such as the ASCVD Risk Estimator+.

The vignette demonstrates the approach. To make it work on your local dataset, most likely you will have to adapt the concept IDs and units of measurement.

If you want to make ASCVD scores available within the OHDSI ecossysem (e.g., use them inside Atlas), your best route is probably to use Phea to produce new records with the scores, then ETL those into your dataset using the concept ID that Christian mentioned.

#4

Thanks Christian and Fabricio, I appreciate the quick responses.

I did find the risk calculator for ASCVD in my search of this forum and I do have that in my team’s CDM (I think). however, I was looking more for a standard/consensus on what set of clinical codes would qualify as indicating the presence of clinical ASCVD.

I have seen AHA/ACC define clinical ASCVD as:

“Clinical atherosclerotic cardiovascular disease (ASCVD) includes acute coronary syndrome (ACS), those with history of myocardial infarction (MI),stable or unstable angina or coronary or other arterial revascularization, stroke, transient ischemic attack (TIA), or peripheral artery disease (PAD) including aortic aneurysm, all of atherosclerotic origin”

Through my own struggling attempt at neatly defining a set of ICD 10 diagnosis codes for these outcomes, I have found that it is not as simple as 1) find SNOMED terms for clinical outcomes and 2) map them to ICD10. For example, some of these clinical outcomes can have non-atherosclerotic origins related to pregnancy.

This was the motivation for my question which is basically, has anyone sifted through diagnosis codes (or possibly procedure codes if that applies) for each of component artery diseases in the definition above. If they have, where can I locate that set of codes or concept IDs they used. Or is there interest in doing that?

#5

Ah, I see. When you used the term “ASCVD,” we thought you meant the ASCVD Risk Estimator Plus, which is the risk calculator (risk score) you mentioned.

What you’re talking about is phenotyping ASCVDs in general, as a group of conditions. A “consensus/standard” way of detecting ASCVDs from claims data has probably never been created.

The usual approach to eliminate competing causes is simply requiring patients to not have them in the time preceding or during the condition, e.g., ACS event must not have happened during or shortly after any pregnancy. This is something Atlas can do.

But notice there is quite a bit of phenotype development in there, to phenotype ASCVDs with rigor and accuracy. You should treat each condition included in the ASCVD bucket as its own phenotype, with its own rules, filters, and tests of accuracy.

If you do develop a phenotype for ASCVDs, please come back to share with the forum.

#6

Thanks Fabricio,

I am becoming increasingly aware of the amount of work this project will require :sweat_smile:. I am curious to know if there is any interest in forming a working group in this forum? Should I repost about that topic specifically?

I am working with a research group that has started a list of these diagnosis codes for papers and studies in the past. When I am done updating it with the cardiologist on our team, I can share what we have for phenotyping using the OMOP CDM. It would be great to hear some other informed opinions about that work for reproducibility’s sake going forward.

Additionally, a working group would be informative for determining what the rules, filters, and tests should be for each bucket. In this project, we also plan to use the OMOP CDM to pull the data for the other risk factors mentioned in the 2022 lipid lowering guidelines. I am certain those will also each create their own buckets for rules, filters and tests even though they don’t all involve diagnosis codes.

Andrew

t