OHDSI Home | Forums | Wiki | Github

Opioid use disorder/substance use disorder

Researchers
#1

Hello,

I am a federally funded investigator who has a background in quantitative sciences and also a board certified addiction psychiatrist. I am looking for collaborators in the OHDSI program who have an interest in collaborating on this topic.

Please E-mail me at:
xsl2101@columbia.edu

-Sean

Sean X. Luo, M.D., Ph.D.

Assistant Professor of Clinical Psychiatry

Division on Substance Use Disorders

Department of Psychiatry

Columbia University

Room 3624

1051 Riverside Dr., Unit 66

New York, NY 10032

(646) 774-6144

xsl2101@cumc.columbia.edu

#2

Hi Sean,

I am a postdoc at CUMC who would be happy to help you with this. I’ve sent you an e-mail and look forward to future correspondence.

-Matt

#3

+1 on this!

I might suggest engaging with your neighbors at Mt. Sinai. Joseph Finkelstein’s team is working on converting opioid treatment data. Not sure he’s on the OHDSI Forum – but can connect dots. @mattspotnitz you know how to get ahold of me if you need it. :slight_smile:

#4

I’m sure many people are aware of this already but there has been real difficulty at my institution doing research on opioid use because of redaction. We use claims data and any claims with a Substance Use Disorder code are currently being redacted and are very difficult to access. It is a major barrier described in the paper “Long-Term Implications Of A Short-Term Policy: Redacting Substance Abuse Data” by Andrea M. Austin and others. There is also state by state variation in how the federal law requiring the redaction is interpreted and applied.

#5

I’m interested as I’m a psychiatrist by training. And I’m leading OHDSI Vocabulary team, so I can be helpful in terminology questions.
So, please, count me in, although I’m not sure how much time can I spend on this.

#6

Is that true, @Adam_Black? Any claims, or just the CMS? I am too cheap to pay the $15, and too lazy, to look into the article, can you tell me what they are exactly redacting, so we could check it out?

#7

Paywalls on scientific literature (particularly anything publicly funded) need to be retired. I can say for sure that Medicare, Medicaid, and commercial claims are affected in the Maine All Payer Claims dataset which is the only claims dataset I am very familar with. I think different datasets are affected differently but this page describes how the Maine All Payer Claims data is affected.
https://mhdo.maine.gov/sec_priv.htm

Substance Abuse and Mental Health Services Administration (SAMHSA)-Confidentiality of Substance Use Disorder (SUD) Patient Records, 42 CFR Part 2

MHDO applies the CMS/ResDAC filter to all data in the APCD including, commercial, Medicare and MaineCare data to redact claim lines containing SUD-related codes prior to releasing MHDO APCD data to authorized MHDO data users; which means we have removed any claim lines that have a code that is included on the redaction list. We leave any portion of a claim that doesn’t include one of these codes. This SUD-related data is stored in its own protected database and may be available to authorized MHDO data users under the terms and conditions of payment, health care operations and other health care related activities.
Notes:
* Commercial payers use their own filters to suppress SUD-related claim lines before submitting the data files to the MHDO. The application of the CMS/ResDac filter is an additional measure taken by the MHDO.
* A listing of the CMS/ResDAC codes used for redaction is available on our All Payer Claims Database (APCD) page.

I think the redaction process varies from state to state. I’m also not sure about the distinction between redacting just certain claims lines or redacting the entire claim when an SUD code appears. Since diagnosis codes are associated with a claim header and not a specific line I think that a SUD dx code on a claim would cause the entire claim to be redacted. I’m definitely not an expert on this but do know it is something to watch out for.

#8

I’m a clinical psychologist by training and very interested in finding our more about this. Will send an email to follow up.

#9

This website has some more info on the topic including the list of diagnosis codes used in the redaction. In the case of the MHDO any claim with one of the SUD diagnosis codes is being entirely removed from the dataset which had a large impact on our analysis of opioid overdoses in Maine. Depending on what you are studying the redaction could make a big difference.
https://www.resdac.org/articles/redaction-substance-abuse-claims

#10

When does that get redacted? In the EHR, in the claims sent to the payer, or in the database of the payer before you get a download?

#11

For the All Payer Claims data the redaction can happen in the comercial payers database before sending the data to the state government. The state goverment then applies the redaction rule again to all claims (Medicare, Medicaid, Commercial) before releasing it to the public. For medicare data it sounds like CMS will release SUD claims directly to data users but I’m 100% not sure about that.

#12

Friends:

Is anything happening? Anybody creating a study and having the OHDSI Network engage?

#13

Hi @Adam_Black
What exactly do you want to study? Any hypothesis to check?
Cc: @agolozar

#14

I’m not involved in any study but I do have colleagues who have worked with SUD claims and ran into trouble. I just wanted to make others aware of the potential pitfall of redacted SUD claims.

#15

+1 for interest on this topic too.

The Health Plan for whom I work has looked at using the RIOSORD screening tool to help assess risk of overdose. It is a good start for triaging our members and targeting outreach, but I expect that it could be improved.

Does anyone in the OHDSI community recommend alternate screening tools or approaches for PLP?

Is the community interested in replicating and extending that study (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5939826/#pnx009-B19) to generate a more accurate and actionable PLP model?

The instrument is a good start, but might benefit from being updated due to the age of the data used (only up through 2013), the lack of some important covariates (such as prescription for naloxone or treatment with metadone/suboxone; plus Social Determinants of Health in general), lack of clarity on exit conditions for the cohort, and relatively small size of the comparison cohort for their propensity matching.

We are just starting our OMOP journey (e.g. have not mapped our data yet), but we could help in the study design specification and evaluation if it could be run on others’ data that has already been mapped.

1 Like
#16

@Adam_Black yeah, sorry, I’ve misread this somehow. (I shouldn’t post on a forum just after take off in the last moment when mobile network is available :slight_smile: )
So it’s the question

for @xsl2101

#17

Looks like the commercial claims databases don’t do that. For example, Alcohol abuse in Pharmetrics over time:

grafik
grafik1683Ă—361 24.3 KB

So, anybody who wants to study this reach out to folks with those databases.

#18

Has anybody wondered about the similarity of the words “ResDAC” and “redact”? A little suspicious possibly?

:slight_smile:

#19

Can we replicate this study and run it also on datasets outside US?

https://www.ajpmonline.org/article/S0749-3797(19)30527-6/fulltext

This was using IBM MarketScan. It has really interesting criteria to make into Atlas phenotype.
And all of it should be doable with claims.

Where in OMOP CDM we know that a patient is on health plan that has dental component?

image
image719Ă—821 139 KB

image
image476Ă—788 66.6 KB

#20

You should be able to find that information in the PAYER_PLAN_PERIOD table, especially with use of the new Plan concepts: http://athena.ohdsi.org/search-terms/terms?domain=Plan&page=1&pageSize=15&query=

2 Likes
t