OHDSI Home | Forums | Wiki | Github

How to Capture pregnancy data? (EDC, gestation length, etc)

Hi All,

We are developing ETL for our EHR data into OMOP v5.3. We are trying to figure out where we could store patient’s pregnancy related information. Our source data captures data such as EDC and gestation length. Does anyone have any suggestions on how we should go about it?

Any help would be appreciated.

I see two possible solutions

  1. There are a set of SNOWMED concepts of the type ‘Gestation period, XX
    weeks’. Where XX goes from 1 to 42 weeks. Using these concepts you would
    create condition_occurrence records.

  2. Find a Measurement concept for gestation period. e.g. 3012266
    (Gestational age). Use this as the measurement concept id, then value as
    number get the number of days or weeks as indicated by the unit concept id.

Thanks. I think mapping it to the Measurement table fits our data better, and we will be able to record the estimated date of delivery in there too.

@guanguo:

Not sure I’d agree with @DTorok’s recommendation. Pregnancy is a Condition like all the others, not an Observation or Measurement. Should go into the Condition table, with the conception date as start_date, and birth as end_date. You can use 4336958 “Term pregnancy”.

We should probably write that into THEMIS conventions. @Asha_Mahesh, @mvanzandt, @MNairn, @Tom_Galia, @PaulPetraro?

@christian_reich

This sounds like a “pregnancy era”. The preferred generic term is, I think, “episode of care”, i.e. a window of time when a person experienced start to end of a certain care/sickness (not all have an end). Measurements, conditions, procedures, drugs over many visits dates may be linked together.

Lot of episodes of care

https://catalyst.nejm.org/bundled-payments-maternity-care/

@christian_reich this may be an opportunity to create a generic episode of care table, that links all tables of an episode together.

(Different thread)

@Gowtham_Rao

Actually, come to think of it: A similar idea is evolving in the Oncology WG: The idea of a “treatment”, combining everything that has to happen to administer a, say, course of chemo. @mgurley and @rimma have created a whole proposal along those lines, and claim that it should be wider than just oncology. We should probably talk.

The era approach makes a lot of sense. It also enables more detailed studies into increasing parity or even gravidity if we consider miscarriages.

Sent from Teng’s iPhone, please excuse typos

On 9 Apr 2018, at 3:48 am, Christian Reich <noreply@ohdsi.orgmailto:noreply@ohdsi.org> wrote:

[http://forums.ohdsi.org/user_avatar/forums.ohdsi.org/christian_reich/45/45_1.png] Christian_Reichhttp://forums.ohdsi.org/users/christian_reich
April 8

@Gowtham_Raohttp://forums.ohdsi.org/users/gowtham_rao

Actually, come to think of it: A similar idea is evolving in the Oncology WG: The idea of a “treatment”, combining everything that has to happen to administer a, say, course of chemo. @mgurleyhttp://forums.ohdsi.org/users/mgurley and @rimmahttp://forums.ohdsi.org/users/rimma have created a whole proposal along those lines, and claim that it should be wider than just oncology. We should probably talk.

Correctly determining when a pregnancy ends, which could include spontaneous at-home miscarriages, abortions, and deliveries at other institutions, is a lot more difficult than most folks appreciate.

Yes… This is very interesting

Happy to be involved

I definitely agree its a challenge, but also am encouraged by the progress
in developing shared solutions. Our team published the following work in
this area, for those who haven’t seen it:
http://journals.plos.org/plosone/article/comments?id=10.1371/journal.pone.0192033.
The algorithm is quite complicated, but was demonstrated to work against
both claims and EHR data in OMOP CDM format, so should be a useful start
for those who are interested. The source code to implement the algorithm
is available here:
https://github.com/OHDSI/PhenotypeLibrary/tree/master/pregnancy%20episodes%20and%20outcomes
.

@Christian_Reich @rimma Yes, I agree an “episode of care” concept is very similar to the TREATMENT concept that we are advocating for in the Oncology Treatment Proposal: a collection of clinical events across a window of time linked together for a clinical purpose.

  • Sometimes your source system links clinical events into an EPISODE/TREATMENT and the ETL process is easy.
  • Sometimes your source system represents an EPISODE/TREATMENT occurred but does not link the clinical events. Compendium-aided algorithmic linkage is possible (either pre or post ETL), making the ETL or post-ETL process harder.
  • Sometimes your source system does not represent an EPISODE/TREATMENT at all (the concept being alone embodied in healthcare workers’ intentions) . Forcing you to apply compendium-aided algorithmic derivation of an EPISODE/TREATMENT: unguided by any pre-made EPISODE/TREATMENT abstractions. An even harder ETL or post-ETL process.
  • Sometimes your source system represents that an EPISODE/TREATMENT occurred but you do not want to attempt to link the clinical events. Making the ETL process easy because derivation of linkages to clinical events is not attempted.

I think we want to support all these scenarios.

Here is a link to our work-in-progress Oncology Treatment Proposal:

Totally agree. But like with the DRUG_ERAs and CONDITION_ERAs: Perfection should not prevent us from creating a standardized solution for everybody (= the enemy of the good). When somebody has a better solution we can upgrade.

Hello - tagging onto this important topic - @jweave17, @Miriam2, @shawndolley and myself met last week as part of the maternal and child health group. We would like to work with Themis to establish a home for storing pregnancy episodes, specifically using the algorithm referenced in @Patrick_Ryan’s post above. We are also interested in finding a home for the results from the maternal child linkage algorithm that was presented on 09/12/2017 OHDSI community call. We are looking forward to discussing in a future Themis working group call.

@Patrick_Ryan would love to cite the OHDSI work on this subject…is there any way to get the color figures from the plos one paper? For some reason figure 2 is shades of grey (4 different shades) - it is unreadable…

Is creating ERAs an ETL problem or a era-construction problem.

E.g. we construct cohorts using cohort definitions. We don’t ETL cohorts. ETL based tables go to the OHDSI CDM schema, while constructed tables go to results schema.

So, is ERAs a RESULTS schema table or CDM ETL schema table?

If the construction algorithm is defined by the analyst, then RESULTS?
if the construction conventions are determined by community using THEMIS, then CDM?

1 Like

@Gowtham_Rao:

Not sure I follow your logic with the schemas. Conceptually, we are talking about three different types of abstracted (derived) data representing higher-level medical events with duration:

This is great table.

I meant, is era construction similar to how we construct cohorts or is it a prescribed/standardized method? It could be a hybrid too – that there are standard ways, and there are ad-hoc ways.

I am in favor of a hybrid way

t