OHDSI Home | Forums | Wiki | Github

Achilles thread


(Vojtech Huser) #1

I spoke with several people at the symposium about drill down feature of Achilles Heel but wanted to post this to the forum as well. (and start a general Achilles thread)

Just to re-iterate, Achilles v1.3 contains an additional table Rule_Drill_Down.csv.
Here:
https://github.com/OHDSI/Achilles/blob/master/inst/csv/rule_drill_down.csv

Some EMIF Achilles users and prior requests asked for ability to see (even be given SQL) to fix a given Heel error.
Note that this already is in current version 1.3. Just the web viewer of Achilles outputs has not caught-up.

The file above (is a preparation for ‘Within-Atlas’-Achilles-Web) feature that allows for drill downs. Note that not all drill downs are “in scope”. Only those that don’t require patient level data (no PHI is involved). (in CSV of type Achilles). For ETL developers, the file will contain (in the future Achilles version 1.4) also SQL that allows drill PHI level drill downs. (e.g., for the purpose of tuning ETL (or marking/documenting source data errors).

Also, notice that any Heel rule may have zero, one or multiple drill downs.


OHDSI Informatics Study: Data Quality
(Taha Abdul-Basser) #2

Thank you for documenting this here @Vojtech_Huser.


(Vojtech Huser) #3

Data Quality in Sentinel model include Vital Sign data quality checks.
I implemented some of them in Achilles Heel but I would like to make sure I capture the correct concepts.

For example, Sentinel DQ check ID 248 looks at Ht, Wt, and BP.

/*******************************************************************************/
/
JIRA QCL-81: Remove CheckID=248 **/
/

  • Create Level 2 flags for invalid conversions (CheckID=248);
  • Check conversion from orig_result to ms_result_n;

proc format;
value Mcf
1 = “Ht, Wt, Dia, Sys all present”

2  = "Sys missing, all others present"
3  = "Dia missing, all others present"
4  = "Wt missing, all others present"
5  = "Ht missing, all others present"
6  = "Sys and Dia missing, Tt and Wt present"
7  = "Wt and Sys missing, Ht and Dia present"
8  = "Wt and Dia missing, Ht and Sys present"
9  = "Ht and Sys missing, Wt and Dia present"
10 = "Ht and Dia missing, Wt and Sys present"
11 = "Ht and Wt missing, Sys and Dia present"
12 = "Only Ht present"
13 = "Only Wt present"
14 = "Only Dia present"
15 = "Only Sys present"

Figure below shows one sample OMOP dataset concepts for those.

If you have EHR data outpatient measurements, which concepts do YOU use in your data? (otherwise, be prepared to get a notification from Achilles Heel until you submit your “favorite” concept_ids for those as reply here)

Example reply might look like this:

body weight: http://www.ohdsi.org/web/atlas/#/concept/4099154
body height measure: http://www.ohdsi.org/web/atlas/#/concept/4177340
BP Systolic: http://www.ohdsi.org/web/atlas/#/concept/4152194
BP Diastolic: http://www.ohdsi.org/web/atlas/#/concept/4154790


(Vojtech Huser) #4

A related CDM proposal is described here:

http://www.ohdsi.org/web/wiki/doku.php?id=documentation:next_cdm:deeper_semantic


(Vojtech Huser) #5

An update on the weight concept: I did an internal survey of several sites on how they capture weight. The implementation of sentinel DQ checks will focus on weight (since there is the best use case; drug dosing importance). The plan is to implement a DQ check for the following concept:

http://www.ohdsi.org/web/atlas/#/concept/3025315
(the LOINC code for body weight)

(it will be as notification since it is only relevant to some datasets (and not all))


(Lili Peng) #6

Hi @Vojtech_Huser, do you know which databases ACHILLES pulls data from?


(Vojtech Huser) #7

Achilles group of people met today and discussed new developments for it.


(Vojtech Huser) #8

Yesterday the Achilles WG met again. Notes to follow from Taha. @t_abdul_basser


(Davidcarnahan) #9

Hey Vojtech, Is there any way we can touch base next week? I’d like to get caught up on where things are and how I can get involved with this given the complementary work I’ll be doing with the FDA/IQVIA.


(Vojtech Huser) #10

Notes from Achilles WG meeting March 13,2019
Present: Vojtech,Ajit,Taha

  • framework for Heel (re-do of it)
  • METADATA tables (use them or not?)
  • describe DQ issues just as text
  • review of issues in github
  • reaching out others (invite more to our meeting)
  • release planning (bug fixes)

(Taha Abdul-Basser) #11

Notes are posted on the Achilles WG page.


t