shyamvis / act-covid-ontology Goto Github PK
View Code? Open in Web Editor NEWACT COVID Ontology
ACT COVID Ontology
Anti-Malarials -> Antimalarials
Anti-virals -> Antivirals
In the Medications tree, most of the leaves have a number of NDC-coded leaves below them. These are "LI " inactive leaves.
Some medications (for instance, Chloroquine Sulfate) do not have these inactive leaves. Should we be expecting the NDC leaves for every medication? Or, not necessarily? Thank you.
In the Care Setting section of the ACT COVID ontology, from April 15 or April 16... Rows 3666 and 3667... the codes seem to match what is listed in Row 3667, but that is not what I expected.
Suggestion: Row 3666 Ambulatory
should have UMLS:C0002423
, but instead it has NULL
.
Suggestion: Rows 3666 and 3667 reflect the test for inout_cd
in the visit_dimension
table, but it seems like, for consistency with the other items in that care-setting tree, these two rows should reflect the test for concept_cd
in the concept_dimension
table.
Row 3668 also has the test for inout_cd
, but it appears to be used in a way that is consistent with the other facts from the visit_dimension
table, so Row 3668 is not a problem.
I'm not an expert, so please excuse me if I am mis-characterizing this as an anomaly. Thank you.
What is the purpose of the SNOMED_3947185011
and UMLS_C0022885
elements in the path? This seems counter to the other components of the ACT ontology that use Annnnnnn
codes as their path organizers.
The concept code for REMDESIVIR is currently ACT|LOCAL:REMDESIVIR
. Is this a standard approach to non-global concepts (vs. just LOCAL:REMDESIVIR
)? In either case, is there a corresponding entry for the schemes
table?
Overview
In the ACT_COVID ontology, there is a concept with an anomalous path. This is ICD-10 Procedure 5A09457. See the attached image where the concept is missing from the parent folder.
Risk
Matching observation facts will be counted when the overall "mechanical ventilation" folder is queried, but not when the 5A094 folder or subfolder are queried.
Analysis
My analysis is that the concept path is anomalous. The path is currently:
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0242656\UML_C1444343\UMLS_C0199470\A28335107\A16738804\
Suggested Resolution
I believe the path ought to be:
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0242656\UML_C1444343\UMLS_C0199470\A28322331\A28335107\A16738804\
I believe that the token A28322331
is missing from the current path.
Kind regards, Mark Abajian
What is the reason for having derived terms within the Level of Care folder? How is "Ambulatory (Derived UMLS:C0002423)" term different than "Ambulatory Visit"? Same question for derived Inpatient and Emergency terms.
Thank you.
In Diagnositic Lab Tests, "Diagnositic" should be "Diagnostic"
In the Medications tree, most upper-level folders have UMLS basecodes on them. These two folders have RXNORM codes instead.
Should they have a UMLS code instead, to make them consistent with the other folders at their level? Note: This issue may be related to Issue #4 .
In the Level of Care (Setting)
tree, there seem to be a number of folders that do not open to any leaves. These have CPT codes like 992xx
. It seems like many of the 992xx
codes (or maybe all of them) need to be leaves (and not folders) in the ontology tree.
Under Course Of Illness/Conditions/Cytokine Release Syndromes Symptoms (CRS/SIRS)/R65:
R65.1 and its Children (R65.10, R65.11)
R65.2 and its children (R65.20, R65.21) - all in the same level. R65.1 & R65.2 appear as folder with no children.
In the April 15th version of the Covid ontology, we have "Level of Care (Setting)", and inside of that are a number of CPT codes, UMLS codes, and the standard visit type codes O
, I
, E
, and EI
. The paths for these 4 codes are different in the COVID ontology than they are in the ACT Visit Details ontology, which is good, so that these concepts can be rolled up into the COVID concept tree.
But these 4 paths for these visit type concepts are not included in the Concept Dimension table, and therefore, the patients who match the Inpatient visit type I
will not be counted in the query for the folders Inpatient (Critical Care)
nor Inpatient (Routine)
. Likewise, the O
, E
, and EI
concepts will not be counted in their respective folders in the COVID ontology.
I recommend adding these four rows to the Concept Dimension table, using the path in the ACT COVID ontology table.
Not sure whether this was done intentionally or not, but the ACT COVID tree in the ontology sits one level "lower" than the other ontology domains.
For instance, the ACT Procedures ICD-10 folder sits at HLEVEL 1, and its constituent folders at HLEVEL 2. Likewise for the other domains in ACT Ontology 2.0.1.
But the ACT COVID folder sits at HLEVEL 2, and its constituent folders at HLEVEL 3.
This does not cause a problem for i2b2/SHRINE users, but for programming and maintenance, I found this to be confusing. Thank you for your consideration.
In the April 15th release of the COVID Ontology, we have a "Level of Care (Setting)" tree. And inside that tree are four "settings" where care is delivered: Ambulatory
, Emergency
, Inpatient (Critical Care)
, and Inpatient (Routine)
.
The four settings have visual attributes of a Collection
, but because of that, we can't drag the care setting into the query window.
I feel that these should be set to Folder
instead, so they can be dragged.
This may already be changed in Version 3 of the ontology, but I am submitting this ticket just in case. Thank you!
In the ACT COVID table (and also the CONCEPT DIMENSION table), there is a disagreement between the Remdesivir codes that appear in the description versus the code itself.
MED
Remdesivir (ACT|LOCAL|MED:REMDESIVIR)
MED
ACT|LOCAL:REMDESIVIR
what is the difference between ACT_COVID_V3.dsv vs ACT_COVID_V3D.dsv in ACT_COVID_V3.zip
Course Of Illness/Medications: Immune Stimulants folder is empty
See under CPT at https://github.com/shyamvis/ACT-COVID-Ontology/edit/master/codes.md
The following folders do not appear to have any codes associated with them. But other folders at their level in the hierarchy seem to have either UMLS or SNOMED codes. Could these four folders be missing codes?
Thank you.
icd10 J80 and umls C0003451 are these duplicates
The following drugs/combinations have RxNorm codes also. But they are coded in ATC in the tree.
Under Loop Diuretics:
Under Anti-Malarial:
In the Level of Care (Setting) folder, there seems like there may be an anomaly in the hierarchy for these concepts...
Leaves CPT 99291 and 99292 appear at the same level as folder CPT 1014309. But it seems like they should be leaves inside of the CPT 1014309 folder. That folder is currently empty.
Thank you for your consideration.
Hello! This item:
SARS-CoV-2 ORF1ab region, respiratory specimen (LOINC:94559-2)
is a "qualitative" test. To be consistent with the other qualitative tests, maybe we should include that indicator in the description:
SARS-CoV-2 ORF1ab region, respiratory specimen (qualitative) (LOINC:94559-2)
Just a suggestion. Thank you.
currently set to vancomycin rxnormcode
In the Medications tree, there are two concepts that have two hierarchy levels, when it appears that they really need only one level.
Lopinavir/Ritonavir
has two folders… 1 with an ATC code, but then an identical folder below with RXNORM code. Do we need the outer ATC-coded folder if we already have an RXNORM-coded folder?
The same is true for Tocilizumab
.
Thank you.
Antiviral and Antivirals can be merged.
The c_visualattributes
for c_fullname
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\UMLS_C1335447\LOINC_94547-7 POSITIVE\
is "HA". It should be "LH"
The mechanism we're using to build a datamart of covid cases actually started as a "2% of production data" testing facility. It works by choosing a random sample of encounters and then finding all the data related to those encounters. (for those with access to the HERON repository, see clarity_structure.csv and heron_sample.py )
So at KUMC, we can straightforwardly add a sample of, say, 10,000 encounters from the last 6 (or 9 or 12...) months to our Covid19 build process.
I understand the need to indicate which concepts are being coded as "Derived" facts. I do not dispute that.
There may be some confusion because the folders that contain the derived facts also contain facts that are not "derived", but are actually coded directly in the patient record. And when the tooltip for the folder says "derived", the user may believe that all the concepts that appear in the folder are derived, not just some of them.
These eight folders mention (Derived) in the tooltip, and that may be misleading if these folders contain both derived and "real" facts:
Also, the concept_path, c_dimcode, and c_fullname for these folders may use the term Derived, but I have not checked that; or, that may be a lower priority, if the user will never see those.
Hello,
I'm not sure whether this is caused by a glitch in the ACT Ontology on the Demo server (http://dbmi-ncats-test01.dbmi.pitt.edu/webclient/), or whether this is due to a flaw in i2b2 v1.7.12a on the Demo server...
When I try to "Search by Codes", I always get a popup that says "No Records Found".
When I try "Search by Names", it seems to work fine.
When I try to "Search by Codes", I do not see any error message, just the "No Records Found" popup dialog box. But when I try the Message Log, it provides me with these details (attached)
SearchByCodes_Failure_MessageLog_20200521ma.txt
I think it used to work all right. I wonder what happened?!
Kind regards, Mark Abajian, Keck/USC
Under Moderate Illness, there is only one leaf node of "Moderate Illness (Derived)". Along with this Supplemental Oxygen could be included.
We need to stay current with the ACT COVID phenotype releases, but don't need intermediate work in progress. Would it be possible to provide some of mechanism (github release being one possibility) to indicate stable milestones that we can pull and use to update our own work?
Tooltips: Huge, huge, huge improvement in Version 3. Thank you!
Suggestion: These two concepts do not have their ICD10 codes displayed in their tooltips, even though other concepts in their same folder do have the ICD10 codes displayed in their tooltips. Add the ICD10 code to the tooltip for these?
Thank you.
In the ACT_COVID file dated April 2nd, the "ANY Positive" lab has concept_code (basecode?)
"UMLS:C1335447 "
which has a blank space at the end. I don't know whether that would cause an issue in practice, since I have not deployed this yet. Just fyi.—Mark Abajian
Course Of Illness/Medications/Angiotensin II Inhibitor:
Angiotensin II which is not an Angiotensin II inhibitor can be removed.
I suggest that at the end of the term we include the coding system and the code in parenthesis, for example:
Extracorporeal Membrane Oxygenation, Continuous (ICD10PCS:5A15223)
Continuous negative pressure ventilation (CNP), initiation and management (CPT4:94662)
Respiratory System Diagnosis With Ventilator Support (DRG:475)
And for derived terms:
Mechanical Ventilation (Derived UMLS:C0199470)
As of now there are some portions of the Covid ontology that have terms that do not show the code in the name of the term.
Angiotensin II Inhibitor -> Angiotensin II Inhibitors
Non-Steroidal Anti-Inflammatory Drug (NSAID) -> Nonsteroidal Anti-Inflammatory Drugs (NSAIDs)
Just a suggestion that we maintain a version number in the COVID-19 ontology tree, like we do for the other ontology domains in ACT.
I have an example here that I have added to our copy of the ontology (metadata and shrine_ont) in SQL Server. Please see attached.
ACT_COVID_Insert_Version_20200428ma.sql.txt
I imagine that this code is database-agnostic, but your mileage may vary.
Based on ACT meeting on 5/26/2020 we want to separate the nuclei acids and the antibodies test values.
Rename harmonized values for nucleic acid test as follows:
ANY Equivocal Lab Test -> ANY Equivocal Lab Test (Nucleic Acids)
ANY Negative Lab Test -> ANY Negative Lab Test (Nucleic Acids)
ANY Pending Lab Test -> ANY Pending Lab Test (Nucleic Acids)
ANY Positive Lab Test -> ANY Positive Lab Test (Nucleic Acids)
Create new harmonized values for antibody tests as follows:
ANY Equivocal Lab Test (Antibodies)
ANY Negative Lab Test (Antibodies)
ANY Pending Lab Test (Antibodies)
ANY Positive Lab Test (Antibodies)
Minor name changes:
SARS-CoV-2 nucleic acids (Respiratory specimen) -> SARS-CoV-2 Nucleic Acids (Respiratory specimen) [capitalize nucleic acids]
SARS-CoV-2 nucleic acids (Unspecified specimen) -> SARS-CoV-2 Nucleic Acids (Unspecified specimen) [capitalize nucleic acids]
Change "Cytokine Release Syndromes Symptoms (CRS/SIRS)" to "Cytokine Release Syndrome (CRS/SIRS)"
The following columns in the act_covid file:
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_120646007\LOINC_94505-5\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_120646007\LOINC_94506-3\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_430304001\LOINC_94509-7\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_430304001\LOINC_94510-5\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_430304001\LOINC_94311-8\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_430304001\LOINC_94312-6\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_430304001\LOINC_94511-3\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0022885\SNOMED_430304001\LOINC_94313-4\
Have a data type of 'N' rather than 'T'.
It looks like there are missing concepts, or issues with the FULLNAME paths for child concepts of Mechanical Ventillation -> Ventilation Procedures -> ICD10 Ventilation Procedures
Same issues exist in both places where mechanical ventilation appears in the hierarchy:
Severe Illness -> Severe Diagnosis and Procedures
Respiratory Therapy Management
Ex: there is no concept between "ICD-10 Ventillation procedures", where fullname =
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0242656\UMLS_C1444343\MECHANICAL_VENT\VENT_PROCEDURES\VENT_ICD10
and "5A0945 Ventilation 24-96 Consecutive Hours" where fullname =
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0242656\UMLS_C1444343\MECHANICAL_VENT\VENT_PROCEDURES\VENT_ICD10\A28322331\A28335107\
In addition, some concepts that appear to be folders (have active child concepts in the hierarchy), have LA for a visual attribute, which may be preventing child terms from displaying through navigation.
ex: 5A0955 Assistance with Respiratory Ventilation, Greater than 96 Consecutive Hours, Continuous Positive Airway Pressure
I've attached a snipped of the ontology to highlight this. I have not explored whether similar issues occur elsewhere.
COVID ontology error - mech vent procedures.xlsx
--
Based on ACT meeting on 5/26/2020 we want to specify that the moderate and severe illness concepts are specific to the respiratory system which is the commonest presentation in Covid. In future, this will allow us to add moderate and severe illness concepts for other systems e.g. renal, neurological.
Changes:
Illness Severity -> Illness Severity (Respiratory)
Moderate Illness -> Moderate Illness (Respiratory)
Moderate Illness (Derived) -> Moderate Illness (Respiratory) (Derived)
Severe Illness -> Severe Illness (Respiratory)
Severe Illness (Derived) -> Severe Illness (Respiratory) (Derived)
In 2 concepts that I've found, there's a trailing space in the C_FULLNAME field. Both concepts are paths for the same procedure (same C_NAME and basecode):
"5A0955 Assistance with Respiratory Ventilation, Greater than 96 Consecutive Hours, Continuous Positive Airway Pressure"
C_FULLNAME values are:
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0242656\DISEASE_SEVERITY\SEVERE_DISEASE\SEVERE_DIAGNOSIS\MECHANICAL_VENT\VENT_PROCEDURES\VENT_ICD10\A28341447\
\ACT\UMLS_C0031437\SNOMED_3947185011\UMLS_C0242656\UMLS_C1444343\MECHANICAL_VENT\VENT_PROCEDURES\VENT_ICD10\A28341447\
I did not find another instances of this issue in the covid ontology. I have not checked the full ACT ontology.
Line 3704 has an embedded line feed preceding ' ;Emergency Department'
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.