See first swipe by @dimshitc here:
https://docs.google.com/spreadsheets/d/1t4MlNxRmCTuEeFockb58boiqNf4xSgwKe2TuybQ5plQ/edit#gid=1970199715

Rimma's instructions:

• Parse into computable format.
• Describe treatment variables.
• Pick candidate sites: breast, brain, prostate
• Pick attributes based on use cases.
• Only import attributes/values that make cut.

@dimshitc
I used this NAACCR website to filter the list of relevant NAACCR items down to a more manageable size based on the Section each item belongs to ('Cancer Identification', 'Stage/Prognostic Factors', 'Treatment-1st Course', 'Treatment-Subsequent & Other')

http://applications.naaccr.org/querybuilder/default.aspx?Version=18

Here is the filtered list:

naaccr_questions_filtered.xlsx

Here is @dimshitc first cut at putting NAACCR into a computable format:

https://docs.google.com/spreadsheets/d/1dEvy9xWhEfy-AnjhItMOe5MHbaUUrMZwSk2rZCzxO0k/edit?usp=sharing

Procedures
general idea:
Map their sites to SNOMED, if not possible - create a new standard concept.
then build relationships.
For example:
we have the procedure code "50" for Total proctectomy in "Rectosigmoid" Site.
we create a new concept with
concept_code = 'Proc_50_Rectosigmoid' = 'Proc_'||code||'_'||schema
concept_name= 'Total proctectomy'
vocabulary_id ='NAACCR'
and build the relationships:
'Proc_50_Rectosigmoid' | Maps to | 235364003 (Total proctectomy - SNOMED)
'Proc_50_Rectosigmoid' | Has Site (NAACCR) | 81922002 Rectosigmoid structure (SNOMED)
Note, I used concept_codes for SNOMED in the example

details:

1. Manually identify the procedures needed to be concatenated with their parent concepts description to make a correct name
   e.g.
   source: Ovary | 27 | "WITHOUT hysterectomy" | parent_concept = 25 "Total removal of tumor or (single) ovary, NOS"
   omop: Proc_27_Ovary | "Total removal of tumor or (single) ovary, NOS, WITHOUT hysterectomy"

2. A lot of procedure codes itself have the same meaning regardless of the site, for example "10 Local tumor destruction, NOS"
   But anyway, we create a separate procedure for each site, right?
   e.g
   source: 10 | Local tumor destruction, NOS | Bladder
   omop: Proc_10_Bladder | "Local tumor destruction, NOS of Bladder"

Histological grades:
NAACCR #3843 Grade Clinical
NAACCR #3844 Grade Pathological
NAACCR #3845 Grade Post Therapy
share the same bunch of values within the site

1 G1: Well differentiated
2 G2: Moderately differentiated
3 G3: Poorly differentiated
4 G4: Undifferentiated
8 Not applicable
9 Grade cannot be assessed (GX); Unknown

with same unsignificant name variations:
1 WHO Grade I : Circumscribed tumors of low proliferative potential associated with the possibility of cure following resection (Brain)
1 G1: Low combined histologic grade (favorable), SBR score of 3-5 points (Breast)
or
2 WHO Grade II: Infiltrative tumors with low proliferative potential with increased risk of recurrence (Brain)
2 G2: Intermediate combined histologic grade (moderately favorable); SBR score of 6-7 points (Breast)

A Well differentiated
B Moderately differentiated
C Poorly differentiated
D Undifferentiated, anaplastic
E Stated as ""Gleason score 7"" with no patterns documented or Any Gleason patterns combination equal to 7 not specified in 2 or 3 - exists only for Prostate

H High grade
L Low grade

M TP53 or CTNNB Mutation - Adrenal Gland
M Nuclear Grade II (interMediate) (in situ only) - Breast
S Sarcomatous overgrowth - Corpus Adenosarcoma only

each site uses only numbers OR only letters with number 9 (unknown), 1-4 is synonymical to A-D though

How to create a concept
for all in this way:
concept_code = 'Grade_1'
concpet_name = 'G1: Well differentiated'

except "M":
concept_code = 'Grade_M_Adranal_Gland'
concpet_name = 'TP53 or CTNNB Mutation'

concept_code = 'Grade_M_Adranal_Breast'
concpet_name = 'Nuclear Grade II (interMediate) (in situ only)'

should we neglect the name difference?
is it the right way to represent the concept codes?

@dimshitc @rimusia

Rimma and I are going to outline two possible ways to ingest NAACCR into the OMOP vocabulary tables. I will document the options in this issue. Then we will meet to discuss the options.

@dimshitc @cgreich @rimusia

We should only ingest into the OMOP vocabulary tables NAACCR data items from the following NAACCR sections:

• 'Cancer Identification'
• 'Stage/Prognostic Factors'
• 'Treatment-1st Course'

This covers 442 NAACCR items. This will help us focus on the good stuff in NAACCR and weed out the cruft.

Some NAACCR items are site-independent and some are site-dependent.

• Site-independent means that the NAACCR item is not bound to an an ICD-O-3 site/histology combination and the NAACCR item's list of possible values does not depend/alter based on the ICD-O-3 site/histology combination of the tumor diagnosis it describes/applies to.
• Site-dependent means that the NAACCR item is bound to an an ICD-O-3 site/histology combination and possibly the NAACCR item's list of possible values does depend/alter based on the ICD-O-3 site/histology of the tumor diagnosis it describes/applies to.

Ingesting Site-independent NAACCR Items into the OMOP Vocabulary Tables

Site-independent NAACCR items should be rather straightforward to ingest into the OMOP vocabulary tables. For example, the NAACCR item #410 'Laterality' should be ingested like so:

CONCEPT

CONCEPT_RELATIONSHIP

Ingesting Site-dependent NAACCR Items into the OMOP Vocabulary Tables

Site-dependent NAACCR items are not as straightforward to ingest into the OMOP vocabulary tables. There are 3 different bindings of site-dependent NAACCR items:

• SEER EOD (Extent of Disease) Data : https://staging.seer.cancer.gov/eod_public/list/1.4/
• SEER TNM (Tumour Node Metastasis) Data: https://staging.seer.cancer.gov/tnm/list/1.9/
• NAACCR SSDI (Site Specific Data Items): https://apps.naaccr.org/ssdi/list/

SEER EOD includes NAACCR SSDI. So we can focus on SEER EOD and SEER TNM. Both of these bindings are bound to schemas. Schemas are high-level names of anatomic sites that bundle combinations of ICD-O-3 site/histology. The schemas for SEER EOD and SEER TNM overlap but are not equivalent. Some schema examples:

Some of the NAACCR items bound to an EOD schema/TNM schema have possible values that do not vary based on schema. Some of the NAACCR items bound to an EOD schema/TNM schema have possible values that do vary based on schema.

Here is an examples of a non-varying NAACCR item from two EOD schemas:

This NAACCR item can be ingested into the OMOP vocabulary tables like so:

CONCEPT

CONCEPT_RELATIONSHIP

Here is an examples of a NAACCR item that has varying possible values across EOD schemas:

This NAACCR item can be ingested into the OMOP vocabulary tables like so:

CONCEPT

CONCEPT_RELATIONSHIP

domain_id ='NAACCR', it can't be. should be Observation or Measurement depending on where you want to go this data to.

@dimshitc That is a typo. Sorry. Fixed. The domain for most NAACCR attributes/questions should be 'Measurement' and for NAACCR values/answers should be 'Meas Value'. The vocabulary_id should be NAACCR. Some treatment-related NAACCR attributes/questions values/answers will be in the new Treatment domain (I will work on the Treatment NAACCR items/item answers this weekend).

The first version of this has been completed.