Project Meeting -- 07/29/2019 @ 15:00

Meeting Date: 07/29/2019 Topic: Biweekly Meeting Attendees: @mgkahn

Proposed Agenda:

Briefly discuss cohort queries, with the goal of understanding whether or not I need someone to verify my queries. Queries can be found here:

ADHD - #100

Appendicitis - #101

Crohn's Disease - #102

Hypothyroidism - #103

Peanut Allergy - #104

Sickle Cell Disease - #105

Sleep Apnea - #106

Steroid-Induced Osteonecrosis - #107

Systemic Lupus Erythematosus - #108

Discuss ADHD results (see first tab of the spreadsheet here) and what we need to do to "confirm" or "validate" the cohorts produced by each method. The tabs for the other phenotypes are present in this spreadsheet, but are not finalized. Please ignore them for now.

UPDATE: Great meeting today! I emailed the queries that need verification (and are referenced above). As a record, those queries are also provided below:

Thanks for being willing to take a look at these (listed in order from most to least confident):

ADHD: https://github.com/callahantiff/PheKnowVec/issues/100
Sickle Cell Disease: https://github.com/callahantiff/PheKnowVec/issues/105
Sleep Apnea: https://github.com/callahantiff/PheKnowVec/issues/106
Systemic Lupus Erythematosus: https://github.com/callahantiff/PheKnowVec/issues/108
Crohn's Disease: https://github.com/callahantiff/PheKnowVec/issues/102
Peanut Allergy: https://github.com/callahantiff/PheKnowVec/issues/104
Steroid-Induced Osteonecrosis: https://github.com/callahantiff/PheKnowVec/issues/107
Appendicitis: https://github.com/callahantiff/PheKnowVec/issues/101
Hypothyroidism: https://github.com/callahantiff/PheKnowVec/issues/103

For each query, would you mind taking a look at each logical statement and it’s corresponding code chunk (example shown below) just to make sure that I have properly covered it and didn’t do something silly?

Logical Statement: age_criteria_1: Patient is 1460 days or older (age at earliest diagnosis of ADHD) Matching SQL code chunk:

age_criteria_1 AS (
SELECT co.person_id, cohort.standard_code_set AS code_set 
FROM 
  {database}.person p, 
  {database}.condition_occurrence co, 
  {database}.ADHD_COHORT_VARS cohort 
WHERE p.person_id = co.person_id 
  AND co.condition_concept_id = CAST(cohort.standard_concept_id AS int64) 
  AND cohort.phenotype_definition_number = 1
  AND cohort.standard_code_set = {code_set_group}
GROUP BY
  co.person_id, p.birth_datetime, cohort.standard_code_set
HAVING
  DATETIME_DIFF(DATETIME(MIN(co.condition_start_datetime)), DATETIME(p.birth_datetime), DAY) >= 1460
)

database and code_set_group are looped over when running the code from python. code_set_group is for each of the mapping strategies we are evaluating. The phenotype_definition_number is a number I have linked to all of the “ADHD codes”, this approached is utilized to help me keep track of all of the different codes.

To help keep track of the different code chunks and how they are combined, I have added a logic table for each cohort, their “chunks” and how those should be combined. For example, here is the table for ADHD:

COHORT	CHUNK	LOGICAL OPERATOR
CASE TYPE 1	age_criteria_1	AND
CASE TYPE 1	dx_case_inclusion_criteria_1	AND
CASE TYPE 1	rx_case_inclusion_criteria_1	AND
CASE TYPE 1	dx_case_exclusion_criteria_1	---

CASE TYPE 2	age_criteria_1	AND
CASE TYPE 2	dx_case_inclusion_criteria_2	AND
CASE TYPE 2	dx_case_exclusion_criteria_1	---

CONTROL	age_criteria_2	AND
CONTROL	visit_criteria_1	AND
CONTROL	rx_control_exclusion_criteria_1	OR
CONTROL	dx_control_exclusion_criteria_1	OR
CONTROL	dx_control_exclusion_criteria_2	—

Closing this issue. Please re-open if there is anything that needs clarification.

callahantiff / PheKnowVec

Project Meeting -- 07/29/2019 @ 15:00 #109