ANSWERED! CTE is the best method, using descriptive dummyvariable names
I have two relevant tables concept
which describes all of the database wide concepts (concept_code
) from different vocabularies (vocabulary_id
), and relates them to a unique, standardized ID concept_id
, and concept_relationship
which only has three columns, concept_id_1
, concept_id_2
, and relationship_id
. Relationship ID describes how ID one relates to ID 2. The database team attempts to standardize concepts into one ID as much as possible using the OMOP standards, so many IDs that are at source from one vocabulary get remapped to their corresponding parent standardized ID using the "Maps to" relationship. I have sets of codes which correspond to IDs that were defined by subject matter experts in specific vocabularies that don't necessarily include what the database team chose as their standardized concept. The end product I want is a table representing both the codes/IDs defined by the SMEs, and the corresponding IDs that they "map to".
########################################################################
The new code that seems to be behaving as intended uses a WITH statement:
WITH query1 AS (
SELECT
*
FROM
concept
WHERE
concept_code IN ("398.4", "I45")
AND (
vocabulary_id IN ("ICD9","ICD10")
)
)
SELECT *
FROM concept
WHERE concept_id IN (
SELECT concept_id_2
FROM concept_relationship cr
INNER JOIN query1 q1 ON cr.concept_id_1 = q1.concept_id
WHERE relationship_id = "Maps to"
)
UNION
SELECT * FROM query1
########################################################################
I have tried the query below:
SELECT
foo.*
FROM
(
SELECT
*
FROM
concept
WHERE
concept_code IN ("398.4", "I45")
AND (
vocabulary_id IN ("ICD9", "ICD10")
)
) AS foo
UNION
SELECT
*
FROM
concept
WHERE
concept_id IN (
SELECT
concept_id_2
FROM
concept_relationship cr
INNER JOIN foo ON cr.concept_id_1 = foo.concept_id
WHERE
relationship_id = "Maps to"
)
But it returns an "INVALID ARGUMENT" error, I'm guessing because I can't pass the subquery foo above below the UNION to the other query. Is there a smoother way to do this? I've included some dummy tables below to attempt reproducibility, which seems to work database-side, but all columns are not included for conciseness.
concept :
| concept_id | concept_code | vocabulary_id |
| ----------- | ------------ |---------------|
| 123 | 398.4 | ICD9 |
| 111 | I45 | ICD10 |
| 145 | 45155841 | SNOMED |
concept_relationship:
| concept_id_1 | concept_id_2 | relationship_id|
| ----------- | ------------ |--------------- |
| 123 | 145 | Maps to |
| 111 | 145 | Maps to |
| 145 | 145 | Maps to |
| 145 | 111 | Maps from |
| 145 | 123 | Maps from |
| 145 | 145 | Maps from |
CREATE TABLE `concept` (
`concept_id` VARCHAR NOT NULL,
`concept_code` VARCHAR NOT NULL,
`vocabulary_id` VARCHAR NOT NULL,
PRIMARY KEY (`concept_id`)
);
INSERT INTO concept (
concept_id, concept_code, vocabulary_id
)
VALUES
("123", "388.4", "ICD9"),
("111", "I45", "ICD10"),
("145", "45155841 ", "SNOMED");
CREATE TABLE `concept` (
`concept_id_1` VARCHAR NOT NULL,
`concept_id_2` VARCHAR NOT NULL,
`relationship_id` VARCHAR NOT NULL,
PRIMARY KEY (`concept_id_1`)
);
INSERT INTO concept_relationship (
concept_id_1, concept_id_2, relationship_id
)
VALUES
("123", "145", "Maps to"),
("111", "145", "Maps to"),
("145", "145", "Maps to"),
("145", "111", "Maps from"),
("145", "123", "Maps from"),
("145", "145", "Maps from");
CodePudding user response:
The problem is you use foo on the other side of the union and it can't see it. but a CTE can help like this
WITH foo AS (
SELECT *
FROM concept
WHERE concept_code IN ("398.4", "I45") AND
vocabulary_id IN ("ICD9", "ICD10")
)
SELECT *
FROM foo
UNION
SELECT *
FROM concept
WHERE concept_id IN (
SELECT concept_id_2
FROM concept_relationship cr
JOIN foo ON cr.concept_id_1 = foo.concept_id
WHERE relationship_id = "Maps to"
)
note the 2nd query after the UNION can be re-written as
SELECT *
FROM concept
JOIN concept_relationship cr ON cr.concept_id_2 = concept.conecpt_id
AND cr.relationship_id = "Maps to"
JOIN foo on cr.concept_id_1 = foo.concept_id
using joins.