OHDSI Home | Forums | Wiki | Github

ATLAS Characteristions Feature Analyses design

Developers
#1

Hi all,
I have some questions about Characterizations on ATLAS :

  1. How could I design new feature analyses using aggregated functions. e.g. patient maximum temperature in the first 24 h of admission

  2. How can I report missing data in characterization?

  3. where could I find SQL code for existing feature analyses? now the #71 Demographics Age Group analysis is in 5 year age groups, I want to edit the age groups to 10 year age. Is there any method to see the original design and edit based on it?

Atlas - characterization - how are condition [era] groups defined?
#2

In 2.8, you can now define distribution statistics, such as max temp within a time window of cohort entry. Assuming your cohort entry is the admission, the following link will take you to a distribution criteria that will find the max temp (C and F separately, there may be other ways to measure temp? (K)?) You add it to your cohort characterization to use:
http://atlas-demo.ohdsi.org/#/cc/feature-analyses/172

Iā€™m not sure what you mean by ā€˜missing dataā€™.

The demographics age group SQL is found here: https://github.com/OHDSI/FeatureExtraction/blob/master/inst/sql/sql_server/DemographicsAgeGroup.sql.

You can make your own SQL and create a custom feature where you provide your own custom SQL script. If you go to ā€˜new featureā€™ and look at the sample sql, you can see how you need to return the appropriate columns in order for the custom sql to work. here is the template:

-- Custom analysis producing same results as Feature Extraction's "One covariate per drug in the drug_era table overlapping with any time prior to index."
SELECT
  CAST(drug_concept_id AS BIGINT) * 1000 + @analysis_id AS covariate_id,
  c.concept_name                                                                  AS covariate_name,
  drug_concept_id                                                                 AS concept_id,
  COUNT(*)                                                                            AS sum_value,
  COUNT(*) * 1.0 / stat.total_cnt * 1.0                                   AS average_value
FROM (
       SELECT DISTINCT
         drug_concept_id,
         cohort.subject_id,
         cohort.cohort_start_date
       FROM @cohort_table cohort
         INNER JOIN @cdm_database_schema.drug_era ON cohort.subject_id = drug_era.person_id
       WHERE drug_era_start_date <= cohort.cohort_start_date
             AND drug_concept_id != 0
             AND cohort.cohort_definition_id = @cohort_id
     ) drug_entries
  JOIN @cdm_database_schema.concept c ON drug_entries.drug_concept_id = c.concept_id
  CROSS JOIN (SELECT COUNT(*) total_cnt
              FROM @cohort_table
              WHERE cohort_definition_id = @cohort_id) stat
GROUP BY drug_concept_id, c.concept_name, stat.total_cnt

Note the place holders of @cdm_database_schema and @cohort_table must be used in your own SQL in order for the program to properly replace the cohort and schemas in your local CDM.

2 Likes
[Atlas] Feature analysis - sql codes for existing features
#3

Thank you!

#4

Hi @Chris_Knoll ,

This was helpful. If I wanted to incorporate this feature, for example, into an estimation study that I downloaded from Atlas, how would I do this?

Thanks

#5

If you used custom sql, then you can follow the instructions here.

#6

Thatā€™s what Iā€™m doing, but I donā€™t know how to get the custom feature added to an estimation package. The package I downloaded from Atlas has all the analyses in cmAnalysisList.json and I want to add this new feature into it. In the package this line loads the analyses:

cmAnalysisList ā† CohortMethod::loadCmAnalysisList(cmAnalysisListFile)

and ten it calls:

results ā† CohortMethod::runCmAnalyses(ā€¦)

to run the analyses. Iā€™m not sure how I would add the covariateSettingsList
and subsequently call getDbCovariateData with runCmAnalyses.

#7

I believe in the instructions I sent you, you can add additional elements to cmAnalysisList before you call runCmAnalyses.

#8

Thatā€™s what I was thinking. Iā€™ll investigate to create an analysis and add the covariate setting to it.

t