Skip to content

Latest commit

 

History

History

metadata

README.md

Metadata

Our workflow is built around a series of explicit metadata tables. We can't fully automate the system, but we're hoping that the metadata files will

  • improve clarity and communication between the investigator and the CDW staff,
  • increase the CDW staff's efficiency of querying and manipulating data (on the first run, and on subsequent runs).

These metadata files are completed by the investigator and by the CDW staff as the project parameters are being defined. These are only example datasets --project-specific datasets should be saved in their private project-specific repository.

In the document below, the first table enumerates the metadata tables available to an investigator. The remaining tables document the columns within each table.

Table of Contents of Metadata Files

File Name Description
metadata-dx ICD-9 (and eventually ICD-10) diagnoses relevant to the investigation.
metadata-medlist The 'medlist' table in Centricity defines many lookup values, such as those for race and ethnicity.
metadata-obs-header Defines the types of observations to be considered.
metadata-order Labs and tests relevant to the investigation.

Terms

  • investigator: the person ... todo: add more terms

Data Dictionaries of Metadata Files

metadata-dx

Variable Description
dx_name English name of diagnosis.
phentotype Phentotype of diagnosis.
icd_9 Exact ICD-9 code. Only one code per row.
desired Indicates if the dx is desired for the specific investigation. This allows someone to reuse the table across multiple investigations, without having to delete or recover rows. Simply toggle its inclusion with a TRUE/FALSE value.

Constraints & Validation Checks:

  • All cells must be nonmissing.
  • dx_name cannot be repeated across multiple rows.
  • dx_name must be found in the CDW (specifically, [ML].[DIAGNOSIS].[DESCRIPTION]). A warning is thrown otherwise.
  • icd_9 cannot be repeated across multiple rows.
  • icd_9 must be found in the CDW (specifically, [ML].[DIAGNOSIS].[ICDCODE] where [ML].[DIAGNOSIS].[CODETYPE]=1 for ICD9) or in the ICD R package. A warning is thrown otherwise.
  • The dx_name and icd_9 combination must match the combination in the CDW.

metadata-medlist

These values show the possible levels within some important Centricity variables. The level_centricity EMR value was selected by the clinician. Many times a study sample will be too small to consider each level separately, so the investigator can specify how the Centricity levels should map to the desired new levels. For instance, the example file preserves the race levels for black, white, and hispanic, while combining the remaining levels into either other or unknown.

The retain_patient field indicates which patients should be included/excluded in the investigation. This permits a clear specification for an investigation focused on a specific target population (e.g., hispanic females).

Many times an investigator's grouping decisions are necessitated by statistical limitations. Depending on the goals, some levels that occur infrequently in our regional population cannot be reliabily estimated. We encourage the investigator to repeatedly try different groupings to strike a good balance. We've designed the Prairie Outpost so that subsequent runs (after the investigator modified the metadata) should require minimal time from our staff.

Variable Description
table_name 'Table Name' in Centricity. This shows the variables' groupings.
medlist_id ID value within Centricity.
level_by_centricity Level assigned within Centricity.
level_by_investigator New level assigned by the investigator.
retain_patient Should patients with this value be included in the investigation?

Constraints & Validation Checks:

  • All cells must be nonmissing.
  • medlist_id cannot be repeated across multiple rows.
  • table_name, medlist_id, and level_by_centricity must be found in the CDW (specifically, TABLENAME, MEDLISTID, and DESCRIPTION in [ML].[DIAGNOSIS]). A warning is thrown otherwise.
  • medlist_id cannot be repeated across multiple rows.
  • The table_name and level_by_centricity combination must be unique.

metadata-obs-header

The 'obs' (short for 'observation') values in Centricity are classified single values; they can be numbers, dates, or text. Obs typically lead to values that are cleaner, better structured, and better classified than values recovered from Centricity's free-text 'documents'.

The investigator should specify minimum and maximum reasonable values (for numbers and dates). Values outside this range are set to missing. This range is one place we can reduce the implausible values from polluting the analysis.

There are thousands of possible obs headers. The Prairie Outpost staff will make an educated guess which headers the investigators might consider to include (based on their initial description). However it is the investigators' responsibility to confirm that this list contains all the desired headers.

Variable Description
hdid The ID for the observation 'header' in Centricity. This specifies the type of observation.
header A short English name of the header in Centricity.
description A longer description of the header, which typically specifies the units.
reasonable_min The minimum reasonable value --assigned by the investigator. (If the value type isn't a number or date, please specify NA.)
reasonable_max The maximum reasonable value --assigned by the investigator. (If the value type isn't a number or date, please specify NA.)
retain_obs Should this type of observation be included in the investigation?

Constraints & Validation Checks:

  • All cells must be nonmissing. NA should be entered as the min & max value associated with a text observation.
  • hdid cannot be repeated across multiple rows.
  • hdid, header, and name_long must be found in the CDW (specifically, HDID, NAME, and DESCRIPTION in [ML].[OBSHEAD]). A warning is thrown otherwise.

metadata-order

category_name,category_id,test,order_code,order_description,desired,comments

Variable Description
test_name English name of the test/lab.
abbreviations Abbreviated notation of the test. Multiple values should be separated by a semicolon.
cpts CPT code. Multiple values should be separated by a semicolon.
desired Indicates if the dx is desired for the specific investigation. This allows someone to reuse the table across multiple investigations, without having to delete or recover rows. Its inclusion is toggled with a TRUE/FALSE value.

Constraints & Validation Checks:

  • All cells must be nonmissing.
  • test_name cannot be repeated across multiple rows.
  • abbreviations elements cannot be repeated across multiple rows.
  • abbreviations elements must be found in the CDW. A warning is thrown otherwise.
  • cpts elements cannot be repeated across multiple rows.
  • cpts elements must be found in the CDW. A warning is thrown otherwise.

####Location of Care Code *In the WHERE section of code, the [NAME] and [ABBREVNAME] will change for each project.

  d.LOCOFCARE 
  , l.[NAME] 
  , COUNT(*) AS document_count 
  , l.[SEARCHNAME] 
  , l.[ABBREVNAME] 
  , l.[ADDRESS1] 
FROM centricity.[DOCUMENT] AS d 
  LEFT OUTER JOIN [centricity_emr].[peek].[LOCREG] AS l ON d.LOCOFCARE=l.LOCID 
WHERE 
  ([NAME] LIKE '%comm%') 
  OR ([NAME] LIKE '%latin%') 
  OR ([NAME] LIKE '%general%') 
  OR ([NAME] LIKE '%adolesc%') 
  OR ([ABBREVNAME* LIKE '%sp_ou%') 
GROUP BY  
  d.LOCOFCARE 
  ,l.[NAME] 
  ,l.[SEARCHNAME] 
  ,l.[ABBREVNAME] 
  ,l.[ADDRESS1] 
ORDER BY document_count DESC