Relating NSVs and Attributions to a Domain

The Supplemental Qualifiers (SUPP--) special purpose dataset model is used to represent both nonstandard variables (NSVs) and attributions with their relationship to records in a parent dataset. SUPP-- will only be used with

There will be only one SUPP-- dataset for a parent dataset. Relationships between records in the SUPP-- dataset and observations in the parent dataset are specified using the key variables STUDYID, RDOMAIN, and USUBJID, along with IDVAR and IDVARVAL (the value of the identifying variable). IDVARVAL will contain the value of the variable described in IDVAR. Single records can be related by using a unique-record-identifier variable such as --SEQ in IDVAR. Groups of records can be related by using grouping variables such as --GRPID in IDVAR.  Using --GRPID can be a more efficient method of representing relationships in RELREC when relating a single or group of records in one dataset to a group of records in another dataset. 

In a SUPP-- dataset, the combined set of values for STUDYID, USUBJID, POOLID (for nonclinical studies only), IDVAR, IDVARVAL, and QNAM will be unique for every record. There will not be multiple records in a SUPP-- dataset for the same QNAM value, as it relates to IDVAR/IDVARVAL for a USUBJID in a domain. Using --GRPID can be a more efficient method to identify individual qualifier values (SUPP-- records) related to multiple general observation class domain records that could be grouped, such as relating an attribution to a group of laboratory measurements.

Subject records or datasets expressing a relationship are specified using the key variables STUDYID, RDOMAIN (the domain code of the record or dataset in the relationship) with IDVAR (the variable that identifies the related record(s)). Each record in the RELREC special-purpose dataset contains keys that identify a record (or group of records) and an identifier for the relationship which is stored in the RELID variable. The value of RELID is chosen by the applicant and it is recommended that applicants use a standard system or naming convention for RELID (e.g., all letters, all numbers, capitalized).

Relationships Between Records for a Subject

Relationships between records for a subject in different datasets are specified using the key variables STUDYID, RDOMAIN, and USUBJID, along with IDVAR and IDVARVAL (the value of the identifying variable). IDVARVAL will contain the value of the variable described in IDVAR. Single records can be related by using a unique-record-identifier variable such as --SEQ in IDVAR. Groups of records can be related by using grouping variables such as --GRPID in IDVAR.  Using --GRPID can be a more efficient method of representing relationships in RELREC when relating a single or group of records in one dataset to a group of records in another dataset. Variable RELTYPE will not be used when relating records for a subject and is only used when representing relationships between datasets. The value of RELID will be identical for all related records within USUBJID.

Nonstandard Variables

SUPP-- represents the metadata and data for each NSV/value combination and relates this to the observation or observations qualified by the NSV in the parent dataset. Each SUPP-- dataset record also includes the name of the qualifier variable being added (QNAM), the label for the variable (QLABEL), the actual value for each instance or record (QVAL), the origin (QORIG) of the value, and the evaluator (QEVAL) to specify the role of the individual who assigned the value when the value is assigned. Gary - How can we describe origin here? Can we simply refer to Define-XML?

Attributions

An attribution is typically an interpretation or subjective classification of 1 or more observations by a specific evaluator. A SUPP-- dataset can contain both objective data (where values are collected or derived algorithmically) and subjective data (attributions where values are assigned by a person or committee). For objective data, the value in QEVAL will be null. For subjective data, the value in QEVAL should reflect the role of the person assigning the value.