Download PDF

discrepancy management

Discrepancy management is also called as query management, is a process of cleaning subject data in the Clinical Data Management Systems(CDMS). Discrepancies are raised/ created during data entry and data validation.

Where data entered does not pass validation rules then a data query may be issued to the investigative site where the clinical trial is conducted to request clarification of the entry. Data queries must not be leading (i.e. they must not suggest the correction that should be made).

For electronic CRFs only the site staff with appropriate access may modify data entries. For paper CRFs, the clinical data manager applies the data query response to the database and a copy of the data query is retained at the investigative site.

Query management is done via Data Clarification Form (DCF).

It is done through:

  1. Manual verification
  2. Computerized verification

 Discrepancy Types:

Univariate: During data entry or batch data load, Oracle Clinical checks data entered as a response to a question against the definition of the question, and generates a discrepancy if the response does not meet the definition’s specifications, such as: wrong data type or length, response not one of a discrete value group (for example, the answer to a Yes/No question is X), mandatory response missing, precision of response greater than allowed (for example, temperature of 98.689F), incorrect partial date (for example, day of month missing where required), or responses to the question that are above the upper bound or below the lower bound.

Multivariate: A response may have to meet the criteria set out in one or more Validation Procedures, which compare responses to one or more other responses for the same patient. The system creates discrepancies when it runs Validation Procedures, either during Batch Validation or manual execution of a Procedure.

Indicator: The response to an indicator question determines which set of the remaining questions require responses. For example, if the response to the indicator question “Do you smoke?” is Yes, then the question “How often?” must also be collected. If the response to “Do you smoke?” is No, then “How often?” must not be collected. If a follow-up question is either not collected when it should be, or collected when it should not be, Oracle Clinical creates an indicator discrepancy during Batch Validation. The discrepancy is logged against the indicator question; for example, “Do you smoke?”, not “How often?”.

Manual: If there is a problem transcribing data from a CRF, you can enter a manual discrepancy. There are two types:

>>Manual Data Point. If there is a problem with the CRF, such as illegibility, you can enter a manual discrepancy either during data entry (see Chapter 1, “Entering Data”) or directly into the discrepancy database (see “Entering Manual Discrepancies”).

>>Manual Header. This type of discrepancy is not directly associated with a single response, but with the CRF in general. You can enter a manual header discrepancy associated with a Received DCM directly in the Maintain Discrepancy Database window.

The system creates univariate, multivariate, and indicator discrepancies. Users create manual discrepancies.

Procedure for query management:

Identify Discrepancy —–> Find the status of Discrepancy ——–> Query Desig ——-> Query form sent to investigator/ Sponsor ———> Identifying the Query by the Investigator ——–>Final resolution ——> Update

The important thing for query design is LSA i.e. LOCATE, STATE, ASK for resolution for the discrepencies.