Example: Using Discrepancies to Clean Data

In order to collect a patient's blood pressure, your organization creates two questions, one for the diastolic and one for the systolic number. To check the validity of the values entered in response to each question, each question is defined as being of data type Number and having an upper and lower limit marking an acceptable, medically plausible, range of values. In addition, your organization defines a Validation Procedure to compare the responses of questions collecting systolic and diastolic blood pressure and to create a discrepancy if the diastolic response value is higher than the systolic response value.

When the system finds a case where, at a particular visit, a patient's diastolic blood pressure is recorded as higher than the systolic blood pressure, it creates a multivariate discrepancy. If the diastolic value is out of the range defined for its question, the system also creates a univariate discrepancy. Both discrepancies appear in the Maintain Discrepancy Database window when you enter an appropriate query.

In the Maintain Discrepancy Database window you can click the More button in order to see the questions and their responses containing the blood pressure figures. If you select the diastolic question and click Details, you see both discrepancies, the multivariate one and the univariate one, listed in the bottom panel of the window.

If your organization has the feature set up, you can go to the Special menu and select Get Image to see an image of the actual paper CRF where the figures were recorded at the patient's visit. If you see that the two values were transposed during Data Entry, and if you have the necessary privileges, you can go to the Special menu, select Update Patient Data, correct the two values, enter a comment, and update the discrepancy's review status according to your organization's policies. Assuming that the diastolic value is now within range, the system closes both discrepancies during the next batch validation.

Alternatively, you could select the DCF Report? box for each question (Procedure variable) involved in the multivariate discrepancy, so that when the discrepancy is included in a DCF Report for an Investigator, both values are displayed. The Investigator then has the responsibility for directing the changes to be made. The changes are entered and the discrepancy review status updated according to your organization's policies.