| Oracle® Health Sciences Data Management Workbench User's Guide Release 2.3.1 Part Number E35217-02 |
|
|
View PDF |
| Oracle® Health Sciences Data Management Workbench User's Guide Release 2.3.1 Part Number E35217-02 |
|
|
View PDF |
Oracle Health Sciences Data Management Workbench (Oracle DMW) is designed for two basic functions:
It is designed for integration with other systems:
Oracle Health Sciences Data Management Workbench (Oracle DMW) lets you load and merge clinical data and metadata from various sources, including InForm and labs, in one location where you can then transform raw clinical data to successive standard structures suitable for review and analysis. These structures are sets of tables called clinical data models that can be reused from a library or from study templates.
The diagram shows a simple study setup in Oracle DMW. All studies have a clinical data model for each data source: here, a local lab, a central lab, and an electronic data capture system—InForm. The tables in these clinical data models are generated with the same structure as in the source system. As Oracle SQL Loader loads data from each source it checks data conformance. Then a transformation program reads the raw source data, derives values, merges and rearranges data as required, and writes to a target model whose purpose is data review. Validation (edit) checks run on data in the review model and identify data discrepancies. Then another transformation program reads data from the review model and writes to a statistics analysis model in SDTM format.
You can create any number of clinical data models and transformations from one to another. You can have parallel models; a single linear chain is not required. Creating a transformation from one model to another is a partially automated process and if you make changes to the source or target model of an existing transformation, you can use the Upgrade feature to upgrade the transformation to reflect these changes.
If you build up a set of validated clinical data models and study templates, you can reuse them in many studies, focusing validation effort on any changes you make in a particular study.
If you reuse standard source and target models without change, you can reuse the transformation between them without change.
Data Lineage The system tracks each data point from tables in the source models through each transformation to the final model. The system recognizes the unique key of each source record, enabling the exchange of comments and data updates between Oracle DMW and source data systems.
Data Blinding You can hide sensitive data by blinding entire tables or providing masking values for specific columns, rows, or cells. Once blinded, sensitive data cannot be viewed, even through multiple transformations, without explicit authorization by a user with a combination of special privileges.
Data Export You can create an Oracle LSH data mart to export data in files. Supported formats include SAS Transport (XPort and CPort), SAS datasets, fixed length or delimited text, or Oracle Export, optionally compressed into zip files.
All Data—One Location Oracle DMW provides a full set of features that allow a data manager to easily review clinical data from various sources at one central location.
Seamless Hand-off Data managers, medical monitors, and biostatisticians can all work in a single environment, handing off work to one another as required.
A discrepancy—called a query in InForm—is associated with a single data point that has been identified as faulty or possibly faulty. Identifying and resolving discrepancies and correcting the underlying data helps ensure that clinical data is complete, accurate, and compliant with the study protocol.
Discrepancies are identified and created in several ways:
InForm queries are imported as discrepancies with their source context intact. Oracle DMW users can add a comment or question to an InForm query discrepancy and send it back to InForm, where a CRA or other InForm user can respond and/or correct the data point and send it back. InForm queries are visible in Oracle DMW but must be resolved in InForm.
You can write validation check programs to examine data, including checking lab ranges, checking CRF data against lab data, and comparing multiple data points across CRFs, and create discrepancies against data points that are, or may be, faulty. You can write a validation check so that it creates either Candidate- or Open-state discrepancies. You can set up validation checks to automatically close discrepancies they created when the underlying data point is corrected, or require manual review before closing.
Users can examine data in the Listings page, write ad hoc queries to identify faulty data, and create discrepancies manually.
The following sections describe Oracle DMW features used for data review and cleaning:
The Listings pages display records in a selected clinical data model:
The Default Listings page displays all data in the model, table by table, consistent with the privileges of the user.
The VC (Validation Check) Listings page displays records containing data identified as discrepant with related data as specified by a validation check.
The Custom Listings page displays records retrieved by queries developed by data managers using the Query Builder. These queries can be saved and made available to all users with the required access.
You set up security so that, for example, data managers have access to the data in the input and Review models and Biostatisticians have access to the data in the Analysis model. Access can be restricted by table within the model, and special privileges are required to see blinded data. These restrictions apply to all three Listings pages.
Data Manager In any of the Listings pages, the data manager can:
Write custom queries to identify faulty data.
Review data identified as faulty by validation checks.
Filter data by flag, category, or value such as subject or subject visit.
Open discrepancies against or apply flags to a single data point or to many at once—for example, to all data identified by a custom query or validation check.
View InForm queries against data points and InForm status flags.
View the data lineage of any data point back to its source system and forward to subsequent clinical data models.
With the required privileges, perform an audited blind break to see real, sensitive data that is otherwise hidden.
Biostatistician A biostatistician can, for example:
Filter data by flags applied by special validation checks created for the purpose of tracking the completeness and cleanliness of data in individual CRFs, labeling those that are ready for analysis.
Write custom queries to check for a given condition and mark the resulting data with a flag.
View and filter data at any time, without waiting for an official hand-off. Data loads can trigger transformation to the Review model, which can trigger transformation to the Analysis model so that statisticians always have the most current data possible.
With the required privileges, perform an audited blind break to see real, sensitive data that is otherwise masked.
View data graphically and interactively in a visualization tool; see "Integration with Data Visualization Tools".
The Discrepancies page displays only data that has been identified as discrepant within a selected clinical data model, including:
Discrepancies raised by validation checks
Discrepancies raised manually
InForm queries imported as discrepancies. These must be resolved in InForm, but Oracle DMW users can send messages to InForm users about InForm queries.
Data Manager and Medical Monitor can both act on discrepancies:
Filter discrepancies by subject, visit, table, item, state, tag, data source, or categories you define.
View the full record that contains the discrepant data point.
View the history of the discrepancy.
Apply an action to change the status of a discrepancy. For example, a data manager might change the status from Candidate to Open and requiring medical review.
The medical reviewer can filter for discrepancies requiring medical review, view the data in context, and change the status to Canceled or Closed, or send a message to InForm or a lab requesting action there.
Send discrepancies on InForm data to InForm.
Export discrepancies on lab data to a spreadsheet to be sent to the lab.
You can create any number of flags, each with any number of states, for use in tracking data review. These flags can be applied by data managers or other users, or by validation checks.
In addition, InForm Subject and Subject Visit status flags are imported and viewable in Oracle DMW.
Resolving a discrepancy can involve multiple users and groups. Oracle DMW comes with a set of discrepancy states, allowed transitions from one state to another, and predefined actions available for users to move a discrepancy from one state to another. These actions constitute a workflow among users that you can use out of the box. However, if you need a more complex, fine grained workflow to satisfy your standard operating procedures, you can define custom actions that apply custom tags that serve as substates, effectively creating a custom workflow.
Oracle DMW provides a bidirectional exchange of study data, including InForm queries and Oracle DMW discrepancies, between InForm and Oracle DMW. Loading data from InForm can trigger all downstream data processing, including validation checks and transformations of new and updated data from one data model to the next.
In each study, the system automatically creates the InForm source clinical data model based on metadata from InForm
The system loads data from InForm according to a configurable schedule
Data loads can trigger running transformation programs to propagate data updates to downstream clinical data models
The system stores the unique InForm context of each data point and recognizes data changes made in InForm.
InForm queries raised in InForm must be resolved in InForm, but can be viewed in Oracle DMW. Oracle DMW users can comment on InForm queries and InForm users can reply. Oracle DMW users and validation checks can raise discrepancies against InForm data, and users can send these discrepancies to InForm as queries.
Oracle DMW can maintain multiple InForm studies in a single database.
Oracle DMW supports partially automated exchange of files with any number of labs. The File Watcher feature checks for files from labs in a specified location and automatically processes them upon detection, loading data into the clinical data model created for this purpose.
Loading data from a lab can trigger all downstream data processing, including validation checks and transformations of the new and updated data from one clinical data model to the next.
Discrepancies can be exported to a file and sent to a contact person at the lab.
The system stores the unique external context of each data point and recognizes data changes made at the lab.
The same features could support integration with any system providing data in SAS or text files.
You can integrate SAS (purchased separately) in order to write custom programs in SAS and execute them, and to create SAS CPort, XPort, and dataset data marts to export data from Oracle DMW.
You can integrate Informatica (purchased separately) in order to write custom transformation programs in Informatica and execute them.
You can use external data visualization tools to allow users to view clinical data graphically and interactively with the protection of Oracle DMW security and blinding access privileges. Oracle Business Intelligence Enterprise Edition is included for this purpose, and third party tools that already have an adapter to Oracle Life Sciences Data Hub (Oracle LSH) are available. You must create a Business Area object in Oracle LSH to take advantage of this functionality.
You can integrate other third-party tools using the Oracle LSH generic visualization adapter.
Oracle DMW is installed on top of Oracle Life Sciences Data Hub (Oracle LSH), which provides its underlying internal data model, execution engine, validation lifecycle, and security system. The Oracle DMW database is an Oracle LSH database. Oracle DMW studies are Oracle LSH domains, and clinical data models are a new Oracle LSH object type, as are codelists, transformation mappings, and more.
You can use Oracle LSH features including:
Data marts to export SAS datasets, CPort or XPort files, text files or Oracle Export files based on the tables in a clinical data model.
Business areas to view and report on data in a data visualization tool.
Snapshot labels; see "Applying Snapshot Labels"
See the first chapter of the Oracle Life Sciences Data Hub Implementation Guide for an overview. See "Object Ownership" for more information on the relationship between Oracle LSH and Oracle DMW objects.
|
 Copyright © 2013, Oracle and/or its affiliates. All rights reserved. Legal Notices |
|
