SemanticBits

caTRIP is focused on the complex problem of outcomes analysis in which clinicians look across data from a cohort of existing patiens with similar characteristics to identify patients that were treated successfully. The ultimate goal is to identify a treatment that will have a high chance of success for a patient the clinician is currently seeing. Take for example John Doe who enters the clinic and has test A and B come back with results X and Y. A clinician would like to investigate the outcomes and treatments of all other patients that have results X and Y. The problem becomes extremely challenging when the clinician begins looking for data across modalities, such as lab results, pathology results, demographics, diagnoses, treatments, outcomes, and even basic science data.

caTRIP provides an engine for querying across disparate data sources, joining and aggregating the data, and presenting to the user in an intuitive way. The clinician is presented with a simple window in which they can selected data elements as filter criteria and the data elements to return. caTRIP constructs the potentially complex distributed query, deconstructs it into the necessary sub-queries, submits them to federated data sources, joins the results, and aggregates the results before presenting them. The user is alleviated from having to know in which systems different pieces of data exist - caTRIP handles all of that behind the scenes.

While initially focused on outcomes analysis, caTRIP has large number of applications in a clinical and research setting.  These include:

• Helping inform treatment and improve patient care
• Enabling the searching of available tumor tissue
• Locating patients for clinical trials
• Investigating the association between multiple predictors and their corresponding outcomes such as survival

Of specific importance is that caTRIP provides a unified view of potentially federated and disparate data.

caTRIP relies on the vast array of existing caBIG applications, including the Tumor Registry, Cancer Text Information Extraction System (caTIES), caTissue Core , caTissue Cancer Annotation Engine (CAE), and caIntegrator.  However, caTRIP is driven completely by the metadata exposed by the domain data grid services. Queries spanning multiple systems are designed using common data elements (CDEs) displayed in dropdown menus. CDEs were also used to establish links between different services. Some of the key features of include:

• Leveraged metadata for checking the semantic equivalence of classes and attributes at runtime
• Storing domain-model metadata of a data service in a XML file, which is generated using the caDSR grid service
• Using the above XML metadata file to retrieve information about various CDEs that are exposed by a service
• Partnership with the caGrid development team to design the metadata extract schema.

caTRIP was developed in partnership with Duke Comprehensive Cancer Center and piloted using BRCA1/BRCA2 breast studies. It is currently being evaluated by a number of organizations in the cancer research community. The underlying technologies implemented in the caTRIP project have been reused in a number of other open source projects, including caGrid and caB2B.

caTRIP 1.0 was released in Summer 2007.

PreviousNext