Despite their potential to reduce cancer burden through early detection, the effectiveness of screening programs is often undermined by systems-based factors and human factors. The broad aims of this application are to develop and evaluate a computer-based technology called TRACS to monitor the process of cancer screening in individual patients and intervene as necessary to ensure that the screening is conducted reliably and effectively. Interventions in TRACS consist of reliable notifications to clinicians and/or patients. However, unlike many other reminder systems, TRACS continues to monitor the care of the patient subsequent to the notification. ? ? In this phase of the application, TRACS will be used to address the reliability of follow up of abnormal or equivocal results from screening mammography for breast cancer. A knowledge module will be created in TRACS knowledge base for the monitoring of the follow up of abnormal screening mammograms. A knowledge module comprises the process definition that specifies the appropriate course of actions for a clinical scenario and logical rules for making decisions within the process. In the next phase, additional knowledge modules will be included for comprehensive support for screening of breast cancer, cervical cancer, and colon cancer. ? ? An important feature of TRACS will be the ability to extract information from clinical text documents using a natural language processing engine called HITex. Reports of several cancer screening tests, such as mammography, cervical smear, and colonoscopy, are in the form of text documents. Thus it is important to be able to extract relevant information from these documents for use by TRACS. HITex will be customized to extract the recommendations of the radiologist from screening mammogram reports. Specifically, it will extract whether the radiologist recommends a non-routine follow up. If that is the case, HITex will extract the recommended follow up (e.g., diagnostic mammogram, ultrasound, or needle biopsy). ? ? TRACS will be deployed and evaluated at the New England Medical Center (NEMC), a large urban medical center in Boston. The first step in the evaluation will be to confirm that information from screening mammogram reports from NEMC can be extracted accurately by HITex. During this phase, the TRACS software and the abnormal mammography follow up knowledge module will be tested in a simulated setting at NEMC. The tests will be conducted using records of patients who had previously undergone screening mammography at NEMC. The behavior of TRACS for these cases will be rated by a panel of experts from NEMC. In the next phase, TRACS will be deployed operationally at NEMC. Project Narrative The software proposed to being developed under this project aims to improve the early identification of a number of commonly occurring cancers. It will do so by following up reliably on the management of those patients that have abnormal results in cancer screening tests. ? ? ?
| Shenvi, Edna C; Meeker, Daniella; Boxwala, Aziz A (2015) Understanding data requirements of retrospective studies. Int J Med Inform 84:76-84 |
