The mission of the NIAID TB Portals Program is to collect and make available de-identified multi-domain TB patient case data. We utilize a uniform data dictionary with generally accepted medical terminology and data field values. The data we collect comes from a range of sources – clinical trials, research studies, as well as routine collection of atypical patient cases receiving medical care. There is no single identifiable data collection protocol that is uniformly enforced. Therefore, TB Portals data are structured and should be regarded as a natural history study, not an epidemiological study.
Many of the cases in the TB Portals have been contributed by the TB Portals Consortium. These cases are chosen based on the scientific interest of the TB Portals Consortium member, usually with a heavy selection focus on multi- and extensively- drug resistant TB. In these cases, information collection begins at the time of first diagnosis, collecting both prospective and available retrospective data. Imaging is conducted during the course of treatment at regular intervals. Genomic samples from select cases are likewise collected at regular intervals. Standardization of these data is enforced using a central data entry form, and all of the clinical data from TB Portals Consortium cases have been curated and validated as accurate by physicians at the time of data entry. The TB Portals also contain patient cases that have been submitted externally, and Application Program Interfaces (APIs) were created to enforce data standardization. As with TB Portals Consortium contributed data, the case selection follows each study or data contributor’s specific protocol.
All data within the TB Portals follows the HL7 Fast Healthcare Interoperability Resources (HL7 FHIR) standard. A central data dictionary enables users to explore the data and conduct analyses across all data sources using TB Portals data descriptors and fields. While this standardization reconciles different naming conventions, variability in case selection and data collection methods remain a feature of the TB Portals dataset.