Data types in STARR
STARR is growing to support AI use cases and is bringing in more Hospital data sources. It has structured and unstructured, raw and “analysis-ready” data. Following data types are accessible either via self-service or a data consultation service.
Electronic Health Record data:
Epic Clarity: Research IT receives a copy of Epic Clarity Electronic Health Records (EHR) data from the two Stanford Hospitals. Learn more about EHR data in STARR. A subset of the EHR data is available via self-service.
HL7 feeds: Research IT also receives HL7 feeds in real time from the two Hospitals such as Admit Discharge Transfer (ADT), and Billing Account Record (BAR). Learn more about HL7 data in STARR. The HL7 data is only available via a technology consultation service.
Imaging: STARR contains historical pediatric and adult radiology DICOMs, has ability to query/retrieve echochardiograms from both pediatric and adult hospitals, has ability to access pediatric Syngo metadata, and is able to query/retrieve adult retinal fundus imaging. Learn more about imaging data in STARR. To access this data, please request a consultation service.
LPCH Bedside Monitoring Data: LPCH archives its Philips PIC iX bedside monitoring data to STARR. Research IT takes that data and recreates the clinical grade bedside monitoring data in a more interoperable format (i.e., Physionet compatible). Learn more about bedside monitoring data in STARR. A subset of this data is available as self-service.
Other: Research IT has brought in databases such as PowerPath, ARIA Radiation Oncology database, Xcelera etc. We have also brought in large datasets such as EEG. These are brought in as needed by research teams or projects. To access such data types, please request a consultation service.
Linking data modalities in STARR
Data linking in research clinical data warehouses (r-CDWs) across sources and across different modalities use different approaches.
- Patient at two hospitals: The pediatric and adult hospital Epic Clarity data are linked by MRN. For a short period of time, it is possible that one patient presents two different MRNs. These are often resolved in a matter of days.
- DICOM and EHR: These two modalities (DICOMs and Epic Clarity) are linked by accession_id at Stanford. This is essentially an order number e.g., physician orders a MRI for the patient. The accession_id is the mechanism to link DICOM image with the radiology report.
- Bedside monitoring and EHR: Bedside monitoring data is a complex modality to integrate with MRN. STARR has pediatric bedside monitoring data. This data often relies on bed number. A patient often goes from bed to bed, as a result, we need ADT feed (Admit/Discharge/Transfer) to link. For more details, please read the manuscript.
- EHR and LADMF death data in STARR Tools: The LADMF has SSA number, Date of Birth (DoB) and Name. There are multiple ways of matching patients in two different sources. In STARR Tools where LADMF is integrated, if there is no SSN match with Stanford patient's SSN, we do not link. If there is a SSN match, we require that either DoB matches OR (First Name, Last Name AND Year of Birth) matches.