Session number: 1038
Abstract title: Merging three distinct VHA Databases (NPCD, PBM and VISTA-LAB): Using clinical data for research
Author(s):
GJ Joseph - SCI QUERI/VIReC, Hines VA Hospital, Hines, IL
FM Weaver - SCI QUERI, Hines VA Hospital, Hines, IL
C Evans - SCI QUERI, Hines VA Hospital, Hines, IL
J Parada - Midwest Center for Health Services and Policy Research, Hines VA Hospital, Hines, IL
N Mosk - VIReC, Hines VA Hospital, Hines, IL
S Sabarwal - Milwaukee VA SCI Center, VA Medical Center, Milwaukee, WI
B Neumchawsky - Hines VA SCI Center, Hines VA Hospital, Hines, IL
Objectives: 1) To study the feasibility and validity of merging clinically relevant data from three VHA sources. 2) To explore antibiotic therapy use for urinary tract infections (UTI) for persons with Spinal Cord Injury (SCI).
Methods: Data from the microbiology (micro) component of the VISTA lab package was also obtained. The VISTA data extractions required several steps including constructing an output ASCII files with additional variables created to sort and identify the data elements. Data from these three sources were then manipulated to create a single SAS dataset.
Results: 328 patients with SCI had a least one UTI diagnosis. The PTF data merged with the VISTA pharmacy and micro-lab revealed that only eight patients (6.83 %) had a urine micro test within 48 hours of admission. One patient started an antibiotic regimen before a urine micro was ordered. For outpatients, we merged OPC, VA PBM and VISTA micro data. Five patients (6.57 %) had urine micro tests done within 48 hours of a visit. Two patients received their antibiotics before a urine micro was ordered.
Conclusions: By merging these data, we able to obtain clinically relevant insights about the timing of key diagnostic and therapeutic measures for this cohort. Using similar methods to merge other laboratory data (chemistries, hematology, additional micro studies, etc.) a more complete and sequential understanding of patient management, care patterns and identification of potential medical errors is possible. The results show that combining clinical data with administrative datasets is feasible and yields results which were previously not obtainable.
Impact statement: This effort overcomes an important critique of administrative databases for health services research: lack of clinical data. It provides empirical evidence that clinical data from the VA systems can be merged with corresponding administrative and pharmacy databases. Researchers can henceforth incorporate clinically relevant variables and results into their research methodology for retrospective data analysis.