Validation of the University HealthSystem Consortium administrative dataset: concordance and discordance with patient-level institutional data.

BACKGROUND: The University HealthSystem Consortium Clinical Database-Resource Manager (UHC CD-RM) is an administrative database increasingly queried for both research and administrative purposes, but it has not been comprehensively validated. To address this knowledge gap, we compared the UHC CD-RM with an institutional dataset to determine its validity and accuracy.
MATERIALS AND METHODS: Age, gender, and date of operation were used to identify patients undergoing pancreaticoduodenectomy from 2009-2011 in both the UHC CD-RM and our institutional pancreatic surgery database. Patient- and intervention-specific variables including perioperative mortality, complications, length of stay, discharge disposition, and readmission were compared between datasets.
RESULTS: A total of 107 UHC CD-RM and 105 institutional patients met inclusion criteria. In both datasets 103 matched cases were present. Between the 103 matched cases, there was concordance with respect to median age (P = 0.87), gender (P = 0.89), race (P = 0.84), overall length of stay (P = 0.46), discharge disposition (P = 0.95), 30-d readmission rate (P = 0.87), and 30-d mortality (P = 0.70). Most comorbidities and complications were captured; however, several disease-specific complications were absent within the UHC CD-RM.
CONCLUSIONS: Most of the clinically significant patient- and intervention-specific variables within the UHC CD-RM are reliably reported. With recognition of its limitations, the UHC CD-RM is a reliable surrogate for institutional medical records and should be considered a valuable research tool for health service researchers.