153 Use of a clinical data warehouse to describe enhanced epidemiology of Clostridium difficile-a validation study

Saturday, April 2, 2011
Trinity Ballroom (Hilton Anatole)
Al-Muntaser Abu Ali, MD , University of New Mexico School of Medicine, Albuquerque, NM
Susan M. Kellie, MD, MPH , University of New Mexico School of Medicine, Albuquerque, NM
Background:  National goals for reduction of Clostridium difficile infection (CDI) call for a 30% decrease in incidence by 2013. Streamlined surveillance would allow more resources to be directed to prevention.

Objective:  We sought to demonstrate the utility and accuracy of a laboratory data warehouse in surveillance of CDI compared to an existing infection control database.  We hypothesized that the data warehouse could enhance data collection by applying National Healthcare Safety Network (NSHN) definitions and identifying unique patients with multiple relapses and recurrences.

Methods: The University of New Mexico Health Sciences Center is a academic tertiary care medical center with a 524-bed hospital and outpatient clinic network in Albuquerque, New Mexico.  A  C. difficile toxin A and B assay is used. The data warehouse is a repository for treatment, payment, and operations data at TriCore Reference Laboratories, our laboratory affiliate.  The data extraction from the warehouse included all toxin assays ordered, patient locations, admission, testing and discharge dates, and results from 6/1/2009-5/31/2010. Patients were de-identified, and all dates were shifted to further mask identities. Infection control provided a de-identified data set from the same time period. UNM IRB approved the study. NHSN Multidrug-resistant Organism LabID definitions were used for the following: CDI episode, duplicate test, Hospital Onset –Health Care Facility -Associated (HO-HCFA), Community Onset- Health Care Facility -Associated (CO-HCFA), Community Onset-Community Associated (CO-CA), indeterminate and recurrent cases. We defined a relapse as a positive test occurring more than 8 weeks after the last positive.  The infection control database uses paper reports, and only hospital-onset or community-onset definitions are recorded. 48 hours of hospitalization was used to define cases until January 1, 2010, when > 3 days of hospitalization was used.

Results: A total of 4536 tests were ordered, of which 1625 were cancelled by physicians or by the laboratory when no specimen was received. Of the remaining 2911 tests, 272 were positive, representing 174 unique patients.  Of the positive tests, 70 were duplicates, yielding 202 episodes of CDI. The warehouse was able to classify 89 episodes as CO-CA, 6 as CO-HCFA, and 107 as HO-HCFA. 15 patients had one or more recurrences (a total of 19 recurrent episodes), and 9 relapsed 8 weeks or more after their last CDI episode. Hospital infection control identified 192 episodes of CDI, with 102 defined as community-onset and 84 as hospital-onset (Table 1).


Conclusions: A data warehouse was comparable to daily data entry by infection control staff in enumerating CDI episodes. The data extract used NHSN definitions and tracked unique patients with recurrences and relapses, offering an enriched epidemiology of CDI and the possibility of easy data transfer to NHSN.