Administrative databases provide inaccurate data for surveillance of long-term central venous catheter-associated infections

BACKGROUND: Efficient methods are needed to monitor infections associated with long-term central venous catheters (CVCs) in both inpatient and outpatient settings. Automated medical records and claims data have been used for surveillance of these infections without evaluation of their accuracy or validity. OBJECTIVE: To determine the feasibility of using electronic records to identify CVC placement and design a system for identifying CVC-associated infections. DESIGN AND SETTING: Retrospective cohort study at an HMO and two teaching hospitals in Boston, one adult (hospital A) and one pediatric (hospital B), between January 1991 and December 1997. Tunneled catheters, totally implanted catheters, and hemodialysis catheters were examined. Claims databases of both the HMO and the hospitals were searched for 10 CPT codes, 2 ICD-9 codes, and internal charge codes indicating CVC Long-term central venous catheters (CVCs) are commonly used for infusion therapy in the inpatient and ambulatory settings.' Use of these devices in ambulatory patients has become even more commonplace as a consequence of decreasing length of hospital stay and the ability to deliver more sophisticated therapies outside Of the hospital. Bloodstream infections are frequent complications of CVC use and are associated with substantial morbidity and resource utilization.(2.7) Thrombosis and mechanical dysfunction are also relatively common. Healthcare systems need surveillance systems to track these complications and to assess the effect of prevention programs. There has been extensive study of surveillance methods for CVC-associated bloodstream infection in the inpatient setting. However, these methods based on chart review and laboratory surveillance are labor intensive and unlikely to perform well in the larger healthcare system. Furthermore, there are limited data insertion. Lists were compared with each other and with medical records for correlation and accuracy. PATIENTS: All members of the HMO who had a CVC inserted at one of the two hospitals during the study period. RESULTS: There was wide variation in the CVC insertions identified in each database. Although ICD-9 codes at each hospital and CPT/ICD-9 combinations at the HMO found similar total numbers of CVCs, there was little overlap between the individuals identified (62% for hospital A with HMO and 4% for hospital B). CONCLUSION: Claims data from different sources do not identify the same CVC insertion procedures. Current administrative databases are not ready to be used for electronic surveillance of CVC-associated complications without extensive modification and validation.