Identification of adverse events related to central venous catheters using two semiautomated methods to review free text computerized medical records

Methods for surveillance of adverse events in clinical settings are limited by cost, technology, and appropriate data availability. Manual chart review is the gold standard methodology against which any new technology must be compared at this time. In this study, two methods for semiautomated review of text records within the Veterans Administration database are utilized to identify adverse events related to the placement of central venous catheters (CVCs) and compared to a sample of manually reviewed records. A phrase-matching algorithm was found to be a sensitive but relatively nonspecific method, whereas a Medical Language Processing program (MedLEE from Columbia University) was significantly more specific but less sensitive. Positive predictive values for each method estimated the CVC-associated adverse event rate at this institution to be 6.4 and 6.1%, respectively. Using both methods together results in acceptable sensitivity and specificity (72.0% and 80.1%, respectively). All methods including manual chart review are clearly limited by incomplete or inaccurate clinician documentation. The automated methods are additionally more likely to be negatively affected by spelling errors, abbreviations and use of proprietary brand names to identify specific catheters. A secondary finding related to administrative data (ICD-9 and CPT codes) used to identify intensive care unit patients in whom a CVC was placed. An estimate of the completeness of this administrative data was made by determining a minimum number of patients in which CVCs must have been placed for vascular and cardiac surgical cases and comparing this to those found by searching administrative data. Administrative data identified only 11% of patients who certainly must have had a CVC placed. This suggests that other methods, including automated methods such as phrase matching, may be more sensitive than administrative data in patients with devices.