External and internal validation of healthcare-associated infection data collected by the Korean National healthcare-associated Infections Surveillance System (KONIS)

Abstract

This study analyzed the validity of healthcare-associated infection (HAI) data of the Korean National healthcare-associated Infections Surveillance System. The validation process consisted of external (EV) and internal (IV) validation phases. Of the 193 hospitals that participated from July 2016 through June 2017, EV was performed for 10 hospitals that were selected based on the HAI rate percentile. The EV team reviewed 295 medical records for 60 HAIs and 235 non-HAI control patients. IV was performed for both the 10 EV hospitals and 11 other participating hospitals that did not report any HAIs. In the EV, the diagnosis of urinary tract infections had a sensitivity of 72.0% and a specificity of 99.3%. The respective sensitivities of bloodstream infection and pneumonia were 63.2% and 70.6%; the respective specificities were 98.8% and 99.6%. The agreement (ĸ) between the EV and IV for 10 hospitals was 0.754 for urinary tract infections and 0.674 for bloodstream infections (P<.001, respectively). Additionally, IV found additional cases among 11 zero-report hospitals. This study demonstrates the need for ongoing validation and continuous training to maintain the accuracy of nationwide surveillance data. IV should be considered a validation method to supplement EV.