Validation of the Belgian surveillance data of ICU-acquired nosocomial infections

Introduction

The Belgian national surveillance of nosocomial infections in intensive care units (ICU) was set up in 1996 in collaboration with the Belgian Society of Intensive Care Medicine. The validation followed in 2001. The objective was to evaluate the sensitivity, specificity, exhaustiveness and accuracy of the nosocomial pneumonia (PN) and bacteraemia (BAC) data reported by Belgian ICUs to the National Surveillance program (NSIH).

Methods

The sample size was calculated at 1300 patient charts (904 negative files and 268 infections) to achieve an estimated sensitivity of 65% (± 10%) and specificity of 99.5% (± 0.5%). Forty-five hospitals were selected randomly from a list of quarterly participants between 1997 and 2001. All reported PN and BAC patient charts were selected. False-negative PNs were estimated on reported noninfected patients (approximately 20% of total negatives). False-negative BACs were identified matching a list of ICU admissions with a list of positive blood cultures. The golden standard was the surveyors' opinion. Surveyors, trained NSIH team members, reviewed charts without prior knowledge of their surveillance status. Discrepant charts were double-checked by another surveyor. If still discordant, they were then discussed with the clinician and/or infection control practitioner to determine sources of discordance.

Results

Up to now 33 hospitals have been visited from 2001 to 2003. Preliminary results are presented in Table 1.

Table 1

Global exhaustiveness was 81.2% for patients staying more than 48 hours in the ICU. False-negative pneumonia was more frequently not documented by microbiology. Also, there was a strong correlation between internal quality of the data (missing or discordant values in surveillance database) and external validity.

Conclusion

Specificity was high although sensitivity was quite low. BAC was more accurately reported than PN. The lack of sensitivity might be related to factors such as the perception of the surveillance as an external control instrument, the professional background of the data collectors, and poor knowledge of the protocol (e.g. the case definitions). Validation is crucial for credibility of data, provides an opportunity for personal technical assistance, and should be implemented on a continuous basis in surveillance programmes.