A nosocomial infection surveillance network in Belgian ICUs: methodology and feedback

Introduction

The risk for nosocomial infections (NI) in ICUs is five to 10 times higher than in other hospital care units, causing extra morbidity, mortality and costs. Previous studies have shown a significant decrease in NI after the set up of effective surveillance programs. If this surveillance is performed in a multicentric setting, the results from the aggregate database can offer an added value to the local results. The most important element is the production of meaningful feedback reports, allowing each ICU to evaluate its own situation as compared with the aggregate results of the whole studied group.

Objectives

The aim was to provide each ICU with its own incidence of nosocomial pneumonia (PN) and blood stream infection (BSI), and their mortality rate after controlling for the relevant risk factors. These include severity of illness (SAPS II) and exposure to specific devices and risks. Additionally, the feedback reports should also provide the results of the same analyses performed on the national database, obtained by pooling the data of all participating units.

Materials and methods

In December 1995, all Belgian ICUs were invited to join a voluntary prospective NI surveillance network, using a common protocol, which was largely based on the consensus obtained in the HELICS project. Participation was for 3-month periods, starting as of January 1996. During the first two trimesters of 1996, in total 64 different ICUs (from 28% of the Belgian acute care hospitals) have joined the network.

Data included

• (i) Patient characteristics at admission: administrative data, type of admission. SAPS II score, prior surgery or antibiotics, impaired immunity, trauma, and infection at entry, (ii) During ICU stay, exposure to ventilation, central catheters and a number of other devices and treatments were recorded on a daily basis, (iii) If a PN or BSI occurred, diagnostic criteria were registered in detail, including microbiological evidence. Finally, (iv) vital status at discharge was registered.

Results

From January to June 1996, over a total of 77 ICU-trimesters. 8475 patients were observed. The same set of feedback tables was produced for each individual ICU and hospital (by trimester and for both trimesters, if applicable). Besides the mortality and the incidence of PN and BSI by major risk categories, each ICU could also compare its own case mix, length of stay and level of utilization of relevant devices (ventilation. central catheters, antibiotics, feeding practices, etc) with the corresponding figures in the national database.

The fact that some hospitals presented high infection rates after control for the relevant risk factors, without any increase in length of stay nor in death rate, poses challenges to the interpretation of the results.

Before high infection rates can he attributed to deficiencies in the process of care, one must carefully control for differences in case mix, risk factors and length of stay, and make sure that the validity of the results is sufficient and homogeneous.

Even so, the availability of standardised surveillance results should be of great help in any ICU willing to critically evaluate and improve its own performance.