taken at insertion were sometimes less when it was already expected that the CVC would be needed for a few days only. Although the subclavian vein was the preferred vein as it was associated with a lower CRBSI risk, the proportion of CVCs inserted here decreased. Insertion via the jugular vein has become popular with the increased availability of ultrasound guidance[15]. We used data from a national surveillance network that facilitated voluntary reporting of compliance with the DHPSP CRBSI intervention bundle, creating an extensive database to study the relationship between the bundle and CRBSI risk. Thus far, studies that investigate the effect of CRBSI bundles outside the ICU are relatively sparse [11‐13]. Our surveillance was not limited to the ICU and comprised non‐ICU wards, which allowed us to determine the bundle effect for these two settings separately. However, as participation in the PREZIES CRBSI surveillance is voluntary, there might have been participation bias. A quarter of the Dutch hospitals did not participate in the CRBSI surveillance during the DHPSP programme. Of the 66 hospitals that did participate in this period, 56 (84.8%) reported (non)‐compliance with the bundle for at least one year. Six hospitals merged to four new centres during the study period. The analysis was also performed using the merger hospitals, with respective longer participation periods, instead of the original sites, with comparable results (not shown). Some hospitals may have been motivated to join the DHPSP programme because they experienced a problem with CRBSI. The high mean CRBSI rate during the first year of the patient safety programme might reflect this. This could result in an overestimation of the observed effect. However, analysis without 2009 led to similar results (data not shown). It is also possible that hospitals that started later, e.g. because they needed more time to prepare the practicalities of the CRBSI and/or bundle compliance registration, initially had higher CRBSI rates as well. It is known that, even in formal study set ups with a prospectively monitored baseline phase, infection rates can decrease before the actual implementation of an intervention. This probably results from increased awareness following the introduction of, and build‐up to, the surveillance/study, a national programme, or political pressure[8, 16]. In our regression analyses we were not able to distinguish a participation effect from a secular trend. However, the effect of each extra participation year was only significant in the first four years, which advocates a true effect of ongoing participation. The cut‐off value of 80% was to some extent arbitrary. Including all hospital years with at least 50% recordings led to a further inclusion of 39 hospital years and comparable results (not shown). Our data were collected in ‘real life’, including many hospitals, thus the results are more generalizable than data from a (limited) study setting. However, a limitation of the 9 229 The effect of a CRBSI prevention bundle in the Netherlands
RkJQdWJsaXNoZXIy MTk4NDMw