specific efficacy of the WHO multimodal promotion strategy [30] and other hand hygiene promotion strategies [22]. Hand hygiene improvement was highest for the HHi arm while the average increase of hand hygiene in the CVCi‐ and COMBi arms was moderate. Together with “HH study fatigue” in one centre, economic constraints resulting in budget and salary cuts, low staffing levels, and high workload were mentioned to play a role in prioritizing CVC‐intervention over HH intervention in the COMBi arm. Nurses performed better than doctors across all study arms as has been reported by many others [29]. The average baseline CRBSI incidence density of 2.4/1000 CVC‐days as seen in our study has become standard in high‐income countries [31‐33], although the hospital specific incidence density ranged from 0 to 10.2/1000 at baseline. This range may reflect not only real variation in CVC insertion and infection prevention practice but also differences in culturing quality and frequency. To minimize this bias we discussed these issues with the on‐site investigators and study nurses during the kick‐off and observer training that took place before baseline measurements began. Meta‐analyses about the effectiveness of CVC bundle or checklist interventions on central line‐associated BSI (CLABSI) identified significant reductions (odds ratio of 0.34 [19], incidence rate ratio 0.45 [18]) but the lowest baseline incidence densities in both meta‐analyses were higher (3.4 and 5.7/1000 CVC‐days) compared to our study. However, it must be taken into account that CRBSI is a more specific definition than CLABSI [34]. CRBSI decreased already during baseline. This trend may have been the result of external factors, but probably also was a result of the concurrent study, and particularly due to direct observations of CVC insertion and hand hygiene, as shown by increasing CVC insertion scores and HH compliance. However, other aspects of care may have improved as well, e.g. catheter care due to general patient safety awareness. Feedback reports were sent only after the formal start of the intervention, and thus, are barely responsible for the observed ‘surveillance effect’ [35‐37]. Alternatively, the improvement of baseline CRBSI rates and CVC insertion scores may also represent a secular trend as reported by the English ”Matching Michigan” program, which was due to pre‐existing or ongoing quality improvement initiatives [38]. In contrast, many of our centers had little or no exposure to national quality improvement initiatives and adopted the PROHIBIT project as an opportunity to improve practice [39]. Measuring process indicators allows testing whether the target of an intervention is achieved, provides insight into the implementation process, and allows evaluation of direct association between process parameters and outcome. We report that one percentage point increase in compliance was associated with a two to five percentage point decrease in CRBSI. However, the magnitude of the observed associations must be
RkJQdWJsaXNoZXIy MTk4NDMw