but not all were complied with). For the partial bundles, the definitions were adapted accordingly. Data selection and statistical analysis We included data from 2009 up to and including 2019. Not all hospitals that joined the national CRBSI surveillance during this period additionally reported bundle compliance during the entire surveillance period. Some hospitals managed to record compliance only for some bundle elements or for part of the CVCs. For the analyses in this study we included only those years of a hospital where bundle recordings were complete for at least 80% of the CVCs. The trend in CRBSI rates and patient and CVC characteristics is additionally described for all hospital data in this period. We used Cox regression to determine the association of bundle compliance as a binary variable with CRBSI risk, while accounting for clustering of data within patients and hospitals, using robust covariance estimation and stratification for hospitals. We performed the analyses for the overall bundle as well as for the insertion and maintenance bundles separately in one model. Sex, age, CVC indication and insertion vein were included as covariates to account for potential confounding. Variables were included in the initial multivariable model if the p‐value was <0.25, but sex and age were retained during manual backward selection. To assess the effect of ongoing participation, results were evaluated both with and without calendar years and the number of years each hospital participated in the surveillance. We additionally analysed the association of bundle compliance with catheter duration, using multilevel analysis with a gamma distribution. Some hospitals performed surveillance in ICUs only, but in 2016 hospital wide surveillance became mandatory when participating. Therefore, the analysis of CRBSI was also performed separately for CVC days in the ICU and ICU‐acquired CRBSI (including CRBSI developed < 48 hours after ICU‐discharge), and for CVC days outside the ICU and non‐ICU‐acquired CRBSI. Peripherally inserted CVCs can be recorded in the CRBSI surveillance from 2014 onwards but were excluded from these analyses. The statistical analyses were performed using SAS Statistical software, version 9.4 (Cary, NC, USA). RESULTS Participation in the surveillance Following the start of the DHPSP the number of hospitals that took part in the national CRBSI surveillance increased from nine in 2009 to 43 (47% of all 91 Dutch hospitals) in 2014 (Figure 1). After the DHPSP ended, participation decreased to 21 hospitals in 2019. In total, 66 hospitals took part in the CRBSI surveillance, for one to 11 years with a median of five years (interquartile range (IQR) 3‐7), for a total of 324 ‘hospital years’). 9 219 The effect of a CRBSI prevention bundle in the Netherlands
RkJQdWJsaXNoZXIy MTk4NDMw