Co‐variables The recorded patient and CVC characteristics were age, sex, type of admission, type of ICU, length‐of‐stay, place of CVC insertion, insertion site, CVC type, number of lumens, indication for CVC insertion, and CVC dwell‐time. Local study teams provided the Simplified Acute Physiology Score (SAPS) II or the Acute Physiology and Chronic Health Evaluation (APACHE) II score at patient admission, where available. Training of study participants Six weeks before the start of the study baseline the PROHIBIT study nurses were trained in the direct observation of CVC insertion and hand hygiene compliance at the University of Geneva Hospitals. Three to six months before the start of the intervention, study nurses and physicians attended a two‐day PROHIBIT workshop for training best practices and implementation science. Parallel to the WHO training material on hand hygiene (http://www.who.int/gpsc/5may/tools/en/), an e‐learning program developed at the University of Geneva Hospitals was adapted to the PROHIBIT protocol and made available publicly at www.carepractice.net. Hospitals were encouraged to adapt the intervention program to their local context. Implementation activities The hospitals adapted CVC insertion procedures and introduced material (e.g. large drapes) as needed and feasible. They promoted strategy elements through educational sessions and bed‐side training, using the WHO material and/or the Carepractice e‐ learning program. Some CVCi hospitals filmed the ‘old and new’ local insertion procedure to be used in educational sessions. All HHi hospitals applied posters and/or other reminders in the workplace and many came up with various additional activities. Detailed intervention activities are described in Supp. Tables 1a and 1b. Statistical methods For sample size calculations, we anticipated a baseline CRBSI incidence density of 3/1000 CVC‐days. We hypothesized that the HHi‐, CVCi‐, and COMBi interventions would reduce CRBSI by 15%, 35% [15], and 50% (alpha=0∙05; approximately 60% power for HHi, and over 80% power for CVCi and COMBi), respectively. χ² and exact tests were used to calculate 95% confidence intervals (CIs) for CRBSI incidence densities and compliance proportions. Differences between medians were tested with the Kruskal‐Wallis test. CVC removal without CRBSI, discharge from the ICU with the CVC in place, and death were modeled as competing events for CRBSI (Supp. Methods and results) [16]. The association of the three interventions (CVCi, HHi and COMBi) with CRBSI incidence
RkJQdWJsaXNoZXIy MTk4NDMw