Marjon Borgert

161 Implementation of the transfusion bundle Study of the intervention Data was collected prospectively from the electronic registration system (Patients Data Management System, PDMS). The occurrence of a RBC transfusion was audited by the researcher three times daily in the PDMS during week days. Transfusions that occurred during the weekends were audited on Mondays. Bundle checklists were used to track compliance (see Supplementary File). Measures Compliance with the completion of each element of the bundle was assessed during the eight study months. Each administered unit of RBC was counted as one inclusion. Compliance was calculated by using the all-or-none (AON)-approach. 24 If one of the interventions was not performed, the nurse was considered as non-compliant. Moreover, if checklists were not found, nurses were considered as non-compliant. The denominator is the total number of RBC units administered per month. The numerator is the total number of applied transfusion bundles per month. Bundle checklists were available in prominent places in the ICU. These places were equal in both teams. Bundle checklists were collected daily by the researcher during weekdays or on Mondays after weekends. Compliance data was entered in a database by the researcher. Compliance levels were calculated at the end of each month per nursing team. Analysis Continuous normallydistributedvariableswill be expressedby theirmeans and standard deviations or when not normally distributed as medians and their interquartile ranges. Categorical variables will be expressed as n/N (%). To test groups Student’s t-test will be used, if continuous data is not normally distributed the Mann-Whitney U-test will be used. Categorical variables will be compared with the Chi-square test or Fisher’s exact test.Thegoal of theprimary analysiswas toquantify thenet e ect of theA&F intervention on transfusion bundle compliance, controlling for other variables. Exploration of interaction (e ect modi cation) and confounding was considered methodologically relevant. We rst focussed on the crude (uncorrected) e ect of A&F (independent variable) on transfusion bundle compliance (dependent variable). Then statistical and clinically relevant covariates were added as an interaction term (implementation and post-implementation period, nurses’ characteristics: age, gender and work experience, and patient characteristics: Apache IV, ICU mortality). If the interaction term appeared to be signi cant ( P < 0.05), this would indicate that the relation between A&F and transfusion bundle compliance could be di erent for various levels of the covariate. This indicates the need for separate models for the levels of the covariate. As a signi cant