Marga Hoogendoorn

77 Statistical analysis We used linear mixed models to analyze the association between the Nursing Activities Score per nurse and NASA-Task Load Index, using the perceived workload with NASA- Task Load Index as the outcome measure. To correct for clustering within type of hospital we included type of hospital (academic or teaching versus non-teaching) as a random intercept in our models. We identified confounding factors for the association between Nursing Activities Score per nurse and NASA-Task Load Index based on expert opinion of a nursing scientist of the Intensive Care, an intensivist, and a data scientist. We used association frameworks to identify variables in the same causal pathway and variables not in the same causal pathway (appendix 5). We analyzed each possible interaction between the different factors per model (see appendix 5 Figure 1.1 till Figure 1.10). Next, we analyzed the association between the different patient-, nurse-, and contextual factors and the NASA-Task Load Index. In every model we adjusted for possible confounding factors determined in the association frameworks. Patient-, nurse-, and contextual factors were considered statistically significant when they had a p-value of < 0.05 after adjustment for possible confounders. All statistical analyses were performed using R version 3.3.3. We used STROBE 27 as a reference to report on this study. Ethical approval The Institutional Research Board of the Amsterdam University Medical Centre reviewed the research proposal and waived the need for informed consent (IRB protocol W17_366). RESULTS We included NASA-Task Load Index data from 228 nurses in 226 different shifts of 8 different hospitals. During these shifts, nurses took care of 389 patients. From the 389 patients we had to exclude 8 patients due to missing Nursing Activities Score -data. Table 1 shows characteristics of the hospitals, as well as patient-, nurse-, and contextual factors. Due to the low number of nurses with 3 or 4 patients per nurse (N = 6) we categorized the results of this contextual factor in two different categories: 1 patient per nurse or > 1 patient per nurse. Due to the low number of nurses with < 2 years’ experience (N = 9) we categorized the results of this nursing factor in two categories: student nurse or certified Intensive Care nurse.