Marjon Borgert

28 Chapter 2 Data collection Two types of data were collected during a study period of three months: 1) all vital signs were recorded during a seven-day period at the end of each study month; 2) all AEs (i.e. cardiopulmonary arrests and unplanned ICU admissions) and the RRT activations were recorded during the whole three study months. MEWS was recorded on paper-based charts supported by a owchart for RRT activation (Fig. 1). Measurements were excluded from data analysis when: 1) taken on non-participating wards e.g. delivery rooms; 2) taken during palliative care; 3) taken on days while the patient was signi cantly absent from the ward e.g. due to surgery; 4) prede ned alterations on four protocolized wards were present (Supplementary File 2) which de ned alternate frequency of measurement of MEWS; 5) deviations from the MEWS threshold (or sub scores) de ned by the primary physician were recorded (Supplementary File 2). The presence of delay in notifying the physician was determined by measuring the time between the rst critical MEWS (nurse documented and/or retrospectively determined) and the noti cation of the physician (Fig. 2). Dates and times at which the physician was noti ed including the critical MEWS were used and all subsequent rst occurrences for these parameters were located. Patients were excluded when thresholds were uncertain (speci c vital signs and/or MEWS) or if the physician raised the threshold for calling, e.g. MEWS of 5 instead of 3. Data analysis and statistics We applied an intention-to-treat analysis. Consequently, patients who were transferred to a di erent ward were analyzed in the original study arm (n=21). Continuous variables that were normally distributed were expressed as means with standard deviations and not normally distributed variables as medians and inter-quartile ranges (IQR). To test two independent groups of not normally distributed continuous variables, the Mann-Whitney U test was used. Categorical variables were expressed as percentages, numerators and denominators and were compared with the Chi-square test or Fisher’s exact test or when appropriate as Relative Risk. Statistical uncertainty was expressed by 95% con dence intervals as appropriate, and statistical signi cance was de ned at < 0.05. All data were entered into a Microsoft Access database and the analyses were performed using SPSS version 19.0 (Chicago, Illinois, USA) and con dence interval analysis software version 2.2.0 (University of Southampton, UK). Ethics This study conforms to the provision of the Declaration of Helsinki in 1975 (revised in 2008). 26 Given the observational nature of the study the hospital medical ethics committee waived the need for informed consent.