Anne van Dalen

90 I Chapter 3 Questionnaire and statistical analyses All survey data collection and statistical analyses were executed by the authors at our academic medical centre (AvD, and SvD) to adjudicate possible conflicts of interest. The founder and equity holder (TG) of Surgical Safety Technologies (SST) Inc., Toronto, Canada was involved in the co-development and delivery of the structured performance outcome reports, but not in set-up nor outcome analysis of study. Following the TOPPER-trial team debriefing sessions, participants completed a standardized questionnaire surveying user satisfaction regarding the performance report and OR Black Box® as a tool for team debriefing. The original questionnaire is written in Dutch and can be found in the Appendix. As the debriefing was also done in Dutch and the questionnaire was analysed by the Dutch study coordinator (AvD, SvD), it was not translated to English. Exploratory factor analysis of the questionnaire was used to measure the satisfaction of the users. This included a principal-axis factor analysis which was conducted on the 23 items (10-point Likert scale questions) with oblique rotation. The Kaiser– Meyer–Olkin (KMO) and Bartlett’s test was used to verify the sample size adequacy of the completed satisfaction questionnaires. The correlation matrix and anti-image matrix (values < 0.5) were used to decide which questions had to be removed, because these questions correlated too highly (> 0.9) or poorly (< 0.2). The questions clustered in the satisfaction factors were tested for reliability by the Cronbach’s alpha test (> 0.7). 30 Linear regression analysis was used to determine whether independent covariates were significantly correlated with the, in the factor analysis identified, different satisfaction factors. Covariates with a threshold p value of 0.20 were entered in the multivariable linear regression model. Multivariable regression analysis was performed to estimate differences in variables associated with the selected satisfaction factors. The multivariable regression model was created using a backward stepwise fashion. Covariates in the multivariable regression model with a threshold p value of 0.05 were considered to be significantly associated with the outcome variable. The B values with 95% confidence intervals (CI) were presented. All statistical analyses were conducted using SPSS statistics 24.0 for Windows.