40 Chapter 3 Figure 1: Flowchart demonstrating the main five steps of the virtual training course. Statistical analysis Statistical analyses were performed using R statistics version 4.1.0 (2021) and IBM SPSS version 27 (2020). Diagnostic accuracy and interobserver agreement (IOA) for each staging item was evaluated for the seven test cases and compared between the pre- and post-course test. As a standard of reference to calculate accuracy, the seven test cases were scored independently by two of the expert radiologists from the faculty (DL, RD); in case of any disagreement, a third faculty member (ML) was consulted to reach final consensus. IOA was calculated using Krippendorff’s alpha. Mixed effect regression was used to test the course impact on the test-case performance, 2 levels were used for this model (the case number and the course participant) to take into account the multiple measurements per participant and case before and after the course. The responses for the linear mixed model were the absolute difference between the participants estimations and the experts. A mixed-effects logistic model was used for binary outcomes, and for continuous outcomes, a mixed-effects linear model where the response was the participant being in agreement with the experts versus the participant being in disagreement with the experts. To evaluate the impact of the participants’ baseline (i.e. pre-course) experience a model including a covariate representing the level of experience of the reader (<5 years vs ≥ 5 years’ experience) and an interaction term
RkJQdWJsaXNoZXIy MjY0ODMw