2.3 Data Analysis
Responses were extracted from the LimeSuervey® application to an Excel® spreadsheet for analysis. Data was cleaned and records were deleted if the respondent did not complete the questionnaire. Responses of the Likert scale were coded from 1 to 5 (1 being the lowest degree and 5 the highest). These were then described in terms of frequency grouped by thematic categories. Univariate Poisson regressions were used to determine any differences among the measured percentages of the coded form (the Likert scale) within subgroup of both answers and questions. The explanatory variable was the group that was being analysed and was inserted as a categorical variable. Wilcoxon rank sum test was used to compare if two groups were equal or not. The Kruskal-Wallis equality-of-populations rank test was used to evaluate the differences of the median. Fisher’s exact test was used to measure association. Significant p-value cut-off was set to 0.05.
Open-ended questions were sorted manually by theme or topic and summarized in an interpretative way. All statistical analyses were conducted using Microsoft Excel® and STATA S.E. 14.2® software (College Station, Texas, USA).