2.3 Statistical analysis plan

Statistical analyses were conducted using SPSS 20 (Statistical Package for Social Science, Chicago, Illinois) and JAMOVI [34]. Firstly, a reliability analysis was carried out on the items included in the Legitimacy questionnaire.

Then, to study medical decision-making patterns in relation to the experimental conditions, we used a generalised linear mixed model. Condition (Normality, Emergency, COVID-19) was modelled as a fixed factor. The agreement scores were modelled as the target variable, for which a multinomial distribution with a cumulative logit function was adopted. A random intercept modelled on the subjects was included.

Thirdly, we investigated the extent to which interoception modulates the perceived legitimacy of non-evidence-based treatments in the case of a COVID-19 scenario compared to a Normal scenario. To this aim, for each subject, we calculated a score for each of the Legitimacy questionnaire condition by averaging responses at items 1,4,7,10,13 for the Normality scenario, items 2,5,8,11,14 for the Emergency scenario and items 3,6,9,12,15 for the COVID-19 scenario. Then, we generated a legitimacy index, calculated as the difference between the COVID-19 and Normality scenario scores. Higher legitimacy index values indicate higher perceived legitimacy in the case of the COVID-19 compared to the Normality scenario. To investigate the impact of interoception on this index, we ran a linear regression analysis with the eight MAIA subscales as predictors and legitimacy index scores as the dependent variable. Participants’ age and gender were included as covariates.

Lastly, we explored the potential effects of other psychological factors on the legitimacy index. To this aim, we ran a Pearson’s partial correlation analysis between the legitimacy index and the (i) number of COVID-19 patients treated, (ii) perceived severity and (iii) preoccupation with the COVID-19 outbreak, (iv) anxiety level, controlling for the participants’ age and gender.