For the multivariate analyses, generalised linear regression models will be used: Poisson regression for outcome variables in the form of scores, multinomial logistic regression for non-dichotomous categorical variables, or logistic regression for dichotomous variables, as appropriate.
The practical implementation of the approach uses an existing software DMU, and therefore the approach can be easily extended to other types of models implemented by that software, in particular multivariate analysis and generalised linear mixed models.
Multivariate analyses utilised generalised linear modelling (GLM) to assess the main effects of gender, high school of origin, region of origin or special entry pathway status entered as predictive factors, ICSEA score entered as a predictive covariate and TER, TES, UMAT or interview scores (or their component parts) entered as dependent variables.
Adjusted for covariates selected from the univariate analysis, the multivariate Poisson regression (generalised linear model) analysis reveals that home ownership is associated with the total number of readmissions over 365 days: patients owning and occupying their homes had fewer readmissions (B -0.46, 95% CI -0.86 to -0.06, p = 0.024).
MaAsLin tests a sparse multivariate generalized linear model against each feature independently.
Data analysis will primarily rely on bivariate and multivariate regression analysis using generalised linear regression models adapted to specific characteristics of each outcome/exposure.
The significant variables were then included in the multivariate analysis using a multiple generalised linear model.
Concerning the primary analysis, the univariate analysis will be completed in the second step by multivariate analysis performed using a generalised linear model.
Multivariate analyses, for example, a generalised linear model with logit link, will be performed to evaluate variables independently associated with chronic cough at day 28(±3).
