Assessment of Hiv-Related Quality of Life in a Representative Sample Of French Patients using Multivariate Multi-Block Statistical Models.

Self-reports were collected during the VESPA2 national survey [1] and included three blocks of data: the 7-dimension PROQOL-HIV questionnaire [2], a 22-symptom checklist [3], and 21 binary indicators about clinical status. Data were analysed using bootstrapped sparse PLS (HRQL and symptoms only), and SGCCA (three blocks). Model hyper parameters (number of components, κ, and L1 penalty, λ) were tuned using 20 x 5-fold cross-validation. Model-based clustering of individual factor scores was used to build a ‘typology’ of the VESPA2 patients. To study the influence of symptom experience and clinical markers on a multidimensional Health-Related Quality of Life (HRQL) questionnaire specific to HIV using sparse PLS regression and Canonical Correlation Analysis (SGCCA).