OBJECTIVES: Factor analysis or structural equation models (SEM) of health-related quality of life (HRQL) data generally assume a reflexive measurement model. Soft-modelling techniques like SEM relying on Partial Least Squares (PLS) offer interesting alternatives. A SEM-PLS model was used to analyse the relation between anxiety/depression, harm avoidance and quality of life among HIV patients in an online survey. METHODS: Data were collected on 517 HIV+ patients (70% males, mean age 48 years) using validated self-reported measures. A structural model was posited a priori to link various domains: HRQL (PROOQL-HIV, 3 dimensions—physical, cognitive, and social health, on a 0-100 points scale), personality traits (TCI-56, 2 dimensions—harm avoidance and novelty seeking), and anxiety/depression (HADS, 2 dimensions). Participants were classified into a precarious (52%) and a non-precarious (48%) group. Two-group comparison were performed using two-tailed Student t and Pearson chi-squared tests. The SEM-PLS parameters were computed on the whole sample and regression weights (b) were compared between the two subgroups using bootstrap resampling (B=500). RESULTS: Univariate analysis suggest that average scores were lower (p<0.001) in the precarious group for all dimensions of PROQOL-HIV (-14 up to -20%) and HADS (-10 and -14%) but not for harm avoidance. The PLS-PM analysis indicates that all pre-specified path coefficients were positive and significant at the 5% level. Depression was strongly associated to the cognitive (b=0.325) and the physical (b=0.615) dimensions of PROQOL-HIV while anxiety was strongly related to harm avoidance (b=0.534) as expected. However, no significant differences were observed between the two groups at the level of the structural model using bootstrap resampling despite interesting variations in regression weights between the two subgroups. CONCLUSIONS: Depression is a strong predictor of lower physical, cognitive and social health but anxiety and personality traits are also potential moderators of HRQL, independent of the level of precariousness.