In the full sample, the overall score for the 27 symptoms was 9.03 ± 10.03.
Years since HIV diagnosis, n (%), M (Q25, Q75)įigure 1 shows the severity of symptoms in the full sample and the 5 subgroups (based on the duration since the HIV diagnosis). Covariates, including age (continuous), gender (male = 1, female = 2), ethnicity (Han = 1, minority = 2), education level (middle school or below = 1, high school or above = 2), employment (employed = 1, otherwise = 2), marital status (married = 1, otherwise = 2), primary caregiver (myself = 1, otherwise = 2), having ART (yes = 1, no = 2), years of having ART (continuous), having comorbidities (yes = 1, no = 2), lgCD4 (continuous), that were significantly different among the 5 subgroups according to ANOVA were included in the network analysis to identify the real relationships among the 27 symptoms after controlling for confounding factors. In this algorithm, force vectors determined the nodes, with the strongest correlations placed in the center of the network, and nodes with similar characteristics were placed relatively more closely. We used the Fruchterman-Reingold algorithm and spring layout to generate undirected association networks. The thickness of the edges represented the magnitude of the relationships. Edges represented the conditional independent relationships between 2 nodes. In the symptom networks, nodes represented symptoms. Spearman correlations were used to estimate the relationships between pairs of symptoms. The Qgraph module was used to perform the network analysis.