Figure 4.

Distribution of BMI variation for different monitoring status. In (A) the density distribution (y-axis) of BMI variation (x-axis) is represented for HA (gray), MA (yellow), and LA cluster (blue), respectively. BMI values were standardized to µ = 0 and σ² = 1. The corresponding box plot, together with the statistical analysis, is represented in (B). Kruskal–Wallis test, used to compare the three groups, shows a statistically significant (p-value = 0.05) difference in BMI variation among clusters. The result of the post-hoc comparison is then reported, revealing that a high-adherence rate of self-monitoring ensures a significant BMI decrease with respect to low-adherence rate (** stands for p-value = 0.0071).