highly significantP < 0.001
Akaike information criterion (AIC) values of segmented regression models for all three factors were significantly lower than those of linear models, and likelihood ratio tests were highly significant ( Supplementary Table 5 , P < 0.001).