基于ROC分析的基本形态指标对青年人"肥胖"的判断效果

Objective: To use ROC analysis to explore the judgment value of height, weight and their regression predictive values and derivative indicators for obesity. Methods: Taking the body fat rate as the "gold standard", using BMI, NBMI, PPBF, Quetelet index, Laurel index, and Levy index to judge "obesity" respectively. Results: (1) When the natural logarithm of height (lnH) is used to predict the natural logarithm of weight (lnW), the regression equation and regression coefficient are both statistically significant (P＜0.05). (2) When using height and weight to predict body fat rate, the regression equation and regression coefficient are both statistically significant (P＜0.05). (3) When judging male obesity by various indicators, the AUC of NBMI, Laurel index, and Levy index were significantly higher than the AUC of BMI (P＜0.05), and the AUC of Quetelet index was significantly lower than that of BMI (P＜0.01) There was no statistically significant difference between the AUC of PPBF and the AUC of BMI (P＞0.05). When judging female obesity, the AUC of Quetelet index was significantly lower than that of BMI (P＜0.01). Conclusion: (1) The constructed "New Body Mass Index" (NBMI) has a good effect in judging obesity in men and women, with AUC of 0.916 and 0.894. NBMI male=W/H2.7；NBMI female=W/H2.3. (2) The regression equations for predicting body fat percentage by height and weight are 〖PPBF〗_male=91.473-0.593H+0.420W，and 〖PPBF〗_female=97.679-0.606H+0.513W. When PPBF judges obesity in men and women, AUC is greater than 0.88. (3) The effect of NBMI, Laurel index, and Levy index in judging male obesity exceeds BMI, and AUC is greater than 0.9. The effects of NBMI, PPBF, Laurel index, and Levy index in judging female obesity are basically the same as BMI, and the AUC is greater than 0.88. The Quetelet index has the worst effect in judging obesity in men and women, with AUC of 0.860 and 0.873, respectively.