(P20-035-24) Variables Related to the Behavior of Body Mass Index Through a Multiple Linear Regression Model

Objectives: To analyze the association of gender, age, muscle mass, and fat percentage with the body mass index (BMI) in the  population from the University Hospital of Guayaquil using a multiple linear regression analysis.

Methods: Descriptive statistics such as the mean, standard deviation, and the box were used to describe the variables that were entered into the model. Multiple linear regression analysis was used. The statistical program was IBM SPSS version 26. The most relevant variables were chosen. Then, using a correlation matrix, the absence of multicollinearity and the maximum values of the correlation coefficient were determined to ensure the assumptions of the multiple linear regression model, two work methods were considered: intro and stepwise.

Once the final model was obtained, a check was carried out to verify the effectiveness of the model.

Results: A linear regression model was evaluated to predict the behavior of the body mass index (BMI) from four variables, the percentage of body fat, the amount of muscle mass, age, and gender, based on data from 748 patients. Of which 487 (65,1%) were women and 261 (34,9%) were men. The BMI had a mean value of 29,35 ± 5,11, the mean fat percentage was 24,86 ± 8,44%, the mean muscle mass was 56,41 ± 10,2 kg and the mean age was 41,77±10,04 years.

A multiple correlation analysis was carried out to detect collinearity in the variables, the analysis showed a value of R = 0,752.

The variables in the final model were the percentage of body fat and gender. The R2 was 0,887. The analysis of the variance was F=2927,104 (p< ,001).

Finally, the LRM was BMI=1,651+%fat×,706+gender×7,522 (p< ,0001). The results show that a change in fat percentage slightly affects the increase in BMI.

The effectiveness of this equation was verified by the difference in the real value of the BMI minus the value obtained by the model. The mean of the differences in these values was 0,0011±1,7176. On the other hand, the extreme values of these differences were -5,26 and 4,95. These values indicate that the model responds adequately and only presents five extreme cases.

Conclusions: The gender and the fat percentage were significantly associated with an increase in BMI in the adult population studied. The fat percentage can be modified with preventive strategies.

Funding Sources: Not applicable