Derivation and validation of a model for predicting waist circumference

Abstract

DERIVATION AND VALIDATION OF A MODEL FOR PREDICTING WAIST CIRCUMFERENCE Objective: To analyze the relationship between waist circumference and the age, gender, height, and weight of patients, and to design a model to predict waist circumference using these variables. Type of study: Observational Setting: Eça Family Health Unit (FHU), Cuidados de Saúde Integrados Seixal and Quinta da Lomba. Health Care Center (HCC) Population: Patients aged 20 years or older attending medical consultations in these units. Methods: A convenience sample of patients was drawn from patients who attended medical consultations at the FHU in May and July 2009, and the HCC between February and April 2010. Data collected from patients included age, height, weight, gender, and waist circumference. The population at the FHU was randomly divided into two groups. Data obtained from two thirds of the patients were used for developing the model and data from the remaining one third were used to determine the internal validity of the model. The model design was based on a multiple linear regression model, using the stepwise method, with a 5% entry probability and a 10% removal probability. The sample of patients from the HCC was used for external validation of the model. Results and Conclusions: The FHU sample included 493 patients, of these 55,6% were from USF Eça, and 64,3% were female. The average age of patients in the sample was 57,7 ± 0,8 years, the mean height was 162 ± 0,4 cm, the mean BMI was 27,3 ± 0,2 kg/m2, and the mean waist circumference was 93,8 ± 0,5 cm. Comparison of the modeling and internal validation groups revealed no significant differences between them. A model for the prediction of waist circumference was designed. All predictor variables were significantly associated with the outcome variable with an R2 = 0,772. The model showed an internal validity of 92,7%. Applying the model to the external validation group, consisting of 240 patients, showed that the model predicted the waist circumference of 93,8% of these patients within the 95% confidence interval predicted by the model. In conclusion height, weight, age and gender can be used to accurately estimate waist circumference.