Evaluation of peer effects on eating behaviors: a cluster analysis approach

Objectives: In this study, we aimed to determine the effects of socio-demographic characteristics and the peer groups on the eating attitude and body mass index (BMI) of students at a medical school in Turkey.

Methods: This cross-sectional study was conducted with the participation of the students of Gulhane Military Medical Academy (GMMA). Ethical permissions of the study were obtained from the GMMA Ethics Committee. The target population of the study consisted of 703 students, 533 of whom (75.8%) agreed to participate in the study. The Eating Attitudes Test (EAT) was administered to the participants and their BMI was noted. The EAT consists of 10 questions that measure the socio-demographic characteristics of the participants and 40 questions that evaluate eating habits. In order to determine the peer groups of the students, each student was asked to provide the numbers of their three closest friends. Three peer groups were generated for each grade by applying cluster analysis and as a result 18 peer clusters were examined.

Results: In this study, the average EAT score was 12.5 ± 6.9, and the mean BMI was 23.1 ± 2.4. It was found that the EAT score of 2.4% of the students was equal to or exceeding 30; 0.4% were obese; 21.0% were overweight; and 2.1% were slim. There was a significant difference between the grade level of the students and sport habits (p values respectively; p < .001, p = .015) in terms of the comparison of the EAT score to socio-demographic characteristics. In the analysis of variance between 18 clusters generated according to the cluster analysis, a statistically significant difference was found in terms of both BMI and the EAT (p values <.001, <.001, respectively). This suggests that students with similar eating habits and similar BMI levels have a tendency to cluster among similar peer groups. The variables that effect the EAT scores and BMI levels of the students were evaluated by the analysis of covariance. It was found that students’ smoking status (p = .039) had a statistically significant effect on BMI after it was adjusted according to peer group and grade. Also, it was found that the grade (p = .011) and peer cluster (p = .021) had a statistically significant effect on eating habits.

Conclusions: The peer groups may affect both eating attitudes and BMIs. In medical literature several studies exist that support these findings. But it is a novel approach to identify peer groups by using clustering algorithms and our study has been able to demonstrate the relationship of peer group and eating habits with this method.