The Risk Factor Analysis for Poor Sleep Quality in Taiwanese Adolescents and the Effect of Sleep Disturbance on Body Mass Index

Background: Sleep and obesity are both major issues in public health which affect wellness of adolescence. The association between them had drawn much attention in recent years. Many previous studies had demonstrated a positive association between short sleep duration and adolescent overweight. However, the relationship between sleep quality and adolescent overweight is inconsistent. Besides, no domestic study had ever been conducted about this issue. Therefore we analyze the representative data from Taiwanese teenagers by employing propensity score matching （PSM）to estimate the effect of sleep disturbance on body mass index (BMI).
Methods: We adopted the database of the 2001 Survey of Social Development Trends–Health Security in Taiwan which was collected with each household as a unit. Poor sleep quality was categorized into difficulty initiating sleep (DIS), difficulty maintaining sleep (DMS) and non-restorative sleep (NRS) according to the results of different items of the Insomnia Self-Assessment Inventory. The BMIs were calculated from self-reported body heights and weights. We first employed multivariate logistic regression analysis with each of the 3 types of disturbed sleep and their aggregate as outcome variable separately to obtain the propensity scores. Then we estimate the average treatment effects of different types of sleep disturbance by applying propensity score matching.
Methods: The prevalence of overall sleep disturbance was 20.9% within this population. The prevalence of overweight and obesity was 8.0% and 7.4% respectively according to the criteria proposed by Department of Health in Taiwan. Under multivariate logistic regression, the significant risk factors of disturbed sleep were: (1) DIS: current habitual smoker (odds ratio: 4.346; 95% confidence interval: 1.980 – 9.539); (2) DMS: chief of household who works at night (OR: 2.335; 95%CI: 1.070 – 5.095); (3) NRS: current habitual smoker (OR: 2.913；95%CI: 1.864 – 4.552). After propensity score matching analysis, only DIS had significant average treatment effect on higher BMIs for those teenagers who suffered from DIS among the 3 types of disturbed sleep. None of the 3 kinds of sleep disturbance had significant average treatment effect for the whole population.
Conclusion: The study revealed that different types of sleep disturbance may have different risk factors and cause unequal effect on BMIs of the adolescence. For those teenagers who have difficulty initiating sleep, improving their sleep initiation may significantly lower their BMIs. This result implies that improving sleep quality may be a novel way to prevent adolescent overweight, at least for those who have disturbed sleep initiation.