Is childhood obesity predictable, and therefore preventable?

Louis Chislett

Statistics

3rd

Year of study:

Is it possible to create a set of early life screening variables for childhood obesity, in order to better predict those most likely to become overweight or obese (and therefore target interventions more fairly and efficiently)? The prevalence of childhood obesity has risen in recent years, in 1990 15% of children in Scotland at the Primary 1 assessment were at risk of being overweight, compared to 22.4% in 2017/2018. Alarmingly, there is increasing evidence of a growing association between deprivation and obesity in childhood. It is believed that targeting intervention towards disadvantaged members of society closes the health gap – however we need strong evidence in order to justify who to target. A linked dataset containing administrative data on children born in Scotland between September 2009 and February 2013 was used for analysis. Routinely collected screening variables were combined into models which could be used to predict those who would be overweight or obese at the P1 assessment. The predictive power of these models was good only when an earlier measure of the child’s BMI was used as a screening variable. This raises the question of whether obesity specific targeted interventions are useful, given how pervasive obesity is across all society, or whether more systemic changes need to be made.

I am a third year statistics student from London. I am currently undertaking a work placement year at the MRC Social and Public Health Sciences Unit in Glasgow as a research assistant, which is where I conducted my research. Through my work I have gained an appreciation of the need for robust statistical evidence in all of our work – and the need to explain what these results mean to a wider audience. My passions lie in the world of data science as a tool for elevating our understanding of public health, and I intend to pursue further study in this area.

Abstract

Bio