What's in a number?

This Cow was uncomplainingly chewing her cud of chocolate croissant and marble cake at breakfast this morning when she read about the Health Promotion Board's National Healthy Lifestyle Campaign: Know Your BMI, Know Your Risks. HPB's press release reveals this campaign to be overwhelmingly about "obesity and overweight issues" and "weight management" and how BMI "is used as a measurement tool to classify a range of health risk categories". There's even a whole website devoted to this magic number.

There are many blogs and websites out there which explore the wider social context of, and experiences arising from the fixation of equating BMI to obesity to health or poorness thereof (Shapely Prose on our link sidebar is an excellent springboard). But this Cow fiddles around with numbers and statistical distributions for a good part of her waking hours, and something about the definition of the risk categories in the BMI, and the definition of obesity in general, strikes her as easy PR without proven improvements in public health outcome.

Definitions of obesity based on percentile measures of a population distribution of height and weight are just that: percentiles. Percentiles are cut-off points which capture an arbitrary proportion of a population that fall above or below the cut-off point. A statement of fact, "X% of people are above Y kg in weight" tells you only that. What makes me healthy or not is a confluence of what genes I was born with and how I live, not my ranking in a statistical distribution that has not been shown to be clearly and simplistically linked to health outcomes. Since health risks and outcomes are affected by more complex factors and interactions than simple measurements of height and weight or the precise mathematical relationship between them that BMI represents, such measures on their own can't tell you much about an individual's health. "Do You Believe in Fairies, Unicorns, or the BMI?" has a more detailed analysis, and is written by an Actual Mathematician(TM)!

Distributions are characterised not just by their mean, but by the variation about the mean. And when distributions measure some characteristics of a population, they change over time. Focusing on one soundbite-y measurement may be convenient for the HPB, but surely an event that promises to "come alive with activities and entertainment for employees and families" has space for ideas on how to be healthy other than trying as hard as you can to not be in that X% that weigh above Y kg?