BMI and Health Status

The flaws of using body mass index as a measure for overall health status are apparent. It is a 19th century technique which ignores the distribution of both muscle mass and bone in the body. In addition, the relationship between and health status is likely to be non-monotonic, most likely quadratic.

However, the use of BMI in research does have some advantages. It is very quick and inexpensive to measure. In addition, it is plausible that BMI is a strong indicator of individual’s health preferences and behaviours. Also, BMI is a continuous metric. Therefore, if we choose to use BMI as a proxy for health status we do not have to constrict ourselves to discrete choice statistics when estimating the conditional distribution for ‘health’.

The way in which BMI is used in estimation strategies needs to be redefined. BMI’s definition of ‘overweight’ is outdated, and does not recognise that the population has become bigger, stronger and healthier in the last 150 years. Bone structures with greater density and increased muscle mass are not the same as body-fat increases. They are health promoting, not health deterring. Obesity is rising, and the negative health effects are undeniable. However, the shift from ‘normal’ BMI to ‘over-weight’ BMI and the negative health outcomes are dubious. For example, Romero-Corral et al. show how coronary deaths amongst ‘over-weight’ BMI cases are lower those individual’s defined as having ‘normal’ BMI.

So how should the eager researcher approach this issue? In my opinion, we should accept that the bounds defined by the BMI scale are now invalid and have no basis acting as a proxy for overall health. The mean of health and BMI has shifted in over the last century, however I would argue that the new mean indicates improvements in health – strongly supported by life-expectancy increases, height increases, etc. – and that it is the deviations away from this mean which give a more precise measure of overall health status. One estimation strategy which maintains the continuous properties of this metric would be to measure BMI in z-scores (deviations from the mean controlling for the size of the standard deviation) or the z-score squared.