Calorie Counter

You are currently viewing the message boards in:

BMI - agree or disagree?

Replies

• Posts: 7,048Member Member
I agree with its proper use which is as a statistical measure for large populations. I don't particularly think that applying rigidly to individuals makes much sense.
• Posts: 176Member Member
cdjs77 wrote: »
I'll post what I posted somewhere else as the "it's meant for populations" statement is a misinterpretation of statistics that really irks me:

People usually misinterpret what is meant when we say a statistical measure is meant for "populations." All statistical measures are meant to determine something about a population, but that doesn't mean they can't or aren't designed to assess risks on an individual level. When statisticians say something is meant to assess risks in a population, what they mean is they have taken a sample from a certain population of people in order to estimate a parameter which can then be applied to assess some probability for other individuals or groups in that population. For example, if we take a sample of university students and estimate their score on a calculus test based on how much they study for it, we have a parameter we can use to predict test scores for the population of university students based on the amount of time they study. This has two important points:

1. Because the sample came only from university students, we can only accurately apply this statistic to the population of university students. We don't know how this applies to other people. Maybe university students already have some sort of knowledge which helps them on the test that the general population doesn't. So studying for one extra hour as a university student may increase your score by 10 percentage points, but only 2 percentage points for those with a lower education level.
2. Just because we say this can only be applied to the population known as university students, doesn't mean we can't use it to assess outcome likelihoods for individuals. We can use it to estimate individual scores with varying degrees of accuracy, but only if the people come from our select population known as university students. Let's say our confidence level is 99%. We can use this to say that, with 99% confidence, Student A who studied for 2 hours will score 10 percentage points higher than Student B who studied for one hour.

BMI works similarly. It estimates parameters for a "population" based on sample parameters, but that doesn't mean it was only meant to assess risks for a population as a whole. It is meant to serve as a easy way to estimate whether or not someone has a healthy body fat percentage based on certain criteria. It is accurate to within whatever confidence level is chosen when estimating these parameters and it applies to whatever individuals come from the population we estimated it from (in this case, adults of European descent). Obviously, there is an error rate, so some people will not fall into the correct categories, but error rates for scientific assessments such as these are usually no greater than 5% total (2.5% at each tail, meaning 2.5% will be classified as overfat when they are not, and 2.5% will be classified as underfat when they are not). The other thing to note is that this estimation was done at a different time, so it's very likely the estimate for a healthy BMI has changed over time, but given the fact that the western world has increased it's calorie consumption and become less active in the past few decades, it's unlikely that BMI overestimates the number of people with an unhealthy body fat percentage and more likely that it underestimates it.

In Short: BMI is a fairly accurate predictor for the average person, and can be applied to assess risk at an individual level. It is also possible that BMI today underestimates body fat percentage categories.
Saying it is a predictor meant for populations is a misinterpretation of how statistics works and can be applied. We use statistics derived for population assessments all the time to assess individual risk, that's how insurance rates are often calculated, and, since insurers do not regularly declare bankruptcy, it's a pretty good sign that these estimates are pretty accurate. If you're not someone who has been an athlete for a few years, BMI is very likely to be an accurate picture of your health risks, and possibly even an underestimation.

Overall, it's an accurate predictor for most people. There is a degree of error as with any statistical prediction. Smokers have a higher risk of lung cancer, but of course not all smokers will get lung cancer. The error rate is also not particularly high. It is also one statistic and, as with any one statistic, it cannot be used as the end all be all for determining health. However if your BMI says you are overweight or obese, it's fairly likely that you are. Whether or not you will face the health consequences associated with this depends on a number of other factors.

Not quite. To continue your analogy, just because someone hasn't studied doesn't mean the teacher gets to fail them without grading the test. YOU HAVE TO GRADE THE TEST to find out the student's grade, whatever the predictive value of studying may be. Businesses and insurance companies are not doing that today, and they should. To be very specific in language, BMI was designed as a quick and dirty tool for identifying at risk people, derived from population statistics. A similar situation might be observing students studying, and students who don't study enough getting a call from the teacher saying you should maybe study if you hope to pass the test - the teacher might not know that an individual student already knows the material from previous exposure and will do just fine.

That's different from the current situation, getting a rise in insurance rates or a cut in your paycheck, based on a statistical prediction which may or may not apply to you.

I think you're reading more into what I wrote than I intended. BMI is useful for what it was intended for: identifying individuals at risk of having a high body fat percentage and the risks associated with it, nothing else. So in that, I agree with you. Being "at risk" doesn't mean anything other than "the chances are higher." Obviously, when someone is given a BMI in the unhealthy range, there needs to be further investigation: looking at activity level, diet, DEXA scan, or any other health indicators. I made no insinuation that a teacher should fail a student for not studying enough, nor did I say insurance companies should use BMI as the end-all-be-all of health metrics. It's expensive to do a lot of tests which indicate actual health risks, such as DEXA scans for body fat or various other tests, so BMI is a way to see if taking a further look is worth the effort.

My problem is that people misinterpret what is meant as a "population statistic." BMI was, and still is, a statistic meant to assess individuals even though it is "based off populations." People use this as a means to dismiss it, and if we could dismiss BMI for this reason, we could dismiss all statistical predictors. However, statistical predictors are actually pretty useful, even on an individual level.
The percentage of misclassified people, according to the most recent studies, was significantly higher than any 2.5. It's true that the majority of people fall where they should, but not so true that the alternative should be dismissed out of hand.

As I noted in my original post: "The other thing to note is that this estimation was done at a different time, so it's very likely the estimate for a healthy BMI has changed over time," so I'm not sure why you brought that up as an argument. Population trends change over time, the error rate today, given our changes in eating habits and activity level is certainly different than the one chosen at the time.

(Another point is that 2.5% is the error rate usually chosen for the tails. That means people will only be classified as overweight when they are not about 2.5% of the time, a type I error. You also have a type II error rate where people would be classified as healthy when they are not, which is somewhat inverse of the type I error, i.e. if you make the type I error rate smaller the type II get's larger. So I'm not sure what the study said but misclassifying more than 2.5% of people doesn't mean that the type I error rate is not 2.5%, but this is statistics I won't bore people with unless they really ask )
• Posts: 229Member Member
There's no agreeing or disagreeing, the BMI is a statistical tool for studying populations. It's not meant to be used to make judgements on individuals, except as a starting point which should be modified by observation of the individual.

This.

People who criticizes it just don't understand it.
I agree with its assessment of my fatness, and if it didn't accurately represent my fatness I would still agree with it as an average and would just consider myself an outlier. I chose a goal weight in the overweight BMI. I consider it a good goal for me, but it doesn't change the fact that I will be overweight likely by both BMI and body fat. If I cared to have a different body composition where I could be overweight but not overfat at that goal, I would still consider myself overweight by BMI (because by calculations I would be), but with the caveat that I'm not overfat.
saintor1 wrote: »
There's no agreeing or disagreeing, the BMI is a statistical tool for studying populations. It's not meant to be used to make judgements on individuals, except as a starting point which should be modified by observation of the individual.
This.

People who criticizes it just don't understand it.
The problem, at least in the US, is that some (especially insurance companies) are using it to make judgements on individuals.

For example: BMI 24.9? You pay X premium. BMI 25? You pay X premium plus a penalty for having a BMI that falls into the overweight range. That extra pound could cost hundreds of dollars a month.
• Posts: 10,631Member Member
CSARdiver wrote: »
Like most systems it provides a decent "first glance" metric. It's a population survey and thereby limited to the constraints of the survey.

The closer you lie to the median height and weight, the more accurate the analysis. I'm 6'4", so automatically on the outskirts of the study as I represent less than 2% of the population. My ego would like to discard this, but if I'm perfectly honest then this largely works for me. I'm currently at 215 and per BMI overweight, but I am carrying 10 lbs of fat I could stand to lose.

This is where I like models like the military uses - using BMI as an initial metric, then moving to more precise means of measurement.

My husband never passed on height/weight, but he always made tape. There definitely need to be additional criteria beyond BMI.

BMI works for me, but I'm a 5'4 female with an average figure. At the top of normal BMI I'm still overfat, to be honest.