Savings in Medical Expenditures Associated with Reductions in Body Mass Index Among US Adults with Obesity, by Diabetes Status

BMI is equal to weight in kilograms divided by height in meters squared.

We also compared the results of the Hosmer–Lemeshow tests with those calculated from an alternative estimation method: ordinary least squares (OLS) of the log of medical expenditures. For men and women pooled, as well as each gender separately, the Gamma GLM with log link provided the best fit.

We obtained similar results using the BMIs of higher-parity biological children, but there was a higher response rate to questions about height and weight for older children and thus we maximized statistical power by using the eldest.

We do not report potential savings associated with weight loss starting from BMI above 45 kg/m² because the number of respondents in the sample who have BMI in that range is so small that marginal effects are imprecisely estimated. In particular, these individuals represent less than 2 % of the pooled sample of men and women with BMI <80 kg/m² and biological children between the ages of 11 and 20.

This approach uses a linear regression in the first stage, which is most appropriate when the endogenous and mis-measured regressor is continuous. While it is not uncommon to estimate IV models with a linear first stage when the endogenous regressor is discrete, the resulting coefficient estimate may be biased. Black et al. [33] show that if the discrete endogenous variable suffers only from nonclassical measurement error, then the true value of the coefficient will generally lie between the OLS and IV estimates in the case of univariate regression. Frazis and Loewenstein [34] show that if the variable is both endogeneous and mismeasured, then the true value of the coefficient lies within bounds applied to the IV estimate.

Specifically, we limited the sample to adults aged 20 and over, but the youngest adult with a child of at least 11 years of age is 24, so 24 is the minimum age of a respondent in our estimation sample.

We dropped 643 women with MEPS clinical classification codes from 177 to 196, indicating that they had a normal pregnancy or delivery, abortion, or pregnancy complication during the year.

The exact wording of the question in the MEPS priority conditions codebook is: “{Other than during pregnancy, (have/has)/(Have/Has)} (PERSON) ever been told by a doctor or other health professional that (PERSON) had diabetes or sugar diabetes?”.

Each of the lines in Fig. 1 is a smoothed curve that is fit through 14 data points plotted for BMI values ranging from 17.5 to 50 kg/m², at intervals of 2.5 BMI points.

The aggregate costs of adult obesity in the US were calculated as follows. Our estimates indicate that, in 2010, obesity raised medical care costs by $US44.3 billion among adults with biological children (i.e. those who constitute our IV sample). Under the (admittedly, strong) assumption that the effect of obesity in our subpopulation generalizes to the full non-institutionalized population of adults aged 18 and older, we scaled the costs in the subpopulation used to estimate our model up to the entire adult population by multiplying the subpopulation aggregate costs by the ratio of the US population of adults to the US population of adults with biological children, or $US44.3 billion *(233.7 million/32.8 million) = $315.8 billion. The population counts are derived from the MEPS sample and sample weights.