Related Experiment Videos
Methods of calculating deaths attributable to obesity
Katherine M Flegal1, Barry I Graubard, David F Williamson
1National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD 20782, USA. kflegal@cdc.gov
American Journal of Epidemiology
|August 3, 2004
Summary
Estimates of obesity-related deaths in the US may be inaccurate. Bias from incomplete confounding adjustment and differing populations can significantly overestimate obesity
Area of Science:
- Epidemiology
- Public Health
- Biostatistics
Background:
- Previous estimates of obesity-attributable deaths used methods with incomplete confounding adjustment and no allowance for effect modification.
- This limitation may introduce bias into the calculated mortality figures associated with obesity.
Purpose of the Study:
- To investigate the potential magnitude and direction of bias in obesity-related death estimations.
- To assess the impact of incomplete confounding adjustment and population differences on these estimates.
Main Methods:
- Utilized hypothetical examples based on 1991 US population data.
- Incorporated published relative risk data to simulate bias.
- Examined the effects of incomplete adjustment for confounding by age and sex.
- Assessed bias arising from differences between derivation and target populations.
Main Results:
- Incomplete confounding adjustment for age and sex resulted in a 17% overestimation of obesity-related deaths.
- Differences in population demographics (e.g., proportion of individuals ≥80 years) led to a 42% overestimation.
- Minor variations in relative risks between cohorts caused a 97% overestimation, nearly doubling the estimated deaths.
Conclusions:
- Obesity-related mortality estimates are sensitive to the accuracy of confounding adjustment and relative risk data.
- Failure to properly account for confounding and effect modification can lead to significant overestimation of deaths attributable to obesity.
- Accurate estimation requires careful consideration of population characteristics and robust statistical methods.