Correction: Actual Causes of Death in the United States, 2000Correction: Actual Causes of Death in the United States, 2000

To the Editor: In March 2004, we published an article¹ that showed that a large proportion of deaths each year in the United States result from modifiable lifestyle-related behaviors. We are writing to inform the JAMA readership that through an error in our computations, we overestimated the number of deaths caused by poor diet and physical inactivity. Our principal conclusions, however, remain unchanged: tobacco use and poor diet and physical inactivity contributed to the largest number of deaths, and the number of deaths related to poor diet and physical inactivity is increasing.

We and independent statisticians reviewed all of the calculations used to make estimates for the other causes of death. The review found some typographical and transcriptional errors in the published article, none of which affected the final estimates. These errors are listed in the accompanying correction. We will make available on request the corrected tables, a detailed list of the data sources, copies of the published articles from which data were taken, copies of the computer code and screen captures of output, and lists of International Classification of Diseases, Tenth Revision codes.

In our original analysis, we used the methods published by Allison and colleagues² to estimate the total number of US deaths in 1991 related to poor diet and physical inactivity by using hazard ratios for overweight and/or obese persons in 6 large cohort studies. We intended to update all of these cohorts by using vital statistics data for 2000³ and new prevalence estimates for obesity from the 1999 and 2000 National Health and Nutrition Examination Surveys,⁴ but the modifications were not copied properly from the first to the next 5 spreadsheet tables.

We have corrected these calculations. As before, to crudely address the temporal lag between onset of obesity and serious health consequences, we counted roughly half of the estimated increase in overweight-associated deaths. We rounded the average of 414 423 (for overweight-associated mortality in 2000) and 280 184 (Allison and colleagues’ estimate for 1991) to 350 000, and we again added 15 000 deaths to account for undernutrition and some obesity-independent effects of physical inactivity. The number of deaths in 2000 caused by poor diet and physical inactivity thus increased by approximately 65 000 (instead of the 100 000 increase we previously reported¹ ) from the 300 000 estimated by McGinnis and Foege for 1990⁵ and accounted for roughly 15.2% of the total number of deaths instead of the 16% previously reported.

Although we have corrected our computational error, several points are worth reemphasizing. Each condition discussed in our article has methodological issues that can affect the estimates. In particular, the method for deriving deaths attributed to poor diet and physical inactivity and their related conditions is less developed than that used to estimate deaths caused by tobacco use. To avoid potentially double-counting deaths attributable to obesity, we examined a conservative constellation of risks and conditions related to poor diet and physical inactivity. For example, although some deaths from hypertension are likely associated with obesity, we did not add those that might be unrelated to obesity but might be related to high-salt diets. Nor did we include deaths related to elevated cholesterol levels that might be due to high-fat diets but not associated with obesity. For the same reason, we also did not include deaths caused by lack of physical activity among adults of normal weight. This difficulty in attributing deaths to interlinked causes makes assessing the health burden of poor diet and physical inactivity challenging.⁶

We recognize that complex questions remain about effectively quantifying the actual causes of deaths and accurately relating them to modifiable lifestyle factors. Accordingly, at the Centers for Disease Control and Prevention’s request, the Institute of Medicine recently convened a meeting of experts with a diverse range of expertise and perspectives to further explore these areas of ongoing scientific debate. Understanding the burden of illness and how that burden can be linked to potentially modifiable risks and conditions is critical to understanding how individuals can protect their health and critical to guiding public policy choices when opportunities exist for intervention.