BACKGROUND A high body-mass index (BMI, the weight in kilograms divided by the square of the height in meters) is associated with increased mortality from cardiovascular disease and certain cancers, but the precise relationship between BMI and all-cause mortality remains uncertain. METHODS We used Cox regression to estimate hazard ratios and 95% confidence intervals for an association between BMI and all-cause mortality, adjusting for age, study, physical activity, alcohol consumption, education, and marital status in pooled data from 19 prospective studies encompassing 1.46 million white adults, 19 to 84 years of age (median, 58). RESULTS The median baseline BMI was 26.2. During a median follow-up period of 10 years (range, 5 to 28), 160,087 deaths were identified. Among healthy participants who never smoked, there was a J-shaped relationship between BMI and all-cause mortality. With a BMI of 22.5 to 24.9 as the reference category, hazard ratios among women were 1.47 (95 percent confidence interval [CI], 1.33 to 1.62) for a BMI of 15.0 to 18.4; 1.14 (95% CI, 1.07 to 1.22) for a BMI of 18.5 to 19.9; 1.00 (95% CI, 0.96 to 1.04) for a BMI of 20.0 to 22.4; 1.13 (95% CI, 1.09 to 1.17) for a BMI of 25.0 to 29.9; 1.44 (95% CI, 1.38 to 1.50) for a BMI of 30.0 to 34.9; 1.88 (95% CI, 1.77 to 2.00) for a BMI of 35.0 to 39.9; and 2.51 (95% CI, 2.30 to 2.73) for a BMI of 40.0 to 49.9. In general, the hazard ratios for the men were similar. Hazard ratios for a BMI below 20.0 were attenuated with longer-term follow-up. CONCLUSIONS In white adults, overweight and obesity (and possibly underweight) are associated with increased all-cause mortality. All-cause mortality is generally lowest with a BMI of 20.0 to 24.9.

Objective We review uses of electronic healthcare data algorithms, measures of their accuracy, and reasons for prioritizing one measure of accuracy over another. Study design and setting We use real studies to illustrate the variety of uses of automated healthcare data in epidemiologic and health services research. Hypothetical examples show the impact of different types of misclassification when algorithms are used to ascertain exposure and outcome. Results High algorithm sensitivity is important for reducing the costs and burdens associated with the use of a more accurate measurement tool, for enhancing study inclusiveness, and for ascertaining common exposures. High specificity is important for classifying outcomes. High positive predictive value is important for identifying a cohort of persons with a condition of interest but that need not be representative of or include everyone with that condition. Finally, a high negative predictive value is important for reducing the likelihood that study subjects have an exclusionary condition. Conclusion Epidemiologists must often prioritize one measure of accuracy over another when generating an algorithm for use in their study. We recommend researchers publish all tested algorithms—including those without acceptable accuracy levels—to help future studies refine and apply algorithms that are well-suited to their objectives.