Date Published: January 25, 2017
Publisher: Public Library of Science
Author(s): Andrew Stokes, Samuel H. Preston, C. Mary Schooling.
The goal of this research was to identify the fraction of deaths attributable to diabetes in the United States.
We estimated population attributable fractions (PAF) for cohorts aged 30–84 who were surveyed in the National Health Interview Survey (NHIS) between 1997 and 2009 (N = 282,322) and in the National Health and Nutrition Examination Survey (NHANES) between 1999 and 2010 (N = 21,814). Cohort members were followed prospectively for mortality through 2011. We identified diabetes status using self-reported diagnoses in both NHIS and NHANES and using HbA1c in NHANES. Hazard ratios associated with diabetes were estimated using Cox model adjusted for age, sex, race/ethnicity, educational attainment, and smoking status.
We found a high degree of consistency between data sets and definitions of diabetes in the hazard ratios, estimates of diabetes prevalence, and estimates of the proportion of deaths attributable to diabetes. The proportion of deaths attributable to diabetes was estimated to be 11.5% using self-reports in NHIS, 11.7% using self-reports in NHANES, and 11.8% using HbA1c in NHANES. Among the sub-groups that we examined, the PAF was highest among obese persons at 19.4%. The proportion of deaths in which diabetes was assigned as the underlying cause of death (3.3–3.7%) severely understated the contribution of diabetes to mortality in the United States.
Diabetes may represent a more prominent factor in American mortality than is commonly appreciated, reinforcing the need for robust population-level interventions aimed at diabetes prevention and care.
The prevalence of diabetes has been rising rapidly throughout the world. Global age-standardized diabetes prevalence increased from an estimated 4.3% in 1980 to 9.0% in 2014 in men, and from 5.0% to 7.9% in women. The United States is no exception to this trend. Using combined criteria of self-reported diagnosis, fasting plasma glucose and hemoglobin A1c, the prevalence of diabetes among adults aged 20+ rose from 8.4% in 1988–94 to 12.1% in 2005–10.[2, 3] Trends are similar when HbA1c is the sole criterion.[4, 5] The prevalence of self-reported diagnoses rose very rapidly between 1990 and 2008 and slowly during the 1980’s and between 2008 and 2012.
We estimated the mortality consequences of diabetes in two nationally representative samples of US adults surveyed in the National Health and Nutrition Survey (NHANES) and in the National Health Interview Survey (NHIS). In both data sets, individuals were linked to deaths in the National Death Index through December 31, 2011, the last date to which the National Center for Health Statistics has performed this linkage. Although NHIS only provides self-reports of the presence of diabetes, it has the advantage of a much larger sample size, allowing us to examine how diabetes’ contribution to mortality varies with certain characteristics. NHANES contains data both on self-reported diabetes and on HbA1c levels, a preferred biomarker for the presence of diabetes. Drawing on both data sources provides a more comprehensive picture of the contribution of diabetes to deaths in the United States than using either source alone.
Table 1 shows the characteristics of individuals surveyed in NHANES and NHIS. The total sample size was more than 10 times larger in NHIS than in NHANES. The distribution of characteristics was very similar in the two data sources with the exception of BMI, which was based on self-reported weight and height in NHIS, whereas it was measured in NHANES. Consistent with a well-documented tendency for people to underestimate their weight and overestimate their height , the proportion obese was higher in NHANES (35.0%) than in NHIS (28.5%).
The study most comparable to ours used cohorts aged 30–74 who were surveyed in NHANES II between 1976 and 1980 and followed them into mortality statistics through 1992. (15) Using self-reported diabetes, this study reported a hazard ratio of 1.9, a prevalence at survey of 4.3%, and a PAF of 3.6%. Adding undiagnosed cases that were detected using fasting plasma glucose, the PAF increased to 5.1%. The main reason why the PAF values in the present study are much higher is that the prevalence of diabetes has risen sharply since 1976–80 (see also [28,29]).
To investigate the proportion of deaths attributable to diabetes, we used two independent data sets and two different criteria for identifying diabetes among individuals aged 30–84. We found a high degree of consistency in the resulting hazard ratios, estimates of diabetes prevalence, and estimates of the proportion of deaths attributable to diabetes. The proportion of deaths attributable to diabetes was estimated to be 11.5% using self-reports in NHIS, 11.7% using self-reports in NHANES, and 11.8% using HbA1c in NHANES. The proportion of deaths attributable to diabetes is much greater than the 3.3–3.7% of deaths in which diabetes is assigned as the underlying cause of death.