Obesity Prevalence Among U.S. Adults During the COVID-19 Pandemic

Abstract

Introduction: As the COVID-19 pandemic unfolded, several studies collected small and relatively homogenous samples to track U.S. adult obesity rates and obesity-related risk factors. In this study, a much larger sample from a nationally representative survey was used to investigate changes in average BMI, obesity prevalence rates, and 4 obesity-related risk factors in the U.S. adult population during the COVID-19 pandemic.

Methods: Using a large nationally representative sample of U.S. adults aged ≥20 years from the 2011-2020 Behavioral Risk Factor Surveillance System, linear regression models estimated intra-pandemic changes in average BMI and obesity prevalence rates as well as 4 obesity-related risk factors.

Results: Relative to the 2019 to prepandemic 2020 period, significantly higher average BMI (+0.6%, p<0.05, N=3,555,865) and obesity prevalence rates (+3%, p<0.05, N=3,555,865) were observed among U.S. adults during the COVID-19 pandemic. Significantly higher rates of any exercise participation (+4.4%, p<0.01, N=3,607,272), average sleep hours in a 24-hour period (+1.5%, p<0.01, N=1,907,798), average alcoholic drink days in the past month (+2.7%, p<0.05, N=3,577,090), and lower rates of smoking at least some days (-4%, p<0.01, N=3,625,180) were also observed.

Conclusions: During the COVID-19 pandemic, U.S. adult obesity rates were higher and worsened the pre-existing epidemic of adult obesity in the U.S. Higher rates of alcohol consumption and lower smoking rates may have contributed to the higher rates of adult obesity in the U.S during the COVID-19 pandemic.

Figure 1

Changes in average BMI, obesity prevalence rates, and obesity-related risk factors during the COVID-19 pandemic. Note: Each marker shows the linear regression coefficient estimate—relative to the 2019 to March 12, 2020, period—associated with an indicator for the COVID-19 pandemic period from a separate linear regression model estimated by Ordinary Least Squares along with the corresponding 95% confidence band. To compute nationally representative estimates, the appropriate BRFSS sampling weights, strata, and primary sampling units were used in all regressions. Each coefficient estimate shown is statistically significant at the 5% level or better. Please see the notes in Table 1 for the sample sizes for the 6 separate regression models. For the continuous dependent variables, to obtain percentage changes, the formula 100 × exp (linear regression coefficient estimate) − 100 must be applied. To obtain percentage changes for the binary dependent variables, the linear regression coefficient estimate must be divided by the sample mean. Regressors included but not shown: age and its square, male dummy, race/ethnicity dummies, educational attainment dummies, annual household income dummies, marital status dummies, number of children under 18 years of age in the household dummies, indicators for survey years 2011, 2012, …, 2018, and indicators for the survey respondent's state of residence.