Abstract Background Inconsistent findings regarding health outcomes associated with Supplemental Nutrition Assistance Program (SNAP) use may be influenced by common biases in observational studies. We used negative control methods to evaluate the association between SNAP and cognition while adjusting for unmeasured bias. Methods This observational study analysed data from six waves (2008–18) of the Health and Retirement Study—a biennial cohort survey of US adults aged ≥51 years. The analytic sample included 11 942 community-dwelling participants without dementia. SNAP use and household income were self-reported. Global cognitive function (0–27) was assessed via telephone interview. Group differences in cognitive performance were estimated by using three approaches: multiple linear regression, negative control outcome (NCO), and double negative controls (DNCs). Results Linear regression models indicated that SNAP users had lower cognitive scores post-SNAP use, with the deficit increasing from –1.48 95% confidence interval (CI): –1.77, –1.18 in 2010 to –1.84 (95% CI: –2.24, –1.44) in 2018. On the contrary, both the NCO and DNC adjustment results indicated minimal associations between SNAP use and subsequent cognitive performance. NCO estimates ranged from 0.16 (95% CI: –0.23, 0.55) in 2010 to –0.04 (95% CI: –0.57, 0.50) in 2018. DNC estimates ranged from 0.22 (95% CI: –0.40, 0.84) in 2010 to –0.17 (95% CI: –1.06, 0.73). Sensitivity analyses produced similar results. Conclusion After common biases were accounted for by using negative controls, SNAP use had little association with cognitive performance. Valid estimates of potential health benefits or harms associated with SNAP may better inform policies and individual decision-making.
Lohman et al. (Thu,) studied this question.