Abstract
Background: Food insecurity may adversely affect cognitive function through pathways involving nutritional deficiencies, chronic stress, and comorbid health conditions, with potentially different effects across cognitive domains. Longitudinal evidence remains limited by time-varying confounding, and it is unclear whether Supplemental Food Assistance Program (SNAP) participation modifies these associations. Objectives: To examine the longitudinal association between food insecurity and cognitive function using marginal structural models (MSMs), and whether SNAP participation buffers these associations for overall cognition, episodic memory, and attention/mental processing. Methods: 30,641 adults aged ≥50 in the 1998-2020 Health and Retirement Study (HRS) contributed 156,066 person-year observations. MSMs with stabilized inverse probability of treatment weights were used to account for time-varying socioeconomic, health, and cognitive confounding affected by prior exposure. Weighted pooled linear regression models estimated marginal associations and interaction effects. Results: Moderate and high food insecurity were associated with lower overall cognition (b = -0.36 and -0.71, respectively; p < 0.001). Similar graded associations were observed for episodic memory (b = -0.22; -0.43) and attention/mental processing (b = -0.15; -0.28; all p < 0.001). SNAP participation significantly attenuated these associations across cognitive domains, with stronger buffering effects among non-Hispanic Black and Hispanic respondents. Effect sizes corresponded to differences equivalent to several years of cognitive aging. Conclusions: Food insecurity is associated with poorer cognitive function across multiple domains, while SNAP participation mitigates these associations. Despite limitations of observational data, these findings highlight the methodological value of MSMs and the potential role of food assistance programs in reducing cognitive health disparities in later life.