Study

SNAP Benefit Cycles and Antiretroviral Therapy Adherence Among Adults With HIV

JAMA
August 5, 2026

AUTHORS

Aaron Richterman, Elizabeth F. Bair, Eliza W. Kinsey, Nicole Ellingson, Charlotte Babbin, Ola Mohamed, Bashir Fadulalmola, Christina A. Roberto, Dylan S. Small, Harsha Thirumurthy

Key Takeaways

12%

Drop in ART adherence by the end of each SNAP benefit cycle.

Adherence rises after SNAP payments land, then erodes as the month goes on — the first study to causally link SNAP timing to daily treatment behavior.

ART adherence climbed after participants received their SNAP payment each month.

Overview

Medication adherence is a primary determinant of chronic disease outcomes, yet remains suboptimal, particularly among low-income populations.1 Scarcity theory posits that economic hardship imposes cognitive burdens that affect decision-making, potentially undermining health behaviors like adherence.2,3 However, much of the existing research on scarcity relies on artificial exposures (eg, priming) and intermediate outcomes (eg, cognitive tasks), with limited replicability.4 Additionally, few studies have rigorously evaluated whether scarcity relief translates into changes in objectively measured health behaviors like daily medication adherence.

To address this gap, we evaluated whether short-term scarcity relief is associated with improved medication adherence. We leveraged 2 features that create a natural experiment within the Supplemental Nutrition Assistance Program (SNAP), which provides low-income households in the US with monthly cash-like benefits to buy food. First, in many localities the timing of SNAP benefit disbursement within the month is randomly assigned at enrollment. Second, there is a predictable cycle of decreased scarcity after each payment followed by recurrence at the end of the benefit cycle (hereafter SNAP cycle).5 Together, these features create a schedule analogous to randomizing beneficiaries to higher or lower scarcity on a given day.

This field is for validation purposes and should be left unchanged.