OBJECTIVES: This research aims to examine whether minority and/or low-income patients are slower to stop use of a particular medical treatment when there is new evidence of safety risks. METHODS: We examine the demand response of 11 chronic medications that received first-time major safety warnings (including black box warnings) from the Food and Drug Administration (FDA) between July 2006 -- December 2009. We restrict the sample to medications that experienced decreased utilization post-warning and followed patients who were users of at least one of the 11 medications. Selected Medicare beneficiaries were 18 years or older (n = 549,645), drawn from a random 20% sample of Medicare Part D claims data. A pooled multivariate logistic regression was performed on the likelihood of these patients stopping use post warning (defined as having no claims of the medication between 180 -- 360 days post warnings), controlling for patient demographics, Medicare plan type, new user status (naive patients or not), medication and state fixed effects. RESULTS: Preliminary analyses suggest that Hispanics and blacks were more likely to stop taking a medication compared to whites (ORs=1.160 and 1.122, respectively, p<0.001). By contrast, Medicare beneficiaries eligible to receive medications at little or no cost, i.e. dual-eligibiles and those who received low income subsidies, were less likely to stop than non-subsidized beneficiaries (ORs = 0.790 and 0.792 respectively, p<0.001). Users who initiated their medications within 6 months before the safety warnings were the most likely to stop use in the post-warning period (OR = 3.759, p<0.001). CONCLUSIONS: The findings suggest that socio-economic status is associated with modestly longer delays in stopping drug therapy with recent medication safety warnings. Specifically, low-income patients who are subsidized beneficiaries are less likely to stop using medications that received safety warnings and therefore may be subject to greater harm from adverse drug effects.