Part of the justification for enacting Medicare Part D was that by subsidizing the cost of prescription drugs the Medicare program would reduce its expenses on hospital costs for conditions for which drugs and other medical care are substitutable. The delivery of Part D coverage occurs through two different plan designs: stand-alone drug plans or integrated Medicare Advantage prescription drug plans (MAPDs). In theory, integrated MAPDs have incentives to internalize the substitutability between drug and non drug medical care by providing patients by providing more generous coverage for drugs that reduce hospital costs. MAPDs also have stronger incentives to avoid selecting patients with high medical costs through strategic formulary design. In this paper, we identify two sets of drugs, one containing drugs that causally lower non-drug medical costs and one containing drugs that are indicative of high medical costs. Using data on every Part D formulary between 2009-2011, we test the hypothesis that integrated drug and medical plans design Part D formularies strategically to reduce total health care spending and to select patients with lower expected medical costs. We find that integrated MAPDs more generously cover drugs that are likely to have high medical offsets, and that they less generously cover drugs that signal high health care costs, consistent with the selection hypothesis. We find that these effects are stronger in price-based measures of formulary design than in non-price forms, such as prior authorization requirements.