Doctor of Philosophy / Department of Computing and Information Sciences / John M. Hatcliff / Medical devices have traditionally been designed, built, and certified for use as monolithic units. A new vision of "Medical Application Platforms" (MAPs) is emerging that would enable compositional medical systems to be instantiated at the point of care from a collection of trusted components. This work details efforts to create a development environment for applications that run on these MAPs. The first contribution of this effort is a language and code generator that can be used to model and implement MAP applications. The language is a subset of the Architecture, Analysis and Design Language (AADL) that has been tailored to the platform-based environment of MAPs. Accompanying the language is software tooling that provides automated code generation targeting an existing MAP implementation. The second contribution is a new hazard analysis process called the Systematic Analysis of Faults and Errors (SAFE). SAFE is a modified version of the previously-existing System Theoretic Process Analysis (STPA), that has been made more rigorous, partially compositional, and easier. SAFE is not a replacement for STPA, however, rather it more effectively analyzes the hardware- and software-based elements of a full safety-critical system. SAFE has both manual and tool-assisted formats; the latter consists of AADL annotations that are designed to be used with the language subset from the first contribution. An automated report generator has also been implemented to accelerate the hazard analysis process. Third, this work examines how, independent of its place in the system hierarchy or the precise configuration of its environment, a component may contribute to the safety (or lack thereof) of an entire system. Based on this, we propose a reference model which generalizes notions of harm and the role of components in their environment so that they can be applied to components either in isolation or as part of a complete system. Connections between these formalisms and existing approaches for system composition and fault propagation are also established. This dissertation presents these contributions along with a review of relevant literature, evaluation of the SAFE process, and concludes with discussion of potential future work.