Accelerating software development in digital therapeutics

Mobile phones drive innovation in digital therapeutics by supporting patients during treatments. In this talk we discuss Voluntis' approach to accelerating software development by integrating healthcare professionals directly into the development process while at the same time complying with quality and process constraints imposed by regulatory bodies. We discuss the motivation, the business case and the technical approach. We conclude by explaining why the approach can be used in other safety-critical domains as well.

Target Audience: Developers, Architects, Product Managers
Prerequisites: None
Level: Practicing

Extended Abstract
This talk is based on a two-year project between Voluntis and itemis to develop a new tool chain for developing medical companion apps. Essentially, these are applications that accompany patients during treatments by tracking side effects and recommending medications. At the heart of the apps are "algorithms" that essentially contain the relevant medical knowledge. Voluntis intends to develop a wide variety of such apps over the next years; to make this feasible, the process of developing the apps must be accelerated. This, in turn, requires increasing the confidence in the algorithms at the heart of the applications early in the development process, because finding errors after regulatory review is expensive, and finding them in the field is potentially dangerous for patients. Voluntis decided to use a domain-specific language (DSL) to "program" the algorithms that is suitable for use by healthcare professionals. The developed tool also allows healthcare professionals to simulate, experiment and test the algorithms, realizing the goal of detecting functional errors early and building confidence in the correctness of the algorithm.
Crucially, however, the algorithms designed and validated this way must be executed correctly on the target phone. Effectively, we have to verify the generation/interpretation infrastructure end-to-end. The challenge is to do this in a way where the associated efforts do not cancel out the benefits of using DSLs and models in the first place.
In the talk we discuss
* the context: what do medical companion apps do
* the business case: why development must be sped up
* the overall approach to modeling and execution
* the risk analysis we performed to inform the verification strategy
* the technical approach to verification (testing, coverage)
* and the results from using the approach in early applications.
Based on early experience with the approach, Voluntis has succeeded. The speedups are impressive, the errors are reduced and preliminary FDA approval has been achieved.
The talk is based on a detailed case study in chapter 6 of this SOSYM paper:
voelter.de/data/pub/MPS-in-Safety-1.0.pdf
We encourage you to take a look at this chapter; it contains all the details and concrete numbers that make this a well-researched and relevant case study.
The use of (non-qualified) DSLs and generators in safety-critical domains has generally been considered impractical because of the regulatory requirements. We consider the experiences made in this project a significant step in the direction of making this approach applicable to safety-critical software in general.