Containerized Model Engines for Operational Disease Forecasting

Operational infectious disease modeling efforts are important to public health response before, during, and after outbreak events. Recent collaborative initiatives have demonstrated the success of hubs, which openly solicit modeling output from contributors. Automation techniques can help improve timeliness of submissions to hubs. Furthermore, we suggest that containerization approaches may yield even more efficient and portable development and maintenance of model engines. Here we provide a case study in development of a containerized infectious disease modeling framework. We describe our successful implementation of a model engine as well as a custom designed, cloud-based orchestration architecture for near-term influenza forecasting. As a next step, we offer a perspective on how containerization techniques may be standardized for adoption in the outbreak analytics domain.