Stocking Hygeia's toolbox: methodological innovation in support of computational epidemiology

点击翻译按钮获取中文摘要

Despite great past achievements, Canada faces pressing threats to public health. While designing appropriate public health policies requires judicious choice of intervention strategies, such choice is complicated by the complex relationships between interventions and their effects -- raising the risk that even well-intended policies may prove ineffective or cause worse problems than they solve. To enable them to better grapple with these challenges, public health decision makers increasingly turn to agent-based simulation models (ABMs) for insight. By letting decision makers simulate the effects of policies in "virtual populations", such models can help inform the design of health interventions that are high leverage, robust, and cost-effective, facilitate quicker reaction to an infectious disease outbreak, aid in understanding of health trends, help prioritize data collection, and assist in communication with diverse stakeholders. Unfortunately, such models are frequently needlessly challenging to build, hard to understand, computationally expensive, and are difficult to share and collaboratively explore in a team environment. Frequently there is also insufficient observed data on individual behavior to be confident about the fidelity of the virtual populations to the real-world population. Within our proposed work, we seek to facilitate the design and construction of ABMs that inform understanding of population health trends and health policy tradeoffs. Specifically, we seek to make such models easier to use and faster to simulate, thereby opening up more time to carefully investigate policy impacts. We will also aid the construction of models that are better designed and more widely shared and understood by building software to improve team access to models and model results, and by supporting clearer ways of describing models. We will provide smartphone applications to collect rich data to inform models' representation of peoples' behaviour, and to permit cross-checking model results against observed data. Finally, to help improve the ABM accuracy, we will provide ways to automatically correct inaccurate models.