Description of Session
The soaring levels of health inequity have prompted experts in various fields to come together and team up to think through innovative ways of providing targeted care. One such consideration is leveraging predictive analytics to determine which individuals are most at risk of negative health outcomes so that health workers can focus efforts on helping prevent those outcomes from occurring or to mitigate their impact when they do. In October 2018, Medic Mobile and Living Goods launched a pilot aimed to improve health worker efficiency and effectiveness by equipping a group of 67 health workers with tools driven by machine learning algorithms and statistical models. In this session, we will outline the design research approaches employed while building a predictive model from an analysis of historical population data, to the validation of data trends with health workers and beneficiaries, to prioritizing risks, to engaging users to provide feedback and input. After development and deployment of the model, a data-driven iteration approach was adopted to guide in refining and optimizing the model. The approach included quantitative analyses of those flagged as high-risk and data accumulated on interventions delivered during the pilot, as well as interviews conducted with beneficiaries and health workers to help us better understand the observed data trends and gather insights on their experiences with the model outputs. We shall highlight how health workers are delivering care to some vulnerable populations and how the built predictive models can be utilized to complement their work. We shall also discuss the important lessons learned and guide the group on how design research complements data efforts.
