Nursing workflows have a need to communicate patient acuity during shift changes. The nursing staff at our institution developed a point based tool to identify Covid-19 patients at risk of deterioration. The predictive analytics unit at our institution partnered with the nursing team to enhance prediction performance.
This talk will lightly address model development but instead we will focus on social-technical issues. Specifically we will discuss connecting with stakeholders at scale through model fact cards, education, integration, explainability, monitoring, and evaluation. These strategies are particularly important to consider for any machine learning model that must operate in concert with humans to improve decision making.