An Evaluation and Dissemination Model for the Machine Learning Embedded
System Lifecycle in Clinical Practice Settings
Abstract
Machine learning (ML) algorithms are gaining popularity in clinical
practice settings due to their ability to process information in ways
that augment human reasoning. While tools that rely on output from ML
algorithms in the healthcare setting are appealing for their ability to
aid in clinical decision making and streamline workflows, their
implementation and effectiveness are not well documented. There is an
abundance of published ML literature that focuses on whether algorithms
can predict an outcome or predict it better than previous algorithms,
but a dearth of effort evaluating their implementation or impact on
patient outcomes. While developing and validating algorithms is an
important first step in research, comprehensive evaluation is needed
before application of ML algorithms in new settings is considered.
Evaluation should examine both the process of implementation and the
outcomes using a mix of qualitative and quantitative methods. This
commentary describes a model we developed to guide our institutional ML
evaluation efforts.