Determinants for successful deployment of clinical prediction models : a design science research in the Dutch healthcare sector

Abstract

Whereas the promises of (predictive) analytics in healthcare are clear and extensively reported, the executive practicalities are not. Mapping the factors that have a hand in the implementation and continuation (i.e. deployment) of such projects improves the execution of prediction models and hence improves diagnostic and prognostic healthcare for patients. This research takes a design science approach to create an artifact aimed at successful deployment of clinical prediction models (CPMs). Through a literature review, various factors that play a role in the deployment of CPMs are categorized. Interviews with an extensive expert panel lead to the development of the CRISP-DM Deployment Extension for CPMs. Next to opinions on the importance of each factor, new in-sights are collected on related topics. A case study at a Dutch hospital allows for the testing of the artifact. A gap analysis is conducted, leading to a practical advice in terms of successful deployment. The research concludes with a proposed deployment strategy and a list of eight recommendations that can be considered the determinants for successful deployment of clinical prediction models.