JMIR Publications

ABSTRACT

Precision medicine involves a paradigm shift toward data-driven knowledge to support personalized clinical decisions. The term “digital twin” tends to refer to any digital representation of a patient. However, the resulting confusion scrambles the technical and regulatory requirements to develop precision medicine. Here, we clarify three key digital representations of patients and show how their distinctions could clarify the roadmap to implement digital twin-related technologies: (1) integrative digital records, which collect and integrate personal health data from heterogeneous raw health records; (2) synthetic patients, yielding shareable reference databases for secondary collective uses; and (3) medical digital twins, integrating patient data and shared resources into predictive computational models to support clinical decisions. We discuss their implementations’ current tensions and suggest how they could be associated within a data architecture to overcome these issues by dedicated transformations of data. Integrative digital records may be formatted according to various data models to favor data portability or actionable structuration for a specific analysis. Synthetic patients must balance privacy and utility for secondary collective usage to be reliable substitutes for real identifying records. We propose to restrict the term “medical digital twin” to personal instantiations of computational models. They have to balance the relevance to the patient’s unique case and empirical validation to be approved as credible decision support. The distinctions between these three key digital representations and the highlight of their tradeoffs enabled us to draw operational recommendations for the implementation of digital twin-related technologies in precision medicine. It should leverage the multiplicity of the types of digital representations of a patient to support a full array of actionable data usage.