Patient digital twins: definition, uses and limits

The data involved depends on the use case. It can include medical reports, lab results, medical images, treatments, medical history, patient-reported symptoms or measurements from connected devices. An anatomical model built to prepare a surgery does not use the same information as a model estimating how cardiovascular risk may evolve.

Volume alone is not enough. Quality, recency, collection context and representativeness all matter. A single measurement can be accurate without reflecting a person’s usual state. Conversely, a series of measurements taken under comparable conditions can reveal a useful trend. The healthcare professional remains essential to connect the data to the clinical context.

In 2025, the European Union adopted the European Health Data Space Regulation, which will apply gradually. According to the European Commission, cross-border exchange of patient summaries and e-prescriptions should be operational in all Member States from 2029, with other categories, such as medical imaging and lab results, following later. This framework does not create digital twins by itself, but it aims to improve the availability and interoperability of the information that digital health uses depend on.

Building a longitudinal view

The first potential benefit is organizational: bringing scattered information together to follow how a person’s health evolves over time. This continuity can make it easier to prepare for a consultation and to talk with care teams. It does not, however, guarantee the quality of the record or identify a medical cause.

Representing anatomy

From a CT scan or an MRI, some tools produce a 3D model of an organ or anatomical region. Surgeons can use it to visualize structures and plan their approach. This kind of model is personalized, but it is not always dynamic or predictive in the strict sense.

Simulating scenarios

A model can compare several scenarios: the likely evolution of a parameter, the expected response to an intervention, or the possible effects of a lifestyle change. A simulation remains an estimate that depends on its assumptions, its data and the domain in which the model was validated. It is not an individual certainty.

Supporting personalized prevention

When several data points are tracked over time, the system can help spot a change or suggest a point of attention to discuss with a professional. Personalized prevention then means guiding attention and next steps, not announcing that a disease will occur.

Scientific reviews describe a field that is developing quickly but remains heterogeneous. Encouraging results concern specific conditions, organs and settings. They cannot be generalized to every person or every medical decision. The main limits relate to external validation, data quality and interoperability, bias, model transparency, security and integration into clinical practice.

Any information that could have medical consequences must be interpreted in context by a qualified professional.

The World Health Organization recommends that AI for health protect human autonomy, safety, transparency, accountability, equity and sustainability. These principles matter all the more when a model aggregates intimate data and produces estimates that can influence decisions.

MyTwin for patients is an app designed to bring health data and preventive and predictive technologies together in a personalized journey, with the option to talk to a healthcare professional. MyTwin does not replace healthcare professionals: each module and use should therefore be assessed according to its purpose, its level of evidence and the regulatory framework that applies to each integrated technology.

To explore the other journeys, see also MyTwin for clinicians and MyTwin Stories.