From national infrastructure to ordinary practice, without losing equity, quality or purpose
Final instalment of the series, this article addresses the most political question: how to move a system from exception to norm, from project to rule, from innovation to implicit evidence — without betraying territorial equity, clinical safety, medical time or the human purpose of care. Scaling is not “deploying everywhere”. It means accepting that every generalisation creates a structural moral risk: what was progress can become a blind constraint.
Any generalised system changes status and ceases to be one tool among others to become an implicit framework. Three structural drifts recur. Silent obligation: what was optional progressively becomes the expected norm, deviation is no longer a freedom but an anomaly to justify. Blind standardisation: a rule designed for the majority can become oppressive for clinical, social or territorial minorities — the Francis Report (Mid Staffordshire, 2013) demonstrates how a compliance-oriented culture can coexist with severe care degradation. Forgotten purpose (Goodhart’s law): indicators are optimised, then what they represented is forgotten.
Generalisation cannot be a leap but a signposted trajectory with explicit stopping points. Phase 1 (2025–2027): controlled extension targeting 2 to 4 million patients, priority to fragile territories, no automatic extension without independent 18-month evaluation. Phase 2 (2028–2031): majority coverage of 6 to 9 million patients, frozen if territorial equity diverges. Phase 3 (2032–2035): structuring common law for 10 to 12 million patients, with real possibility of non-participation without implicit sanction. Generalisation is valid only if each phase meets its enforceable criteria.
Each phase is conditioned on four KPI families: clinical, organisational, economic and equity. Central principle: no family compensates for another’s failure. An ineffective, iatrogenic or inequitable programme must be frozen or stopped, even if popular. Territorial digital twins can make governance more legible by aggregating medico-economic indicators and enabling scenario exploration, without eliminating political choice but making it explicit.
A generalised system must never become coercive without saying so. The triple guarantee protects all three stakeholders. Clinician guarantee: enforceable clinical autonomy, the physician remains the decision author. Patient guarantee: genuine refusal, genuine understanding, status as subject rather than object of the system. Territorial guarantee: subsidiarity, right to slow down, right to raise alerts. The tension between generalisation and individual freedom is not a system defect — it is acknowledged and framed by democratic safeguards.
Premature generalisation: deployment before sufficient validation, producing incidents and systemic trust erosion. Inequitable generalisation: replicating the digital divide — already under-resourced territories compound physician absence with predictive tool absence. Oppressive generalisation: generalised surveillance, over-alerting, decision fatigue, loss of clinical meaning. Common thread: deferred but certain failure — damage appears not at deployment but through prolonged use.
A broad and durable coalition including clinicians, patients, territories and funders, whose legitimacy does not rest on technical expertise alone. Genuine social acceptability built through transparency and citizen oversight — the Care.data precedent (NHS UK, 2016, £150M abandoned) shows that no public interest justifies forcing through if social legitimacy is broken. Institutional capacity to say NO: system credibility rests on its ability to stop if conditions are not met.
The purpose is not to surveil but to prevent avoidable degradation. Not to optimise dashboards but life trajectories. Not to replace care but to protect its intelligence — so that clinical time becomes once again a time for interaction, understanding, shared decision-making and accompaniment. Three decision tests: the shutdown test (can the system be stopped within 72 hours without care disruption?), the equity test (are under-resourced territories prioritised rather than served last?), the democratic test (is the system prepared to stop in the face of irreducible opposition?).
This series is not solely about predictive prevention. It is about how a society chooses to articulate industrialisation and care, efficiency and equity, technology and human dignity. Scaling without betrayal does not mean refusing scale — it means accepting that, even at scale, not everything should be mechanised, and that the ultimate value of a healthcare system remains what it enables humans to do together. Article originally published on LinkedIn (in French).