Synopsis of the series 'Medical abandonment and clinical drift' (Articles 1 to 5.5)
This document synthesizes the complete arc of the series “Medical abandonment and clinical drift” (Articles 1 to 5.5), from the diagnosis of a healthcare system structurally unable to follow complex chronic patients, to the proposal of a predictive prevention infrastructure embodied by PREDICARE.
André is 74, multi-morbid, without a primary care physician, living far from urban centres. He does not die from a lack of technology but because he vanishes from the clinical follow-up radar — not economically interesting, too complex, too time-consuming, too engaging. He deteriorates in silence, invisible, until the acute decompensation when paramedics bring him to the emergency department — too late, too degraded, losing autonomy and quality of life. André does not disappear by chance: he disappears because the system was not designed for him.
Medical drift is not a bug — it is a systemic property. Physician demographics, fee-for-service, hospital-community fragmentation, clinician cognitive overload, and the “one problem = one specialty” logic manufacture invisibility. Complex patients are not poorly followed — they are unfollowable within the current framework.
The system performs planned maintenance where predictive maintenance is needed. In chronic diseases, degradation is slow, silent, subclinical. It occurs during 99 % of the time, in the patient’s daily life, outside consultations, outside hospitals, off-radar. Weak signals exist — they are simply scattered across data that nobody has the time or the tooling to connect.
The bottleneck is not technological but economic and institutional. Prevention costs more today than treatment — for the actor who intervenes. The savings accrue to the global payer, later. Result: nobody invests seriously in predictive prevention at population scale. Industry, energy, and transport have already made this shift. Healthcare has not. Not for lack of AI — for lack of architecture and incentives.
PREDICARE does not replace the physician nor acute medicine. It surfaces what is invisible: silent clinical degradation, risk trajectories outside clinical time, weak signals preceding decompensation. It enables continuous preventive medicine, capable of articulating with acute medicine: intervening earlier, avoiding preventable crises, consuming fewer heavy resources, and improving clinical outcomes.
Article 5.1 — Shifting the operational paradigm. Moving from acute medicine to predictive prevention does not mean monitoring everyone all the time. It means identifying, at population scale, who will decompensate, when and why, to prioritise human attention where it is genuinely needed. This requires dynamic risk stratification, non-decisional digital twins, and an anticipation rather than reaction logic.
Article 5.2 — Funding what avoids costs. Prevention is structurally penalised: it costs those who act and benefits others later. The proposal: fund predictive prevention as a collective performance infrastructure, share risk between public and private actors, and remunerate based on measured savings (avoided hospitalisations, clinical stability) rather than acts performed.
Article 5.3 — Governing the infrastructure. Mandatory interoperability, model auditability, data portability, solution reversibility. The challenge is not having the best AI but avoiding structural dependency on uncontrolled black boxes.
Article 5.4 — Industrialising without endangering the patient. Models are assistive, explainable, non-prescriptive. The clinician remains the author of the decision. The tool reduces cognitive load, secures follow-up, and detects weak signals. A clinical safety logic, not substitution.
Article 5.5 — Scaling without betrayal. The major risk is not technological failure but drift: generalised surveillance, over-alerting, loss of trust, bureaucratic inflation. Scaling must be progressive, evaluated, framed by ethical and operational safeguards, and focused on patients genuinely at risk.
This series does not advocate for “more AI in healthcare”. It advocates for a complete re-architecture of chronic disease management, aligning technology, economics, governance and clinical practice. Predictive prevention is not a futuristic promise. It is the only realistic path to prevent thousands of patients like André from silently disappearing until the one crisis too many. Article originally published on LinkedIn (in French).