Article — Position paper · ○ Open access

From planned to predictive maintenance — Part 2: how and why it works

Series: 'Medical desertification and care wandering' — Article 3 (Part 2)/5

Jérôme Vetillard · · LinkedIn Pulse · 14 pages · 1 min read
🇫🇷 Lire en français ↓ Download PDF

Series “Medical abandonment and clinical drift” — Article 3, Part 2

In this second part, the clinical scenario of André’s digital twin concretely illustrates how predictive monitoring detects a cardiac decompensation trajectory three weeks before visible symptoms, and how early intervention avoids hospitalisation.

D-21 to D-15: weak signals converge

At D-21, rapid weight gain (+1.2 kg in 5 days) constitutes an early signal of hydrosodic retention. At D-18, abnormal glycaemic variability is correlated with this weight gain. At D-15, rising heart rate and deteriorating sleep form a convergent decompensation pattern, with probability estimated at 62 %.

D-13 to D-5: alert, intervention, stabilisation

At D-13, confirmed convergence crosses the critical threshold (78 %): a structured medical alert is transmitted to the attending physician. At D-12, a 20-minute teleconsultation supported by a longitudinal synthesis pre-structured by the digital twin enables immediate therapeutic adjustment. At D-5, parameter normalisation confirms clinical stabilisation. Result: decompensation avoided, hospitalisation avoided (€5–8k), autonomy preserved.

The evidence base

Recent meta-analyses converge: RPM significantly reduces heart failure hospitalisations, provided sensors are integrated into an active decision loop. The discriminating factor is not the sensor but clinical orchestration.

The scalability question

The decisive question is scalability. Human PSAD (ETAPES-type) is effective but costs approximately €650/patient/year — national deployment for 12 million chronic patients would represent €7.8 billion per year. The alternative lies in an AI companion app at approximately €80/patient/year, capable of capturing 60–70 % of the human PSAD value, with an estimated national ROI of 8.9×. The PREDICARE project, using Qualees/TweenMe’s Sentinelle IA, aims to demonstrate this scalability with initial results expected in Q4 2026. Article originally published on LinkedIn (in French).

Read the document