Keith Deutsch, IMI's co-founder and CTO, explains why healthcare's information architecture has a missing layer, why continuous monitoring makes the gap impossible to ignore, and what we are building at IMI to fill it.
The idea at the centre of the talk
Understanding a patient's condition increasingly depends not only on individual observations, but on the evolving relationships among observations, treatments, interpretations, and time.
Imaging, labs and monitoring answer specific questions. The EHR keeps the conclusions. Keith names what neither was designed to keep, and why it matters for every patient with more than one condition.
An AV block, a pacemaker, a beta blocker, a cardiac MRI, an ED visit for dizziness, occasional palpitations. Six accurate entries. The patient behind them is Keith himself.
Watch them grow into a living network of tentative signals, caregiver notes, device trends and clinician hypotheses. It is the clearest picture of the context layer we have.
Observation, analysis, conclusion assumes a bounded episode. Wearables, implantables and caregiver reports never close the episode. Conclusions become working models instead of endpoints.
Personalisation inside one app is not a clinical patient model. Keith sets out the stricter burden: multi-source, provenance-aware, uncertainty-aware, and led by clinicians.
AI trained on the settled record can automate documentation and workflow. The closing section explains what clinical reasoning needs instead, and why AI is also what makes that possible.
Keith Deutsch
Co-Founder & CTO, IMI · LinkedIn
MIT-trained engineer with 16 patents and 30+ years building frontier technology in IoT and health tech. He originated the "Active Context Graph" architecture for GE's Predix platform, the lineage behind IMI's context layer. After his own pacemaker and a cardiac MRI whose findings had nowhere to live, he set out to build the layer this talk describes.
Book time to talk through the architecture, or read how IMI's context layer works in practice. More talks and papers will be added to Knowledge as we publish them.
