Digital twins

In one line

A digital twin is software state that mirrors a physical entity in near-real time — last known values, desired values, and a model of behaviour — so systems can read, command, and simulate without touching the physical thing directly.

The idea

A physical thing — an infusion pump, a ward, a body — is hard to query directly: it may be offline, busy, or unsafe to experiment on. A digital twin is a persistent software model kept in sync by telemetry, so you interact with the copy: read its current state, command a change, replay its history, or simulate a scenario — none of which risks the real asset.

Reported state vs desired state

The twin holds two views, and reconciling them is the core mechanic:

• Reported (current) state — updated by incoming telemetry (often via MQTT): the last known values.
• Desired state — what an application wants (e.g. "set pump rate to X"); the platform reconciles it with the device when the device is reachable.

Because the twin persists, you can query a device that's currently offline, replay its history, and run what-if simulations against the model instead of the physical asset.

Levels of twin

• Device twin — every infusion pump's twin shows status, firmware, location (fleet view).
• Systems / operations twin — a ward or hospital: bed flow, OR scheduling, simulated under different staffing.
• Patient twin (research frontier) — a patient-specific physiological model for dosing and treatment planning.

Where it shows up in digital health

Device-fleet management; hospital operations modelling; and, increasingly, physiological twins in research. In this platform, the simulation bridge treats a training mannequin as exactly this: a twin whose vitals the scenario engine reads and writes — a safe, synthetic stand-in for a patient.

Common pitfalls

• Twin drift — if telemetry lags or gaps, the twin diverges from reality; trust degrades.
• Treating a model as the patient — a physiological twin is a model, with all a model's limits; clinical decisions need human judgement.
• Security of command paths — if writing "desired state" can actuate a device, that path is safety-critical and must be locked down.

Key takeaways

• A digital twin is a live virtual replica, synced by telemetry — read, command, replay, simulate without touching the real thing.
• It holds reported and desired state; reconciling them drives the device.
• Twins scale from a single device to a ward to (eventually) a patient.
• Powerful for fleets, operations, and simulation — but only as trustworthy as its telemetry.