Safety signals
Sentinel
Sees risk before it escalates.
Sentinel reads the signals already in the record, changes in mood, sleep, engagement and observations, and surfaces them so a clinician can prioritise attention. It supports the decision. The clinician always makes it.
The problem
On the ward, the signal is there before the harm. It is just buried.
Inpatient mental health is where patients are least safe, and the warning signs are often present in notes and observation logs days before an incident. They are scattered across systems and shifts, where no single person can see the pattern. Hindsight finds it. By then it is too late.
How it works
From scattered signals to prioritised attention.
- 1
Aggregate
Sentinel brings together signals already in the record across notes, observations and time.
- 2
Surface and explain
Patterns worth a clinician's attention are surfaced, each with a plain explanation of why.
- 3
Clinician reviews
A named clinician reviews every signal. Sentinel never acts on its own.
- 4
Prioritise observation
The team decides how to prioritise observation and care.
A first look at Sentinel.
An illustrative design, not live software. Synthetic data, and a clinician stays in control.
Signals from the record, surfaced and explained, so a clinician can prioritise observation. Not a prediction of individual events.
Capabilities
What Sentinel does.
Signal aggregation
Brings together mood, sleep, engagement and observation changes across time.
Explainable by design
Every signal comes with a clear reason it was surfaced. No black box.
Observation prioritisation
Helps the team focus observation where attention is most needed.
Equity and bias monitoring
Performance is audited across ethnicity, age and language as a release criterion.
Human-in-the-loop
A clinician reviews and acts on every signal. Always.
Record-integrated
Works from the existing record rather than asking staff to enter more data.
Built safe
The guardrails, by design.
- Decision support, never decision
Sentinel surfaces and explains. A named clinician decides and acts, every time.
- No suicide prediction
We do not claim to predict individual events. The evidence does not support it, and we will not say it.
- Explainable, not a black box
Every signal carries a reason a clinician can interrogate.
- Audited for equity
Bias monitoring across groups is a condition of release, not an afterthought.
Regulatory position. Sentinel is being developed as MHRA Class IIa software as a medical device, under ISO 13485, with a clinical safety case to DCB0129 and DCB0160, a clinical safety officer, DPIA and DSPT. It is in development, not a registered or certified device, and makes no diagnostic or predictive claim.
The rest of the platform
Sentinel is one of four.
Each product stands alone and grows stronger connected to the others through a shared clinical intelligence layer.
See Sentinel in a live demo.
We walk verified NHS and independent-provider teams through a clickable demo of the platform. Tell us about your organisation and we will set one up.
This page describes software in development. It is not a medical device and not for clinical use. Product screens are illustrative concepts using synthetic data.