Longitudinal Health Monitoring
See sustained change in each patient, with the evidence behind it.
A system that reads wearables, labs, self-reports and clinical records over time, compares each person to their own baseline, and brings clinicians the few cases worth reviewing.
What it is
Most patient data is checked one reading at a time, against a population range. A slow change that stays inside that range can go unseen, while one-off spikes bury clinicians in alerts.
This system looks at each person over time instead. It learns what is normal for them, watches for change that lasts, and raises a case only when related measures move together.
Detection is plain, deterministic rules. AI does the work rules cannot: reading free-text notes, writing summaries and answering questions about a record. Every output is checked against the source data, and a clinician makes every decision.
How it works
01
Every source, one record
Wearables report daily, labs arrive irregularly, self-reports come weekly, and clinical records come in FHIR R4. All of it lands in one longitudinal record, with units converted along the way.
02
Each person against their own normal
Recent weeks are compared to a personal baseline, not a population range, so a drift can matter even while values look in range. One-off spikes pass, and context such as travel or fasting can explain a change away.
03
Findings, not alerts
One metric moving raises a flag. A finding needs related metrics moving together in the same window, so the review queue stays short and each case has a reason behind it.
04
Notes become structured data, with a person in the loop
A model reads clinic notes and lab reports and proposes structured fields. Each field must quote the document word for word or it is withheld, and nothing enters the record until a clinician confirms it.
05
Ask the record
Clinicians ask questions in plain language and get answers drawn only from that person’s record. Every statement cites the records it relies on, and each citation opens the source.
06
Every result can be reproduced
Each detection run keeps its code version and configuration, and each model call keeps its full trace. Audit re-runs a result and checks it matches, and the whole system can be read as it stood on any past date.
07
Runs inside your environment
Models run locally, on your own hardware, so patient data does not leave the environment it lives in. That makes the system a fit for organisations where data cannot go to an outside API.
Who it’s for
Hospitals and clinics
Teams following patients over time who need a short, evidence-backed review queue.
Remote care programmes
Programmes with wearable data at a volume nobody can read reading by reading.
Insurers and health systems
Organisations with records and notes spread across systems that never meet.
Digital health teams
Product teams who want audited, validated AI in a regulated setting.
Bring this to your organisation
Tell us what data you have and who reviews it, and we will show you how this would work with your records.
Every organisation’s data and clinical workflow are different, so we adapt the system to each one. It supports clinical review and does not diagnose. It is not a medical device, and regulatory requirements depend on how it is deployed.
