Skip to content
Get in touch

The data model

Every value comes with context

On its own, “4.1” tells nobody anything. Potassium, 4.1 mmol/L, for this patient, on this stay, drawn at 08:15, from the central laboratory, final and uncorrected: that is something a clinician can act on and an application can rely on. monaOS attaches that context to every value before anyone reads it, and the API returns the value with all of it.

attributes on every value, listed below
9
timestamps on every value: when it was true, and when it arrived
2
to one day: the resolution of the time-series API
1 min
values ever deleted: a wrong value is marked, never removed
0

Without context, a value is a liability.

A laboratory result on the wrong stay, a timestamp that records when the message arrived rather than when the blood was drawn, a value in the wrong unit that looks normal, an artefact that quietly drags a trend and a score. Every application that reads raw hospital data has to solve these again, and each mistake reaches a patient. monaOS settles them once, in the record, before any application sees the value.

What travels with every value

Seven things every value carries

From the moment it enters the platform to the moment an application reads it.

  1. 01Binding

    Patient and stay

    Whose value, and on which stay.

    Every value is bound to its patient and, within that patient's history, to the encounter it belongs to. A creatinine from an admission two years ago and one from this morning describe different situations; the record keeps them apart and lets a stay be read as a whole.

  2. 02Timeline

    Clinical time

    Plotted at the moment it was true, not the moment it arrived.

    A specimen drawn at 08:15 whose result arrives at 08:30 belongs at 08:15. Results that arrive out of order are placed where they belong on the timeline, so the trend shows the course in the order it happened.

  3. 03Normalisation

    Unit and meaning

    The unit is part of the value.

    The same analyte comes in different units from different laboratories, and a value in the wrong unit can look normal while being critical. Values carry their unit and their concept, and a manual correction converts units automatically rather than trusting a retyped number.

  4. 04Setting or measurement

    Device context

    What the ventilator was told and what it measured are different values.

    A ventilator reports its settings and its measured result: set against measured tidal volume, set against measured pressures, and the measured pulmonary mechanics. monaOS keeps them apart, because a setting read as a measurement changes the clinical meaning of the chart. Advanced haemodynamic values from specialised monitoring hardware are captured the same way. monaOS receives data from medical devices without controlling them.

  5. 05Nothing is deleted

    Entered in error

    A wrong value is marked, never deleted.

    A loose probe produces a saturation of 62 percent that does not reflect the patient. Nothing in the record is ever deleted: the value is marked as entered in error, with a clinical reason that is mandatory. From then on, trends, fluid balances, scores and reports leave it out. The value itself stays traceable: anyone reviewing the record can see it, who marked it, when and why.

  6. 06History

    Corrections

    A correction keeps both values, and says who made it.

    A manual override preserves the original and the corrected value. Every entry carries a version history: the field that changed, its value before and after, when, and the practitioner who changed it. Review mode opens the record of a discharged patient.

  7. 07Time-series API

    Resolution

    From one minute to one day, through the time-series API.

    High-frequency device data is kept at the resolution it was measured and served from a time-series store built for it. Through the time-series API, an application asks for minute-level detail in an acute situation or for hourly and daily overviews, without losing what happened in between.

Anatomy of one value

What the API returns for one laboratory result

Attribute by attribute, and why each one matters downstream.

Patient
Who the value belongs to. On the wrong patient, a correct value becomes a safety risk.
Encounter
The stay it belongs to, which separates this admission from the patient's history.
Observation
The concept and the kind of statement: a measurement, a device setting, a calculated score or an assessment.
Value and unit
4.1 mmol/L. The unit travels with the value and is converted explicitly rather than assumed.
Clinical time
08:15, when the specimen was drawn. This is where the value is plotted.
Arrival time
08:30, when the result reached the platform. Recorded, but not plotted.
Source
Central laboratory, results feed. Provenance for interpretation and for troubleshooting.
Status
Final, uncorrected. Invalidated values stay in the history and leave trends and scores.
History
Every change with the field, the value before and after, the time and the practitioner.

FAQ

Questions developers ask first

Do we get raw data or prepared data?

Prepared. Every value arrives bound to its patient and stay, with its unit, clinical time, source and status, and local codes already mapped to international code systems. Values marked as entered in error are flagged, so an application never has to guess.

How do we get high-frequency device data?

Through the time-series API, at the resolution an application needs, from one minute to one day. The clinical record keeps what was documented, with its version history.

Is data ever deleted?

No. A value that should not count is marked as entered in error, never removed. Calculations and reports leave it out, and it stays in the record for review, with who marked it, when and why.

Can we see what a value was before it was corrected?

Yes. The version history keeps every earlier value with the field-level change, the time and the practitioner, and a value marked as entered in error stays in the audit trail.

How are late results handled?

They are placed at the clinical time they describe rather than the time they arrived, so late or out-of-order results end up where they belong on the timeline.

Does a correction change the scores?

Yes. An invalidated value is excluded from trends, fluid balances and score calculations, so everything computed afterwards is computed on what was true.

Is the data standards-based?

Yes. monaOS is FHIR-native: the record is held as standard HL7 FHIR resources, and diagnoses carry ICD-10-GM and SNOMED CT codes.

See the API on your own data

Bring a stay you know well. We show you how each value got there, what it is bound to, what the API returns and what happens when somebody corrects it.