SOFUS Outputs

This section describes the outputs produced by SOFUS.

SOFUS outputs are derived clinical products generated by interpreting inputs from Journl Streamed. Outputs are represented using standard FHIR resources and are intended to be consumed by downstream systems, workflows, and user interfaces.

SOFUS produces multiple types of outputs, each serving a distinct clinical and analytical purpose. The primary output types are:

• Status outputs (current state assessment)

• Prediction outputs (forward-looking risk estimation with explanations)

Output representation

All SOFUS outputs are represented as FHIR Observation resources.

Each output:

• Is time-stamped

• Is linked to a patient and encounter

• Contains a primary value

• May include structured components and extensions

Outputs are not raw observations and should not be treated as such.

Status outputs

Status outputs describe the current clinical state based on available inputs at a given point in time.

SOFA status

The SOFA status output represents the current SOFA assessment, including both the aggregated status and the underlying components used to derive it.

The output includes:

• A primary SOFA status value

• Structured components representing sub-scores and intermediate values

• Contextual metadata related to notifications and feedback

The SOFA status output is intended to support situational awareness and transparent inspection of how the current status was derived.

Example: SOFA status output

Clinical concept	FHIR value type	Example codes
SOFA status (overall)	Quantity	sofa
Circulatory sub-score	Quantity	sofa_kredsloeb_score
Systolic blood pressure (value)	Quantity	sofa_kredsloeb_systolisk_value
Systolic blood pressure (score)	Quantity	sofa_kredsloeb_systolisk_score
Mean arterial pressure (value)	Quantity	sofa_kredsloeb_map_value
Mean arterial pressure (score)	Quantity	sofa_kredsloeb_map_score
Respiratory sub-score	Quantity	sofa_respiration_score
PF ratio (value)	Quantity	sofa_respiration_PF_Ratio_value
PF ratio (score)	Quantity	sofa_respiration_PF_Ratio_score
PaO₂ (value)	Quantity	sofa_respiration_Pa02_value
PaO₂ (score)	Quantity	sofa_respiration_Pa02_score
Consciousness sub-score	Quantity	sofa_bevidsthed_score
Consciousness (value)	Quantity	sofa_bevidsthed_value

Notes:

• Status outputs include both scores and supporting values

• Components are included to support transparency and inspection

• Status outputs reflect the current state, not future risk

---

Prediction outputs

Prediction outputs describe forward-looking assessments based on patterns observed in the inputs.

Predictions are always accompanied by explanations that describe contributing factors and their relative influence.

SOFA prediction

The SOFA prediction output represents a predicted SOFA-related risk, expressed as a prediction score.

The output includes:

• A primary prediction value

• Explanation components (e.g. SHAP values)

• Raw transformed feature values used as input to the prediction

This structure supports traceability and interpretability without exposing internal model logic.

Example: SOFA prediction output

Clinical concept	FHIR value type	Example codes
SOFA prediction score	Quantity	sofa-prediction
Explanation – age	Quantity	Alder
Explanation – biological sex	Quantity	Mand?
Explanation – respiration rate	Quantity	Respirations_frekvens_Value_time
Explanation – pulse	Quantity	Puls_Value_time
Explanation – oxygen saturation	Quantity	Saturation_i_%_Value_time
Explanation – oxygen flow	Quantity	Iltbehandling_i_l/min_(Flow_i_l/min)_Value_time
Explanation – temperature	Quantity	Temperatur_Value_time
Explanation – laboratory values	Quantity	e.g. P-Kreatinin_Value_time, P-Bilirubiner_Value_time
Raw transformed input – vital signs	Quantity	Puls, Respirations_frekvens, Temperatur
Raw transformed input – laboratory	Quantity	P-Kreatinin, P-Karbamid, P-Laktat

Notes:

• Prediction outputs always include explanations

• Explanation values indicate relative contribution, not causality

• Raw transformed inputs are included to support audit and inspection

Differences between status and prediction outputs

Aspect	Status output	Prediction output
Purpose	Describe current state	Estimate future risk
Temporal focus	Present	Forward-looking
Primary value	Status score	Prediction score
Components	Sub-scores and values	Explanations and raw inputs
Explanations	Implicit (via components)	Explicit (e.g. SHAP values)

Output responsibility and use

• SOFUS outputs are decision-support artifacts, not clinical decisions

• Outputs are intended to be interpreted by downstream systems and clinicians

• Clinical responsibility remains with healthcare professionals