A model is a hypothesis.
Evidence is the test.

Why use biomathematical models?

Models provide a consistent way to examine how specified inputs and assumptions affect a defined output. They can help organise questions about duty patterns and cumulative exposure.

Their interpretation depends on what the output represents, the evidence supporting it and the setting in which it is used. A score is one input to a broader assessment of operational risk. [3]

What wearables may contribute.

Suitable devices can support repeated observations of sleep timing, movement and other available signals. They may provide a richer longitudinal record than retrospective recall alone.

Device estimates need scrutiny. Algorithms, firmware, sensor quality, adherence and the population being studied can affect results. Consumer outputs are best interpreted in the context of the measurement method used. [4]

A validation program built for scrutiny.

Prospective work records predictions before the observations used to evaluate them are available. The model version, input data and assumptions are frozen for each comparison.

1. Define the research question, outcomes, protocol and required approvals.
2. Retain prospective predictions and versioned input assumptions.
3. Collect suitable observations independently, with informed consent.
4. Align records in time and report completeness and measurement limitations.
5. Assess error, calibration and performance on held-out data.
6. Record changes separately and evaluate them on new evidence.

This approach makes performance, limitations and improvements inspectable rather than assumed.

Test the implementation. Evaluate the model.

Software testing confirms that the implementation behaves consistently for defined inputs. Model evaluation asks a different question: how well its indications compare with independent observations.

MAFRM’s evaluation program is designed around prospective comparisons, versioned inputs and evidence collected independently of the predictions. Until that work is complete, outputs are presented as model-based indications under stated assumptions.

Regulatory context.

CASA describes an FRMS as an actively managed operator process involving policy, procedures, hazard assessment, assurance and change management. CASA approval applies to an operator’s overall FRMS process. [5]

MAFRM is designed to support that broader fatigue-management process rather than replace it. Use of MAFRM does not by itself constitute or confer regulatory approval.

CASA guidance treats biomathematical models as supportive inputs that should be considered alongside operational data, experience and applicable requirements. [6]