Methodology

Approach

The methodological grounding of the project is best understood as the intersection of three established research traditions: ecological interface design, evidence-anchored physiological modelling, and the "knowledge-first" paradigm increasingly recognised in clinical AI. The constraint-state-space formulation is what emerges when all three are applied simultaneously to the problem of bedside ICU decision support.

The clinical problem

The project began as a structured response to a recurring clinical problem: a single ICU patient generates more streams of monitored data than clinicians can integrate continuously and accurately, particularly under the cognitive load of overnight cover, multi-patient handover, or simultaneous deteriorations. The working memory available for integrating multiple variables in real time is bounded originally estimated at 7 by Miller (1956), with more recent estimates lowering this further. Modern ICU bedside data routinely and substantially exceeds those bounds.

Ecological interface design

The ecological-interface-design tradition was initiated by Rasmussen and Vicente in the early 1990s for high-stakes process-control environments such as nuclear power plant operation. The central EID claim is that interfaces should make the constraints of a given problem perceptually available. This means not just the data, but the structural relationships between data, are displayed in a form that the user can engage with directly.

Applied to critical care, a patient's clinical state can be represented as a position within a constraint space, where the axes are the physiological dimensions along which the patient is being supported. Polygon visualisation makes these constraints simultaneously visible. Worst-metric highlighting and zone bands map abstract physiological thresholds into a new visual paradigm. A 24-hour temporal loop view extends the static snapshot to a trajectory, allowing the clinician to perceive direction and rate of change as well as the current state.

CURRENT PRACTICE Hourly blood gas + ventilator readout BLOOD GAS VENTILATOR pH 7.28 ↓ PaCO₂ 58 mmHg ↑ PaO₂ 64 mmHg ↓ HCO₃⁻ 26 mmol/L Base excess +1.2 Lactate 1.8 mmol/L P/F ratio 98 ↓↓ SpO₂ 91% ↓ FiO₂ 0.65 PEEP 10 cmH₂O RR 22 /min VT 420 ml Ppeak 32 cmH₂O ↑ Pplat 28 cmH₂O ↑ ΔP 18 cmH₂O ↑ Compliance 23 ml/cmH₂O ↓ Resistance 12 cmH₂O/L/s Mech. power 18.6 J/min ↑ Vd/Vt 0.52 ↑ P(a-ET)CO₂ 8 mmHg ↑ 20 values per cycle ↺ Repeated every 1–2 hours throughout ICU admission → CONSTRAINT VIEW Ventilation Constraint Model MP ↑ P/F ↓↓ pH ✓ Mechanical Power Oxygenation Vent. Adequacy Within target Borderline Deranged Dominant constraint: immediately apparent Direction of change: readable from trajectory ↺ Updated on each blood gas cycle
The EID principle applied to the ICU: the same clinical state represented as discrete values (left) and as a constraint-space position (right)

The EID tradition is described in the foundational papers by Rasmussen and Vicente (1992) and Vicente (2002). See the References section for the full citations.

Evidence-anchored physiological modelling

Each constraint axis is grounded in published evidence: zone bands are calibrated against landmark trials and established physiological thresholds, for example:

  • Driving pressure cutoffs follow Amato et al. (2015)
  • Mechanical power values follow Serpa Neto et al. (2018) and Gattinoni's 2023 review
  • Tidal volume per kilogram aligns with ARDSnet (2000)
  • Vd/Vt is computed by the Enghoff modification of the Bohr equation

Suggestion rules generated by the model encode published differential-reasoning patterns rather than improvised heuristics. Reference material is provided alongside each suggestion and the basis for any boundary is modifiable as evidence evolves.

Knowledge-guided machine learning

The longer-term framing of the project draws on knowledge-guided machine learning (KGML) This is an approach to domain knowledge where established constraints are embedded directly into Machine Learning (MLO architecture, rather than leaving the model to derive all structure from data alone. This contrasts with the currently dominant paradigm in clinical AI: imitation learning from retrospective outcome-labelled datasets, in which models reproduce the decisions found to be associated with better outcomes in the training cohort. This approach intrinsically inherits the variation, biases, and dated practices of that cohort together with a fundamental opacity about how those decisions and suggestions were reached.

Case Example — AlphaFold2

The Google DeepMind project AlphaFold2 (Jumper et al., 2021) achieved a step-change in protein structure prediction by embedding evolutionary, physical, and geometric constraints directly into the neural network architecture. The model reasoned within an explicit domain-knowledge framework rather than from data alone. Protein folding constraints are mathematically precise and physically invariant, whereas ICU physiology is heterogeneous, measurement-noisy, and patient-specific. The constraints in the ICU are probabilistic thresholds anchored to landmark trial evidence rather than absolute physical laws but we believe the same reasoning applies: encode what is known explicitly within a validated model then let ML operate within that structure. This is what the project terms a constraint scaffold, the architectural principle being that domain knowledge should scaffold the model's reasoning structure. This novel approach is what the Critical Care State Space Project intends to explore in the future.

Bayesian networks

The formal language underlying this approach is Bayesian. Zone thresholds in the constraint model function as evidence-based priors calibrated against landmark trials rather than local convention. Clinical reasoning at the bedside is itself a process of updating beliefs as new data arrives. Bayesian networks are therefore the natural candidate architecture for the ML layer when that stage arrives: they encode prior knowledge explicitly, propagate uncertainty transparently and produce auditable outputs at each inferential step. This is precisely what is lost in imitation learning, and what the constraint scaffold is designed to preserve.

Trajectory

The project's first priority is to validate the constraint models as a synthesis display tool for human clinicians. This will be described in methodology papers followed by detailed analysis of the model's accuracy and usability on a retrospective set of real-world ICU data using the MIMIC-IV and eICU-CRD databases. The "scaffold-for-AI" framing becomes relevant when that validation evidence is in hand. At that stage, Bayesian network integration is the planned ML step: operating natively within the constraint scaffold, encoding uncertainty explicitly and preserving the full interpretability chain from prior evidence to posterior clinical recommendations.

CURRENT Constraint Scaffold Prototype development Methodology Paper Peer-reviewed publication Retrospective Validation MIMIC-IV / eICU-CRD Bayesian ML Layer Knowledge-guided augmentation
Project development roadmap — current stage highlighted

Principal investigator

Dr Sean Edwards

The project is led by Dr Sean Edwards, a consultant cardiac anaesthetist at University Hospitals Plymouth NHS Trust. The work is conducted alongside clinical practice and is informed by daily exposure to the cognitive integration problem it addresses.

Correspondence is welcomed from clinicians, methodologists, statisticians, human-factors researchers, and clinical-AI researchers whose work intersects with constraint-state-space modelling, ecological interface design, explainable AI in healthcare, or critical-care decision support.

Get in touch