{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__medicine_healthcare","trajectory_id":"R","attempt_index":0,"archetype_slug":"invariant_mode_decomposition_design","domain_slug":"medicine_healthcare","decision":"CANDIDATE","problem_id":"ward_deterioration_hidden_coupled_trajectory","causal_lever_id":"unstable_joint_physiology_mode_review_trigger","proposal":{"problem":"HYPOTHESIS: On adult hospital wards, impending physiologic deterioration may appear first as a repeatable, weakly damped combination of vital-sign and laboratory changes while individual measurements and aggregate early-warning scores remain below escalation thresholds.","actors_substrate":["adult inpatients on general wards","bedside nurses and physicians","rapid-response clinicians","longitudinal vital signs, laboratory values, observations, treatments, and missingness indicators","clinical-governance and health-data teams"],"observable_state":"Time-aligned patient-state vectors show joint trajectories whose estimated short-horizon transition repeatedly amplifies a particular combination of variables; conventional thresholds have not yet fired.","consequence":"Delayed clinical review can permit reversible deterioration to progress toward emergency escalation, ICU transfer, or other serious harm.","affected_objective":"Increase timely, clinically useful review of deteriorating patients without unacceptable alert burden or autonomous treatment decisions.","structural_mapping":[{"archetype_element":"Coupled transformation","domain_realization":"A locally estimated transition from one ward-patient state vector to the next over a fixed observation interval.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Invariant directions","domain_realization":"Approximately preserved joint physiologic trajectory directions under the fitted local transition.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Scalar response","domain_realization":"Estimated persistence, damping, or growth of each trajectory mode.","claim_kind":"INFERENCE"},{"archetype_element":"Action-relevant dominance","domain_realization":"Modes selected by prospective outcome sensitivity and reliable separation, not magnitude alone.","claim_kind":"INFERENCE"},{"archetype_element":"Residual and drift governance","domain_realization":"Reconstruction error, subgroup error, mode rotation, and spectral separation determine whether the model remains usable.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Transformation Scope","status":"adapted","domain_realization":"Fixed-cadence, short-horizon transitions for adult ward patients within prespecified clinical regimes."},{"component":"State-Vector Definition","status":"direct","domain_realization":"Standardized vitals, selected labs, consciousness and urine observations, treatment inputs, time-since-measurement, and missingness indicators."},{"component":"Invariant Mode Basis","status":"adapted","domain_realization":"Locally invariant or approximately invariant directions of the estimated transition operator."},{"component":"Modal Gain Spectrum","status":"direct","domain_realization":"Per-mode eigenvalue magnitude and phase, with uncertainty intervals."},{"component":"Dominant Mode Selection Rule","status":"direct","domain_realization":"Retain modes only when stability, deterioration-outcome sensitivity, reproducibility, and residual criteria all pass prespecified thresholds."},{"component":"Stable/Unstable Mode Partition","status":"adapted","domain_realization":"Classify modes as damped, marginal, growing, or oscillatory within the validated window."},{"component":"Modal Intervention Map","status":"adapted","domain_realization":"Map elevated patient modal coordinates to a request for clinician review and display contributing original variables."},{"component":"Reconstruction Residual Check","status":"direct","domain_realization":"Measure held-out state and outcome-relevant residuals, including structured subgroup residuals."},{"component":"Mode Drift Monitor","status":"direct","domain_realization":"Track changes in mode direction, gain, ordering, residuals, and subgroup performance."},{"component":"Interpretation Scope Contract","status":"direct","domain_realization":"Document that modes are local predictive summaries, not diagnoses, physiologic causes, or treatment prescriptions."},{"component":"Mode-Coupling Register","status":"direct","domain_realization":"Record near-degenerate, non-orthogonal, treatment-coupled, and regime-dependent modes."},{"component":"Local Linearization Window","status":"direct","domain_realization":"Prespecified care settings, sampling cadence, prediction horizon, acuity range, and treatment conditions."},{"component":"Spectral Gap Threshold","status":"direct","domain_realization":"Bootstrap-supported minimum separation between retained and omitted modes; failure disables dominant-mode interpretation."}],"mechanism_dispositions":[{"slug":"eigendecomposition_workflow","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Decompose the explicit fitted local transition operator, with conditioning and uncertainty checks.","counterfactual_removal":"No invariant directions or growth factors would connect coupled state changes to the proposed lever."},{"slug":"modal_sensitivity_sweep","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Perturb modal coordinates in the fitted model and test association with deterioration outcomes and cross-mode effects.","counterfactual_removal":"Mode magnitude could be mistaken for clinical leverage, invalidating prioritization."},{"slug":"modal_stability_analysis","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Classify discrete-time modes within the validated local regime, carrying eigenvalue uncertainty.","counterfactual_removal":"The proposal could not distinguish transient variation from persistent or growing joint trajectories."},{"slug":"mode_shape_testing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Deliberate excitation of patients is inappropriate; passive clinical observations and held-out validation are used.","counterfactual_removal":"No change; it is not in the causal chain."},{"slug":"network_spectral_centrality_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"The state is a physiologic trajectory, not a node-importance problem.","counterfactual_removal":"No change; centrality rankings are unnecessary."},{"slug":"power_iteration_probe","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A dominant-only estimate would hide clinically relevant secondary or near-degenerate modes.","counterfactual_removal":"No change because the bounded operator is intended to receive a fuller decomposition."},{"slug":"principal_component_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"High variance is not equivalent to dynamic persistence, instability, or clinical consequence.","counterfactual_removal":"No change; covariance modes do not supply the claimed transition mechanism."},{"slug":"reduced_order_model","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"The first test needs a warning model, not a fast surrogate for a costly simulator.","counterfactual_removal":"No change to the proposed review trigger."},{"slug":"residual_reconstruction_test","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Evaluate out-of-sample reconstruction and outcome-relevant residual structure across patients and subgroups.","counterfactual_removal":"Omitted clinically important behavior could remain invisible while the modal model appears adequate."},{"slug":"singular_value_decomposition","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Singular vectors may diagnose amplification but are not invariant repeated-transition modes; conditioning is handled as a validity check.","counterfactual_removal":"No change to the invariant-mode causal claim."},{"slug":"spectral_decomposition_report","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Produce a clinician-facing scope contract, variable traceability, uncertainty statement, and coupling register.","counterfactual_removal":"The mathematics remains, but safe interpretation and governance are materially weakened."},{"slug":"spectral_gap_monitor","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Re-estimate mode separation, identity, residuals, and subgroup performance on scheduled recent-data windows.","counterfactual_removal":"The system could continue using a simplified basis after its separation and identity have failed."}],"causal_chain":["Estimate a bounded local transition from longitudinal patient states while representing treatment and observation processes.","Decompose it and identify reproducible joint modes with reliable persistence or growth estimates.","Retain only modes that predict the defined deterioration outcome and pass residual, coupling, gap, and subgroup checks.","When a patient's coordinate along a retained risk mode crosses a locked threshold, request clinician review and expose contributing measurements.","Earlier useful review may enable ordinary assessment and treatment before emergency escalation; this final benefit is HYPOTHESIS."],"baseline":"Usual ward monitoring with single-measurement thresholds and the institution's existing aggregate early-warning score and escalation workflow.","nearest_rival":"A calibrated nonlinear deterioration-risk model using the same longitudinal inputs but without invariant-mode structure or modal traceability.","authority_safety":{"affected_parties":["patients whose records are analyzed or whose care could later generate alerts","bedside staff receiving review requests","rapid-response services absorbing escalation workload"],"decision_authority":"Clinical governance, privacy oversight, and responsible ward leadership authorize data use and any later workflow change; licensed clinicians retain all patient-care authority.","authorized_first_step":"Run a locked retrospective validation followed, only if prespecified criteria pass, by an eight-week silent shadow test on one adult ward; generate no clinician-visible alerts or care changes.","excluded_actions":["autonomous diagnosis or treatment","withholding or delaying existing alerts","patient randomization before silent validation","deployment outside the validated population or regime","using modal coordinates as asserted biological causes"],"halt_rollback":"Stop the shadow test and invalidate the model version for excessive residuals, unstable modes or gaps, subgroup performance breach, data-pipeline error, or privacy incident; retain usual care as the unchanged rollback state."}},"negative_tests":{"strongest_counterevidence":"A well-calibrated nonlinear rival or existing score may equal or outperform the modal model, while fitted modes may vary across units, treatments, sampling patterns, or time and therefore lack a usable interpretation window.","analogy_break":"Human physiology is nonlinear, partially observed, treatment-responsive, and nonstationary; an estimated transition may reflect clinician actions or measurement practice rather than autonomous physiologic dynamics.","failure_condition":"No reproducible spectral separation, highly ill-conditioned modes, structured outcome-relevant residuals, or unacceptable subgroup error makes modal interpretation unsafe.","problem_falsifier":"On held-out data, impending deterioration is adequately captured by single-variable thresholds or shows no repeatable coupled temporal structure beyond measurement and treatment artifacts.","intervention_falsifier":"Reproducible coupled modes exist, but their review trigger does not improve prespecified warning utility—lead time and sensitivity at a fixed alert burden—over both usual monitoring and the nonlinear rival.","risks":["alert burden and workflow displacement","false reassurance from apparently stable modes","confounding by treatment and informative measurement","subgroup performance disparities","mode-label instability near degeneracy","privacy loss from longitudinal clinical data"]},"null_rationale":null,"classification":{"candidate_kind":"TESTABLE_CONJECTURE","prior_art_status":"UNSEARCHED","evidence_maturity":"HYPOTHESIS"},"revision_change_log":{"revision_kind":"ORIGINAL","prior_problem_id":null,"prior_causal_lever_id":null,"problem_changed":false,"causal_lever_changed":false,"conceptual_changes":[],"operational_changes":[],"repairs_addressed":[]},"confidence":0.78,"generator_notes":"Closed-book structural transfer. Empirical effectiveness, prevalence, and novelty were not established."}